Energy Research and Development Division

FINAL PROJECT REPORT

Emissions, Indoor Air Quality Impacts, and Mitigation of Air Pollutants from Natural Gas Appliances

California Energy Commission Edmund G. Brown Jr., Governor

October 2017 | CEC-500-2017-034

PREPARED BY: Primary Author(s): Brett C. Singer William W. Delp Brennan D. Less David M. Lorenzetti Randy L. Maddalena Nasim A. Mullen Vi H. Rapp Lawrence Berkeley National Laboratory 1 Cyclotron Road Berkeley CA 94720-1710 Phone: 510-486-4779 | Fax: 510-486-5928 http://www.lbl.gov Contract Number: 500-09-042 Prepared for: California Energy Commission Yu Hou Contract Manager Aleecia Gutierrez Office Manager Energy Generation Research Office Laurie ten Hope Deputy Director ENERGY RESEARCH AND DEVELOPMENT DIVISION Robert P. Oglesby Executive Director

DISCLAIMER This report was prepared as the result of work sponsored by the California Energy Commission. It does not necessarily represent the views of the Energy Commission, its employees or the State of California. The Energy Commission, the State of California, its employees, contractors and subcontractors make no warranty, express or implied, and assume no legal liability for the information in this report; nor does any party represent that the uses of this information will not infringe upon privately owned rights. This report has not been approved or disapproved by the California Energy Commission nor has the California Energy Commission passed upon the accuracy or adequacy of the information in this report.

i

ACKNOWLEDGEMENTS

The primary funding source for the research described in this report was California Energy Commission Contract 500-09-042. Additional support was provided by the U.S. Department of Energy Building America Program via Contract DE-AC02-05CH11231, by the U.S. Department of Housing and Urban Development Office of Healthy Homes and Lead Hazard Control through Interagency Agreement I-PHI-01070; and by the U.S. Environmental Protection Agency Indoor Environments Division through Interagency Agreement DW-89-92322201-0. The research team also thanks A.O. Smith and Rheem companies for providing water heaters for testing.

The research team is grateful to the participants of the two field studies – the California Indoor Air Quality Study of 2011-2013 (Chapter 1) and the study of coincident exhaust fan use and spillage in homes with natural draft wall furnaces (Chapter 4) – and the people who allowed us to use their homes for the measurements of cooking burner pollutants described in Chapter 2. The team particularly appreciates the efforts of the participants of the 2011-13 study, who were so careful in their execution of the sampler deployment and packaging protocols, and so generous with their time answering survey questions and completing daily activity logs.

Technical support for the Indoor Air Quality Study of 2011-2013 was provided by Tosh Hotchi, who helped design and test the sampling materials, Colette Tse who contributed to sampling package preparation and sample analysis, and Jina Li who conducted statistical analysis.

Technical support for the study of time-resolved pollutant concentrations from cooking burners was provided by Tosh Hotchi and Marion Russell.

The field work for the study of backdrafting, spillage and coincident exhaust fan use in apartments with natural draft wall furnaces was conducted by Sebastian Cohn, Brian Finn and Andrew Brooks of the Association for Energy Affordability (AEA). The AEA team also contributed text to the reporting of that task.

The research team thanks Tosh Hotchi and Doug Sullivan for contributions to chamber set-up for range hood performance experiments and Marcella Barrios and Omsri Bharat for carefully conducting the cooking experiments.

Technical contributions to the study of pollutant emissions from advanced technology water heaters was provided by Doug Sullivan, Tosh Hotchi, Peter Grant, and Anna Liao, who assisted with installation, software setup, and data collection for several of the water heaters. Thanks also go to Robert Cheng, who shared laboratory space and collegially coordinated experimental schedules, and to Jim Lutz and Marla Mueller, who reviewed the complete technical report of that work.

ii

PREFACE

The California Energy Commission Energy Research and Development Division supports public interest energy research and development that will help improve the quality of life in California by bringing environmentally safe, affordable, and reliable energy services and products to the marketplace.

The Energy Research and Development Division conducts public interest research, development, and demonstration (RD&D) projects to benefit California.

The Energy Research and Development Division strives to conduct the most promising public interest energy research by partnering with RD&D entities, including individuals, businesses, utilities, and public or private research institutions.

Energy Research and Development Division funding efforts are focused on the following RD&D program areas:

• Buildings End-Use Energy Efficiency

• Energy Innovations Small Grants

• Energy-Related Environmental Research

• Energy Systems Integration

• Environmentally Preferred Advanced Generation

• Industrial/Agricultural/Water End-Use Energy Efficiency

• Renewable Energy Technologies

• Transportation

Emissions, Indoor Air Quality Impacts, and Mitigation of Air Pollutants from Natural Gas Appliances is the final report for the Healthy Homes – Exposure to Unvented Combustion Gases project (contract number 500-09-042) conducted by Lawrence Berkeley National Laboratory. The information from this project contributes to Energy Research and Development Division’s Energy-Related Environmental Research Program.

For more information about the Energy Research and Development Division, please visit the Energy Commission’s website at www.energy.ca.gov/research/ or contact the Energy Commission at 916-327-1551.

iii

ABSTRACT

This project advanced understanding of factors that impact exposures to air pollutants from unvented and improperly vented residential natural gas appliances. Researchers identified strategies to reduce exposures and health risks to Californians through the following tasks:

• A study of 352 California homes involving measurements of combustion-related air pollutants and surveys to characterize appliances and usage

• Intensive measurements of combustion pollutant concentrations from gas cooking burner use in nine homes

• An expert meeting that identified potential improvements to combustion appliance safety test protocols in California energy efficiency programs

• A study of backdrafting, spillage, and coincident exhaust fan use in sixteen apartments with natural draft wall furnaces identified as susceptible to improper venting

• A laboratory study to quantify the relative effectiveness of range hoods at capturing pollutants from cooking burners and fine particles from cooking

• Measuring the effect of gas quality on nitrogen oxide emissions from advanced technology residential water heaters

The 352-home study found time-averaged nitrogen oxides and nitrogen dioxide and the highest short-term carbon monoxide concentrations were higher in homes with natural gas versus electric cooking and increased with the amount of gas cooking. Elevated concentrations of nitrogen oxides and nitrogen dioxide were also found in homes with gas pilot burners on cooking appliances or floor or wall furnaces. In households that cooked 4 hours or more with gas, reported use of kitchen exhaust was associated with lower pollutant concentrations. Homes with gas appliances did not have higher formaldehyde or acetaldehyde concentrations. Moderate cooking burner use produced nitrogen dioxide concentrations that exceeded the 1-hour outdoor air quality standard in four out of nine homes and was more than half the standard in two others.

Keywords: Combustion pollutants, Carbon monoxide, Nitrogen dioxide, Kitchen ventilation, Residential indoor air quality

Please use the following citation for this report:

Singer, Brett C.; William W. Delp; Brennan D. Less; David M. Lorenzetti; Randy L. Maddalena; Nasim A. Mullen; Vi H. Rapp. (Lawrence Berkeley National Laboratory). 2016. Emissions, Indoor Air Quality Impacts, and Mitigations of Air Pollutants from Natural Gas Appliances. California Energy Commission. Publication number: CEC-500-2017-034.

iv

TABLE OF CONTENTS

Acknowledgements ................................................................................................................................... i

PREFACE………………………………………………………………………………………………….ii

ABSTRACT……………………………………………………………………………………………...iii

TABLE OF CONTENTS ......................................................................................................................... iv

LIST OF FIGURES ................................................................................................................................ viii

LIST OF TABLES ...................................................................................................................................... x

EXECUTIVE SUMMARY ........................................................................................................................ 1

Introduction ........................................................................................................................................ 1

Project Purpose ................................................................................................................................... 2

Project Process and Results ............................................................................................................... 2

Project Benefits ................................................................................................................................... 6

CHAPTER 1 : The California Healthy Homes Indoor Air Quality Study of 2011-2013 ............ 7

1.1 Introduction ................................................................................................................................ 7

1.2 Materials and Methods .............................................................................................................. 8

1.2.1 Participant Recruitment .................................................................................................... 8

1.2.2 Data Collection Instruments and Methods .................................................................... 9

1.2.3 Data analysis ..................................................................................................................... 11

1.3 Results and Discussion ............................................................................................................ 15

1.3.1 Demographics of Sample ................................................................................................ 15

1.3.2 Quality Assurance Results .............................................................................................. 16

1.3.3 Measured Pollutant Levels in Kitchen, Bedroom and Outdoors .............................. 17

1.3.4 Impact of Appliance Types on Indoor Pollutant Levels ............................................. 19

1.3.5 Impact of Pilot Burners on Indoor Pollutant Levels.................................................... 23

1.3.6 Impact of Cooking Burner Use on Indoor Pollutant Levels ....................................... 26

1.3.7 Impact of Kitchen Exhaust Ventilation on Indoor Pollutant Levels ......................... 29

1.4 Conclusions ............................................................................................................................... 32

1.5 Introduction .............................................................................................................................. 33

v

1.6 Methods ..................................................................................................................................... 34

1.6.1 Overview ........................................................................................................................... 34

1.6.2 Study Homes .................................................................................................................... 35

1.6.3 Cooking Burners............................................................................................................... 36

1.6.4 Range Hoods ..................................................................................................................... 36

1.6.5 Burner Operation and Simulated Cooking................................................................... 37

1.6.6 Execution of Experiments ............................................................................................... 38

1.6.7 Air Quality Measurements ............................................................................................. 39

1.6.8 Range Hood Performance Characterization ................................................................ 41

1.6.9 Attributing Concentration Profiles to Individual Cooking Events ........................... 43

1.6.10 Estimating Emission Factors from Ambient Concentrations ..................................... 45

1.7 Results ........................................................................................................................................ 46

1.7.1 Measured Range Hood Performance ............................................................................ 46

1.7.2 Experiments Conducted .................................................................................................. 47

1.7.3 Measured Pollutant Concentrations .............................................................................. 49

1.7.4 Time-Integrated Pollutant Concentrations under Base Conditions .......................... 53

1.7.5 Repeatability ..................................................................................................................... 57

1.7.6 Effect of Range Hood Use ............................................................................................... 58

1.7.7 Effect of FAU Use ............................................................................................................. 60

1.7.8 Spatial Variations ............................................................................................................. 62

1.7.9 Emission Factors ............................................................................................................... 64

1.8 Summary and Conclusions ..................................................................................................... 66

1.9 Recommendations .................................................................................................................... 68

CHAPTER 2: Workshop on California Combustion Appliance Safety Procedures ................. 69

2.1 Overview ................................................................................................................................... 69

2.2 Participants ............................................................................................................................... 69

2.3 Vision and Framework ............................................................................................................ 71

2.4 Issues and Recommendations ................................................................................................ 71

vi

2.5 Identfied Data and Research Needs ...................................................................................... 77

CHAPTER 3: A Field Study of Wall Furnace Venting and Coincident Exhaust Fan Usage in 16 Northern California Apartments ..................................................................................................... 80

3.1 Introduction .............................................................................................................................. 80

3.2 Methods ..................................................................................................................................... 84

3.2.1 Building and Participant Recruitment .......................................................................... 84

3.2.2 Occupant Survey and Daily Logs .................................................................................. 85

3.2.3 Guidance on Use of Windows and Exhaust Fans ........................................................ 86

3.2.4 Diagnostic Testing Protocols .......................................................................................... 86

3.2.5 Short-Term Monitoring Protocols .................................................................................. 88

3.2.6 Data Processing and Analysis ........................................................................................ 91

Stovetop Temperatures ........................................................................................................................... 91

Kitchen Range Hood Anemometer ....................................................................................................... 93

Bathroom Fan Motor Operation ............................................................................................................ 94

Furnace Flue Temperatures .................................................................................................................... 94

Identification of Furnace Cycles and Burner Operation ..................................................................... 95

Identification of Furnace Spillage .......................................................................................................... 97

3.2.7 Identification of Furnace Downdrafting ....................................................................... 99

3.3 Results ...................................................................................................................................... 104

3.3.1 Description of Apartment Units................................................................................... 104

3.3.2 Diagnostic Testing .......................................................................................................... 105

Envelope Airtightness ........................................................................................................................... 105

Exhaust Device Airflow Testing .......................................................................................................... 105

Step-Wise Depressurization and Spillage Testing ............................................................................ 106

Combustion Appliance Flue Carbon Monoxide Levels ................................................................... 107

3.3.3 Field Monitoring ............................................................................................................ 108

Outside Conditions ................................................................................................................................ 108

Living Space Conditions ....................................................................................................................... 109

Temperature and Relative Humidity .................................................................................................. 109

vii

Carbon Dioxide ...................................................................................................................................... 111

Carbon Monoxide .................................................................................................................................. 111

Summary Statistics for Equipment Use, Spillage, and Downdrafting ........................................... 111

Burner Cycles 112

Downdrafting 113

Spillage 114

Bathroom Exhaust Usage ...................................................................................................................... 115

Cooking and Kitchen Exhaust Fan Usage .......................................................................................... 116

Coincident Operation Assessments ..................................................................................................... 121

3.4 Discussion ............................................................................................................................... 124

3.5 Recommendations .................................................................................................................. 126

CHAPTER 4: Capture Efficiency of Cooking-Related Fine and Ultrafine Particles by Residential Exhaust Hoods ................................................................................................................. 128

4.1 Introduction ............................................................................................................................ 128

4.2 Materials and Methods .......................................................................................................... 129

4.2.1 Overview ......................................................................................................................... 129

4.2.2 Range Hoods ................................................................................................................... 129

4.2.3 Experimental Setup ........................................................................................................ 130

4.2.4 Airflow and Mixing Verification Experiments .......................................................... 132

4.2.5 Pollutant Measurements ............................................................................................... 132

4.2.6 Cooking Procedures ....................................................................................................... 133

4.2.7 CO2-Based Capture Efficiency ...................................................................................... 134

4.2.8 Experimental Schedule .................................................................................................. 134

4.2.9 Capture Efficiency Calculations ................................................................................... 135

4.2.10 Uncertainty in Calculated Capture Efficiencies ......................................................... 138

4.3 Results and Discussion .......................................................................................................... 138

4.3.1 Measured Concentrations ............................................................................................. 138

4.3.2 Capture Efficiency Results ............................................................................................ 142

4.3.3 Airflow and Mixing Verification Experiments .......................................................... 145

viii

4.3.4 Comparisons to Published Studies on Cooking Hood Effectiveness ..................... 146

4.4 Conclusions ............................................................................................................................. 146

CHAPTER 5: Effect of Fuel Wobbe Number on Pollutant Emissions from Advanced Technology Water Heaters .................................................................................................................. 147

5.1 Introduction ............................................................................................................................ 147

5.2 Methods ................................................................................................................................... 148

5.3 Results ...................................................................................................................................... 149

5.4 Conclusions ............................................................................................................................. 151

5.5 Chapter 7 Recommendations ............................................................................................... 152

GLOSSARY…………………………………………………………………………………………….153

REFERENCES ........................................................................................................................................ 156

APPENDIX A…………………………………………………………………………………………..A-1

LIST OF FIGURES

Figure 1: NOX and NO2 Measured in Kitchen and Bedroom, and Measured or Assigned Outdoor Concentrations, Ordered by Concentrations in Bedroom .................................................................. 13

Figure 2: Formaldehyde and Acetaldehyde Measured in Kitchens, Bedrooms, and Outdoors of Study Homes, Ordered by Bedroom Concentrations ......................................................................... 14

Figure 3: Indoor Pollutant Concentrations by Type(s) of Appliances Inside Home ...................... 21

Figure 4: Indoor Pollutant Concentrations by Cooktop (CT) Fuel and Presence of Pilot Burners on Cooktop or Furnace (F) ...................................................................................................................... 24

Figure 5: Indoor Pollutant Concentrations by Cooktop (CT) Fuel and respondent-Reported Total Cooking Time during Monitoring Period ............................................................................................ 27

Figure 6: Indoor Pollutant Concentrations by Kitchen Exhaust Fan Use During Cooking in Homes with Gas Cooktops and >4 h Cooking ..................................................................................... 30

Figure 7: Example of Pressure-Balanced Flow-Hood Airflow Measurement (H8) ........................ 41

Figure 8: Measurement of Range Hood Capture Efficiency in H8 ................................................... 42

Figure 9: Disentangling NO Data for Experiment H608 (Middle Peak) in the Kitchen of H6 ...... 44

Figure 10: Air Pollutant Concentrations Measured on First Day of Testing at House H3 under base conditions (no range hood or FAU operation) ............................................................................ 51

ix

Figure 11: Air Pollutant Concentrations Measured on First Day of Testing at House H5 under Base Conditions (No Range Hood or FAU Operation) ...................................................................... 52

Figure 12: Highest 1h Time-Integrated Concentrations in Kitchen and Bedroom Resulting From use of Natural Gas Burners in Simulated Cooking Activities ........................................................... 54

Figure 13: Highest 4h Time-Integrated Concentrations in Kitchen and Bedroom Resulting from Use of natural Gas Burners in Simulated Cooking Activities ............................................................ 55

Figure 14: Percent Reductions in Highest 1h Kitchen Concentrations Calculated by Comparing Experiments with Range Hood Use to Analogous Experiments without Range Hood Use ........ 60

Figure 15: Ratios of Highest 1h time-Integrated Concentrations in Kitchen and Bedroom .......... 63

Figure 16: Ratios of Time-Integrated Concentrations in Kitchen and Bedroom over 4h after Cooking Burner Use Commenced ......................................................................................................... 64

Figure 17: Emission Factors Calculated from Ratios of Highest 1h NO2, NOX, and PN to Highest 1h CO2 ........................................................................................................................................................ 65

Figure 18: Placement of Sensors Around and Inside a Wall Furnace ............................................... 90

Figure 19: Placement of Thermocouples at Draft Diverter and Vent ............................................... 90

Figure 20: Example of a Stovetop Burner Cycle, Using Differenced Time-Series to Identify Cycle Start and Stop Points, Apt A_12_1 ......................................................................................................... 92

Figure 21: Example of an Oven Burner Cycle, Using Differenced Time-Series to Identify Cycle Start and Stop Points, Apt A_12_1 ......................................................................................................... 92

Figure 22: Example of a More Complex Cooking Burner Event That Required Manual Editing of the Burner Index, Apt A_12_1 ................................................................................................................ 93

Figure 23: Characteristic Plot of Wall Furnace Temperatures in Apartment A_1_2 ...................... 95

Figure 24: Example of a Furnace Burner Cycle in Apartment A_1_2 ............................................... 97

Figure 25: Example of Unambiguous Wall Furnace Spillage Event during AEA Diagnostic Testing, Apt A_4_1................................................................................................................................... 98

Figure 26: Example of Disagreement between Furnace Temperature Sensor Data and Reports of Visually Identified Spillage by AEA during Diagnostic Testing in Apt B_15_1 ............................. 99

Figure 27: Wall Furnace Temperature Measurements Displaying Downdrafting in Apt B_15_1 .............................................................................................................................................. 100

Figure 28: Example of a Downdrafting Events Identified in Apt B_15_1 ...................................... 101

Figure 29: Example of Downdrafting Events Identified in Apt B_15_1 ......................................... 101

Figure 30: An Example of Bathroom Operation (Gray Bands) Impacting Draft Diverter and Vent Temperatures in Apt A_11_1, with Possible Implications for Spillage of Pilot Exhaust Gases .. 103

x

Figure 31: An example of Kitchen (Turquoise Shading) and Bathroom (Grey Shading) Exhaust Fan Operation Impacting Vent and Draft Diverter Temperatures in Apt A_7_2 ......................... 103

Figure 32: Example of Increasing Vent Temperature during Bath Fan Operation in Apt A_12_1 .............................................................................................................................................. 104

Figure 33: Use of Kitchen Exhaust Fan, Cooking and Coincident Use for Each Apartment ....... 120

Figure 34: Experimental Configuration of the Room Used to Measure Gas and Particle Phase Capture Efficiency .................................................................................................................................. 131

Figure 35: Concentrations of CO2 in the Hood Exhaust and Particles in Room Measured by the CPC and OPC Resulting from (a) Pan-Fry and (b) Stir-Fry Cooking Activities ........................... 139

Figure 36: Concentration or Particles ≥6 nm as Measured by the CPC from a Sampling Point Near the Exhaust Outlet of Room ........................................................................................................ 140

Figure 37: Capture Efficiencies (CE) Calculated Using CO2 Measured in Hood Exhaust and Particles Measured in Room for Pan-Frying Hamburger ................................................................ 143

Figure 38: Capture Efficiencies (CE) Calculated Using CO2 Measured in Hood Exhaust And Particles Measured in Room for Stir-Frying Green Beans ............................................................... 144

Figure 39: Capture Efficiencies (CE) Calculated Using CO2 and Particle Measurements for Stir Frying on the Front or Back Burners ................................................................................................... 145

LIST OF TABLES

Table 1: Summary of Pollutant and Environmental Monitoring Instruments Used in Study ...... 10

Table 2: Self-Reported Race and/or Ethnicity, Household Income, Highest Education Level, and Number of Residents Living in Households Included in this Study ............................................... 16

Table 3: Summary Statistics for Measured Pollutant Concentrationsa ............................................. 18

Table 4: Coefficient of Determination (R2) Between Pollutants Measured at Different Locations in Homes a,b ............................................................................................................................................... 19

Table 5: Sample Characteristics and Median Pollutant Concentrations (ppb; except CO) in Homes Grouped by the Type of Gas Appliance(s) in the Living Space........................................... 22

Table 6: Sample Characteristics and Median Pollutant Concentrations (ppb; except CO in ppm) in Homes Grouped by Presence of Pilot Light(S) in the Living Space ............................................. 25

Table 7: Sample Characteristics and Median Pollutant Concentrations in Homes Grouped by Cooking Fuel Type and Amount of Cooking During the Week of Sampling ................................. 28

xi

Table 8: Median Pollutant Concentrations in Homes That Cooked with Gas for More Than 4-h in during the Monitoring Period, Grouped by the Self-Reported Frequency of Kitchen Exhaust Fan Use ...................................................................................................................................................... 31

Table 9: Study Home Descriptions ........................................................................................................ 35

Table 10: Natural Gas Cooking Appliances in Study Homes ............................................................ 36

Table 11: Range Hoods in Study Homes .............................................................................................. 37

Table 12: Burner Operating Procedures Used for Most Experiments in Homes H2 to H9 ........... 38

Table 13: Dates and Periods That Homes Were Used for This Study .............................................. 39

Table 14: Description of IAQ Devices and Location Specific Packages ............................................ 40

Table 15: Measured Performance Parameters of Range Hoods in Study Homes ........................... 46

Table 16: Measured Capture Efficiency of Range Hoods in Study Homes ..................................... 46

Table 17: Experiments Conducted in study homes H1-H3 ................................................................ 47

Table 18: Experiments Conducted in Study Homes H4-H6 .............................................................. 48

Table 19.:Experiments Conducted in Study Homes H7-H9 .............................................................. 49

Table 20: Result for CO and PM2.5 in Experiments Meeting Criteria of CO rising at Least 9 ppm and PM2.5 Rising at Least 20 µg m-3 ........................................................................................................ 53

Table 21: Relative Deviations (RD) and Relative Standard Deviations (RSD) of highest 1h Time-Integrated Concentrations for Replicated Conditions ........................................................................ 57

Table 22: Relative Deviations (RD) and Relative Standard Deviations (RSD) of Highest 4h Time-Integrated Concentrations for Replicated Conditions ........................................................................ 58

Table 23: Highest 1h Concentrations in Paired Experiments to Investigate the Impact of Operating the Forced Air Unit (FAU) Mixing Fan Compared to Base Conditions ........................ 61

Table 24: LBNL CAS Testing Harmonization Stakeholder Meeting Participants List ................... 70

Table 25: Furnace Spillage Order of Test Conditions ......................................................................... 87

Table 26: Spillage Classifications ........................................................................................................... 88

Table 27: Monitoring Sensor and Data Acquisition Equipment Summary ..................................... 89

Table 28: Characteristics of Tested Apartment Units ........................................................................ 105

Table 29: Summary of Exhaust Fan Airflow Measurements and Maximum Installed Exhaust Capacity in Each Apartment ................................................................................................................ 106

Table 30: Summary of Wall Furnace Spillage Testing in Each Apartment Unit ........................... 108

xii

Table 31: Monitoring Start and End Dates (Year-Month-Day) for Each Apartment Unit, along with Mean Heating Degree-Days per Day for the Period using a 65°F Base Temperature ........ 109

Table 32: Summary Statistics of Measured Indoor Temperatures .................................................. 110

Table 33: Summary Statistics of Measured Indoor Relative Humidity .......................................... 110

Table 34: Summary Statistics of Measured 5-Minute CO2 Concentrations in the Main Living Room of Each Apartment ..................................................................................................................... 111

Table 35: Overall Summary Statistics for Activities Recorded across All Apartments ................ 112

Table 36: Summary Statistics for All Heating System Cycles in All Apartment Units ................ 113

Table 37: Summary Statistics for Downdrafting Cycles in All Apartment Units ......................... 114

Table 38: Summary Statistics for Wall Furnaces Spillage in Apartments with Any Spillage Identified ................................................................................................................................................. 115

Table 39: Summary of Bathroom Exhaust Fan Use Building A Apartments ................................. 116

Table 40: Summary of Kitchen Exhaust Fan Use in Each Apartment Unit .................................... 117

Table 41: Summary of All Cooking Burner Activities in Each Apartment Unit, Where at Least One Cooking Burner Was Being Used ................................................................................................ 118

Table 42: Summary of All Cooking Activities in Each Apartment Unit, as Characterized by Either Cooking Burner Operation or Kitchen Fan Operation .......................................................... 119

Table 43: Summary of Coincident Cooking Burner and Kitchen Fan Operation ......................... 120

Table 44: Summary of Coincident Operation of the Heating System, Cooking (Either Cooking Burners or Kitchen Exhaust Fan), and Bathroom Exhaust Fan ....................................................... 122

Table 45: Summary Statistics for Coincident Cooking+Bathfan Events in Each Apartment Unit (i.e., Worst-Case Depressurization Conditions) ................................................................................ 123

Table 46: Summary Results for Hamburger Pan-Frying Experiments1 .......................................... 141

Table 47: Summary Results for Green Bean Stir-Frying Experiments1 ........................................... 141

Table 48: Summary of Experiments and Results for Storage Water Heaters1, 2 ............................. 150

Table 49: Summary of Experiments And Results For On-Demand Water Heaters, Adjusted for Effect of Water Flow Rate1, 2 .................................................................................................................. 151

1

EXECUTIVE SUMMARY

Introduction The combustion products of natural gas burners include pollutants that can negatively impact indoor air quality when released into residences in substantial quantities. The California building code requires combustion products from most natural gas appliances be vented directly to the outside. The exception is cooking burners, which are addressed in the code through a requirement of exhaust ventilation over the cooktop or elsewhere in the kitchen. Unvented combustion-powered heating appliances (marketed as “vent-free”) are not legal for sale in California because of the hazard they present to indoor air quality.

Combustion of natural gas primarily produces carbon dioxide (CO2) and water vapor, but also can produce air pollutants. High temperature in the flame initiates reactions involving nitrogen and oxygen molecules in air to produce nitrogen oxides (NOX) including nitric oxide (NO) and nitrogen dioxide (NO2). Incomplete combustion can produce carbon monoxide (CO), formaldehyde (CH2O), and ultrafine particles. Nitrogen dioxide is a respiratory irritant that has been shown to exacerbate asthma in children among other effects. Carbon monoxide temporarily impacts the oxygen carrying capacity of blood, creating health risks for individuals with compromised cardio-pulmonary function and at very high concentrations, presenting a life-safety hazard. The United States Environmental Protection Agency set National Ambient Air Quality Standards (NAAQS) for CO and NO2 to protect the general population including sensitive sub-populations. The standards for CO are 35 parts per million (ppm) averaged over 1 hour, and 9 ppm over 8 hours. The standard for short-term exposure to NO2 is 0.1 ppm over 1 hour. Formaldehyde is a mucosal irritant and a human carcinogen. California’s Office of Environmental Health Hazard Assessment has a reference exposure level for formaldehyde of 7.3 parts per billion (ppb) for both 8 hours and long-term exposures for non-cancer health hazards. These reference level values are benchmarks for safe exposure; however, exceeding the NAAQS presents a hazard. Although there are currently no standards for ultra fine particles exposure, an expert elicitation review of the available literature suggested there is a medium to high likelihood of increased short-term ultra fine particles exposure causing adverse health effects.

In-home exposure to pollutants from natural gas appliances can result from three general scenarios:

• Improper or ineffective venting of exhaust gases from appliances required to be vented

• Using cooking burners without venting or with ineffective venting

• Using illegal vent-free heaters or fireplaces.

Improper venting refers to vent configurations that are inconsistent with current code due to design (such as incorrectly sized vent duct), bad installation, or degradation. Ineffective venting can result in some cases when exhaust fans operating coincident with the combustion appliance depressurize the house (create a negative pressure or vacuum) enough to impact the upward draft of the hot exhaust gases.

2

A suite of measurement-based studies conducted in California and elsewhere in the U.S. in the 1980s and early 1990s reported that NO2 concentrations were higher in homes with natural gas cooking appliances compared to homes with electric-only appliances. Higher NO2 was also found in homes having appliances with pilot burners and in some homes with a natural gas wall furnace or floor furnace. Prior to the field study described in this report, there had not been a large-scale study to measure combustion pollutant concentrations in California homes with natural gas appliances in more than two decades. During that time, there was a substantial decrease in the number of in-use appliances with pilot burners in addition to changes in cooking burner designs that could impact NO2 and also CO. As a result, there was an important gap in the knowledge about the concentrations of combustion pollutants that presently occur in California homes on both chronic (long-term) and acute (short-term) exposure time scales.

It was essential to understand those factors that impact indoor concentrations of combustion pollutants, including using kitchen exhaust ventilation and depressurization induced spillage of combustion products from natural draft appliances. A prior study conducted with the California Energy Commission’s Public Interest Energy Research program found range hoods installed in homes often do not achieve the airflow that is indicated on the product specifications. Furthemore, many capture a minority of the combustion gases emitted at the cooktop or in the oven (this study is decribed in a 2012 report and scientific paper by Singer and colleagues).

Depressurization results when one or more exhaust fans are operated in a home that is closed to the outdoors and the airflow of the exhaust fans substantially exceeds any supply fan airflow. The frequency and magnitude of depressurization are higher in homes that are more airtight (as occurs in energy efficiency retrofits) and as exhaust fan airflow rates increase. Substantial depressurization can interfere with buoyancy-driven venting of natural draft appliances, resulting in spillage of combustion gases and associated pollutants into the home.

In addition to potential indoor air quality concerns, NOX emissions from natural gas appliances are of concern as contributors to ambient ozone. Prior research funded by the Energy Commission examined the potential impacts of liquefied natural gas use on NOX emissions from appliances and found that on-demand (tankless) water heaters were the most susceptible (at the time) to increases in NOX emissions when using liquefied natural gas..

Project Purpose This project advanced the understanding of factors that contribute to indoor air pollutant exposures from unvented and improperly vented combustion appliances to reduce health risks to Californians. The study also examined the effect of gas quality on pollutant emissions from advanced technology residential water heaters.

These objectives correspond to component studies that were conducted as project tasks:

Project Process and Results Brief summaries are provided of the research methods, results, and conclusions from component studies conducted to achieve these technical goals. Recommendations arising from

3

all of the component studies are presented after the summaries. Published reports and papers that present more information are listed in the Appendix A.

1. The California Healthy Homes Indoor Air Quality Study of 2011-2013 (Objectives 1 and 2)

A study of combustion pollutant concentrations and their determinants in California homes was designed around a mail-out approach requiring substantial participant involvement. Pollutant concentrations were measured with sampling equipment that was mailed to participants’ homes with deployment instructions. Residence and household characteristics and activity data were collected via two phone surveys and an activity log.

Using these methods, data were collected inside and outside of 352 homes. Passive samplers measured time-resolved CO and time-integrated NOX, NO2, formaldehyde, and acetaldehyde over a period of about six days in each home during November 2011 – April 2012 and October 2012 – March 2013.

Researchers found that using natural gas cooking burners substantially increases the risk of elevated CO, and gas cooking and the presence of pilot burners on cooking and heating appliances within the living space are associated with elevated NOX and NO2. These findings are consistent with prior studies and demonstrate these indoor air quality challenges in California (and likely other U.S.) homes must be addressed. The impacts of pollutant emissions from unvented cooking and pilot burners are more pronounced in smaller homes.

2. Pollutant Concentrations and Emission Rates from Scripted Natural Gas Cooking Burner Use and Pilot Testing of Range Hood Benefits in Nine Northern California Homes (Objective 3)

The short-term indoor air quality impacts of using natural gas cooking burners were investigated through controlled experiments in nine residences in Northern California. Cooktop, oven, and broiler burners were operated in a prescribed manner intended to simulate use during typical cooking activities, while avoiding emissions associated with food preparation. Experiments were conducted in all homes under the base condition of closed windows, no use of the forced air heating system, and no mechanical exhaust. Additional experiments were conducted while operating a forced heating system and/or venting range hood. CO2, NO, NO2, and PN data from sequential experiments were analyzed to quantify the contribution of burner pollutants concentrations in each room.

In four of the nine homes, NO2 concentration in the kitchen after one hour of cooktop use exceed the national ambient air quality standard of 100 ppb. Two other homes had 1 hour NO2 exceed 50 ppb in the kitchen, and three had 1 hour NO2 above 50 ppb in the bedroom, suggesting substantial exposures to anyone at home when burners are used. Range hood use substantially reduced cooking burner pollutant concentrations both in the kitchen and bedroom of several homes.

Results generally confirm the finding of a 2014 simulation study by Logue et. al that using natural gas cooking burners without venting commonly produces short-term NO2

4

concentrations in homes that approach or exceed the federal ambient air quality standard of 100 ppb.

3. Workshop on California Combustion Appliance Safety Procedures (Objective 4)

This was the first of two component studies addressing combustion safety diagnostics. On February 2, 2015, LBNL convened a meeting of stakeholders and experts to discuss challenges and explore potential improvements to combustion appliance safety (CAS) assessment protocols used in California’s single-family energy retrofit programs. This meeting was 1) brought together the designers and implementers of CAS protocols for all of the major California single-family retrofit programs, 2) provided historical and technical background on the subject of CAS testing in retrofits, 3) initiated discussion current challenges and potential improvements to CAS protocols, and 4) identified research and data necessary to support protocol changes.

The meeting also discussed research needs and opportunities. Two ideas had broad support: 1) Simplifying test procedures and harmonizing protocols across programs, and 2) Basing safety protocols on risk management with clear safety objectives.

4. Field Study of Wall Furnace Venting and Coincident Exhaust Fan Usage in 16 Northern California Apartments (Objective 4)

This was the second component study to address combustion safety diagnostics (Objective 4). Researchers monitored the indoor environmental conditions, wall furnace operation, using exhaust fans and cooking activity over three-week periods in 16 affordable apartment units in Northern California. Almost all of the studied apartments failed a key component of the combustion appliance safety (CAS) test procedure used for energy retrofits in California — their natural draft wall furnaces spilled combustion pollutants to varying degrees under “worst-case” depressurization conditions of all exhaust fans operating on their highest settings. Since the apartments had only two exhaust fans, the worst-case condition was the kitchen exhaust fan on high and bath fan operating at the single available speed (or the high speed in apartments that had a two-speed bath fan). In many of the apartments, the wall furnaces also spilled combustion products with the kitchen fan on low and bath fan on, and some with just the kitchen fan on high or low. Residents were asked to use exhaust fans when bathing or cooking; and in units that spilled under natural test conditions, they were instructed to open windows during kitchen fan use. Using exhaust fans is considered best practice but does not represent ‘typical’ behavior.

Findings suggest that robust visual inspection and conducting combustion appliance tests under more realistic challenge conditions can identify truly hazardous installations. Results also support the idea that the worst-case test condition commonly used in CAS testing might be irrelevant to occupied residences even when exhaust fans are used as recommended (with all cooking and bathing) and certainly with how they are actually used in most residences. In the apartments measured in this study, it was quite rare for the both kitchen and bath fans to be operated together, at the worst-case depressurization condition used in CAS testing.

5

5. Capture efficiency of cooking-related fine and ultrafine particles by residential exhaust hoods. (Objective 5)

This laboratory experimental study assessed if combustion gases and particles from cooking are captured at the same rates by residential range hoods. If so, the same test method could be used to inform the public about the effectivess of range hoods for pollutants from both aspects of cooking.

Experiments involving actual and simulated cooking with natural gas cooktop burners were conducted in a large room with a ventilation system that provided filtered air with extremely low concentrations of fine and ultrafine particles. Two cooking activities were used: stir-frying green beans in a wok on high heat on the front burner and pan-frying a hamburger on medium heat on the back burner. Capture efficiency for cooking particles was determined by measuring concentrations in the air being exhausted from the room and comparing experiments with and without range hood use. Capture efficiency for combustion pollutants was measured using CO2 produced by the natural gas burners during cooking activities and during separate experiments in which burners were used to heat water in pots or pans. This study reinforces previous findings that exhaust hoods can be much more effective in capturing pollutants when cooking occurs on the back burners, compared to the front cooktop burners. Results indicate that capture efficiencies measured for burner CO2 are not predictive of those for cooking-generated particles under all conditions; however, they may be suitable to identify devices with efficiencies above 80 percent both pollutant sources. This research led to developing a standard test method for capture efficiency submitted to the American Society for Testing and Materials (ASTM).

6. Effect of Fuel Wobbe Number on Pollutant Emissions from Advanced Technology Residential Water Heaters (Objective 6)

This study was part of a larger effort to evaluate the potential air quality impacts of using liquefied natural gas in California. The study focused on pollutant emissions from advanced technology residential water heaters when these devices are operated with fuels having the higher Wobbe numbers associated with many liquefied natural gas supplies. Wobbe number is a measure of the energy delivery rate for appliances that use orifice- or pressure-based fuel metering, the approach used in most residential appliances.

The effect of varying natural gas blends and associated Wobbe number on pollutant emissions from residential water heaters was evaluated in controlled experiments. Experiments were conducted on eight storage water heaters, including five with “ultra low-NOX” burners, and four on-demand (tankless) water heaters, all of which featured ultra low-NOX burners. Pollutant emissions were quantified as air-free concentrations in the appliance flue and fuel-based emission factors in units of nanogram of pollutant emitter per joule of fuel energy consumed. Emissions were measured for CO, NOX, NO, formaldehyde and acetaldehyde as the water heaters were operated through defined operating cycles using fuels with varying Wobbe number.

6

The controlled experiments conducted in this study indicate that for some advanced technology water heaters, emissions of some pollutants are sensitive to the Wobbe number of the fuel. The most prominent sensitivity was an increase in NOX emissions with increasing Wobbe number.

Project Benefits The primary benefits of this project are the scientific papers and technical reports identifying the hazards resulting from using unvented natural gas cooking burners in California homes, and the potential for effective kitchen ventilation to reduce the hazard. Project results helped to motivate the necessity for a test method for capture efficiency and advanced its development. The project also helped spur improvements and efforts to harmonize CAS test procedures used in various residential energy efficiency programs including multifamily.

Project findings have been disseminated through presentations at more than twenty professional conferences – mostly geared toward residential energy efficiency, articles in Home Energy magazine, and via popular media. Media highlights include reports on local and national public radio; an interview on the nationally syndicated Leonard Lopate radio show; articles in the Los Angeles Times, San Jose Mercury News and New York Times Health blog; and a news feature in the high impact journal Environmental Health Perspectives.

7

CHAPTER 1 : The California Healthy Homes Indoor Air Quality Study of 2011-2013 1.1 Introduction Residential natural gas appliances can produce pollutants including carbon monoxide (CO), nitrogen dioxide (NO2), formaldehyde and ultrafine particles (UFP) (Afshari et al., 2005; Brown et al., 2004; Dennekamp et al., 2001; Moschandreas et al., 1986; Singer et al., 2010a; Traynor et al., 1996; Traynor et al., 1985). When the exhaust from a gas appliance enters the living space, indoor air quality (IAQ) can be compromised. Many gas appliances, including water heaters and furnaces, are designed to vent their exhaust directly to the outdoors. If the venting is not operating correctly (for example, because it is broken or not designed and installed correctly, or when depressurization in the indoor space exceeds the draft capacity of the appliance), combustion products including pollutants spill into the indoor space. Combustion products of cooking appliances and “vent-free” (unvented) heating appliances are released indoors by design. Venting range hoods (extractor fans) and other kitchen exhaust fans are intended to remove some of the pollutants emitted by cooking burners before they mix throughout the home (Delp and Singer, 2012; Singer et al., 2012a). However, surveys suggest that regular use of kitchen ventilation during cooking is infrequent, even when it is available (Klug et al., 2011; Mullen et al., 2013a; Piazza et al., 2007).

Numerous studies have found that homes with gas cooking burners and/or gas appliances with pilot burners tend to have indoor concentrations of combustion-related pollutants that are higher than similar homes without gas appliances, and that sometimes exceed U.S. national and California state ambient air quality standards (AAQS) (Garrett et al., 1999; Ryan et al., 1988; Schwab et al., 1994; Spengler et al., 1994; Spengler et al., 1983; Wilson et al., 1986; Wilson et al., 1993). A recent simulation study estimated that among Southern California homes that cook at least once per week with natural gas and do not regularly use a venting range hood, more than half have 1-h NO2 concentrations exceeding 100 ppb and roughly 5% have short-term CO concentrations that exceed the concentration thresholds of acute ambient standards on a weekly basis in winter (Logue et al., 2014). Homes that use unvented gas heaters and fireplaces can have particularly high concentrations of combustion pollutants, often exceeding AAQS thresholds (Dutton et al., 2001; Francisco et al., 2010; Ryan et al., 1989). In homes with gas appliances, smaller home size and the presence of floor and wall furnaces have been associated with higher combustion pollutant levels (Wilson et al., 1986).

There is a large literature showing associations between exposure to pollutants generated by gas appliances and adverse health impacts, with many of the studies focusing on nitrogen dioxide (Belanger et al., 2006; Franklin et al., 1999; Garrett et al., 1998; Hansel et al., 2008; Morales et al., 2009; Neas et al., 1991; Nitschke et al., 1999; Pilotto et al., 1997; van Strien et al., 2004). The most recent EPA assessment for carbon monoxide concluded that “a causal relationship is likely to exist between relevant short-term exposures to CO and cardiovascular

8

morbidity, whereas the available evidence is inadequate to conclude that a causal relationship exists between relevant long-term exposures to CO and cardiovascular morbidity” (US EPA, 2010). Formaldehyde is a known human carcinogen (International Agency for Research on Cancer, 2006) and exposures at levels that occur in homes have been linked to respiratory pathology (Franklin et al., 2000; Roda et al., 2011). A recent study found higher lung function and lower odds of asthma, wheeze, and bronchitis among children whose parents reported using kitchen ventilation when cooking with gas compared to children living in homes in which kitchen ventilation was not used with gas stoves (Kile et al., 2014).

More than two decades have elapsed since the last large-scale studies that focused on the impacts of natural gas appliances on IAQ in California homes (Spengler et al., 1994; Wilson et al., 1993). During this time, there have been many changes to the population of homes and gas appliances. Burner and appliance designs have advanced and attention to IAQ by appliance manufacturers, utilities and the home renovation industry may have reduced the frequency of improper appliance operation or venting, leading to fewer homes with elevated concentrations. Air sealing retrofits and the construction of new homes with airtight envelopes for energy efficiency should translate to lower outdoor air exchange rates during winter conditions when windows are closed; this could produce higher concentrations of any pollutants that are released into the home.

The California Healthy Homes Indoor Air Quality Study of 2011-2013 was designed to investigate the extent to which gas appliances still negatively impact IAQ in California homes. The study targeted homes with one or more gas appliances that could be a source of indoor air pollutant emissions, including gas cooking burners and venting appliances contained in the living space. There was oversampling of homes with previously identified risk factors, such as smaller floor area, frequent cooking with gas burners, presence of a wall or floor furnace, and lower household income, as these households can less frequently update or upgrade appliances. This paper presents analyses examining the impact of the types of appliances present in the home, the presence of pilot burners, the frequency of cooking with gas or electric burners and the use of kitchen exhaust during cooking.

1.2 Materials and Methods The core data collection methods of the study entailed monitoring inside and outside of homes using passive measurement devices while also conducting telephone interviews with participants to collect information about the homes. Mullen et al. (2013a) provides a thorough description of experimental methods, participant communication materials and all interview questions. The sections below provide summary descriptions. The study protocols were approved by LBNL’s Institutional Review Board.

1.2.1 Participant Recruitment The study was publicized by direct outreach to organizations associated with ethnically, economically and geographically diverse sub-populations in California. Recruitment efforts in the first year focused on the northern coastal region of California. The second year focused on the southern and inland regions of the state. Organization representatives were asked to pass

9

along information about the study to their constituents. Interested individuals were directed to a project web site and telephone number to obtain more information and complete a screening survey. The web site noted the incentive of $75 and a report about the air quality in the participant’s home to be provided at the completion of participation. The screening survey asked questions about the building and appliances, household demographics, and activities related to appliance use. Responses were used to calculate a risk score for IAQ hazards from gas appliances based on the algorithm described in Mullen et al. (2013a). The following factors were considered: frequency of use of gas cooking burners; which gas appliances were inside the living space or connected spaces and whether they were vented; size of the home; year the home was built (recognizing that newer homes are generally tighter with less infiltration air exchange); household income; and whether the home had been weatherized to increase airtightness. Twenty-four homes constructed or retrofitted for low energy use were included as part of a supplemental study of IAQ in high performance homes (Less, 2012; Less et al., 2014). There was intentional sampling of some homes without gas appliances to serve as controls (n=38). Homes were selected for sampling in geographic clusters. When a home was identified as desirable for inclusion, the individual who submitted the screening survey was contacted by telephone for consent and scheduling. The content of the web site, outreach materials and screening survey is provided in Mullen et al. (2013a).

1.2.2 Data Collection Instruments and Methods Measurement devices were deployed in homes to determine pollutant concentrations, temperature (T) and relative humidity (RH) in two indoor locations and at an outdoor site nearby to each residence. Furnace and water heater operation were also monitored. A structured interview was conducted by telephone before monitoring to collect more detailed information about the building, appliances, household demographics and general activities. A post-monitoring structured interview collected data about activities during the monitoring period and about general practices relevant to gas appliance impacts on IAQ. Some questions were not asked until after the monitoring period, so as to avoid affecting occupants’ behaviors and attitudes related to their gas appliances.

Measurements were conducted using a package of passive samplers and monitors that were mailed to 323 participant homes and delivered by researchers to 29 homes. Participants receiving the package by mail set up the samplers using written and pictorial instructions provided with the package. A researcher contacted each participant by telephone to check if the materials were clear and to help resolve any difficulties. Monitoring was planned to occur in each home for six days. The standard schedule was for the package to be sent on Monday morning to arrive at the home by Tuesday afternoon. The request was for the samplers to be set up within 24 h of receipt and then repackaged and mailed back the following Tuesday. Participants were asked to package samplers in pre-addressed return shipping envelopes on Monday night or Tuesday morning. In 29 homes, equipment was deployed and retrieved by a researcher who visited the homes. Sampling was conducted in two phases from late November 2011 to mid-April 2012 and from late October 2012 to mid-March 2013. During those periods, 5 to 14 homes were sampled per week during most weeks. Sampling did not occur during the

10

weeks in which the Thanksgiving, Christmas and New Years holidays were observed in the United States.

Table 1: Summary of Pollutant and Environmental Monitoring Instruments Used in Study

Parameter Manufacturer, model Data resolution Location of deployment

Formaldehyde, Acetaldehyde a

Waters, Sep-Pak XPoSure DNPH Cartridges

Integrated over sample period

Bedroom, kitchen, outdoor a

NOX, NO2 a Ogawa NOX/NO2 sampler Integrated over sample period

Bedroom, kitchen, outdoor a

CO (ambient) Lascar, USB-EL-CO300 1 minute Kitchen T, RH (indoors) HOBO, U10 1 minute Bedroom, kitchen Furnace operation (by T) HOBO, U10 1 minute Furnace supply

register Water heater operation (T) HOBO, U12-014 1 minute Water heater exhaust

flue Water heater spillage (T) HOBO, U12-014 1 minute Adjacent to draft

hood T, RH (outdoors) a HOBO, U23 Pro v.2 1 minute Outdoors

a Outdoor sampling occurred at a subset of homes.

The monitoring package included samplers and instruments listed in Mullen et al., 2013b.Pollutant concentrations, T, and RH were measured in the kitchen and a bedroom (child’s bedroom, if available) of each home, and outside of selected homes to define outdoor concentrations for a cluster of similarly located homes. NO2, NOX, volatile aldehydes and CO were measured in the kitchen, and all pollutants other than CO were measured in a bedroom. Volatile aldehydes were measured with a sampler that is typically used for active sampling, based on passive uptake rates determined for 5-10 d deployment periods (Mullen et al., 2013b). NOX and NO2 were measured using Ogawa passive sampling equipment (Singer et al., 2004), with NO calculated as the difference between the NOX and NO2 results.

A thermocouple placed on the water heater and a thermistor placed on a heating supply register monitored the operation of these appliances. Temperature, RH, CO, and appliance monitors all had on-board data loggers. Participants were asked to take photos of the samplers deployed in the homes and to send the photos via email or text message to the study director to ensure proper placement. Most sent relevant photos. Roughly half the homes received either a duplicate sampler that was to be placed in the bedroom or a field blank. Participants were called as a reminder the night before they were expected to return the package.

The post-monitoring telephone interview collected data on activities in the home during the sampling period, including frequency of appliance use, occupancy patterns and other potential pollutant sources inside and outside of the home. The interview included questions that might have affected resident behavior if asked prior to the sampling periods, e.g., about the frequency of kitchen exhaust fan use, reasons why the kitchen exhaust fan was not used, and the condition

11

of the stovetop and oven (flame quality, operational problems etc.). The post-monitoring interview was the last task for participants to complete.

1.2.3 Data analysis Passive samples were analysed using methods described in (Mullen et al., 2013a). Passive samplers that were returned unsealed were flagged as invalid. Photos and analytical results were reviewed to identify obvious errors such as a samplers being deployed with caps in place or switching of samples and blanks. Data from the CO, T, RH and appliance monitoring data loggers were downloaded and compiled into a database and analysed to calculate mean, as well as the highest 1-h and 8-h averages for the sampling period in each home.

The potential for depositional losses of NO and NO2 inside the two designs of outdoor sampling enclosures was evaluated in six side-by-side deployments with the open samplers used in Singer et al. (2004); details are reported in Mullen et al. (2013a). Adjustment factors of 1.22 and 1.18 were determined for NO2 sampling in the outdoor enclosures used in the first two weeks and all subsequent weeks, respectively. The data did not show any clear bias in NO measured in the outdoor enclosure, so no adjustments were made for NO. Outdoor NOX was calculated as the sum of the adjusted NO2 and the unadjusted NO (Mullen et al., 2013a).

Recognizing that outdoor NOX and NO2 concentrations have a major impact on indoor levels, researchers used concurrently measured outdoor concentrations to estimate the indoor levels that could be attributed to indoor sources. This adjustment was made for NOX and NO2 since outdoor concentrations were of similar magnitude to indoor concentrations (Figure 1). Outdoor air contributed a minority of indoor aldehydes (Figure 2); analyses were thus conducted on the directly measured levels of these pollutants in kitchens and bedrooms. The highest 1-h and 8-h CO in kitchens also were analysed as measured since outdoor levels are typically much lower than short-term indoor peaks in homes with a CO source. Homes without outdoor monitoring were assigned the outdoor NOX and NO2 concentrations measured at the closest home within the cluster or the closest ambient monitoring station, when either the cluster sample was not available or the central monitoring site was deemed more representative based on land-use. Indoor concentrations attributed to indoor sources were calculated as follows. For NO (NOX-NO2), outdoor levels were subtracted from those measured indoors. For NO2, researchers multiplied the assigned outdoor value by an infiltration factor F=0.4 to obtain an estimate of the indoor NO2 that can be attributed to outdoor sources. This value is obtained as the air exchange rate (λ – accounting for entry from outdoors to indoors – divided by the sum of the air exchange rate and indoor deposition rate (λ + kd) – which is the rate at which NO2 is removed from indoors. The value of 0.4 was estimated based on consideration of published data on air exchange rates in California homes (Wilson et al., 1993; Wilson et al., 1996; Yamamoto et al., 2010) and reported NO2 indoor deposition rates (Noris et al., 2013; Spicer et al., 1989; Spicer et al., 1993; Wilson et al., 1986; Yang et al., 2004). Indoor NOX attributed to entry from outdoors was calculated as the sum of NO and NO2 from outdoors. Figure 3 shows after the estimated outdoor contribution is subtracted, the median bedroom NO2 and NOX in all-electric homes were both close to zero, as would be expected for homes with no indoor sources.

12

The impacts of gas appliances on IAQ was explored by comparing distributions of calculated pollutant concentrations noted above grouped by the following characteristics: (a) the type(s) of gas appliance(s) inside the living space, (b) cooking burner fuel type and which appliances, if any, had pilot burners, (c) cooking burner fuel type and frequency of use and (d) use frequency of kitchen exhaust ventilation in homes that reported cooking for 4 or more hours during the monitoring period. Analyses were conducted using the measured, time-integrated concentrations of aldehydes, the highest 1-h and 8-h CO, and the estimated indoor source-attributed concentrations of NOX and NO2.

Recognizing that the impact of emissions from a combustion appliance or pilot burner will scale inversely with the dilution volume, researchers scaled the indoor-attributed concentrations of NOX and NO2 and measured CO to a common home size of 130 m2 (1400 ft2). This scaling was done after the first series of bivariate analyses revealed that homes with gas cooking appliances and with pilot burners had significantly higher concentrations of these pollutants than homes without gas cooking. This analysis was designed to assess if any between-group differences in un-scaled concentrations were caused by differences in homes sizes.

13

Figure 1: NOX and NO2 Measured in Kitchen and Bedroom, and Measured or Assigned Outdoor Concentrations, Ordered by Concentrations in Bedroom

Data displayed for 343 homes with bedroom measurements; results for each home aligned vertically. Outdoor concentrations were measured in this study or taken from a nearby regulatory air monitoring station. Tables present arithmetic means, geometric means, and geometric standard deviations.

14

Figure 2: Formaldehyde and Acetaldehyde Measured in Kitchens, Bedrooms, and Outdoors of Study Homes, Ordered by Bedroom Concentrations

Data displayed for 344 homes with bedroom measurements; results for each home aligned vertically. Outdoor concentrations measured only at a subset of homes. Data from each location – kitchen, bedroom and outdoors - were lognormally distributed. Outdoor acetaldehyde data were statistically indistinguishable from field blanks.

15

1.3 Results and Discussion 1.3.1 Demographics of Sample Data were collected from 352 homes, including the high performance home sub-sample (Less, 2012). The overall sample mostly comprised homes with gas appliances in the living space and homes that used gas cooking appliances: 90% of study homes had at least one gas appliance and 82% had gas cooking burners. A gas cooktop was used more than 7 times during the sampling period in 53% of study homes and 26% of study homes used a gas cooktop more than 14 times (all by self-report). Participants reported that they either did not have a kitchen exhaust fan or that they rarely or never used it in 64% of homes.

The sample included many older appliances, as reported in Mullen et al. (2013a). Table 19 of that report indicates that 24% (40/165) of central furnaces and 65% (35/54) of wall and floor furnaces with estimated ages were more than 15 years old. Table 25 of the report indicates that 20 of 150 water heaters (13%) with age estimates were more than 15 years old. And Table 38 of the report indicates that 20% (62/310) of cooktops with age estimates were more than 15 years old.

The demographics of the mail-out sample population are presented and discussed by Mullen et al. (2013a) and summarized in Table 2. The study sample had a similar breakdown of renters and homeowners (46/54%) compared to California overall (43/57%) (RASS, 2009). The sample had more homes with floor areas under 93 m2, fewer homes larger than 186 m2 and similar percentages of 93-186 m2 homes compared to the California stock. The study sample was under-represented in the lowest household income brackets (<$50,000 per year), with 19% in the sample compared to 44% for the state. Although researchers could not find directly comparable statewide data, it seems likely that the educational attainment of the study sample was skewed relative to the general population. The racial distribution of the sample was reasonably similar to that of the California population, allowing for uncertainty related to the US Census not tracking “Hispanic” as a race and considering that census data is tabulated per individual whereas statistics on the study population are tabulated per household. Relative to California, there were fewer households in the study containing children or seniors.

16

Table 2: Self-Reported Race and/or Ethnicity, Household Income, Highest Education Level, and Number of Residents Living in Households Included in this Study

# in study % in study % in CA a Types of appliances present Home rented 147 46% 43% Home owned 176 54% 57% Floor Area of home (sq. ft.) <1000 110 34% 22% 1000-2000 143 44% 46% >2000 47 15% 32% Did not answer 23 7% Number of residents 1 – 2 164 51% 55% 3 – 4 116 36%

45% b 5 or more Did not answer

42 1

13% <1%

Presence of minors and seniors At least one resident <18 years old 51 16% 37% At least one resident >64 years old 20 6% 25% All residents between 18-64 years old 252 78% 38% Highest education level of ANYONE in household c Less than Bachelors degree 60 19% NA Bachelors degree 90 28% NA Graduate degree 172 53% NA Did not answer 1 <1% Ethnicities represented by residents d Native American 7 1% 2% Hispanic/ Latino 36 5% 38% Black, African-American 45 14% 7% Asian or Pacific Islander 80 30% 14% White, Caucasian 219 76% 74% Combined Gross Income Less than $25,000 50 6% 22% $25,000 - $49,999 47 13% 22% $50,000 - $74,999 53 15% 17% $75,000 - $99,999 36 14% 12% $100,000 - $150,000 67 25% 14% >$150,000 36 18% 13% Prefer not to say 34 6%

a Home floor area data obtained from Residential Appliance Saturation Survey, 2009 (www.energy.ca.gov/appliances/rass/). Remaining data obtained from www.census.gov. b Percent of households with 3 or more persons in CA. c Educational attainment statistics were not available on a per household basis for the CA population. d All race/ethnic categories that partially/fully characterize an individual/household are weighted equally, therefore percentages sum to greater than 100%. However, statistics for the study population are tabulated on a per household basis, whereas CA statistics are tabulated per individual.

1.3.2 Quality Assurance Results The available evidence – including survey completion and sampler return rates, submitted photographs of sampler deployment locations, inspections of returned sampler packages and

17

results of quality assurance replicates and blanks – indicates that most participants followed the instructions to deploy samplers as intended (Mullen et al., 2013a). In only one instance did a participant report that a sampling package mailed from LBNL did not arrive. Sampler packages were mailed back by all participants who received them. In seven cases, data were lost from all passive samplers sent to a home, either because participants returned the package with delays of more than a month, or participants did not seal time-integrated samplers in the provided airtight bags before mailing. Two additional homes had invalid NOX and NO2 data because of an error in sampler preparation before shipment to one home and improper sealing of the samplers from the other home. The mean relative deviations for all pairs of duplicate samplers were 3% for NOX, 7% for NO2, 5% for formaldehyde and 5% for acetaldehyde. The percent of field blanks with concentrations above the analytical LOQ were 8%, 5%, 16% and 45% for NOX, NO2, formaldehyde and acetaldehyde. Field blanks had mean concentrations of 0.37 ppb NOX, 0.25 ppb NO2, 0.6 ppb formaldehyde and 1.7 ppb acetaldehyde for an assumed 6-day deployment period. Reported measured concentrations were not adjusted for the values measured on field blanks. Additional quality assurance results and participant compliance notes are presented and discussed in Mullen et al. (2013a).

1.3.3 Measured Pollutant Levels in Kitchen, Bedroom and Outdoors Summary statistics for measured pollutants and pairwise correlations are provided in Table 3 and Table 4.

The time-integrated concentrations of NOX and NO2 measured indoors and measured or assigned outdoors at each home are presented in Figure 1. Each series (outdoor, bedroom and kitchen) followed a lognormal distribution. Outdoor NO2 concentrations were higher than the 30 ppb threshold of California’s annual average ambient air quality standard (CAAQS) for 9% of study homes. Measured NO2 exceeded 30 ppb in about 24% of kitchens and 12% of bedrooms, and indoor-attributed NO2 was above 30 ppb in about 14% of kitchens and 6% of bedrooms. These statistics result from monitoring over periods of only about 6 days in each home and over-sampling of homes with potential indoor sources of NO2. Figure 1 and Table 4 show that concentrations of NOX (r2=0.90) and NO2 (r2=0.86) were highly correlated between kitchens and bedrooms. Many homes had higher NOX in bedrooms and kitchens than outdoors, indicating indoor source(s). In the absence of indoor sources, indoor NO2 should be substantially lower than outdoor NO2 owing to indoor deposition. The homes with the lowest values of bedroom NO2 had indoor concentrations that were on the order of half of outdoor levels. At higher bedroom NO2 concentrations, the ratio of indoor to outdoor NO2 generally was higher. For NO2, there was a clear trend of higher concentrations in the kitchen than in the bedroom: arithmetic (AM) and geometric (GM) mean levels of NO2 in kitchens (23.2 and 16.9 ppb) were 31% and 26% higher than NO2 in bedrooms (17.7 and 13.4 ppb). Kitchen NOX was also higher than bedroom NOX, with AM and GM ratios of 13% and 14%.

18

Table 3: Summary Statistics for Measured Pollutant Concentrationsa

Parameter (integrated over sample period, except where noted)

N AM GM GSD 90th %-ileb

95th %-ileb

Kitchen NOX 343 73 51 2.5 150 192 Bedroom NOX 344 65 47 2.4 124 168 Outdoor NOX, measured c 180 32 24 2.2 61 70 Outdoor NOX, assigned d 345 32 26 2.1 57 84 Kitchen NO2 343 23 17 2.3 108 147 Bedroom NO2 344 18 13 2.2 32 41 Outdoor NO2, measured c 180 16 13 1.9 29 33 Outdoor NO2, assigned d 345 17 14 1.8 28 32 Kitchen highest 8-h CO 304 3.4 2.2 2.6 6.8 10 Kitchen highest 1-h CO 304 6.4 3.8 2.8 13 18 Outdoor CO, assigned d 334 0.5 0.5 1.5 0.9 1.0 Kitchen formaldehyde 340 17 15 1.7 29 34 Bedroom formaldehyde 340 17 15 1.7 30 36 Outdoor formaldehyde, measured c 179 2.4 2.0 2.1 3.4 3.9 Kitchen acetaldehyde 340 9.7 8.0 1.8 16 23 Bedroom acetaldehyde 340 9.7 7.9 1.8 17 23 Outdoor acetaldehyde, measured c 178 1.8 1.4 2.4 3.2 4.6

a See also Figure 1 and Figure 2. b Percentiles from measured or assigned values, not fitted distributions c Measured in this study d Statistics of values assigned to all homes in the study. Assignments based on measurements conducted in this study and values obtained from compliance monitoring sites, as described in text and Mullen et al. 2013.

Broadly, NO2 concentrations measured in the Healthy Homes study of 2011-2013 were lower than those reported for California homes in large studies conducted in the 1980s and early 1990s (Spengler et al., 1994; Wilson et al., 1986; Wilson et al., 1993), with decreases in outdoor pollutant levels accounting for much or all of the difference.

The highest 1-h and highest 8-h CO levels were log-normally distributed across homes that had CO exceed the instrument quantitation limit of 0.5 ppm. Of the 316 homes with CO data in the current study, roughly 5% had short term concentrations exceed California ambient air quality standards of 20 ppm over 1 h or 9 ppm over 8 h. Arithmetic and geometric mean values of highest 1-h CO were 6.4 and 3.8 ppm in the current study.

19

Table 4: Coefficient of Determination (R2) Between Pollutants Measured at Different Locations in Homes a,b

NO2-K NO2-B NOX-K NOX-B NO-K NO-B CO-1h CO-8h FA-K FA-B AA-K AA-B

NO2-K 1.00

NO2-B 0.74 1.00

NOX-K 0.79 0.61 1.00

NOX-B 0.61 0.74 0.83 1.00

NO-K 0.08 0.06 0.04 0.04 1.00

NO-B 0.04 0.04 0.02 0.03 0.72 1.00

CO-1h 0.17 0.15 0.17 0.16 0.06 0.04 1.00

CO-8h 0.18 0.17 0.19 0.19 0.07 0.08 0.76 1.00

FA-K 0.00 0.00 0.00 0.01 0.00 0.00 0.03 0.04 1.00

FA-B 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.52 1.00

AA-K 0.00 0.00 0.02 0.03 0.00 0.00 0.03 0.06 0.12 0.05 1.00

AA-B 0.00 0.00 0.02 0.03 0.00 0.00 0.02 0.04 0.10 0.13 0.76 1.00 a NO2 and NOX are estimates of the indoor concentrations resulting from indoor sources; calculated by subtracting 40% and 100%, respectively, of the simultaneous outdoor concentration. Indoor NOX concentrations adjusted by summing the adjusted NO2 and NO. b Letters “B” and “K” following pollutant abbreviations are used to indicate measurements made in bedrooms and kitchens, respectively. (FA = formaldehyde, AA = acetaldehyde).

The time-integrated concentrations of formaldehyde and acetaldehyde measured indoors and measured or assigned outdoors at each home are presented in Figure 2. Roughly 95% of homes had indoor formaldehyde levels above the Cal-EPA Chronic Reference Exposure Level (CREL) of 7.3 ppb. Indoor aldehyde concentrations were higher than outdoor concentrations in almost all homes with data for both locations. Concentrations of each pollutant measured in bedrooms and kitchens of the same homes were somewhat correlated with r2=0.52 for formaldehyde and r2=0.76 for acetaldehyde (Table 4). Formaldehyde and acetaldehyde were not highly correlated with each other, with r2=0.34 and r2=0.36 for measurements in kitchens and bedrooms. Figure S3 illustrates the lognormal distributions of formaldehyde and acetaldehyde concentrations measured in kitchens, bedrooms and outdoors. As a group, outdoor acetaldehyde concentrations were indistinguishable from field blanks.

Concentrations of NOX and NO2 were not highly correlated with CO or either aldehyde; and the aldehydes were not highly correlated with CO (Table 4).

1.3.4 Impact of Appliance Types on Indoor Pollutant Levels Figure 3 presents summary statistics for highest 1-h CO in the kitchen, as well as indoor-attributed NO2 and NOX and formaldehyde and acetaldehyde in the bedroom, grouped by the type(s) of appliances inside the home. P-values in the figure represent the likelihood that other

20

groups’ distributions are drawn from the same distribution as the “All Electric” group, based on the Kruskal-Wallis test. Table 5 presents additional results for this analysis. In comparison to homes without gas appliances, there was a large and statistically robust increase in indoor-attributed concentrations of bedroom and kitchen NOX and NO2 and highest kitchen 1-h CO for homes that used gas cooking burners, whether or not there were also venting gas appliances in the home. Indoor-attributed NOX and NO2 concentrations were higher in kitchens than in bedrooms for the two groups with gas cooking appliances. Table S5 shows that in comparison to homes with gas cooking but no venting appliances, homes with both cooking and venting appliances had significantly higher indoor-attributed NOX and NO2. Highest 1-h kitchen CO was not different between these groups.

Some of the differences in NOX and NO2 between the last two groups of Figure 3 and Table 5 may result from differences in home volumetric dilution rates. The outdoor air dilution rate (e.g. in units of m-3 h-1) is the product of the residence air volume and the air exchange rate. An air pollutant source of fixed size, such as a cooking burner, will have less dilution in a smaller home compared to a larger home with the same outdoor air exchange rate. When concentrations were scaled to home size (by floor area), the difference in NOX and NO2 between the last groups disappeared (see last 4 rows of Table 5). This suggests that much / all of the difference between those groups may result from cooking burner pollutant emissions occurring in smaller spaces with less outdoor air dilution.

There were no statistically robust differences in formaldehyde or acetaldehyde levels associated with gas appliances (Table 5).

21

Figure 3: Indoor Pollutant Concentrations by Type(s) of Appliances Inside Home

Highest 1-h CO from kitchen and indoor-attributed NOX and NO2 from bedroom measurements. Formaldehyde and acetaldehyde from bedroom measurements. Indoor-attributed concentrations calculated by subtracting estimated outdoor contribution from the indoor measured value. See text for additional details. Boxes show inter-quartile range (IQR). Whiskers span to 1.5 IQR. Filled circles show all data >1.5 IQR. P-values indicate likelihood that data from other groups are drawn from same distribution as the “All Electric” group, based on the Kruskal-Wallis test.

22

Table 5: Sample Characteristics and Median Pollutant Concentrations (ppb; except CO) in Homes Grouped by the Type of Gas Appliance(s) in the Living Space

Parameter

No gas appliances

(Ref. group)

Only vented

gas appliances

Only gas cookingc

Gas cooking + vented gas appliancesc

Compare last two groupsd

Homes (N) 38 28 144 142 Mean floor area (m2) a 115 159 148 111 Median floor area (m2) a 105 163 128 105 Measured concentrations b Highest 1-h CO (ppm) 1.7 1.5 4.8** 4.9** Bedroom NOX 20 23 54** 69** p=0.03 Kitchen NOX 18 21 60** 75** p=0.09 Bedroom NO2 6.7 6.0 14** 16** p=0.01 Kitchen NO2 6.5 7.6 18** 22** p=0.05 Bedroom formaldehyde 13 18^ 16 15 Kitchen formaldehyde 14 16 16^ 16 Bedroom acetaldehyde 7.9 8.1 7.7 7.4 Kitchen acetaldehyde 8.0 8.3 7.3 7.3 Indoor-attributed b Bedroom NOX -0.8 3.2 31** 42** p=0.02 Kitchen NOX 3.9 5.6 38** 53** p=0.05 Bedroom NO2 1.7 2.6 7.9** 9.6** p=0.02 Kitchen NO2 1.4 3.2 12.3** 16.9** p=0.03 Scaled to 130 m2 home a Highest 1-h CO (ppm) 1.2 1.3 4.4** 3.8** p=0.07 Indoor-attributed & scaled a,b Bedroom NOX -1.2 2.1 32** 35** Kitchen NOX 3.2 4.8^ 39** 42** Bedroom NO2 1.1 2.7^ 8.0** 7.7** Kitchen NO2 1.5 3.0^ 12.2** 17.0**

a Floor area assigned to each home for purpose of this calculation was the midpoint of the size bin selected by participant during telephone interview. Size bins were 46-70, 70-93, 93-116, 116-139, 139-186, 186-232, 232-279, >279 m2. b Estimated by subtracting the assigned outdoor NO and 0.4x the outdoor NO2 from corresponding kitchen and bedroom measured values, then calculating indoor-attributed NOX as NO+NO2. c Symbols indicate statistical discernibility using Kruskal-Wallis test of likelihood that other groups’ distributions are drawn from the same distribution as the Reference group: **p<0.01; *0.01≤p<0.05; ^0.05≤p≤0.15. d Last column indicates discernible differences between the last two groups; only shows p≤0.15.

23

1.3.5 Impact of Pilot Burners on Indoor Pollutant Levels Figure 4 and Table 6 present summary statistics for the same pollutants displayed in Figure 3 and Table 5, this time grouped by cooktop fuel type and the presence of pilot burners on cooktops or furnaces located inside the home. The homes were divided into five groups: (1) electric cooktop, no furnace pilots; (2) gas cooktop without pilot, no furnace pilots; (3) gas cooktop with pilot, no furnace pilots; (4) gas cooktop without pilot, furnace(s) with pilot(s); (5) gas cooktop, furnace(s) with pilot(s). The fourth group includes seven homes with floor furnaces and 41 homes with wall furnaces, four of which did not have valid NO2 and NOX data. The fifth group includes 3 homes with floor furnaces and 24 with wall furnaces, one of which did not have valid NO2 and NOX data. Each group of homes was compared to homes that had gas cooking but no pilot burners, using the Kruskal-Wallis test, with p-values shown in the figure. Researchers also compared the third and fifth groups to further explore whether homes with furnace pilots have higher pollutant concentrations than those without.

All three groups with any pilot burner had indoor-attributed NOX and NO2 in bedrooms (Figure 4) and kitchens (Table 6) that were significantly higher than homes with gas cooktops but no pilots (Group 2). Higher concentrations in Group 4 compared to Group 2 (with p-values of 0.02 to <0.01) suggests that furnace pilots significantly increase NOX and NO2 throughout the home. The impact of furnace pilot burners is further indicated by higher concentrations in homes with both cooking and furnace pilots (Group 5) compared to homes with gas cooking pilots only (Group 3); Table S6 shows that differences in indoor-attributed NOX and NO2 between these groups are significant with p-values of <0.01 to 0.03 for three of the parameters and p=0.09 for kitchen NOX.

The three groups of homes with pilot burners also appear to have had higher values of the highest 1-h kitchen CO compared to the gas cooktop homes with no pilots, with p-values of 0.01-0.09. Bedroom formaldehyde was lower in the last two groups with p-values of 0.06 and 0.09.

Much of the apparent impact of furnace pilots appears attributable to these appliances being present in smaller homes, which may have lower volumetric dilution rates as noted earlier. The last 4 rows of Table S6 show that differences in NOX and NO2 between homes with gas cooktops and only furnace pilots (Group 4) and gas cooktops with no pilots (Group 2) largely disappear when indoor-attributed concentrations are adjusted by the size (floor area) of the home. The effect of furnace pilots in homes that also have cooktop pilots – comparing Groups 3 and 5 – persists for bedroom NO2 (p=0.03), but not for bedroom NOX or kitchen NOX and NO2, when adjusting for floor area.

24

Figure 4: Indoor Pollutant Concentrations by Cooktop (CT) Fuel and Presence of Pilot Burners on Cooktop or Furnace (F)

Refer to Figure 3 caption and text for descriptions of calculations for indoor source attribution and definitions of boxes and whiskers. P-values indicate likelihood that data from other groups are drawn from the same distribution as the “Gas CT, no pilots” group, based on the Kruskal-Wallis test.

25

Table 6: Sample Characteristics and Median Pollutant Concentrations (ppb; except CO in ppm) in Homes Grouped by Presence of Pilot Light(S) in the Living Space

Parameter Electric cooktop,

no pilotsc

Gas cooktop, no pilots

(Ref. group)

Gas cooktop

w/pilot, no furnace

pilotc

Gas cooktop w/o pilot, furnace

pilotc

Gas cooktop w/pilot, furnace

pilotc

Compare 3rd &

5th groupsd

Homes (N) 60 182 29 48 27 Mean floor area (m2) a 138 165 115 86 79 Median floor area (m2) a 128 128 81 81 81 Measured concentrations

Highest 1-h CO (ppm) 1.6** 4.1 7.0* 6.3^ 4.8^ Bedroom NOX 20** 51 78** 70* 108** p=0.07 Kitchen NOX 19** 55 119** 74* 138** Bedroom NO2 6.5** 12 26** 18** 32** p=0.03 Kitchen NO2 6.0** 15.4 41** 22** 58** p=0.14 Bedroom formaldehyde 14 16 17 12^ 11^ Kitchen formaldehyde 15 16 17 12* 13 Bedroom acetaldehyde 8.0 7.6 8.1 7.4 6.2 Kitchen acetaldehyde 7.9 7.4 8.9 7.2 6.2 Indoor-attributed b Bedroom NOX -0.1** 26 62** 41** 109** p=0.02 Kitchen NOX 5.2** 31 102** 51** 116** p=0.09 Bedroom NO2 1.9** 6.5 16** 9.9** 28** p<0.01 Kitchen NO2 1.8** 10.3 33** 16* 52** p=0.03 Scaled to 130 m2 home a Highest 1-h CO (ppm) 1.3** 4.3 4.7 3.7 3.9 p=0.13 Indoor-attributed & scaled a,b

Bedroom NOX -0.4** 28 42** 26 58** Kitchen NOX 3.8** 36 66** 29 62** Bedroom NO2 1.6** 7.1 11** 5.9 16** p=0.03 Kitchen NO2 1.9** 11 24** 8.6^ 22**

a Floor area assigned to each home for purpose of this calculation was the midpoint of the size range that the participant selected during a telephone interview. Size ranges were 46-70, 70-93, 93-116, 116-139, 139-186, 186-232, 232-279, >279 m2. b Estimated by subtracting the assigned outdoor NO and 0.4x the outdoor NO2 from corresponding kitchen and bedroom measured values, then calculating indoor-attributed NOX as NO+NO2. c Symbols indicate statistical discernibility using Kruskal-Wallis test of likelihood that other groups’ distributions are drawn from the same distribution as the Reference group: **p<0.01; *0.01≤p<0.05; ^0.05≤p≤0.15. d Last column indicates discernible differences between the compared groups; only shows p≤0.15.

26

1.3.6 Impact of Cooking Burner Use on Indoor Pollutant Levels Figure 5 and Table 7 present summary pollutant statistics for homes grouped according to cooking appliance fuel and cooking time during the monitoring period. Cooking time was estimated as the sum of self-reported cooking activity by meal. Highest 1-h kitchen CO and indoor-attributed NOX and NO2 measured in both kitchens and bedrooms increased with more gas cooking but not with more electric cooking. This trend was seen with and without scaling for floor area. Formaldehyde and acetaldehyde in homes that cooked more frequently with gas appliances were statistically indistinguishable from those that cooked less frequently with gas or cooked with electric appliances at any frequency. These results add to the weight of evidence that natural gas cooking burners are substantial and statistically significant sources of CO, NOX and NO2 in many homes.

27

Figure 5: Indoor Pollutant Concentrations by Cooktop (CT) Fuel and respondent-Reported Total Cooking Time during Monitoring Period

Cooking time from daily log. Refer to Figure 3 caption and text for descriptions of calculations for indoor source attribution and definitions of boxes and whiskers. P-values indicate likelihood that data from other groups are drawn from same distribution as the “Gas CT, <4 h” group, based on the Kruskal-Wallis test.

28

Table 7: Sample Characteristics and Median Pollutant Concentrations in Homes Grouped by Cooking Fuel Type and Amount of Cooking During the Week of Sampling

Parameter Elec. CT,

<4h/wkc

Elec. CT, 4-8h/wkc

Elec. CT, >8h/wkc

Gas CT, <4h/wk

(Ref. Group)

Gas CT, 4-8h/wkc

Gas CT, >8h/wkc

Homes (N) 17 19 28 93 94 99 Mean floor area (m2) a 105 138 164 120 130 141 Median floor area (m2) a 81 114 163 105 105 128 Measured concentrations Highest 1-h CO (ppm) 1.6** 2.0* 1.5** 3.9 5.6** 5.0* Bedroom NOX 17** 26* 19** 54 67 63* Kitchen NOX 14** 24** 18** 61 71^ 74** Bedroom NO2 6.6** 7.3** 5.2** 13 14 17^ Kitchen NO2 5.5** 9.3** 6.0** 18 19 24^ Bedroom formaldehyde 19 15 13 14 16 15 Kitchen formaldehyde 16 15 14 14 16 17^ Bedroom acetaldehyde 7.8 9.2 7.6 7.5 7.9 7.4 Kitchen acetaldehyde 7.2 9.4^ 7.4 7.1 7.4 8.5 Indoor-attributed b Bedroom NOX -1.6** -0.9** 0.6** 24 46** 47** Kitchen NOX 4.5** 5.2** 3.7** 28 52** 59** Bedroom NO2 2.7** 1.9** 1.7** 7.2 8.8 11** Kitchen NO2 1.3** 3.2** 2.1** 12 13 17* Scaled to 130 m2 home a Highest 1-h CO (ppm) 0.9** 1.5* 2.0* 3.3 4.4** 4.8** Indoor-attributed & scaled a,b Bedroom NOX -1.3** -1.2** 0.5** 17 39** 45** Kitchen NOX 3.1** 4.5** 3.4** 26 42** 56** Bedroom NO2 0.7** 2.2** 1.6** 5.7 7.8^ 10** Kitchen NO2 0.8** 2.6** 2.3** 9.8 13^ 17**

a Floor area assigned to each home for purpose of this calculation was the midpoint of the size range that the participant selected during a telephone interview. Size ranges were 46-70, 70-93, 93-116, 116-139, 139-186, 186-232, 232-279, >279 m2. b Estimated by subtracting the assigned outdoor NO and 0.4x the outdoor NO2 from corresponding kitchen and bedroom measured values, then calculating indoor-attributed NOX as NO+NO2. c Symbols indicate statistical discernibility using Kruskal-Wallis test of likelihood that other groups’ distributions are drawn from the same distribution as the Reference group: **p<0.01; *0.01≤p<0.05; ^0.05≤p≤0.15.

29

1.3.7 Impact of Kitchen Exhaust Ventilation on Indoor Pollutant Levels The final bivariate analysis investigated the impact of using kitchen exhaust fans when cooking. For this analysis, homes that reported cooking with gas for more than 4 h total during the week were grouped according to self-reported frequency of kitchen exhaust fan use. The Kruskal-Wallis test was used to compare homes in which kitchen exhaust was used some times or most times when cooking with gas against homes that cooked with gas but either never used or did not have a kitchen exhaust fan. Figure 6 presents summary statistics for highest 1-h kitchen CO and indoor-attributed kitchen NO2 and NOX; additional results are presented in Table 8. Measured aldehydes were not included in the analysis, since the prior analyses showed they were not significantly influenced by gas cooking in the homes. The results suggest that even occasional use of a kitchen exhaust fan reduces peak CO in the kitchen and time-integrated NO2 and NOX throughout the home. The effect broadly persists but at lower significance levels (higher p-values) when indoor-attributed concentrations are adjusted for home floor area (Table 8). The lack of a clear progression from infrequent to frequent use could be related to how decisions are made about exhaust fan use. For example, occasional use may occur during the most intensive cooking events, having a disproportionate effect on both peak and time-integrated concentrations in the home. The very wide range in pollutant removal effectiveness for range hoods installed in existing California homes (Singer et al., 2012a) might also have obscured the expected relation between pollutant concentrations and frequency of range hood usage, such that consistent usage in some homes may have very low efficacy.

These results provide empirical evidence that regular use of a kitchen exhaust fan when cooking with gas burners helps reduce concentrations of combustion pollutants in the kitchen. The effectiveness of range hoods in these homes presumably was reduced by the fact that 35% of participants (among those having fans) reported using it on medium or low speed, and 70% of participants reported cooking primarily on front burners. Research on range hood effectiveness indicates that the effectiveness is substantially lower when the hoods are operated at lower speeds and when cooking occurs on the front burners (Delp and Singer, 2012; Lunden et al., 2014; Singer et al., 2012a).

30

Figure 6: Indoor Pollutant Concentrations by Kitchen Exhaust Fan Use During Cooking in Homes with Gas Cooktops and >4 h Cooking

Cooking time from daily log. Exhaust fan use reported by respondent. Refer to Figure 3 caption and text for descriptions of calculations for indoor source attribution and definitions of boxes and whiskers. P-values indicate likelihood that data from other groups are drawn from same distribution as the “Gas CT, <4 h” group, based on the Kruskal-Wallis test.

31

Table 8: Median Pollutant Concentrations in Homes That Cooked with Gas for More Than 4-h in during the Monitoring Period, Grouped by the Self-Reported Frequency of Kitchen Exhaust Fan

Use

Parameter No gas cooking (Elec.)

Fan used most or all of the time

Fan used about half of the time

Fan used infrequently

Fan used rarely or

never (Ref. group)

Homes (N) 64 46 31 61 55 Mean floor area (m2) a 139 146 137 154 107 Median floor area (m2) a 128 163 128 128 81 Measured concentrations Highest 1-h CO (ppm) 1.6** 3.3** 4.4* 5.9^ 7.6 Bedroom NOX 20** 57 54* 62* 93 Kitchen NOX 19** 57** 64* 70** 105 Bedroom NO2 6.3** 14** 16* 14** 24 Kitchen NO2 6.6** 16** 22** 20** 34 Bedroom formaldehyde 15 18 16 15 15 Kitchen formaldehyde 15 17 17 17 14 Bedroom acetaldehyde 8.1 9.9^ 8.4 7.1 6.7 Kitchen acetaldehyde 8.1^ 9.4 8.6 7.2 6.6 Indoor-attributed b Bedroom NOX -0.1** 43** 30** 39** 70 Kitchen NOX 4.7** 47** 40** 45** 90 Bedroom NO2 1.9** 8.4** 8.3** 8.8** 17.3 Kitchen NO2 2.1** 12** 13** 13** 26 Scaled to 130 m2 home a Highest 1-h CO (ppm) 1.3** 3.3^ 4.1 5.2 4.7 Indoor-attributed & scaled a,b Bedroom NOX -0.7** 46 31* 38^ 50 Kitchen NOX 3.5** 44** 39* 42* 64 Bedroom NO2 1.5** 8.5* 7.9^ 9.5* 11 Kitchen NO2 1.7** 11** 13* 15** 19

Results for homes with electric appliances included for comparison.

a Floor area assigned to each home for purpose of this calculation was the midpoint of the size range that the participant selected during a telephone interview. Size ranges were 46-70, 70-93, 93-116, 116-139, 139-186, 186-232, 232-279, >279 m2. b Estimated by subtracting the assigned outdoor NO and 0.4x the outdoor NO2 from corresponding kitchen and bedroom measured values, then calculating indoor-attributed NOX as NO+NO2. c Symbols indicate statistical discernibility using Kruskal-Wallis test of likelihood that other groups’ distributions are drawn from the same distribution as the Reference group: **p<0.01; *0.01≤p<0.05; ^0.05≤p≤0.15.

32

1.4 Conclusions Pollutant measurements over multiple day monitoring periods in 352 California homes demonstrate that associations still exist between the presence and use of some gas appliances and elevated concentrations of CO, NOX and NO2. The largest impacts were associated with use of gas cooking appliances. More cooking led to higher concentrations in homes with gas cooking appliances but not in homes with electric cooking. In homes with gas cooking, the presence of additional appliances with venting was associated with higher concentrations of indoor-attributed NOX and NO2. However, when indoor-attributed concentrations were scaled to home size – to account for pollutants from cooking burners possibly reaching higher concentrations in smaller homes owing to less overall dilution – the effect of vented appliances on NOX and NO2 disappeared. Cooktop and furnace pilot burners were each associated with higher concentrations of time-integrated, indoor-attributed NOX and NO2 and highest 1-h CO when not scaled for home size. When pollutant concentrations were scaled to a common home size, the impacts of furnace pilot burners largely disappeared. Formaldehyde and acetaldehyde concentrations were not significantly impacted by any of the gas appliances examined in this study. Homes that cooked frequently with gas burners and reported using kitchen exhaust ventilation had lower concentrations of highest 1-h CO and time-integrated NOX and NO2 compared to homes that never use kitchen exhaust ventilation when cooking with gas burners.

33

Pollutant Concentrations and Emission Rates from Natural Gas Cooking Burner Use and Range Hood Performance in Nine Northern California Homes

1.5 Introduction The combustion products of natural gas cooking burners (NGCBs) include pollutants that can impact indoor air quality (IAQ). While complete combustion directly produces water vapor and carbon dioxide (CO2), the high flame temperatures also produce nitrogen oxides (NOX), including nitrogen dioxide (NO2), a respiratory irritant. Incomplete combustion can produce non-negligible emissions of carbon monoxide (CO), formaldehyde (CH2O), and nanometer-sized particles that form from condensation of partially oxidized organic compounds. These tiny particles grow through coagulation and condensation into particles that are tens of nm in diameter (Rim et al., 2012; Wallace et al., 2008), but remain within the <100 nm diameter threshold that defines ultrafine particles (UFP).

The U.S. EPA sets national ambient air quality standards (NAAQS) for carbon monoxide and NO2, in order to protect both the general population and sensitive sub-populations (US EPA, 2010; US EPA, 2016). The NAAQS for CO are 35 ppm averaged over 1h, and 9 ppm averaged over 8h. The NAAQS for short-term exposure to NO2 is 0.1 ppm (100 ppb) over 1h. Formaldehyde is both a mucosal irritant and a human carcinogen (Kaden et al., 2010). Health-based guidelines for short-term formaldehyde exposure span a range of values (Salthammer et al., 2010). An expert elicitation review of the available literature rated the likelihood of increased short-term UFP exposure causing health effects as medium to high (Knol et al., 2009). Another review noted the substantial experimental evidence and plausible mechanisms for respiratory and cardiovascular effects of UFP intake, but deemed the evidence as “not sufficiently strong to conclude that short-term exposures to UFPs have effects that are dramatically different from those of larger particles” (HEI Review Panel on Ultrafine Particles, 2013). The particles emitted from NGCBs typically don’t have sufficient mass to be governed by health standards for ambient fine particulate matter (PM2.5). There are no standards or guidelines for UFP or fine particle number concentrations.

Emission factors of CO, NO2, and formaldehyde from NGCBs have been measured in laboratory and field studies (Moschandreas and Relwani, 1989; Singer et al., 2010a; Singer et al., 2010b; Traynor et al., 1996). Several studies have reported emission factors and/or indoor concentrations of ultrafine particles resulting from NGCB use (Bhangar et al., 2011; Rim et al., 2012; Singer et al., 2010a; Wallace et al., 2008).

Many studies have reported elevated concentrations of CO and NO2 in homes with natural gas cooking burners, compared to homes with electric cooking (Garrett et al., 1999; Ryan et al., 1988; Schwab et al., 1994; Spengler et al., 1994; Spengler et al., 1983; Wilson et al., 1986; Wilson et al., 1993). A recent study of 350 California homes reported that NO2 and NO concentrations increased with increasing (self-reported) use of NGCBs (Mullen et al., 2016). Several studies have reported substantial ultrafine particle emissions associated with use of NGCBs (Singer et al., 2010a; Wallace et al., 2008).

34

While several measurement-based studies have reported time-resolved CO, e.g. (Mullen et al., 2016), only a few have reported time-resolved or peak NO2 concentrations resulting from NGCB use (Fortmann et al., 2001; Franklin et al., 2006; Moschandreas and Zabransky Jr, 1982).

A recent simulation study assessed the impact of NGCB use on concentrations of CO, NO2, and formaldehyde in a representative sample of Southern California homes (Logue et al., 2014). The study used data on home sizes, cooking activities, and home ventilation rates, along with emission factors measured and reported by Singer et al. (2010a). The study concluded that the weekly highest 1h mean NO2 concentrations exceed 100 ppb in the majority of simulated homes in which NG cooking burners were used without kitchen exhaust ventilation.

The primary strategy for mitigating exposure to pollutants from cooking burners is to use a venting range hood or other kitchen exhaust ventilation (Stratton and Singer, 2014). Recent assessments of range hoods in the U.S. indicate wide performance variations across devices, and across airflow settings and burner configurations for many devices tested (Delp and Singer, 2012; Lunden et al., 2015; Rim et al., 2011; Singer et al., 2012b). Several of these studies used capture efficiency, CE, as the performance metric. CE indicates the fraction of pollutants generated at the cooking appliance that are removed or exhausted by the range hood before they can mix into the air of the home. These studies found that for many range hoods, CE is much higher for the back than for the front cooktop burners. The Logue et al. (2014) modeling study of Southern California homes found that routine use of a venting kitchen range hood with a 52% CE (reflecting performance of a common hood for front burner cooking) should dramatically reduce the percentage of homes with 1h mean NO2 exceeding 100 ppb.

The primary objective of the research reported here was to quantify time-resolved concentrations of NO2 resulting when NGCBs are used under realistic conditions, and specifically to investigate if the threshold of 100 ppb over 1h is commonly exceeded. Researchers also sought to measure concentrations of NO, NOX, CO2, CO, CH2O, PM2.5 (estimated by light scattering) and the number of particles with diameters ≥ 6 nm (most of which are UFP) following controlled burner use. Another objective was to conduct a pilot study of the benefits of using venting range hoods to reduce in-home concentrations of pollutants emitted by NGCBs. This report focuses on results for NO, NO2, CO2, and particle number concentration. Limited results are presented for CO and PM2.5. Additional details are provided in an LBNL report (Singer et al., 2016a).

1.6 Methods 1.6.1 Overview For this study, researchers operated NGCBs and measured the resulting pollutant concentrations in nine homes in the San Francisco Bay area. Experiments were conducted, by permission, when residents were away from the home. Researchers controlled the operation of cooking appliances, ventilation, and forced-air heating systems. The NGCB operation sequences were designed to represent common cooking patterns. To avoid generating pollutants from food preparation, pots containing tap water were used as heat sinks. Air pollutants – including NOX, NO, number concentrations of particles ≥6 nm (PN), formaldehyde, CO, CO2, and

35

estimated PM2.5 (using a light-scattering instrument) – were measured in the kitchen and a hallway or bedroom that was far from the kitchen. CO2 was also measured in a common room between the other two locations but generally closer to the kitchen. NO2 was inferred as the difference between NOX and NO, even though that value likely includes non-negligible amounts of nitrous acid (HONO) (Spicer et al., 1994).

The base set of experiments included operation of each available burner set (cooktop, oven bottom burner, and separate broiler burner where available) with windows closed, no mechanical air mixing (i.e., the forced air heating system, if present, was turned off), and no mechanical exhaust. Additional experiments were conducted with the forced air system operated in fan-only mode when this setting was available, and with a venting range hood when available.

1.6.2 Study Homes The nine homes varied in size and layout, as described in Table 9. They included seven detached houses, one flat (first floor of two-flat duplex), and a small apartment. There were three homes with open floor plans and no walls enclosing the kitchen. Four of the homes had kitchens that were distinct rooms, connected to other rooms in the home via standard interior doorways. Two homes had semi-open kitchens. One of these (labeled H6) had a small galley kitchen with both a floor-to-ceiling passage and a large pass-through connecting the kitchen to the adjacent dining room. The other (H9) had two wide, open passages between the kitchen and adjacent rooms.

Table 9: Study Home Descriptions

ID Floor area (m2)

Levels BR/Ba Year built

Kitchen design

Flooring1 Gas burners2

Venting range hood?

FAU fan on for

mixing? H1 134 1 2/2 1910 Closed Hard CT/O/B Y Y H2 124 1 4/2 1963 Open Hard CT/O/B Y Y H3 117 1.53 2/2 1904 Closed Hard CT/O N N H4 26 1 1/1 <19904 Closed Hard CT/O N No FAU H5 108 1 3/1 1925 Closed Hard CT/O/B Y No

H6 119 2 2/2.5 1991 Semi-open

Hard L1, carpet on stairs & L2 CT/O/B Y Y

H7 226 2 5/3 1990 Open Hard L1, carpet on stairs & L2 CT N Y

H8 219 2 4/3 1990 Open Hard L1, carpet on stairs & L2 CT Y Y

H9 139 2 3/2.5 19865 Semi-open

Hard K, other rooms carpeted CT/O/B Y Y

1 L1 = level 1 or first story; L2 = level 2 or second story; K = kitchen. 2 CT = cooktop; O = oven; B = broiler (top of oven compartment) 3 Small room below kitchen connected via stairwell at back of kitchen (house on hill). 4 Building was renovated and expanded in 1990. 5 Home has been retrofitted for energy efficiency including extensive air sealing, insulation and windows; thus has characteristics of new, energy efficient home.

36

Cooking appliances and ventilation equipment varied across homes. Five homes had a gas range with cooktop, oven bottom burner and waist-high burner; two homes had a gas range with only a cooktop and oven burner; and two homes had a gas countertop cooktop separate from an electric oven. A venting range hood was present in six homes. Six homes had forced air systems that could be operated in fan-only mode.

Study home access was arranged with owners or renters who were paid $200 for each day and $200 for each overnight period that a home was unoccupied and made available for experiments, up to a total allowable payment per home of $600. A single day of experiments required 11h of access to the home without occupants.

1.6.3 Cooking Burners Table 10 summarizes the natural gas cooking appliances in each study home. The burner firing rates were obtained from the nameplate tag found on the appliance.

Table 10: Natural Gas Cooking Appliances in Study Homes

Burner firing rate (kbtu/h) ID Cooktop

type Cooktop

burner type Left front

(LF) Left rear

(LR) Right rear

(RR) Right front

(RF) Oven Broiler

H1 Range Sealed 9.5 16.2 14.2 5 18 15.5 H2 Range Sealed 9.5 17 14 5 18 15.5 H3 Range Open 9 9 9 9 18 None H4 Range Open 9 9 9 9 18 None H5 Range Sealed 12 12 9.2 9.2 18 15 H6 Range Sealed 9.5 9.5 9.5 9.5 16 13.5 H7 Counter Sealed ND ND ND ND Elec. Elec. H8 Counter Sealed 9.5 14.2 11 5 Elec. Elec. H9 Range Sealed 9.5 12 9.5 5 16 12 ND = not determined.

1.6.4 Range Hoods Table 11 summarizes the kitchen exhaust fans in the study homes. Six of the homes had exhaust devices above the cooktop. Two of the venting hoods were “microwave over range” (MOR) appliances that combine the functions of a microwave and externally venting exhaust fan. Home H3 and H4 had no range hoods of any kind. H7 had a non-venting (recirculating) range hood that was operated during two experiments.

37

Table 11: Range Hoods in Study Homes

Airflow from product literature (L/s)

Sound level from product literature (Sones)

ID Type Make /model Low speed

Medium speed

High speed

Low speed

Medium speed

High speed

H1 Vented range hood

Vent-A-Hood B200 MSC

(2) (2) 531 (2) (2) 6.5

H2 Micro-wave exhaust

Electrolux FGMV174KFB 85 (2) 142 (2) (2) (2)

H5 Vented range hood

Zephyr Cyclone AK6500 142 (2) 142 1.5 (2) 5

H6 Micro-wave exhaust

GE Space-maker JVM140

(2) (2) 230 (2) (2) 5.5

H7 Non-vented hood

Broan QS3 (2) (2) 109 0.3–0.54

(2) 4.5–5.54

H81 Vented range hood

Kenmore 233.516891 50 (3) 76 (2) (3) 6

H91 Vented range hood

Broan 42000E 50 (3) 76 (2) (3) 6

1The hoods in H8 and H9 are the same product, sold under different nameplates. 2 Not provided with product literature 3 Setting not available; only 2 settings on this model. 4 Varies with vent geometry.

1.6.5 Burner Operation and Simulated Cooking A procedure was developed to simulate common usage scenarios for the cooktop, oven, and broiler burners. The procedures, used in homes H2 to H9, are described in Table 12. The “Boil/Simmer” and “Sautee/Simmer” activities were combined into a single “Cooktop” procedure.

The procedures in Table 12 were not finalized until after experiments were completed in H1; experiments conducted in H1 thus included variations and combinations of the procedures. In H1, the simulated Roast activity involved 60 min total of appliance operation, corresponding to roughly 40 min of operation after the pot was added. The simulated Broil activity lasted for 40 min total, including 20 min of preheat and 20 min after the pot with water was added (compared to 15 min of simulated broil in H2–H9). Cooktop experiments were conducted either with two fry pans or with two boil/simmer pots. An experiment that was done only in H1 combined a slight variation of the Roast (oven set to 450 °F instead of 425 °F) and the standard Cooktop (Boil + Sautee) activities. In H1, H3, and H4, an additional cooktop experiment was conducted in which two or more cooktop burners were operated at high setting without pots.

38

Cooking appliances generally were operated as found. In a couple of cases the appliances were wiped with wet paper towels to remove large debris.

Table 12: Burner Operating Procedures Used for Most Experiments in Homes H2 to H9

Simulated Activity Description

Boil/Simmer on Cooktop

Add 4L cold tap water to 5L stainless steel pot with 23 cm diameter base. Place uncovered on largest cooktop burner At start, set burner to high. When water reaches rolling boil, cover and turn down burner to just below boiling. Maintain this condition to complete 30 min total burn.

Sautee/Simmer on Cooktop

Place stainless steel fry pan with 30.5 cm diameter base on second largest burner. At start, set burner to high. At 2 min from start, slowly add 1L cold tap water. At 6 min from start, turn burner to medium setting. Adjust burner down as needed to avoid boiling. Stop burner at 30 min total burn.

Roast with Oven Burner

Remove top oven rack and place bottom rack to allow 5L stainless steel pot to fit into oven. At start, set oven temperature to 218 ºC (425 °F). When oven reaches setpoint (or after 20 min if setpoint not yet reached), place uncovered pot containing 1L water on bottom oven rack. Maintain this condition for 30 min; then turn oven off and remove pot. (Note: total burn time varies with time to reach setpoint temperature.)

Broil with Broiler Burner

Remove top oven rack and place bottom rack such that top of 5L stainless steel pot on rack is approximately 13 cm below broiler burner at top of oven. At start, set oven to “broil” with oven door ajar. After 20 min, place 5L stainless steel pot containing 1L water into oven. With door ajar, maintain this condition for 15 min. Turn burner off, remove pot, and close door.

1.6.6 Execution of Experiments Upon arriving to each home, researchers reviewed the planned experimental procedures with the host (homeowner or renter) and obtained her/his signature for the agreed usage periods. Together with the host, researchers conducted a walk-through to identify potential hazards, locate controls for the forced air unit (if relevant), and confirm acceptability of the planned placement of monitoring equipment (for electrical capacity). The host then left the home to return at the agreed time.

Study homes were used for varying duration as summarized in Table 13. In the homes where researchers had access during consecutive days but not the intervening night, all instruments were shut down when the host returned at the end of the first day, and restarted when the host left the home the following morning. When the home was accessible for a night-day sequence, monitoring instruments operated continuously overnight.

The protocol prior to starting cooking experiments included set-up and airflow checks of air quality instrumentation, recording information about the home and appliances, measurements to characterize performance of the range hood (if it vented to outdoors), and a check that all windows were closed and interior doors open. There were two exceptions to the general approach of operating homes with all windows closed and interior doors open. It was discovered mid-way through experiments at H4 that there was a bathroom window open about 15 cm; the window was left in this position for all experiments in H4. And in H8 there was a

39

large opening in the master bathroom ceiling, related to a home construction project; this was addressed by keeping the door to the master bath closed throughout experiments in H8.

Table 13: Dates and Periods That Homes Were Used for This Study

Home ID Dates Sequence1 H1 Feb 2-4 N-D-N-D H2 Feb 17-18 D-D H3 Feb 19-20 N-D H4 Feb 21-22 N-D H5 Feb 25-26 D-N-D H6 Feb 27-28 D-N-D H7 Feb 29 D H8 Mar 1 D-D H9 Mar 4-6 D-N-D

1 D=Day; N=Night

Each experiment included the following elements. (1) Air mixing in the home was either driven by the ambient temperature and pressure drivers, or by operation of the forced air unit (FAU) mixing fan. (2) If present, the venting range hood was either operated or not. (3) The simulated cooking procedure was followed. (4) Following the end of the cooking procedure, and range hood use if it was part of the experiment, ventilation and mixing in the home were held in the same condition for 60-90 minutes or more, to enable analysis of mixing and decay. When the FAU mixing fan was used, in most cases it was started approximately 10 min prior to the start of the cooking activity, and operated until the mixing condition was established for the next experiment. For experiments involving range hood use, the hood was started approximately 1 min prior to the start of cooking, and remained in operation until 5 min after the cooking procedure ended. Range hoods were most commonly operated on the highest setting, though lower settings were used in several experiments. Excluding one experiment in H1 in which the range hood was started 15 min after the end of a cooking procedure, mixing and decay periods varied from a minimum of 54 minutes to a maximum of overnight.

Summary descriptions of conditions for all experiments conducted in each home are provided in a series of tables in the Results section. The sequence of experiments varied across homes.

1.6.7 Air Quality Measurements Researchers continuously monitored the concentrations of carbon dioxide (CO2), carbon monoxide (CO), nitric oxide (NO), nitrogen oxides (NOX), formaldehyde (CH2O), total number concentration of particles greater than 6 nm (PN), number concentrations of particles in 6 size bins from 0.3 µm to 10 µm, and the estimated mass of PM2.5 via forward light scattering. Temperature and relative humidity were also monitored. The instruments used to measure each of these parameters are noted in Table 14. The values reported for NO2 are based on the difference between measured NOX and NO, and likely include HONO (Spicer et al., 1994); for

40

simplicity, the NO2 value reported by the instrument is referred to as NO2 throughout the remainder of this report.

Table 14: Description of IAQ Devices and Location Specific Packages

Target Metric Symbol Units Instrument or device (Measurement principle)

Locations 1

Nitric oxide Nitrogen dioxide Nitrogen oxides

NO NO2 NOX

ppbv Thermo Scientific Analyzer Model 42 TSI-API Analyzer Model 200E

(Chemiluminescence; catalytic NO2 reduction)

L1 L3

Nitrogen dioxide2 NO2 ppbv Aeroqual Series 500 with NO2 sensor2 (Electrochemical sensor)

Varied

Carbon Monoxide CO ppmv Lascar EL-USB-CO data logger (Electrochemical)

L1, L2, L3

Carbon Dioxide CO2 ppmv ExTech SD800 (Infrared absorption)

L1, L2, L3

Temperature Relative humidity

T RH

°C %

HOBO UX100-003 and Extech SD800 L1, L2, L3

Formaldehyde CH2O ppbv Shinyei FMM-MD (Colorimetric) 2 L1, L2, L3 Estimated PM2.5 (mass)

PM g/m3 TSI DustTrak II Model 8530

(Forward light scattering) L1, L3

Number conc. of particles >6 nm

PN #/cm3 TSI 3781 Condensation Particle Counter (Growth by H20 condensation; laser counting)

L1, L3

Number conc. of particles 0.3–2.5 um

P(2)ange #/cm3 MetOne BT-637S 2 (Laser particle counter)

L1, L3

1 L1=kitchen, L2=central; L3=distant. 2Data collected with these instruments has not been analyzed and is not presented or discussed in this report.

Air quality monitoring occurred in two primary locations and one secondary location in each home. The primary locations were the kitchen and a room or hallway far from the kitchen. The distant location was used to determine an approximate lower bound of combustion pollutant concentrations and potential exposures. The kitchen and distant location had nearly identical collections of instruments, mounted on mobile carts. The kitchen cart was placed no closer than two meters from the cooktop in all homes other than the small apartment H4, in which the cart was as far from the stove as possible. The devices on the cart sampled air at heights of roughly 1.4-1.65 m. The third monitoring package was installed on a table in a common room (living room, dining room, great room) closer to the kitchen, or connected to the kitchen in an open floor plan (L2 or “central” location is defined in Table 14). Images of monitoring configurations are provided in a separate report (Singer et al., 2016a).

Air quality data were recorded at 1 min or more frequent intervals for the majority of analytes. The formaldehyde sensor recorded a reading every 30 minutes.

Particle instruments were operated side-by-side in a Lab test chamber to confirm consistency prior to deployment. CO2 instruments were calibrated after the field study and calibration factors were applied to the raw data. CO instruments were used as received new from the manufacturer, without an independent calibration. The NOX and NO2 analyzers were single-point calibrated at the start of the field study, then checked for zero and span calibration in the

41

laboratory prior to most home deployments. Despite this, it was observed that in most homes, the NO and NO2 (NOX–NO) concentrations were offset between instruments, at times when the home was thought to be well mixed. The observed differences between the two NOX analyzers were used to determine an offset adjustment. Because the NO concentrations reported by the bedroom instrument were consistently in line with data obtained from nearby regulatory monitoring sites, at times when the homes were not impacted by indoor NOX emissions, the offset was typically applied to the NOX analyzer on the kitchen cart.

1.6.8 Range Hood Performance Characterization Measurements were made to quantify airflow and sound characteristics of all the venting range hoods in the study. The capture efficiency (CE) of cooking burner combustion products was measured for five of the six venting hoods. CE could not be determined for the hood in H1 because the range hood vent was inaccessible.

Exhaust air flow from the hoods was measured using a balanced-pressure flow hood method described by Walker et al. (2001). The method uses a calibrated and pressure-controlled variable-speed fan (Minneapolis Duct Blaster, Energy Conservatory1) connected to either the exhaust inlet (preferred approach) or outlet. The Duct Blaster is connected using a customized transition that was fabricated / adapted at each site using cardboard and tape. An example, in H8, is shown in Figure 7. Using a pressure sensor, the Duct Blaster fan is controlled to match the flow of the exhaust fan while maintaining the pressure at the exhaust inlet at its normal value when the Duct Blaster is not installed. The pre-calibrated speed versus flow relationship of the Duct Blaster provides the flow through the exhaust fan.

Figure 7: Example of Pressure-Balanced Flow-Hood Airflow Measurement (H8)

1 www.energyconservatory.com

42

Sound levels were measured using the AudioTools app (version 8.9.X) from Studio Six Digital

(www.studiosixdigital.com) on an iPhone6. Researchers used the “Real Time Analyzer” tool, a 1/3-octave band analyzer, which provides sound pressure (in decibels, dB) as a function of frequency. The sound pressure distribution was measured for background conditions (hood off) and for each available fan speed when the house was in quiescent condition, i.e. with no air quality monitoring devices operating. The A-weighted total sound pressure (dBA) was reported by the app as a summary statistic. Additionally, the procedure used to determine the sound level reported by the Home Ventilation Institute (HVI) in their Certified Home Ventilating Products Directory was applied; the calculation procedure is described in HVI Publication 915, available on the HVI web site (www.hvi.org).

Capture efficiency (CE) refers to the fraction of pollutants emitted from the cooking burner (and cooking, when applicable) that are removed by the venting range hood before mixing into the air of the kitchen. CE can be estimated by calculating both the mass flow of CO2 exiting through the range hood, and the mass generation rate based on fuel composition and the assumption of complete combustion (Singer et al., 2012b). In this study, a simpler approach was used that compares the flow of CO2 through the hood under the normal operating condition to the flow of CO2 when a foil curtain is used to extend the hood over the cooktop was used to ensure perfect or nearly perfect capture. This approach assumes no change in airflow between the conditions, meaning the CO2 mass flow changes proportionally with the CO2 concentration. CE is calculated using CO2 concentrations measured under the normal operating condition (CN) and with the hood extended to create nearly perfect capture conditions (C100), and background concentrations with the cooking burners off (C0), as shown in Equation 1.

CE = (CN – C0) / (C100 – C0) (1)

CO2 concentrations in the exhaust from the range hood were measured using a PPSystems EGM-4 analyzer drawing from the ducting above the range hood. An example of the setup for this procedure, from H8, is shown in Figure 8.

Figure 8: Measurement of Range Hood Capture Efficiency in H8

Left panel shows hood with foil curtain to achieve 100% capture. Right panel shows foil curtain lifted and taped to cabinetry above to measure CE without curtain.

43

1.6.9 Attributing Concentration Profiles to Individual Cooking Events The primary objective of this study was to quantify the impact of using NGCBs on pollutant concentrations. However, in order to complete as many experiments as possible in the limited time available in each house, the interval between experiments was not long enough to allow concentrations to return to background levels before the next experiment. Therefore the analysis included a disentanglement procedure, in which the data for each analyte were decomposed into estimated concentration profiles for each cooking event.

Disentangling the data for a particular cooking event involved four steps: (1) fitting a first-order decay model to the experimental data; (2) using this model to predict how the previous cooking event would have evolved, had the current event not taken place; (3) using the same model to predict how the current cooking event would have evolved, had the next event not taken place; and (4) finding the difference between the two, in order to estimate what the current event would have looked like in the absence of other events.

Figure 9 shows an example. Here researchers sought to find the concentration profile for the second of the three cooking events shown (the 8th experiment in H6, identified as H608). First, the first-order decay rate for the current event was estimated using the data recorded after the cooking itself ended. This involved fitting a decaying exponential model to a span of data starting after the range hood turned off, and ending before the FAU turned off. In selecting intervals for fitting data, researchers considered the movement of pollutants across zones of the house, as indicated by the dynamics of the two measured locations. In general, researchers waited until the two series “came together” – ideally by reaching or approaching the same concentration, and alternately by starting to follow the same downward trend. For the distant location, the decay interval often extended past the start of the simulated cooking event.

The decay rate was estimated using a least-squares fit to the measured data, with an assumed background concentration. The background represents the indoor concentration that would result from outdoor pollutant entry and indoor losses (including air exchange for all pollutants, and deposition for NO2 and PN). When choosing the background concentration, researchers performed least-squares fits to find both decay rates and backgrounds and considered all experiments that had the same airflow conditions (e.g. central mixing fan on or off) during the same day, in addition to considering measurements at the start and end of the day or at any other time that informed the background estimate. For Experiment H608, the background was set at 10 ppb and the resulting fit gave a decay rate of 0.17 h-1.

After fitting a decay model to the data following a generation event, researchers used it to project the profile from the prior cooking event. In Figure 9, this projection has two stages: a short period, from the start of the current cooking event, up to the time the range hood turns off; and a longer period, during which only the FAU is on (dashed purple line). The decay rate for the first stage, during which both the hood and the FAU operate, was estimated by adding the decay rates for the FAU and the hood in quadrature, i.e., as the square root of the sum of the squares of the individual rates (the decay rate for the hood was estimated as the hood flow rate divided by the house volume). Note that it was not possible to estimate the decay rate for this

44

first stage directly from the data, because there is simultaneous generation with a source of unknown strength.

The projection of the prior cooking event represents a prediction of how the prior experiment would have evolved, had the current event not taken place. Similarly, the same decay model was used to project the current event, past the first change related to the following experiment (dashed orange line). In this experiment, that first change was turning off the FAU; in other experiments, it was the beginning of the cooking itself. Extending the current experiment was necessary to estimate the four-hour time-integrated concentration due to the event.

Note that both the projection of the prior experiment, and the projection of the current experiment, used the same decay rate and background, as estimated from the data collected immediately after cooking ended for the current experiment. This ensured the disentanglement for a given experiment most closely reflected the measurements made during that experiment. It also means that a single span of data could be processed using two different decay rates. Consider, for example, projecting the results of H608 past the beginning of the next event, H609. When predicting exposures due to H608, researchers used the decay rate observed for this event. However, when predicting exposures due to H609, the decay rate observed for that experiment was used.

Figure 9: Disentangling NO Data for Experiment H608 (Middle Peak) in the Kitchen of H6

The experiments occur too closely in time to distinguish individual concentration profiles in the measured data (solid blue line). Therefore researchers projected (dashed purple line) the profile from the preceding cooking event, past the start of the event being analyzed (here, Experiment H608). This projection accounts for the higher removal rate that occurred from range hood use

45

during the first part of H608. Similarly, researchers projected the profile for H608 (dashed orange line), to remove the effects of the following experiment. The difference between the projections gives the estimated profile for H608 (solid green line), disentangled from the preceding and following experiments. This profile was used to estimate the integrated concentrations that would have occurred had H608 taken place without confounding experiments before and after.

Subtracting the projection of the previous experiment from the measured and projected data for the current experiment yielded the estimated profile for the current experiment, i.e., a prediction of what the current experiment would have yielded, had there been no cooking events immediately before or after it (solid green line). These adjusted data also have the assumed background concentration added in, since otherwise the current event would start with zero concentration.

This extrapolation and subtraction procedure was applied separately for CO2, NO, NO2, and PN, in the kitchen and bedroom, as data were available. The time intervals used for decay fits are shown as shaded areas in data plots provided for each home in the report Appendix. Inferred baselines and decay rates are provided for the four analytes in the Appendix.

The measurements and extended concentration profiles were analyzed to determine the highest 1h average concentration of each pollutant at each location, and also to find the integrated concentration over 4 h.

1.6.10 Estimating Emission Factors from Ambient Concentrations The highest 1h concentrations was used to estimate fuel-normalized emission factors for NO2, NOX, and PN, using the method described in (Singer et al., 2010a). Briefly, this approach calculates the emission rate of CO2 from natural gas combustion, and uses the measured concentration ratios for each pollutant to CO2 to calculate an emission factor. For NOX and NO2, the calculation proceeds according to Equation 1, where the first term on the right side is the ratio of NO or NOX to CO2, and the second term is a property of the fuel; 1.1 mol CO2 / MJ fuel was used based on Singer et al. (2010a).

(1)

Equation 2 was used to calculate PN emission factors.

(2)

46

1.7 Results 1.7.1 Measured Range Hood Performance

Measurements of range hood performance parameters are presented in Table 15. Consistent with a prior study by the research group (Singer et al., 2012b), the measured airflows were substantially below rated values for five of the six installed hoods. Interestingly, the estimated sound ratings (in sones) were lower than the rated values for many of the hood settings.

Table 15: Measured Performance Parameters of Range Hoods in Study Homes

Measured flows [L/s] (% of rated flow)

Bkg [dBA]

Measured sound [dBA] (Calculated Sones)

Home ID Type Low

speed Medium speed

High speed

Low speed

Medium speed

High speed

H1 Hood 66

(2) 108 (2)

148 (59%) 30.1 57.3

(3.0) (2) 59.2

(3.7)

H2 Microwave oven/ exhaust

66 (78%)

(2) 76 (54%) 36.8 63.9

(4.6) (2) 72.1

(6.2)

H5 Hood 135 (98%)

(2) 153 (2) 28.9 58.9

(4.0) 62.0 (4.1)

66.7 (4.8)

H6 Microwave oven/ exhaust

43 (2)

(2) 49 (45%) 30.7 59.2

(3.6) (2) 62.5

(4.6)

H81 Hood 20 (40%)

(3) 30 (40%) 32.9 54.0

(2.2) (3) 58.2

(3.7)

H91 Hood 39 (79%)

(3) 19 (64%) 39.8 54.1

(2.0) (3) 61.4

(3.7) 1The hoods in H8 and H9 are the same product, sold under different nameplates. 2 Setting available but performance information not provided with product literature 3 Setting not available; only 2 settings on this model. 4 Varies with vent geometry.

Table 16 presents the estimated capture efficiencies measured with the new field test method described in Section 2.2.8. Consistent with a prior field study (ibid), the performance of several of the hoods was dramatically different for the front and back cooktop burners. For most ranges, the performance of the back cooktop burners is a good indicator of capture for oven emissions.

Table 16: Measured Capture Efficiency of Range Hoods in Study Homes

Low speed High speed Home ID Hood type Front burners Back burners Front burners Back burners

H1 Hood NM1 NM1 NM1 NM1 H2 Microwave 25% >95% 35% >95% H5 Hood 61% 68% 72% 84% H6 Microwave 31% 88% 31% 93% H81 Hood 59% 68% 65% 80% H91 Hood 25% 74% 36% 75%

1Not measured; there was no way to access the range hood exhaust duct without aesthetic damage.

47

1.7.2 Experiments Conducted All experiments conducted in the study homes are listed in Table 17,

Table 18, and Table 19.

Table 17: Experiments Conducted in study homes H1-H3

Expt ID

Date Burners & use Burn start time

Burn end time

Decay - conditions

unchanged (min)

Range hood

setting

Mixing fan

H101 2-Feb CT: 3 burners, no pots

20:10 20:30 Overnight

H102 3-Feb BR: Broil 10:09 10:49 60 min H103 3-Feb CT: Fry 2 pans 11:50 12:20 97 min H104 3-Feb OV: Roast 13:57 14:57 68 min H105 3-Feb CT: Boil 2 pots 16:05 16:35 62 min H106 3-Feb OV+CT 17:37 18:37 Overnight H107 4-Feb BR: Broil 7:01 7:41 54 min High H108 4-Feb CT: Fry 2 pans 8:40 9:11 54 min High H109 4-Feb OV: Roast 10:11 11:11 54 min High H110 4-Feb CT: Boil + fry 12:11 12:41 54 min High H111 4-Feb OV+CT 13:41 14:41 60 min High H112 4-Feb BR: Broil 15:41 16:21 15 min Off / High H113 4-Feb BR: Broil 18:15 18:55 60 min On H114 4-Feb OV+CT 19:55 20:55 Overnight On H201 17-

Feb OV: Roast 10:15 10:56 90 min

H202 17-Feb

CT: Boil + fry 12:26 12:52 90 min

H203 17-Feb

BR: Broil 14:22 14:57 90 min

H204 17-Feb

CT: Boil + fry 16:27 16:52 90 min

H205 18-Feb

CT: Boil + fry 8:46 9:11 85 min High

H206 18-Feb

OV: Roast 10:41 11:22 80 min Medium

H207 18-Feb

CT: Boil + fry 12:57 13:24 76 min On

H208 18-Feb

BR: Broil 14:45 15:20 100 min High

H301 19-Feb

CT: 4 burners, no pots

22:15 22:20 Overnight

H302 20-Feb

OV: Roast 8:10 9:00 90 min

H303 20-Feb

CT: Boil + fry 10:30 11:05 90 min

H304 20-Feb

OV: Roast 12:35 13:25 90 min

H305 20-Feb

CT: Boil + fry 14:55 15:30 100 min

48

Table 18: Experiments Conducted in Study Homes H4-H6

Expt ID

Date Burners & use Burn start time

Burn end time

Decay - conditions

unchanged (min)

Range hood

setting

Mixing fan

H401 21-Feb

CT: 2 burners, no pots

20:45 21:10 Overnight

H402 22-Feb

OV: Roast 5:30 6:20 90 min

H403 22-Feb

CT: Boil + fry 8:15 8:45 92 min

H404 22-Feb

OV: Roast 10:40 11:32 88 min

H405 22-Feb

CT: Boil + fry 13:25 13:55 125 min Bath fan

H501 25-Feb

OV: Roast 15:30 16:10 90 min

H502 25-Feb

CT: Boil + fry 17:40 18:12 90 min

H503 25-Feb

BR: Broil 19:43 20:18 87 min

H504 25-Feb

CT: Boil + fry 21:45 22:15 Overnight

H505 26-Feb

CT: Boil + fry 5:30 6:02 88 min High

H506 26-Feb

OV: Roast 7:30 8:13 100 min Medium

H507 26-Feb

CT: Boil + fry 9:53 10:29 95 min Medium

H508 26-Feb

OV: Roast 12:05 12:47 92 min

H601 27-Feb

OV: Roast 11:30 12:10 90 min

H602 27-Feb

CT: Boil + fry 13:40 14:10 90 min

H603 27-Feb

BR: Broil 15:40 16:15 105 min

H604 27-Feb

CT: Boil + fry 18:00 18:30 125 min On

H605 27-Feb

OV: Roast 20:45 21:25 Overnight

H606 28-Feb

CT: Boil + fry 5:25 5:55 89 min High

H607 28-Feb

OV: Roast 7:30 8:12 85 min High

H608 28-Feb

CT: Boil + fry 9:52 10:22 83 min Med On

H609 28-Feb

OV: Roast 12:10 12:51 119 min

49

Table 19.:Experiments Conducted in Study Homes H7-H9

Expt ID

Date Burners & use Burn start time

Burn end time

Decay - conditions

unchanged (min)

Range hood

setting

Mixing fan

H701 29-Feb

CT: Boil + fry 9:13 9:40 89 min

H702 29-Feb

CT: Boil + fry 11:10 11:38 87 min High*

H703 29-Feb

CT: Boil + fry 13:20 13:47 92 min On

H704 29-Feb

CT: Boil + fry 15:20 15:47 88 min High* On

H801 1-Mar CT: Boil + fry 9:00 9:30 90 min

H802 1-Mar CT: Boil + fry 11:30 12:00 75 min High On

H803 1-Mar CT: Boil + fry 13:40 14:10 84 min High

H804 1-Mar CT: Boil + fry 15:50 16:20 70 min On

H901 5-Mar OV: Roast 10:10 11:00 90 min

H902 5-Mar CT: Boil + fry 12:30 12:58 93 min

H903 5-Mar CT: Boil + fry 14:41 15:11 83 min On

H904 5-Mar BR: Broil 16:34 17:09 Overnight

H905 6-Mar CT: Boil + fry 8:04 8:34 65 min High On

H906 6-Mar OV: Roast 9:46 10:36 78 min High

H907 6-Mar CT: Boil + fry 12:06 12:36 84 min Low On

H908 6-Mar OV: Roast 14:00 14:48 82 min On

*Recirculating (non-venting) range hood.

1.7.3 Measured Pollutant Concentrations To elucidate some of the major themes seen in the data, sample results for base conditions are presented in Figure 10 and Figure 11. These plots present data for NO2, NO, PN, and CO2, in both the kitchen and the distant bedroom monitoring locations. The same parameters are presented for each full day of experiments in a series of plots in the Appendix of Singer et al. (2016a).

Figure 10 presents data from Day 1 in H3. As expected, kitchen CO2, NO, NO2 and PN increased quickly as burners fired at the maximum settings: cooktop burners set to the highest flame, or oven or broiler burner firing continuously. Concentrations remained elevated throughout the simulated cooking events as the cooktop burners were set to medium-low, oven burners cycled to maintain temperature, and broiler burners continued to fire continuously. The kitchen traces show more short-term variability owing to their proximity to the source. After the burners were switched off at the end of a simulated cooking event, concentrations in the kitchen started to decay as pollutants mixed throughout the house and were removed by ventilation and infiltration. NO2 and PN also were removed by deposition. With each burner use, concentrations in the bedroom started to rise after a short delay, representing the transport / mixing time from the kitchen to the distant location.

50

Bedroom concentrations increased over a time interval that was similar in duration to burner use, but shifted later in time. For NO and CO2, concentrations decayed slower in the bedroom than in the kitchen, presumably due to lower rates of outdoor air directly entering the bedroom, compared to the kitchen. For PN and NO2, each of which deposit indoors at rates that are fast enough to compete with air exchange as a removal process, the kitchen and bedroom concentrations decayed at the same rate in the two rooms.

Day 1 in H5 shows a somewhat different pattern (Figure 11). In this house, the distant bedroom had less indoor air exchange with the kitchen. Following the start of an emission event without the FAU operating, the bedroom concentrations increased more slowly, and reached peak levels that were much lower than in the kitchen. Concentrations in the kitchen and bedroom converged for NO and almost converged for CO2 only in the first and fourth experiments. Decay rates in the kitchen and bedroom were similar.

The absolute concentrations, relative dynamics and peak concentrations, and the effect of FAU operation on the relative dynamics of pollutants in the kitchen and bedroom varied widely across homes. In general, as the delay increased, so did the difference between the kitchen and distant room peak concentrations. The closest coupling without FAU use occurred in H4, H7, and H8. The coupling in H4 is explained by it being a small, 2-room apartment. H7 and H8 are newer homes with open floor plans that resulted in closer connections between spaces. FAU operation substantially increased PN decays in H9, which had a high-performance (MERV13) filter installed in the FAU. In the single FAU experiment in H6 (on Day 2), NO2 decays were much faster than they were in the experiments without the FAU operating; this suggests removal of NO2 in the air handler.

51

Figure 10: Air Pollutant Concentrations Measured on First Day of Testing at House H3 under base conditions (no range hood or FAU operation)

52

Figure 11: Air Pollutant Concentrations Measured on First Day of Testing at House H5 under Base Conditions (No Range Hood or FAU Operation)

53

1.7.4 Time-Integrated Pollutant Concentrations under Base Conditions Summary results for all experiments with base conditions of no mechanical mixing (no FAU operation) and no use of range hood or other exhaust fan are provided in Figure 12 and Figure 13. These figures show the highest 1h and 4h time-integrated concentrations in the kitchen and bedroom locations following each simulated cooking event. For this presentation, the homes are grouped according to the opening between the kitchen and the rest of the home, and ordered by home size. Data are presented for NO, NO2, and PN.

Overall, concentrations of each pollutant varied widely both across and within homes. As expected, bedrooms had substantially lower pollutant concentrations than kitchens, across all homes except for the two-room apartment, H4, and the open floor plan house, H8. For NO and NO2, there was no trend across homes associated with the burner used in the simulated cooking event (cooktop, oven or broiler). For PN, there were several homes that had much higher concentrations when using the oven or broiler burner compared to the cooktop, but no homes in which the cooktop produced substantially higher PN. Unsurprisingly given its very small size, H4 had the highest concentrations of NO and NO2. That H4 did not have higher PN concentrations than other homes suggests that the variance in PN emission rates had a larger impact than variations in home size. The data also indicate a trend of higher kitchen concentrations in homes with enclosed kitchens (H1, H3, H4, and H5) relative to homes with semi-open kitchens (H6 and H9) or open floor plans (H2, H8, H9). The trio with open floor plans included the two largest homes, which contributed to the generally lower concentrations observed in those homes.

The plot of highest 1h concentrations shows the NAAQS benchmark of 100 ppb NO2 over 1h. Four of the nine homes had kitchen levels exceed this value, and two other kitchens had 1h NO2 concentrations of at least half this value. Three of the nine homes had bedroom NO2 levels exceed 50 ppb. This suggests significant exposures may occur for anyone at home when natural gas burners are used for even a single, substantial cooking event.

For the vast majority of experiments, there were negligible increases in CO and PM2.5. Researchers thus limited quantitative analysis for these pollutants only to those cases in which concentrations were observed to increase by approximately 9 ppm for CO and 20 µg m-3 for PM2.5. These criteria were satisfied by 3 experiments for CO, and three for PM2.5, and those only in the kitchen. Results for these experiments are provided in Table 20.

Table 20: Result for CO and PM2.5 in Experiments Meeting Criteria of CO rising at Least 9 ppm and PM2.5 Rising at Least 20 µg m-3

Expt Cooking FAU Parameter [units] Highest 1h Highest 4h H402 OV No PM2.5 [g m

-3 h] 238 378 H404 OV No PM2.5 [g m

-3 h] 86 131 H603 BR No PM2.5 [g m

-3 h] 24 53 H402 OV No CO [ppm h] 9.5 21.7 H403 CT No CO [ppm h] 8.6 17.7 H405 OV No CO [ppm h] 7.4 11.5

54

Figure 12: Highest 1h Time-Integrated Concentrations in Kitchen and Bedroom Resulting From use of Natural Gas Burners in Simulated Cooking Activities

55

Figure 13: Highest 4h Time-Integrated Concentrations in Kitchen and Bedroom Resulting from Use of natural Gas Burners in Simulated Cooking Activities

56

Whereas short-term CO concentrations in homes using gas and electric cooking burners have been reported in several studies, including the recent work of (Mullen et al., 2016), researchers are aware of few published reports of short-term NO2 or NO measurements associated with cooking burner use in homes. In a UK-based study with limited relevance to the US owing to differences in cooking equipment, (Franklin et al., 2006) used an innovative approach in asking participants to open a passive sampler whenever the oven or hot plate was used at least 15 min. A companion passive sampler was open continuously through the study period. Data were collected in 24 homes with flued gas cookers, 8 homes with unflued gas cookers, and 21 homes with non-gas cookers. Peak samplers were open an average (SD) of 3.7 (2.6) h and long-term samplers were open for 74.1 (5.8) h. Peak NO2 concentrations were significantly higher in homes with flued and unflued gas cooker compared to homes with non-gas cookers. In the homes with unflued gas cookers, the geometric mean (95% confidence interval) peak concentrations were 44 (26–75) ppb. In the homes with non-gas cookers the peak concentrations were 13 (10–19) ppb. (Note that results have been converted from the published units of µg m-3 to ppb).

In an extensive study of air pollutant emissions associated with cooking, (Fortmann et al., 2001) measured time-resolved concentrations of CO, NO, and NO2 outdoors and in the kitchen, living room, and master bedroom of a 76.6 m2 (824 ft2) single story house, which appears from the floor plan to have had an enclosed kitchen. Boiling water on the natural gas cooktop burner over 1h produced concentrations (across the three rooms) of 2.8–3.5 ppm CO, 219–293 ppb NO, and 53–74 ppb NO2. Operation of the oven for 2h without food produced concentrations of 2.0–2.1 ppm CO, 272-280 ppb NO, and 79–94 ppb NO2 across the rooms. Outdoor concentrations during the cooktop and oven experiments were 1.2 and 0.5 ppm CO, 56 and 24 ppb NO, and 19 and 17 ppb NO2. Concentrations over the cooking events exceeded 9 ppm CO during a fish broil (8.6–10.1 ppm) and oven cleaning (14.8–19.9 ppm) with gas burners and were also elevated (7.5–7.9 ppm) during electric oven cleaning. Average NO2 concentrations during the cooking period exceeded 90 ppb in at least one room during fish broiling, making French fries, baking lasagna, preparing a full meal, and oven cleaning with gas burners. The highest NO2 with an electric burner was during oven cleaning: LR concentrations of NO2 reached 42 ppb with outdoor concentrations at 22 ppb.

Researchers identified only a few studies that reported the impact of gas cooking burners, distinct from food preparation, on particle number concentrations in homes (Fortmann et al., 2001; Wallace, 2006; Wallace et al., 2008). In a 76.6 m2 (824 ft2) single story house, which appears from the floor plan to have had a closed kitchen, (Fortmann et al., 2001) reported PN (>30 nm) of 44 x103 cm-3 when operating the gas cooktop with a pot of water over 1h and 88.5 x103 cm-3 when the gas oven was operated for 2h without food. Concentrations in the kitchen were 5.5 x103 cm-3 and 3.9 x103 cm-3 before cooktop and oven use, respectively. In a 400 m3, 3-story townhouse, (Wallace, 2006) reported concentrations of PN (>10 nm) of 5.8 x103 cm-3 during 36 events of tea preparation using the gas stove, compared to PN of 3.4 x103 cm-3 during 888 periods when no indoor sources were present. Results from (Wallace et al., 2008) indicate that the PN levels reported in these two prior studies, and even in the current study, are likely under-reported: the median diameter of PN emitted from their gas stove was is in the range of 4-7 nm for a naked cooktop burner (no pots, and grate removed), 5.5–20 nm when a pot or pan

57

was used to boil water or cook, and 4–24 nm when the over or broiler burner was used. For the stovetop without the grate, peak concentrations of 2-64 nm particles were 290–2200 x103 cm-3 in the kitchen and 90–740 x103 cm-3 in the master bedroom. Peak concentrations of 2-64 nm particles were 48–450 x103 cm-3 in the master bedroom.

1.7.5 Repeatability Results for replicate experiments are provided in the following two tables. Table 21 presents the highest 1h time-integrated concentrations, and Table 22 presents the highest 4h results. Replicates were most consistent for NO and least consistent for PN. For NO, NO2, and PN, replicates showed similar consistency for 1h and 4h results. By contrast, CO2 had much more variability in 4h vs. 1h results.

Table 21: Relative Deviations (RD) and Relative Standard Deviations (RSD) of highest 1h Time-Integrated Concentrations for Replicated Conditions

Replicates CO2 NO NO2 PN B K B K B K B K H202 CT 385 653 315 124 80241 272426 H204 CT 222 428 No data 202 No data 86 58288 160641 RD 54% 42% 44% 37% 32% 52% H302 OV 676 769 260 443 79 188 136108 462680 H304 OV 607 774 241 382 109 195 111229 292114 RD 11% 1% 8% 15% 32% 3% 20% 45% H303 CT 656 815 306 458 105 168 75519 171467 H305 CT 723 707 281 467 148 300 117536 243596 RD 14% 9% 2% 34% 56% 44% 35% 35% H402 OV 2105 933 519 745969 H404 OV No data 1836 No data 944 No data 546 No data 678747 RD 14% 1% 5% 9% H501 OV 439 692 213 411 36 148 107478 345440 H508 OV 339 607 166 400 31 130 71321 271881 RD 26% 13% 25% 3% 15% 13% 40% 24% H502 CT 291 408 149 276 20 82 46321 144795 H504 CT 264 392 130 227 27 78 60987 201738 RD 10% 4% 13% 20% 26% 4% 27% 33% H601 OV 474 756 130 231 73 187 120050 419191 H605 OV 460 795 139 251 85 203 112871 342922 H609 OV 432 747 130 234 86 193 88439 284396 RSD 5% 3% 4% 2% 9% 4% 15% 5% Mean 19% 13% 12% 12% 23% 17% 30% 29% SD 18% 14% 8% 15% 11% 20% 11% 19% N 6 7 5 7 5 7 6 7

58

Table 22: Relative Deviations (RD) and Relative Standard Deviations (RSD) of Highest 4h Time-Integrated Concentrations for Replicated Conditions

Replicates CO2 NO NO2 PN B K B K B K B K H202 CT 816 1335 582 185 121702 350038 H204 CT 321 690 No data 326 No data 129 72353 195312 RD 87% 64% 57% 36% 51% 57% H302 OV 2089 2178 685 871 174 331 233139 578513 H304 OV 1634 1974 610 838 228 373 189379 398404 RD 24% 10% 12% 4% 27% 12% 21% 37% H303 CT 1828 1958 797 942 223 319 121421 222836 H305 CT 2047 1918 724 887 315 484 180079 316888 RD 11% 2% 10% 6% 34% 41% 39% 35% H402 OV 3789 1565 724 881460 H404 OV No data 3293 No data 1516 No data 784 No data 808337 RD 14% 3% 8% 9% H501 OV 1112 1711 556 885 59 248 171779 464093 H508 OV 835 1426 398 777 60 208 111801 364780 RD 28% 18% 33% 13% 1% 18% 42% 24% H502 CT 630 989 325 610 33 147 70646 199512 H504 CT 605 819 290 437 43 119 90334 254752 RD 4% 19% 11% 33% 25% 21% 24% 24% H601 OV 1543 2089 410 598 178 420 245526 620264 H605 OV 1437 1898 431 592 211 425 218415 490175 H609 OV 1353 2057 394 588 208 428 172328 419537 RSD 2% 1% 1% 1% 4% 2% 8% 3% Mean 26% 18% 13% 17% 18% 20% 31% 27% SD 32% 21% 12% 20% 15% 15% 16% 19% N 6 7 5 7 5 7 6 7

1.7.6 Effect of Range Hood Use The effects of operating a venting range hood during cooking are presented in Figure 14. This figure presents the percentage reduction in the highest 1h concentration, calculated as the difference between experiments with range hood use and analogs without range hood use. Included in this figure are the calculated reductions from using the bath fan as the only available exhaust device in H4, and the recirculating range hood in H7.

Broadly, these results indicate that use of range hoods can yield substantial reductions in cooking burner pollutant concentrations both in the kitchen and throughout the house. The positive impact of the range hood was larger than the variability of the experimental method, producing net reductions in all cases. The most benefit was seen in H1, which had a range hood with large capture volume and a measured airflow of 108 L/s. This hood, which produced reductions mostly in the range of 80–95%, was of similar design to hoods that showed very high capture efficiency in prior studies: hood B5 in the field study by (Singer et al., 2012b), and hood P1 in the lab study by (Delp and Singer, 2012). The next most effective hood, in house H5, also

59

had characteristics reported by (Singer et al., 2012b) to be important to performance: it extended to fully cover the front burners and it had airflow substantially above the benchmark of 95 L/s.

The only other hood with reductions mostly exceeding 50% was the over-the-range microwave with exhaust fan in H2, with an exhaust fan that moved 76 L/s at high speed. Prior studies (Delp and Singer, 2012; Lunden et al., 2015; Singer et al., 2012b) have found over the range microwave exhaust fans to vary widely in their capture efficiency, in large part explained by variations in airflow but likely also relating to disadvantageous geometry. The reductions observed for H5 were consistent with those measured for hood H2 operating at similar airflow in (Singer et al., 2012b). The range hoods in H6, H8 and H9 had calculated reductions of 0–50% with more than half of the results falling in the top half of that range. These hoods had substantially lower airflows and did not extend to cover front burners fully. The bath fan in H4 reduced concentrations by 15–40% across the measured species, suggesting a modest benefit that could have been caused by method variability. The recirculating range hood in H7 showed small net reductions (≤10%) for NO2, NO, and CO2 (which had the same reduction as NO), and a larger reduction for PN (~30%). All were within the variability of replicate experiments.

The relationships of range hood effectiveness for PN, airflow, and burner position have been discussed by (Rim et al., 2012) and (Lunden et al., 2015). In measuring particles down to 2 nm, Rim et al. found that removal effectiveness was lower for 2–6 nm particles than for particles >6 nm. Since a large number of particles are in the lower size range, that effect is expected to reduce the overall effectiveness reported by Rim et al. relative to what would be reported for 6 nm and larger particles, as measured in this study. The “A” hood in the Rim et al. study was similar in design and airflow to the hood in H8 and H9 in this study. The roughly 40% reductions in cooktop-emitted PN calculated with range hood operation in those homes is roughly midway between the 31(6)% for front burners and 54(9)% for back burners reported by Rim for Hood A. The “B” hood tested by Rim et al. was similar in design to the hood in H5 in this study and the measured airflows of the hood in H5 were between the medium and high flows measured for Hood B in Rim et al. Yet the effectiveness for PN was substantially lower in H5 than reported for by Rim et al. for their Hood B. The apparent difference may simply result from the greater variations in PN associated with the experimental approach of this study.

60

Figure 14: Percent Reductions in Highest 1h Kitchen Concentrations Calculated by Comparing Experiments with Range Hood Use to Analogous Experiments without Range Hood Use

Shaded grey areas show the range – back burner to front burner – of capture efficiency measured for the hood at the flow rate used in the experiment which is overlaid.

1.7.7 Effect of FAU Use Table 23 presents results for the experiments in which the air handler of the forced air heating system (FAU) was operated starting approximately 10 min before cooking and through the decay period. The experiments with FAU use are compared to experiments with the same cooking activities with no FAU use. From the basic physical consideration that it increases mixing, researchers expected the FAU to reduce kitchen concentrations and increase bedroom concentrations relative to the same cooking activity with no mixing – unless there are losses in the forced air system (e.g. removal by a furnace filter). The trend of a more positive increase in the bedroom relative to the kitchen is evident in the highest 1h CO2, NO, and NO2 in two H1 experiments, the experiment in H6, and the CO2 and NO2 data in H8. Concentrations in the kitchen either decreased or increased very slightly (which is expected from variability in emissions), whereas concentrations in the bedroom were dramatically higher with FAU operation. The PN results for FAU use in H1 don’t follow this trend, but in H6 there was still a

61

much larger increase in bedroom PN vs. kitchen PN when the FAU operated. The trend was not apparent in the FAU experiments in H7 and H9. The experiment in H2 had only CO2 and PN data to compare. Overall, there was not a consistent trend of FAU impacts in this study.

Table 23: Highest 1h Concentrations in Paired Experiments to Investigate the Impact of Operating the Forced Air Unit (FAU) Mixing Fan Compared to Base Conditions

Cooking CO2 NO NO2 PN Expts FAU B K B K B K B K BR H102 Off 352 673 196 464 28 81 97905 389076 H113 On 452 573 222 337 46 87 67338 215539 FAU effect 28% -15% 13% -27% 63% 8% -31% -45% OV+CT H106 Off 622 838 272 496 104 205 93620 218952 H114 On 741 851 315 426 128 211 56138 148643 FAU effect 19% 1% 16% -14% 24% 3% -40% -32% CT H202 Off 385 653 - 315 - 124 80241 272426 H204 Off 222 428 - 202 - 86 58288 160641 H207 On 377 576 165 303 47 114 78438 270065 FAU effect 24% 6% - 17% - 9% 13% 25% CT H602 Off 313 779 170 390 36 108 45193 170603 H604 On 568 627 283 320 59 85 107997 238636 FAU effect 81% -20% 67% -18% 64% -22% 139% 40% CT H701 Off 362 468 155 225 44 84 173800 341730 H703 On 286 447 132 230 46 91 133297 242676 FAU effect -21% -5% -15% 2% 5% 9% -23% -29% CT H801 Off 381 431 215 278 41 65 176137 239857 H804 On 292 403 164 217 28 60 124629 - FAU effect -23% -6% -24% -22% -31% -7% -29% OV H901 Off 209 346 68 192 15 78 30641 89573 H908 On 190 305 72 162 21 75 5105 29406 FAU effect -9% -12% 6% -16% 37% -4% -83% -67% CT H902 Off 196 331 103 225 14 49 73952 169964 H903 On 262 393 137 274 18 69 40350 175894 FAU effect 34% 19% 33% 22% 29% 42% -45% 3%

The lack of a clear trend across all the data collected for this study does not mean that FAU use does not impact pollutant spatial distributions in homes. Rather, the results suggest that the mixing effect was of the same order of magnitude or less than variations caused by other factors, including emission rate and variability of non-mechanical mixing, for the homes studied.

62

1.7.8 Spatial Variations The calculated time-integrated concentrations in the kitchen and bedroom were compared for each experiment to quantify the potential spatial variations in cooking pollutant exposures in homes; results are shown in Figure 15 and Figure 16. Across the sample, the least pronounced spatial variations occurred in the small two-room apartment (H4) and two of the homes with open floor plans (H8 and H7). Spatial variations of highest 1h NO in these three homes were almost all in the range of 1.3 to 2.1 (there was one outlier in H8). The greatest spatial variations of 1h NO2 were in H9, the two-story 1990s home retrofitted for deep energy savings: K/BR ratios in this home were 3.3 to 6.6. For many of the homes, the kitchen to bedroom ratios were somewhat higher for the highest 1h compared to the 4h time-integrated concentrations. This is consistent with the difference between the two locations being largest during the period during and just after cooking. Within each home, spatial variations for NO (and CO2, not shown in the Figures) were smaller than for NO2 and PN. This results because deposition loss rates for NO2 and PN were competitive with mixing times in all of the homes.

63

Figure 15: Ratios of Highest 1h time-Integrated Concentrations in Kitchen and Bedroom

64

Figure 16: Ratios of Time-Integrated Concentrations in Kitchen and Bedroom over 4h after Cooking Burner Use Commenced

1.7.9 Emission Factors Calculated emission factors for NO2, NOX, and PN are shown in Figure 17.

65

Figure 17: Emission Factors Calculated from Ratios of Highest 1h NO2, NOX, and PN to Highest 1h CO2

Mass emission rates for NOX calculated using a molecular mass of 46 g/mol by convention.

66

The emission factors derived from the simulated cooking experiments can be compared with those presented for samples of burners in two earlier reports: (1) a compilation of data from studies completed prior to 1990, mostly from the 1970s and early 1980s (Traynor et al., 1996); and (2) a study that reported emission factors for previously used appliances including 13 cooktops, 12 oven burners and 6 broiler burners that were first sold in 1992 to 2007 and tested at ages 2 to 17 years (Singer et al., 2010a). Note that NO2 emission factors from the 25 gas ranges reported by (Moschandreas and Relwani, 1989) are included in the Traynor review. (Traynor et al., 1996) presented results for NOX and (Singer et al., 2010a) presented results for NOX, NO2, and PN.

NOX emission factors from the current study ranged from 28 to 64 nanograms per Joule (ng/J) with most of the data between 30 and 45 ng/J, and a geometric mean of 40. These are a bit higher than the emission factors reported by (Singer et al., 2010a), which ranged from 17 to 47 ng/Jwith most of the data between 30 and 36 ng/J, and the results reported by (Traynor et al., 1996), which had a geometric mean of 32 ng/J.

Most of the NO2 emission factors calculated in the current study were between 5 and 15 ng/J, with geometric mean 10.3 ng/J. The prior study by (Singer et al., 2010a) reported similar results with 28 of the 31 burner sets having NO2 emissions within the range of 5-15 ng/J.

The biggest difference between the current results and previously reported emission factors is for PN. In this study, the calculated PN emission rates ranged from 2.5x109 J-1 to 2.2x1010 J-1 with a geometric mean of 1.0x1010 J-1. In the results reported by (Singer et al., 2010a), PN emissions were much more variable. The highest emission factors for each burner type were in the same range as those determined in the current study: 1.4, 0.5, and 2.6 x1010 J-1 for CT, OV and BR burners respectively. But for each burner type reported in the earlier study, there were many more burners with PN emission factors below the lowest values reported for the current study. This difference may result from the cleaning, pre-conditioning, and more repeat experiments in the earlier study. This observation derives from the hypothesis that the particles were formed by volatilization of organics compounds that were deposited on cooktop, oven or broiler burner surfaces, as reported by (Wallace et al., 2015) for electric burners and other hot surfaces.

1.8 Summary and Conclusions The short-term indoor air quality impacts of using natural gas cooking burners were investigated through controlled experiments in nine residences in Northern California. Cooktop, oven, and broiler burners were operated in a prescribed manner intended to simulate use during typical cooking activities, while avoiding emissions associated with food preparation. Homes were set to defined ventilation and mixing conditions with a baseline configuration of windows closed and no operation of the air distribution system of the forced air heating unit (FAU). The impact of FAU operation was investigated in a limited way through experiments conducted in six of the homes. Air pollutants were monitored continuously in the kitchen and the bedroom area to quantify concentrations of NO, NOX, CO2, CO, and the number of particles with diameters ≥6 nm as an indication of ultrafine particles (UFP). The difference between NOX and NO was calculated and reported as NO2 even though that value likely

67

includes non-negligible amounts of nitrous acid (HONO). Estimated PM2.5 was measured by light scattering. A pilot investigation of the benefits of range hoods and kitchen exhaust ventilation was conducted with experiments in eight homes, including six with venting range hoods, one with an exhaust fan in a room adjacent to the kitchen, and one with a recirculating range hood. A total of 65 experiments were conducted during 15 days and 7 overnight periods.

The data were analyzed to disentangle the time-concentration profile for each experiment from the effects of prior and subsequent experiments. This enabled estimation of the highest 1h and 4h time-integrated concentrations for CO2, NO, NO2, and PN in each experiment. The results were used to quantify the pollutant concentrations in each room and relative levels of pollutants in the two locations over time. The highest 1h concentrations of NOX, NO2, PN, and CO2 were used to estimate fuel-normalized emission factors (ng J-1 and particles J-1). Differences between experiments with range hood use and analogous experiments – defined by cooking burner and home mixing condition – without range hood use were used to calculate pollutant reductions from using the range hoods.

Pollutant concentrations varied widely across and within homes. Concentrations were much lower in bedrooms versus kitchens, except for a small two-room apartment, H4, and an open floor plan house, H8. In several homes PN concentrations were much higher when using the oven or broiler burner compared to the cooktop. No clear trends by burner were observed for NO or NO2 across apartments. The data show larger deviations between kitchen and bedroom concentrations in homes with enclosed kitchens (H1, H3, H4, and H5) relative to homes with semi-open kitchens (H6 and H9) or open floor plans (H2, H8, H9). The trio with open floor plans included the two largest homes, which contributed to the generally lower concentrations observed in those homes. Four of the nine homes had kitchen levels exceed the national ambient air quality standard threshold of 100 ppb NO2 over 1h, and two others had 1h NO2 concentrations of at least half this value. Three of the nine homes had bedroom NO2 levels exceed 50 ppb. This suggests significant exposures may occur for anyone at home when natural gas burners are used for even a single, substantial cooking event.

Results from the pilot of study of kitchen ventilation indicate that range hoods can substantially reduce cooking burner pollutant concentrations, both in the kitchen and throughout the house. The positive impact of the range hood was larger than the variability of the experimental method, producing net reductions in all cases. An 80-95% reduction was observed in H1, which featured a range hood with large capture volume and a measured airflow of 108 L/s. The hood in H5 also had characteristics important to performance: it extended to fully cover the front burners, and it had airflow substantially above the benchmark of 95 L/s. When operated on high speed, it reduced 1h concentrations by 56–78%. The only other hood with reductions mostly exceeding 50% was the over-the-range microwave with exhaust fan in H2. The range hoods in H6, H8 and H9 had reductions of 0–50% with more than half of the results falling in the top half of that range. These hoods had substantially lower airflows and did not extend to cover front burners fully. The bath fan in H4 reduced concentrations by 15–40% across the measured species, indicating a modest benefit. The recirculating range hood in H7 showed small net reductions (≤10%) for NO2, NO, and CO2, and a larger reduction for PN (~30%) that were all within the variability of replicate experiments.

68

The results of this limited field study generally confirm the finding of a recent simulation study (Logue et al., 2014) that using natural gas cooking burners without venting commonly produces short-term NO2 concentrations in homes that approach or exceed the federal ambient air quality standard of 100 ppb.

The results also show that operation of a venting range hood with airflow and geometry described by Singer et al. (2012) substantially reduces concentrations of cooking related pollutants throughout the home.

1.9 Recommendations Based on the findings of this field study and the related, prior work referenced herein, the authors offer the following policy recommendations.

Efforts should be made to increase awareness (a) that natural gas cooking burners are a source of air pollutant emissions into homes, and (b) that these pollutants can be controlled with an appropriately-sized venting range hood or other kitchen exhaust ventilation. Since cooking with electric burners also produces pollutants, kitchen exhaust ventilation should be available in all homes, and operated as a precaution whenever cooking occurs. Since the performance of most hoods is much better when cooking is done on the back cooktop burners, this practice should be encouraged to improve safety. Gas utilities could play a valuable role in publicizing these messages.

Building standards should require that range hoods have airflows of at least 95 L/s and cover front burners or preferably demonstrate performance through a standard test. Such a test is currently under development by ASTM.

69

CHAPTER 2: Workshop on California Combustion Appliance Safety Procedures 2.1 Overview On February 2nd 2015, Lawrence Berkeley National Lab (LBNL) convened a meeting of stakeholders and experts to discuss challenges and explore potential improvements to combustion appliance safety (CAS) assessment protocols used in California’s single-family energy retrofit programs. The meeting had the following objectives: (1) bring together the designers and implementers of CAS protocols for all of the major California single-family retrofit programs, (2) provide historical and technical background on the subject of CAS testing in retrofits, (3) initiate discussion current challenges and potential improvements to CAS protocols, and (4) identify research and data needed to support protocol changes.

The workshop opened with an LBNL presentation about the potential hazards associated with combustion appliances and proposed specific risk-management objectives for CAS assessment. In-person and remote participants then introduced themselves and noted concerns, suggestions, and thoughts on best practices in current CAS protocols. LBNL then delivered a brief technical talk outlining some of the basic physics for combustion appliance venting and the risks they pose to occupants. This was followed by a presentation by Richard Heath and Associates (RHA) that outlined the historical research and development of the CAS procedures used in the California low-income retrofit programs. Informal stakeholder discussions continued over the lunch hour, and the meeting reconvened with a short description by the Gas Technology Institute (GTI) of relevant combustion appliance safety field research being done in the U.S. Dept of Energy Building America program. The remainder of the afternoon was a mixture of presentations and open discussion. LBNL compared existing CA protocols with one another and outlined some concrete suggestions for changes to current protocols. These changes were then subject to open group discussion and refinement, and the meeting finished with discussion of next steps, data/research needs and opportunities for collaboration to move the process forward. Several themes emerged from the discussions. (1) There was widespread enthusiasm for simplicity in test protocols and great interest in potential harmonization across programs. (2) The risk management framework and development of clear safety objectives resonated with attendees. And (3) there appear to be many opportunities for simplifying protocols and procedures.

2.2 Participants Participants attended either in-person or by-phone, and they represented stakeholder groups including public policy, research, retrofit contractors, trainers and program management/development consultants. The following is a list of participants, grouped by institution.

70

Table 24: LBNL CAS Testing Harmonization Stakeholder Meeting Participants List

Last Name First Name Affiliation/Organization Attendance

Wahl Andy AC Home Performance In-Person

Bayba Russell Build It Green In-Person

Dellosso Daniel Build It Green In-Person

Glidden Torsten Build It Green By-Phone

Montalbo Javier Build It Green In-Person

Marsolais Elizabeth CAEATFA By-Phone

Ziaja Sonya California Energy Commission In-Person

Gibbs Syreeta CPUC By-Phone

Eastburn Jeff CSD By-Phone

Doll Jeff Energy Commission By-Phone

Brand Larry Gas Technology Institute By-Phone

Fischer Marc LBNL In-Person

Lorenzetti David LBNL In-Person

Rapp Vi LBNL In-Person

Singer Brett LBNL In-Person

Walker Iain LBNL In-Person

Wray Craig LBNL In-Person

Perez Tomas Pacific Gas & Electric In-Person

Magnuson Leif PG&E In-Person

Amader Joseph PG&E sub contractor In-Person

Fagilde Gary PG&E Training Center-Stockton In-Person

Dezell James RHA, Inc. In-Person

Aronson Nathan RHA, Inc. In-Person

Grieco Andre RHA, Inc. In-Person

O'Bannon James RHA, Inc. In-Person

Willaims Barbara RHA, Inc. In-Person

Whitmore Ed SDG&E By-Phone

Allen Craig SoCalGas In-Person

Gonzalez Hugo SoCalGas In-Person

Caudle Sylvester Ron SoCalGas In-Person

Graves Duane State of CA CSD By-Phone

71

2.3 Vision and Framework LBNL convened the workshop with the vision that CAS test procedures should be based on a risk management framework. In this approach, combustion appliance safety hazards are identified then assessed and prioritized based on risk. CAS procedures are designed and carried out in a way that efficiently minimizes, monitors and controls the risks posed by these hazards. To whatever extent possible, CAS procedures should be simplified and harmonized across state programs, such that complexity and cost are minimized and more resources are available for mitigation measures.

The vision also conceives that all CAS protocols should provide a robust assessment with “must address” elements to achieve a baseline level of protection. CAS procedures already recognize that transient spillage events cannot be entirely eliminated and the allowance of gas cooking appliances without automatic venting is tacit acceptance that limited spillage of exhaust gases into the home is tolerable. Researchers propose the risk management targets that a home passing a CAS assessment should have an effectively zero risk of a catastrophic spillage event and near-zero risk that substantial spillage will occur more than infrequently under normal operating conditions. (In this context, “substantial” means spillage that creates problematic moisture or indoor air pollutant levels.) In other words, there should be assurance that venting appliances will properly vent and there will be sufficient ventilation available to address cooking burner use when the home is operated under “normal” conditions. The protection should be robust such that it will persist through typical equipment degradation. Robustness also means that commonly occurring engineering or administrative failures (e.g. blocked air vent, failure to use kitchen ventilation) should not result in a catastrophic spillage event. In other words, the combustion safety of a home “touched” by a retrofit program should be robust, and appliance operation should be resilient in the face of things that commonly happen in homes, such that one thing going wrong will not cause harm.

Harmonized CAS procedures should identify critical hazards accurately, consistently and efficiently. Accurate means that procedures consistently identify hazards and avoid flagging safe appliances. Consistent means that different contractors/inspectors get the same results. Efficient means that the protocols do not use more time or effort than are required to identify a hazard.

2.4 Issues and Recommendations The workshop agenda and discussions were aligned with this vision. Participants offered specific suggestions and feedback on LBNL suggestions. The following list of issues and recommendations were developed by LBNL based on prior work and input provided at the workshop. The recommendations should be considered as starting points for future discussion and honing.

Issue: Safety objectives of many CAS procedures are not clearly enough defined.

While CAS tests used in California retrofit programs were developed to ensure and document that program measures do not create hazards, the procedures are understood by many contractors and homeowners to provide an overall assessment / assurance of safety.

72

Recommendations:

• Explicitly define safety objective(s) for each CAS protocol.

• Add safety improvements as an explicit program objective to all CA retrofit programs; use retrofit programs to improve safety of natural gas distribution network.

• As a starting point, consider the safety objectives noted in Vision statement of this document.

Issue: Variations in CAS protocols are confusing; harmonization would have benefits.

Several participants noted the challenge faced by contractors who conduct work under multiple retrofit programs. As they are using the varying protocols, they must track the idiosyncrasies and tolerances in the face of unclear results that are taught during training for each program protocol. And contractors more than occasionally encounter homes that pass a test that is required for one program, yet would fail under a different one of the existing protocols.

Recommendations:

• Key stakeholders for California retrofit programs should collaboratively explore the potential develop harmonized protocols.

• Harmonized protocol could be structured around a basic set of minimal requirements that all programs adopt and enhanced or additional elements that could be adopted by programs seeking enhanced safety. Minimal requirements should be set to achieve the safety objectives noted in the Vision statement.

Issue: The design of the ESA program is so different from other state retrofit programs that it requires a special procedure.

The Energy Savings Assistance (ESA) program is designed for wide reach among low-income homes in IOU service territories. It is designed for streamlined, basic retrofits that can be achieved at the scale of hundreds of thousands of homes per year. The NGAT test was developed specifically for this program. As long as the program remains in its current form, it likely will not be able to support a CAS procedure that is as thorough and robust as the procedures used for the more substantial and comprehensive retrofit programs.

Recommendations:

• If a harmonized protocol features a basic set of minimal requirements, try to align NGAT program procedure with those basic requirements.

• Since ESA program is likely not air sealing homes to levels that are tight enough to cause depressurization-inducted spillage problems, consider changes to focus NGAT much more on basic safety and operability.

• If ESA ever pilots or otherwise expands to include serious air sealing, reconsider NGAT specifics.

73

Issue: Gas cooking appliances present a number of unique challenges.

Natural gas cooking appliances are not directly vented and measurements in homes have shown that their use leads to elevated levels of combustion pollutants. Simulation based analysis for the population indicates that their use can cause sub-critical acute and chronic hazard conditions in homes. Cooking appliance emissions are difficult to robustly measure/assess during CAS inspection, even with protocol revisions. This is due to the impact that the presence of a cooking pot can have on emissions of CO and other combustion pollutants (e.g., NO2 and ultrafine particles). Emissions of CO typically are much higher when pots are placed over burners, compared to open burners. Many retrofit programs lack of funds to install kitchen exhaust ventilation. And even when installed, effectiveness of kitchen venting is often unreliable due to occupant use patterns. While not a topic of the workshop, multifamily residences pose special additional difficulties in providing kitchen ventilation.

Recommendations:

• Programs should prioritize addressing hazards created by gas cooking appliances and recognize that CO is not the only, nor the most common, potential problem pollutant from unvented combustion. Protocols should not treat a low CO exhaust measurement as an indication that there is no hazard associated with gas cooking burners (which are always unvented).

• To the extent feasible, programs should treat installation of venting range hoods or other kitchen ventilation as a necessary health and safety expense.

• Programs should incorporate administrative controls for exhaust from kitchen appliances. These controls should include educating customers about the hazards associated with the exhaust from gas cooking appliances, as well as pollutants associated with cooking of food, and advise the regular use of a venting range hood or other ventilation when cooking.

• Eliminate center-of-kitchen CO testing for kitchen appliances (as it is not robustly effective).

• Specify measurement of oven burner CO in the oven flue. Reconsider threshold for oven CO. (ANSI limits for new appliances are likely too high and BPI limit may be too low).

• A robust and easily implementable test for cooktop CO should be developed. The test should include pots on burners as the presence of pots can dramatically increase CO production.

Issue: Worst-case depressurization (WCD) testing is problematic for many reasons.

Setting up current WCD testing is laborious and highly dependent on current weather conditions, particularly when pressure verification is required. WCD testing is a major source of confusion and time for inspectors. The confusion and difficulty of these tests grows with increasing home size and number of forced air systems, exhaust fans, combustion appliance zones (CAZ) and appliances in each zone. Depressurization limits used in the CSD Wx and

74

EUC programs are not reliable indicators of appliance safety. In most homes, the worst-case condition is one that never occurs under normal use, and should it ever occur, it would likely occur for limited time. Furthermore, the extreme ventilation conditions that are used to create maximum depressurization bring high air exchange with outdoors; these airflows translate to a dilution protection against any spillage that could be induced. In the workshop, there was disagreement over what the prescribed challenge condition should be. The operation of exhaust fans in CA homes is unknown, such that a “perfect” challenge condition cannot be reliably specified.

Recommendations:

• Eliminate pressure-verified WCD testing in favor of a simple, prescribed and more relevant “Challenge” condition.

• Do not customize the Challenge condition for each CAZ.

• Specify appropriate Challenge condition for each type of appliance (e.g. forced air furnaces should be treated differently than wall furnaces).

• Based on input from workshop participants, researchers recommend the following as key features of a generic challenge condition:

o Use NGAT door configurations.

o Operate any of the following exhaust devices (up to 3) that are present in the home: clothes dryer, kitchen fan (on low speed) and the most used bathroom fan.

o Test with and without Forced Air Unit (FAU)

• If program includes an air-tightness measurement (blower door test), limit challenge testing to homes that meet some threshold for airtightness (i.e., do not test homes that are unlikely to depressurize other than at very high exhaust airflow rates).

Issue: Not all tests contribute meaningfully to appliance and occupant safety.

Some tests in California CAS protocols are essential to identifying hazardous appliances, such as visual inspections of appliance and flue vent integrity and CVA evaluations. Other tests, such as thresholds for draft pressure and CAZ depressurization are much less useful, and national standards that are the basis for the parts of the EUC protocols (e.g., BPI) have recognized this and removed these tests from their protocols. Additionally, current protocols require testing in non-operational configurations (e.g., with house-garage door open, or with interior furnace closet door open) that do not reflect appliance performance in-use. Finally, current protocols test appliances in attached spaces under WCD, even though this condition is often irrelevant to appliance functioning in these spaces.

Recommendations:

• For appliance operational tests, measure flue CO and assess spillage at 5 minutes.

• Eliminate CAZ depressurization and draft pressure limits.

75

• Wherever feasible, use appliance flue CO tests rather than appliance ambient CO.

• Eliminate redundant testing of flue CO and appliance ambient CO (CSD).

• Do not conduct challenge depressurization tests on appliances located in garages and attics unless those spaces show evidence of regular occupancy. Challenge tests may be warranted for appliances in sealed basements / crawlspaces as these spaces could be the conduit for most or all of the air coming into the home during cold weather, “stack effect” airflow.

Issue: Risk varies by appliance type, location within home, other home parameters and use.

Current CAS protocols do not test appliances according to their risk factors, and as a result some appliances are “over” tested (e.g., appliances in attached spaces or fan-assisted furnaces) and others are “under” tested (e.g., gas cooking appliances).

Recommendations:

• Reduce testing criteria for appliances located in attached spaces.

o For appliances in vented attics or outside, perform visual inspection only.

o Appliances in other attached spaces should be tested under as-found conditions for flue CO and spillage at 5 minutes.

o Appliances in conditioned space should be tested under reduced Challenge condition, but must pass in order to proceed.

• Key questions for appliances in attached spaces are:

o Does space show signs of occupancy?

o Is space vented outside approximately to code?

o Is there a good seal between the attached space and living space?

o Can the space serve as the primary path for all house air?

• Consider using Building America criteria2 for determining if an appliance is “outdoors”.

• Reduce or eliminate testing for fan-assisted furnaces. Always perform visual inspection.

• Reduce testing for water heaters, because their operation schedules and smaller burner sizes make them less risky than heating appliances. Possibly challenge condition with only 1 fan operating?

Issue: Many existing homes do not meet current mechanical codes.

2http://apps1.eere.energy.gov/buildings/publications/pdfs/building_america/measure_guide_combustion_safety_appliancezone.pdf

76

This issue is most commonly encountered in relation to codes for kitchen ventilation, combustion ventilation air, vent design and materials, and clearance to combustibles.

Recommendations:

• Prioritize inspection to ensure basic vent integrity (connected, no gaps or damage, etc.)

• Ensure CVA to code.

• Regarding venting that does not meet (National Fuel Gas) Code, develop guidance to distinguish non-code compliant venting that must be addressed from non-code compliant venting that can be flagged but not fixed.

• Distinguish between hazardous defects (e.g., disconnected flue vent, negative vent slope) and code compliance issues (e.g., clearance to combustibles in interior furnace closet).

• Use administrative controls as appropriate to address code compliance issues.

• Directly fix hazardous defects.

Issue: It is confusing when conflicting results may be obtained from multiple or varied CAS inspections on the same home.

Even when the protocols are carried out “correctly”, there are many sources of confusion and inconsistency in retrofit CAS testing. For example, participants noted that when GSRs are called to a job site, as required by some CAS protocols, they often conclude that the situation is not hazardous. This sometimes creates confusion and/or frustration for the contractors and/or household occupants. Other sources of inconsistency include varying WCD set-ups by different contractors, a problem that is exacerbated by the overly detailed set-up and verification of WCD. Similarly, house, room and appliance ambient CO tests can yield highly variable results depending on the local airflow patterns and circumstances at the time of testing. Finally, variable weather can lead to very different tests results, both for depressurization testing and draft/spillage testing.

Recommendations:

• Gas Service Representatives

o Consider aligning CAS procedures and standards with those of the GSRs, or vice versa.

o Inform weatherization contractors and inspectors of GSR procedures when conducting an inspection tasks are within their training.

o Re-evaluate and reduce the conditions that require a GSR call.

o Clarify the exact purpose of the GSR call, so that when appropriate, the CAS inspector can assess.

77

o If not already clear, clarify safety liability when GSR approves a condition that would not otherwise remain unaddressed as part of CAS procedures.

• Contractor confusion

o All programs should target implementation of CAS testing using a standardized form (e.g., CASIF for CSD). CSD is currently pilot testing a tablet based interactive form, which could be used to provide prompts and facilitate complexity while reducing risk.

o See proposed prescribed Challenge condition without pressure measurement.

• Ambient CO tests

o Eliminate house and room ambient CO measurements

o Use flue CO measurements wherever feasible.

• Weather variability

o Consider possible value of developing a tool that would normalize the test conditions to general conditions based on annual weather.

Miscellaneous Issues and Recommendations

• How should protocols account for dilution airflow when assessing potential hazard depressurization induced spillage? Could a calculation tool be developed for contractors that asks for characteristics of home, appliance and airflow, then assesses hazard.

• How to account for the effect of a fireplace when checking appliance draft? How to account for the high exhaust / ventilation airflow rates induced by fireplaces?

• For common vented appliances, first do Challenge test on the appliance with lower heat output (not both together).

• Consider opportunities to use technology to identify problems when they happen, rather than using tests to predict future potential problems. Extension of smoke/CO alarm approach.

• Most CO alarms only protect for life-safety and may give a false sense of security, as they are not measuring other combustion related hazards. Consider proposing a new threshold for CO alarms to indicate a sub-lethal, but potentially important health hazard.

2.5 Identfied Data and Research Needs Many of the challenges and suggestions listed above can be addressed and implemented without additional research, i.e. there is sufficient justification based on available information. Others were regarded as requiring justification through research and/or demonstration. Two main pathways for developing justification for potentially contentious issues were discussed: (1) analysis of data that has been or is being collected by state retrofit programs and (2) a field

78

study designed to investigate specific hazards and/or demonstrate proposed changes to protocols. Below are LBNL’s proposed ideas for existing program data analysis and new field research.

Existing program data. California retrofit programs are currently gathering large amounts of data on each home that they assess and retrofit, and this data could be useful in answering questions related to issues and potential improvements. Data collection varies by program, but can include house characteristics, appliance characteristics, retrofit measures and CAS tests results, such as appliance failures by type/location/reason. Provided below are some examples of characterization and research questions that could be informed through analyzing program data.

• Characterize generally the current hazards being identified by CAS inspections:

o How frequently do homes/appliances fail CAS tests?

o What specific tests are identifying hazardous appliances?

o Which appliances are failing?

o Where are failing appliances located within the home?

o How many appliances fail that are in attached spaces?

o How commonly do specific issues arise, e.g., cracked heat exchangers?

• Below are some examples of specific research questions that could be addressed:

o What CO levels are produced by appliances in California homes?

o How often do homes fail house or room ambient CO tests?

o Do ESA/CSD homes ever fail due to depressurization? Are some homes so leaky that depressurization testing does not need to be done at all?

o Is there any evidence that infiltration measures in the ESA program actually lead to depressurization induced spillage/problems?

o What is the distribution of worst-case depressurization levels in CA homes touched by CSD Weatherization programs (obtained from CASIF)? What is the impact of FAU operation on depressurization, in isolation and in conjunction with exhaust fans?

o Can estimates be made of the numbers of appliance failures that would not fail under proposed, harmonized protocols? Estimate the costs of over-testing?

Proposed Field Study. There are many protocol elements and potential changes that could be investigated through targeted data collection in homes. The field study that was initially envisioned to occur as part of this research was supposed to focus on the effectiveness of a modified set of CAS diagnostics at identifying problem conditions. The planned study would align with a study with the same general objective that is being conducted by two Building

79

America teams under the leadership of GTI. The key elements of the study would be the following:

• Target homes would fail an existing protocol but barely pass or barely fail a proposed alternative protocol. (Inclusion of homes that barely fail is important to see if those homes actually experience sustained and substantial spillage in practice).

• Appliances, exhaust devices, other household operational parameters, indicators of spillage, and both time-resolved and time-integrated pollutant concentrations would be monitored for periods of weeks during one or more seasons.

80

CHAPTER 3: A Field Study of Wall Furnace Venting and Coincident Exhaust Fan Usage in 16 Northern California Apartments 3.1 Introduction The assessment of combustion appliance safety (CAS) is a cornerstone of the health and safety provisions embedded in nearly all of the residential energy retrofit programs in the United States (US). Although no specific risk reduction objectives are identified, the general understanding is that testing is designed to protect residents against exposures to air pollutants that can be generated in the combustion process, to keep excess moisture from being introduced into the residence, and of course to identify gas leaks.

Combustion products are primarily composed of carbon dioxide (CO2) and water vapor; the large quantities of moisture that get introduced into the residence when larger combustion appliances are not venting can degrade materials and increase risk of dampness and mold-related illness. Combustion also produces nitrogen oxides (NOx), including some in the form of the respiratory irritant nitrogen dioxide (NO2), and can produce carbon monoxide (CO) and “ultrafine” particles at levels that present hazards to occupants.

All combustion safety testing protocols allow a brief period of spillage during cold start-up at the beginning of a heating cycle, but the time within which venting must begin varies from one to five minutes. Likewise, the requirements for assessing venting robustness vary, both programmatically and in practice. Starting with a smoke pencil placed adjacent to the draft diverter, some implementers consider venting acceptable when the majority of the smoke is pulled up into the vent, even if it meanders and some escapes. Others require that all of the smoke be pulled into the vent directly. The quantitative connection between these standards and the fraction of combustion gases that will spill in practice is not known.

The majority of combustion appliances in U.S. residences vent properly (i.e., no prolonged spillage of combustion by-products) during “natural” condition tests3. For appliances that use indoor air for combustion – in contrast to “direct vent” appliances that draw combustion air directly from outdoors – depressurization of the combustion appliance zone (CAZ) can weaken or reverse flow in the vent pipe, resulting in spillage of combustion products. The CAZ is the interior space that contains the combustion appliance and in many cases can be limited or expanded by opening or closing interior doors. Depressurization can be impacted by weather, exhaust fan usage, use of vented clothes dryers, interior door positions and forced air system operation. Accordingly, all CAS procedures include some sort of challenge test, in which the

3 Appliances failing natural condition tests often have clear defects, in that they violate code requirements for combustion ventilation air, flue pipe design, etc. or they have developed defects in use, such as clogged flue vent pipes, due to bird nests or the like.

81

house is intentionally set up to depressurize the CAS with respect to outdoors and venting / spillage is evaluated. The current standard is to set up a “worst-case” depressurization (WCD) challenge condition in which all exhaust devices that can “communicate” with the appliance by being in a continuous airflow path are turned to their highest settings. Some test protocols require the implementer to find the condition of maximum depressurization between the CAZ and outdoors by operating all exhaust fans on highest settings then opening and closing interior doors and operating and not operating the forced air system. Other tests prescribe door position and whether the forced air system is used to set up a “worst-case” challenge.

In the past several years, some consensus has developed around the belief that worst-case testing is overly conservative and that it may result in many “safe” residences failing CAS inspections (Rapp, Less, Singer, Stratton, & Wray, 2015). These purported false-positives lead to reductions in the efficacy of energy retrofit programs, either through diversion of program funds towards remediation of perceived (but not real) combustion hazards, or by limiting energy-saving air-sealing in apartments deemed to be at risk of failing a CAS test. Many multifamily energy retrofit programs require that initial testing and remediation be performed before any energy retrofit work can commence; however, these programs do not provide the funding for the remediation work. As a result, it is not uncommon for otherwise strong energy retrofit candidate projects to drop out of these programs because they cannot, or do not want to incur the remediation costs.

National consensus standards for CAS testing (e.g., BPI Standard 1100, ACCA BSR/ACCA 12 QH -201x) have recently eliminated depressurization thresholds for the CAZ, such that appliances only fail worst-case tests if they actually spill combustion products during depressurization testing or fail based on other criteria (e.g., flue CO concentration). This is a large improvement, as the depressurization thresholds were not robust predictors of appliance spillage: some appliances spilled below the threshold while others continued to draft when the CAZ was depressurized beyond the threshold. Other improvements to test procedures are being considered and adopted in standards and being practiced by energy retrofit implementers.

California energy efficiency programs employ a variety of CAS tests and procedures. The programs vary in scope and scale, from providing a basic, prescribed package of efficiency measures to low-income residences to energy-savings-based rebate programs for homeowners. The Department of Community Services and Development (CSD) is the primary agency responsible for low-income weatherization programs in California. These include State of California weatherization programs—the Low-Income Heating Energy Assistance Program (LIHEAP)4 and the Low-Income Weatherization Program (LIWP)5—as well as federal weatherization funded by the U.S. Department of Energy (USDOE). Weatherization programs administered by CSD use the Combustion Appliance Safety Inspection Form (CASIF) and 4 Funded by the California Department of Health and Human Services.

5 Funded by proceeds from the California cap and trade program administered by the California Air Resources Board (CARB).

82

instructional supplements in its CAS testing. This includes a revised Combustion Appliance Safety Protocol released in April of 2015. These California low-income weatherization programs touch approximately 2,000 (USDOE Wx), 12,000 (LIHEAP) and 11,100 (LIWP) residences per year. The Energy Savings Assistance (ESA)6 program also serves low-income residences throughout the state. This program is administered by the investor-owned utilities (IOUs), and touches an order of magnitude more residences (i.e., more than 364,000 in 2014 and 2 million residences from 2009 to 2014). The ESA programs have lower per-house budgets, and they use the Natural Gas Appliance Test (NGAT) procedures for ensuring combustion appliance safety. Market-rate programs retrofitting residences in the state include the Advanced Home Upgrade7 program (formally Energy Upgrade California) administered by the investor-owned utilities. CAS testing protocols used by these programs in the state include the Natural Gas Appliance Test (NGAT) and the Building Performance Institute (BPI) CAS protocols, with specific details of testing and implementation varying by utility.

Multifamily energy retrofit programs in California that have CAS testing requirements include all of the investor-owned utility, Regional Energy Network, and community choice aggregator administered Multifamily Energy Upgrade Programs, as well the CSD administered Low Income Weatherization Program – Large Multifamily (LIWP-LMF). These are all “whole building” retrofit programs, and most of them provide funding to offset some portion of the energy efficiency work, including the initial energy assessments and CAS testing, but they do not provide funding for any remediation work that emanates from that testing.

In considering combustion safety, it is important to recognize that spillage hazards have both physical and statistical features. The physical aspect is the suite of home and appliance operational conditions, combined with weather, that produce sustained (as opposed to brief or intermittent) combustion product spillage when an appliance is operated under these conditions. The statistical consideration is in the frequency and duration of the conditions occurring. The overall hazard in a home is the confluence of these aspects. For example, if an appliance – e.g. a wall furnace – is vulnerable to spillage only when the kitchen range hood is used on the highest speed and the dryer is running and the bath fans are operating and outdoor temperatures are not too cold, it is exceedingly unlikely that those conditions will persist for a long-enough period to allow hazardous levels of combustion products to build up in the home. By contrast, if spillage would occur whenever any two exhaust fans are used irrespective of weather, the statistical likelihood is that this will occur much more frequently and the hazard is greater. To date, no research has documented the frequency of worst-case equipment operation conditions across a population of homes under actual occupancy patterns.

The goal of this research study was to inform efforts to improve CAS testing to identify truly hazardous situations that require remediation, leading to a larger fraction of program dollars being spent on efficiency while also reducing risk. A guiding premise of this research was that CAS testing based on worst-case conditions is an inefficient approach to assessing hazard. 6 Funded by the California Public Utilities Commission (CPUC).

7 Managed by the California Energy Commission (CEC).

83

There are two bases for this premise. First is that the occurrence of WC conditions is extremely infrequent in most homes. Second is the assessment that the equipment operational conditions that create a hazard of depressurization-induced spillage typically involve high rates of outdoor air ventilation that serve to dilute the concentrations of any pollutants emitted in the spilling combustion products, thus providing protection against the hazard of high pollutant concentrations and exposures. Researchers therefore reason that the most important hazards to identify are appliances that spill under less challenging conditions, certainly including when a single, commonly used exhaust fan operates and potentially including combinations of fans that are commonly used at the same time. Appliances spilling under natural or reduced challenge conditions are likely to spill much more frequently and with much lower rates of outdoor air dilution, leading to increased hazard and risk for occupants.

In place of worst-case testing, one could use a more commonly occurring challenge condition to identify appliances that will spill with enough frequency and/or duration to present a true health hazard. Such a test will produce many fewer “false-positives” that fail appliances that will rarely if every spill in practice.

The approach of straightforward combustion safety testing with less than “worst-case” challenge conditions has been demonstrated in a field study of eleven homes in the Midwest US (Brand et al., 2015). The study found that even for those homes that failed under much less challenging conditions, there were only two cases presenting prolonged and excessive spillage in practice. And in both of these residences, the venting (ducting) was out of compliance with the National Fuel Gas Code. Spillage in other homes was either non-existent or occurred only briefly after some burner starts. The results reinforced that visual inspection of appliances and vent ducting is a crucial aspect of any CAS assessment.

In this study researchers sought to identify and characterize venting issues with wall furnaces and track the operation of exhaust fans in apartment units that fail worst-case testing, but pass a less extreme challenge condition. Equipment performance assessments included monitoring of exhaust fans to determine statistics of coincident use, monitoring of wall furnaces to quantify usage and coincidence with exhaust fan usage, and identification and recording of downdraft and spillage events. Researchers also monitored cooking activity, so that the impacts of operating range hood exhaust fans whenever cooking occurs could be assessed, as is recommended. The equipment usage data allows an estimation of how common it is for the households to be operated at various challenge conditions. The study focused on wall furnaces as one of the more hazardous natural draft appliances in California apartments. The collected data can be used to generate frequency distributions of coincident exhaust device operation. With these distributions in hand, researchers can begin to estimate how combustion spillage and hazard are affected by a variety of house parameters, including envelope airtightness, exhaust airflow capacity, and use of venting range hoods.

This chapter provides a summary description of the study methods and key findings. A more complete accounting of this research is available (Singer et al., 2016b).

84

3.2 Methods The study plan, described below, was reviewed and approved by LBNL’s Human Subjects Committee (Institutional Review Board). The Association for Energy Affordability (AEA), under subcontract to LBNL executed the recruitment, diagnostic measurements, and set-up and recovery of in-home monitoring equipment.

3.2.1 Building and Participant Recruitment Potential study sites were identified by AEA using their large database of past and present clients who have participated in one or more of the whole building multifamily energy efficiency programs that AEA has implemented in California. Candidate properties had apartments with wall furnaces and kitchen exhaust fans that were unlikely to not pass the worst-case combustion venting test. Wall furnaces were all located in the living rooms of the apartment units, and they used the room air volume for combustion air. As such, no permanent combustion ventilation air (CVA) openings were required.

AEA was able to identify a number of properties that had been previously tested and had passed the spillage test under natural conditions, but failed under worst-case conditions. As enforced by most of the multifamily programs throughout the state, properties are allowed to proceed with energy retrofits when they fail worst-cases testing, but pass natural condition tests. If any efficiency measures are installed that affect the pressure dynamics of the unit, then CAS tests are repeated post-retrofit, to ensure that appliances still pass natural condition tests.

AEA reviewed results of the industry standard worst-case venting tests and considered how other units located within a site were likely to compare to the measured apartments. AEA used phone and email to contact the Property Owner or Asset Manager of two candidate sites that met the screening criteria. AEA described the research project context, objectives and methods, and obtained verbal consent from the owner or manager to contact and engage with the onsite Property Manager, to coordinate and assist in the resident outreach efforts, and the residents themselves. From that point on, AEA worked closely with the on-site Property Managers to engage directly with the residents. Two properties were identified, henceforth described as Building A and Buildng B.

In Building A, AEA recruited twelve units that had already undergone combustion safety tests as part of an energy efficiency program, and all twelve units passed under natural conditions and failed at worst-case. The units are labeled as A_N1_N2, where N1 is a value of 1-12 and N2 is the number of residents in the unit. Using this information and considering the appliances and apartment configurations in the building, AEA predicted that many of the other apartments in the building would also fail under worst-case and pass under natural draft conditions (as other apartments in the buildings failed during prior energy retrofit testing). Additionally, a number of apartments that were assessed during the rebate program process were found to have intentionally disengaged spill switches. The spill switch is a safety feature that disables furnace operation when the temperature at the draft diverter becomes too high, as this typically indicates spillage. Some rebate program participants reported during those assessments that the

85

safety device was disabled by either building maintenance staff or the building’s outside plumber in response to multiple “no heat” service calls.

Building B had a mix of wall furnaces and ducted FAUs in the apartments. A sampling of units had been tested as part of the rebate program’s initial site assessment, and a number of failures were identified, primarily at the FAUs. Those failures, as well as the fact that new windows were being installed as part of the program, triggered a 100% test-out requirement upon project completion. At the initial test-out many of the FAUs exceeded the CO limits, and as a result ended up being replaced prior to a subsequent test-out. Some of the wall furnaces registered higher than optimal CO levels but still lower than the failure limit (between 26-100 ppm), and it was recommended these units be serviced and cleaned. All of the wall furnaces passed spillage under natural conditions, but failed under worst case. Notably, all apartments had passive air inlets, also called “trickle vents”. Most of these were found to be open during the CAS inspection. These passive inlets were not permanent openings, and as such cannot be counted in combustion ventilation air (CVA) calculations. Furthermore, all of these apartments used the living room volume for combustion air, so they provided no CVA openings to outside. Consistent with this, the passive air inlets were closed during worst-case testing. AEA recruited four units from Building B; these were given identifiers analogous to those used in Building A.

Connections to individual participants were made through a community meeting at Building A, and leaving flyers in both buildings. Phone calls were made to interested residents with apartments deemed likely to meet the study criteria. During the call, AEA answered questions and scheduled the initial testing and equipment installation site visit. AEA used the approved recruitment scripts during these phone calls to provide information about the study.

At the start of the initial site visit an AEA analyst thoroughly reviewed the study consent form and obtained written consent before proceeding.

3.2.2 Occupant Survey and Daily Logs Each participant completed an occupant survey that aimed to characterize his/her perception of the indoor and outdoor air quality and apartment comfort by asking to rate satisfaction levels on a scale from “Very Dissatisfied” to “Very Satisfied”. The form also asked questions about window operation, range hood and bathroom fan operation and usage patterns, bathing and moisture producing appliance operation (dishwasher, washing machine etc.), and interior door operation. The window, door, and fan operation related questions were designed to help characterize the likelihood and frequency of worst-case depressurization events occurring.

Daily log sheets were given to participants with the guidance to complete one log table for each day of the study. These logs were intended to capture information about the activities that were thought to most impact the study objective: number of people in the home; number of baths/showers; usage of cooktop, kitchen or bath exhaust fan, and window opening. A daily log form was provided for each planned day of the study. The participant was asked to complete the form on a daily basis, preferably at the end of the day, and to enter their best estimates for each day and time period.

86

3.2.3 Guidance on Use of Windows and Exhaust Fans The study plan included guidance to participants to keep windows closed and to use the kitchen range hood or bathroom exhaust fans whenever cooking or bathing occurred. This guidance was initially provided to the first four participants in Building A. However, after it was determined that all four of the apartments had mild spillage occur under natural conditions, it was decided that using exhaust fans in this way could cause frequent spillage and potentially result in undesirable levels of occupant exposure to pollutants. AEA staff thus returned to those four units and revised the instructions to occupants. If the unit had a wall furnace that spilled under natural conditions AEA staff instructed the participant to open their windows when cooking and/or using their range hood. If the apartment did not spill under natural conditions, AEA staff instructed the participant to operate windows as they normally do. Occupants were not instructed to open windows during wall furnace operation.

For the next 12 apartments, AEA staff used this same conditional instruction. If the apartment showed any signs of spillage under natural conditions, the participant was instructed to open their windows when operating their range hood. If the tested apartment was not spilling under natural conditions, the participant was instructed to operate their windows as they normally do. AEA made clear to participants that the incentive would be issued however they chose to operate their windows.

These instructions, which were required for safe participation in the study, meant that researchers were not able to study spillage and downdrafting under ‘normal, as-found’ occupied conditions. Rather, occupants were instructed to use their exhaust fans when cooking or bathing, and some occupants were instructed to open windows during kitchen fan use. Window operation was not monitored, which makes these results inconclusive. Any time a window was opened in accordance with these instructions, it affected the potential for downdrafting and/or spillage. As a result, researchers also present the coincident operation assessments of exhaust fans (and cooking) with the wall furnace operation. When paired with the diagnostic testing performed by AEA, these results can serve as an imperfect proxy for spillage or downdrafting, if all windows had been closed.

3.2.4 Diagnostic Testing Protocols Diagnostic testing was carried out in each apartment to determine the airflow through ventilation fans and the draft performance of the wall furnace under various levels of depressurization caused by operating varying exhaust fan and door combinations. CAS testing was conducted on the furnace to ensure that its use during the study would not present a hazard. CAS testing consists of checking natural gas lines for leaks, testing for flue gases spilling into the room at natural and worst-case conditions, and determining the concentration of carbon monoxide (CO) in the flue gases.

Testing of flue gases spilling into the living space (“spillage”) involved using a “smoke stick” to determine whether gases were entering or exiting the draft diverter. In addition, indoor-outdoor pressure differences were measured to assess the magnitude of induced depressurization during exhaust fan operation. A DG-700 Pressure and Flow Gauge (made by The Energy Conservatory) was used.

87

The spillage test procedure is detailed below: 1. Close all doors and windows in the apartment; ensure bathroom and kitchen fans are

turned off (if possible).

2. Record the difference between outdoor (reference) and indoor (input) pressures.

3. Turn on kitchen hood and bathroom fan to highest settings to create “worst-case” conditions, record the new pressure differential.

4. Turn on furnace, use the smoke stick to measure degree to which spillage occurs, if at all. Continue to test with the smoke stick every minute for five minutes (standard CAS protocol calls for two minutes of sampling).

5. If spillage still occurs after five minutes, reduce speed of kitchen fan, note pressure differential, and retest spillage for one minute.

6. Continue to test spillage at reduced challenge conditions, as indicated in Table 25 below, until natural conditions or no spillage are achieved.

7. Characterize spillage using the criteria in Table 26.

In Building B, the apartments included a bathroom exhaust fan that operated continuously on a low setting, and it was boosted to a higher airflow setting based on a motion detector. The inspectors could not measure the low airflow, as any attempt to do so automatically boosted the fans to high speed. As such, during spillage testing, these bathroom fans operated on the high setting during all test conditions listed in Table 25, including ‘natural’ conditions.

Table 25: Furnace Spillage Order of Test Conditions

Order number

#1

#2 #3 #4 #5 #6

Primary Condition

Kitchen fan high & Bathroom fan on

Kitchen fan high & Bathroom fan off

Kitchen fan low & Bathroom fan on

Kitchen fan low & Bathroom fan off

Kitchen fan off & Bathroom fan on

Kitchen fan off & Bathroom fan off

Secondary Conditions

Bathroom door: Open / Closed

Bathroom door: Open / Closed

Bathroom door: Open / Closed

Bedroom door: Open / Closed Passive Vent: Open / Closed

88

Table 26: Spillage Classifications

Classification Description

Drafting (0) Smoke travels directly up the flue, regardless of where it is placed within the diverter

No spillage (1)

Little to no air movement is present in diverter. Smoke may go up the flue, but only if the smoke pen is placed toward the back of the diverter. Near the front of the diverter there may appear to be spillage, but it is primarily eddy currents due to the heat of the furnace. Flue gases are not necessarily spilling but the furnace may fail a traditional CAS test due to the need to have the smoke pen far back within the diverter in order to have any evidence of drafting.

Light Spillage (2)

Smoke moves as though caught in eddy currents regardless of location in diverter. Some flue gases are entering room, and so the space in front of the diverter may feel somewhat warmer than usual.

Medium Spillage (3)

Smoke has a distinct direction of coming out of the diverter, regardless of placement in the diverter. At the corners of the diverter there will be additional turbulence. Air coming from the diverter is hot.

Heavy Spillage (4)

Smoke travels extremely quickly away from the diverter. A lit match or lighter will go out if moved in front of the diverter, and the heat makes it painful to keep your hand there for more than a few seconds.

Flow rates of the kitchen range hood and bathroom fans in each apartment were measured using a Retrotec (www.retrotec.com) duct blaster and a DM-2 Mark II dual-channel digital micro-anemometer and control. A cardboard transition was used to connect the duct blaster to the inlet of the range hood and a loop of soaker hose connected the space inside the transition to the DM-2 controller. With the fan turned on researchers were able to use the controller to drive the duct blaster to reach a speed that balanced the air being pulled up through the range hood, as evidenced by a neutral pressure in the transition. The airflow through the calibrated duct blaster was then taken as the flow through the range hood.

All of the selected units in Building B had side-vented Williams wall furnaces with spill switches. In apartments that spilled even at low exhaust air ventilation rates, tenants commented on the switches being tripped frequently during normal use. As expected, spill switches were tripped during AEA diagnostic tests, requiring that technicians change to air leakage tests while the spill switches cooled, before continuing with the spillage tests. At least one resident was resetting her spill switch herself every time the unit shut off. AEA informed her of the purpose of the device and why resetting it repeatedly is a bad idea. AEA also told her that she should keep her passive vent open anytime she uses her wall furnace, and that if it continues to shut off with the vent open, then she should contact property management staff. Units A_6_1, A_8_1, A_10_1 and A_12_1 of Building A had wall furnaces with spill switches.

3.2.5 Short-Term Monitoring Protocols In order to monitor the patterns of combustion appliance and exhaust devices used under normal occupancy, monitoring equipment was installed for a three-week duration. Sensors were installed by AEA at the start of each monitoring period. Due to data storage limitations some of the sensors were not capable of logging data for the full three-week period; therefore, a

89

mid-term visit was conducted at each apartment in order to download and re-launch those sensors. At the completion of the study all sensors and loggers were retrieved and data were downloaded.

The measured parameters and monitoring equipment are summarized in Table 27. Temperature sensors were placed adjacent to each stovetop burner and the middle of the stovetop to detect when any burner was used. An anemometer measured air speed at the inlet to the range hood fan to determine use and potentially to distinguish setting. Carbon monoxide, carbon dioxide, temperature, and relative humidity were monitored in the living room of each unit. Carbon dioxide and temperature were also logged at the inlet of the draft diverter for the wall furnace. Temperature also was monitored at two locations in the furnace: within the diverter, approximately 2/3 of the way toward the diverter inlet away from the flue, and within the vent approximately 2 feet up from the diverter. An example is provided in Figure 18 and Figure 19. A sensor was placed on the bathroom exhaust fan to monitor its use and a temperature and outdoor grade humidity sensor (which functions to 100% RH) were placed in the bathroom.

Table 27: Monitoring Sensor and Data Acquisition Equipment Summary

Measurement Sensor/Logger Sampling Frequency

Rated Accuracy

Temperature/RH in bathroom1

HOBO U23-001 5 minutes ±0.21 °C (0-50°C) ±2.5% RH from 10–90%2

Bath fan operation HOBO UX90-004 Records each state change (on/off)

Cooking burner use iButton DS1922L 1-minute ± 0.5 °C Range hood operation Digi-Sense 20250-22 Data

Logging Anemometer 1-minute ± 3%

Carbon monoxide in living room

El-USB-CO 5-minute ±2 ppm

Temperature/RH in living room

HOBO U10-003 5- minute ±0.53 °C (0-50°C) ±3.5% RH (from 25 to 85%)

CO2 in living room Vaisala GMW115 w data logged to HOBO UX120-006M

5-minute ±2% of range +2% of reading

Furnace vent temperature

HOBO UX120-014M w/ Omega K-type Thermocouple

1-second ±1.6 °C

CO2 in room just above draft diverter (to identify spillage)

Vaisala GMW115 w data logged to HOBO UX120-006M

1-second ±2% of range +2% of reading

1 To detect excess moisture as indicator of times when bath fan should have been used but was not. This device is designed to withstand condensation (100% RH) conditions but the protected internal sensor has a relatively slow response time of 40 min. 2 Maximum of ±3.5% including hysteresis

90

Figure 18: Placement of Sensors Around and Inside a Wall Furnace

The device on top of the cover is the CO2 sensor and logger. The device attached to the front of the cover logs the thermocouples placed at the draft diverter and up into the vent.

Figure 19: Placement of Thermocouples at Draft Diverter and Vent

Close-up from Figure 18.

91

3.2.6 Data Processing and Analysis All data analysis was performed using the R for Statistical Computing open source software package version 3.2.3 (2015-12-10). The following specific packages were used: data.table (version 1.9.4) for large data sets, xts (version 0.9-7) and zoo (version 1.7-12) for time series manipulation.

Stovetop Temperatures The stovetop temperature data were processed with the goal of identifying cooking events with any active gas burner operation. Researchers did not attempt to identify which burners were operating or how many operated coincidently. As with other measurements in this study, simply using a threshold temperature was insufficient, because of the growth-decay cycle inherent in such environmental measurements. To further complicate analysis, the five measurement locations on each stovetop influence one another, such that waste heat from the front left burner is reflected in the center measurement point and to a lesser degree at other burner locations. This makes identifying individual burner operation difficult. So, burner on/off predictions were made for each of the five burners, and then merged these results into a single cooking index.

In analyzing stove data, researchers first merged data from all five stovetop locations together by time index. For all five temperature locations, researchers calculated the differences from time-step to time-step to find the temperature change for each minute (°C/minute). These data are referred to as ‘differenced’ in this report. Each burner location was then assessed for either rapid positive or negative rates of change in temperate, indicating the start or end of a burner cycle, respectively. Cooking events were then assigned to the time periods between a cycle start and a cycle end signal. In general, changes of +/- 1 °C per minute were indicative of the start or end of a cooking cycle. Some customization was required in order to consistently and cleanly identify burner operation (i.e., to not miss the ‘off’ signal, the negative threshold was sometimes set to -0.5 °C per minute). With on/off predictions for each burner, a combined cooking index was created that was ‘on’ if one or more burners were on and was otherwise ‘off’. This combined index value is shaded in the plots below that show a simple stovetop burner event (Figure 20), an oven burner event (Figure 21) and a more complex event that required manual editing of the data (Figure 22). All five of the burner locations are plotted, along with the differenced series for the back right burner (color black, second y-axis), which tended to be at the highest temperature.

92

Figure 20: Example of a Stovetop Burner Cycle, Using Differenced Time-Series to Identify Cycle Start and Stop Points, Apt A_12_1

Figure 21: Example of an Oven Burner Cycle, Using Differenced Time-Series to Identify Cycle Start and Stop Points, Apt A_12_1

Note that the cycle ‘off’ signal is at -0.5°C, rather than -1.0°C in this case. This more gradual temperature decay is more difficult to correctly identify.

93

Figure 22: Example of a More Complex Cooking Burner Event That Required Manual Editing of the Burner Index, Apt A_12_1

Note that the back right burner remains at a high temperature even after its initial sharp decay around 14:20; this is assumed to be continued operation at a lower burner setting, rather than the burner being turned off.

Collection of temperature data from stoves was challenging, as a sizeable percentage of “iButton” temperature sensors read entirely 0°C for the duration of the data sampling period. At initial midpoint visits these files were not saved, although at later visits these files were saved. Additionally multiple files at Unit A_7_2 were lost due to user error. Unit A_8_1 had one sensor that was non-functional at the midpoint visit (would not download data or re-launch); so this sensor was removed and residents were asked to avoid using that burner if possible. All buttons that were found to record only zeroes at the prior deployment were checked by launching, collecting data for 5-10 minutes and reading and checking the collected data to confirm functionality before re-launching for the next deployment. Despite this precaution, multiple sensors installed in Apartments B_16_1 and B_15_1 still gave non-usable data upon collection. Nevertheless, the approach used to assess if cooking occurred may be robust against some missing sensor locations, because locations where cooking was not happening are still affected by nearby cooking (see Figure 20 and Figure 21). In particular, the center location can serve as proxy for all of the four burner locations. These secondary effects may still have indicated ‘cooking’ occurred.

Kitchen Range Hood Anemometer Kitchen range hood fan data were processed to produce an on-off index based on the anemometer output. All units produced the maximum 4 m/s velocity value at almost all times that the recorded value was not 0 m/s. It is much easier to identify fan operation than burner

94

operation, because there are no signal growth or decay periods. A threshold velocity of 2 m/s was used to identify the start and end of a fan cycle. This threshold captured fan operation but avoided identifying occasional signal noise as fan operation. The 30-second sensor data was averaged to one-minute values for merging with the other sensor data streams. The velocity and the on-off index values were averaged, as well.

Bathroom Fan Motor Operation Most participating apartments had bathroom exhaust fans monitored for operation using fan motor sensors. Bath fan data were first translated into one-minute time series values using simple averaging. Any fan motor fraction values that were greater than zero were assigned a bath fan index value of one, indicating that the fan operated that minute. This approach likely overstates bathroom fan usage, as some minutes had only fractional values, indicating runtime less than 60 seconds. Bath fan monitoring did not record useful data in the apartment units in the second building. Fans in this building were set to operate continuously at a low speed – presumably to provide baseline, dwelling unit ventilation – and they were increased to a higher airflow based on an occupancy sensor.

Furnace Flue Temperatures All participating apartment units in this study used natural gas vented wall furnaces. Each appliance was instrumented with two temperature sensors, one located in the vent above the draft diverter and another on the wall adjacent to the appliance, just above the inlet to the draft diverter. The sensors were placed in this way so as to allow identification of appliance burner operation, as well as to identify spillage and downdraft events. An example of the installation is shown in Figure 18 and Figure 19. An example plot of the two temperatures is provided over a three-day period in Apartment A_8_1 in Figure 23.

95

Figure 23: Characteristic Plot of Wall Furnace Temperatures in Apartment A_1_2

The plot includes the temperature at the draft diverter (red) and in the vent (blue). The vent sensor is characterized by much higher temperatures, up to 250°F in this case.

Identification of Furnace Cycles and Burner Operation Wall furnace burner operation occurs in cycles that appear as a rapid increase in vent temperature followed by a variable length period of elevated temperature and ending with a decay that is less steep than the rise. This clean pattern is disrupted when the exhaust gases spill into the living space instead of venting to the outdoors. Furnace burner cycles were identified using methods similar to those outlined in Section 4.2.6.1 for cooktop burner temperatures. Identification of a burner cycle from a wall furnace flue temperature also requires distinguishing temperature growth and steady-state periods (when the burner is on) from the decay periods (when the burner is off).

All furnace temperature data were first cleaned of obviously erroneous values (e.g., 888.88). The remaining data were then converted from one-second data to one-minute data using simple averaging. This was done to make the data analysis and processing faster and to smooth out noise in the one-second signals. With one-minute data in-hand, the time series were differenced one time-step, and five-minute rolling means were calculated to create ‘smoothed, differenced’ time series. A right-adjusted rolling mean (reflecting concentrations over the prior 5 min) was used to identify the cycle start point: a left-adjusted mean (reflecting concentrations over the next 5 minutes) was used to identify the cycle end point. The rolling mean approach smoothed out fast up-down spikes and ensured that cycle identification algorithms were able to identify clear and continuous burner cycles. The start and end of each cycle was identified based on positive and negative spikes in the smoothed, differenced series. An example of a burner cycle in Apartment A_1_2 is provided in Figure 24. Draft diverter and vent temperatures (red and

96

blue) show three clear burner cycles, with burner operation highlighted in grey. The vent temperature climbs quite high, while the temperature at the draft diverter also increases due to its being secondarily heated by the furnace. Note the large positive and negative spikes in the smoothed, differenced series (green and purple lines centered around zero). Using the smoothed, differenced one-minute data, the team typically identified the start and end of cycles using +/- 15°F per minute threshold values.

The ideal cycle identification algorithm would clearly identify the start and stop of contiguous, real burner cycles. It would never miss an on- or off-signal. Yet, draft diverter and vent temperatures in wall furnaces can be highly variable, both between appliances, as well as between cycles in the same appliance. In order to produce the cleanest and most believable burner cycling index, a combination of human visual review was required, along with some algorithmic fixes. In general, automatic cleaning of the cycle start and end indices was performed first. For example, for a matched pair of start and end time indices, the start index must be lower than the end index. In cases where this was not the case, the offending index was removed. Similarly, a new start index was not allowed if it was less than the end index of the prior cycle (indicating two ‘starts’ in a row). In addition, cycles were removed that were longer than 24-hours, because these indicated that an end index was missed. When start or end indices were missed, it generally required an adjustment of the +/-15°F per minute threshold mentioned above. Identifying a new suitable threshold involved plotting the temperatures, along with the smoothed, differenced values, and a new cut-off was chosen based on the visible patterns in the data. For example, sometimes the decay at the end of the cycle was at a slower rate, so a cut-off of 7 or 10°F per minute was required to capture the end index. With the cleaned start and end indices in-hand, the values between start and end were filled-in with ones, and all other values were set to zero, giving an on-off furnace burner signal.

97

Figure 24: Example of a Furnace Burner Cycle in Apartment A_1_2

The beginning of a furnace cycle is characterized by rapid increases in vent and draft diverter temperatures (blue and red), and the end of the cycle by decaying temperatures. Purple and green series are right and left adjusted five-minute running means of the differenced time series used to identify cycle start and stop times. Furnace operation occurred during gray bands.

Identification of Furnace Spillage The algorithm outlined in Section 4.2.6.5 will only identify burner cycles where the combustion products travel up the vent. Spillage can produce a very different pattern. As described in Section 4.2.4, AEA staff performed a sequence of appliance tests in which flue gas spillage was assessed under different exhaust fan operation and door configurations. These tests were performed after all of the sensors were installed and began logging data. These test results provide us with training data for identifying spillage events in the sensor data streams. After reviewing diagnostic testing notes from AEA, alongside visualization of the sensor data taken during diagnostic testing, it was found that spillage could fall roughly into one of two categories: (1) unambiguous spillage and (2) possible but uncertain spillage.

The spillage assessments in this work involved four steps: (1) on-site diagnostic spillage testing and visual labeling and characterization of spillage events by the AEA team, (2) manual labeling of unambiguous spillage events in the sensor data taken during diagnostic tests by the LBNL team, (3) training of classification learning models on this test data, and (4) use of trained models to predict spillage in sensor data taken during normal occupancy.

The LBNL team first identified the periods in the sensor data during which AEA field-testing was taking place, and these values were identified with an index value of 1. Furnace temperature data was plotted from these periods for visual inspection and comparison with AEA visual inspection notes. Mostly there was agreement between the wall furnace temperature data and AEA’s visual identification of spillage (by smoke pen). Unambiguous

98

spillage events were identified in most apartments with data, though sensor data was available during diagnostic testing for only 10 of 16 units. When the temperature data did not indicate clear spillage but the AEA field technician indicated some spillage (usually “minor”), the event considered uncertain.

A manual index was created for indicating unambiguous flue gas spillage during the AEA test periods (1 for spilling and 0 for not spilling). Examples of unambiguous spillage events are provided in Figure 25. These events are characterized by the draft diverter temperature doing two things: (1) increasing to an abnormally high temperature, and (2) becoming much hotter than the vent temperature. When draft is established, the vent and draft diverter temperatures flip-flop, with the vent taking the characteristically higher temperature, while the draft diverter remains only secondarily heated by the furnace. These spillage events obviously contrast with more typical non-spilling furnace burner events, such as those pictured in Figure 24. As with the furnace burner cycle identification described in Section 4.2.6.5, care was taken in this manual labeling to only include periods when temperatures were increasing (decay periods not included).

Figure 25: Example of Unambiguous Wall Furnace Spillage Event during AEA Diagnostic Testing, Apt A_4_1

Spillage events manually identified in shaded grey regions.

As noted above, possible but uncertain spillage events were also found in the sensor data taken during diagnostic testing. These events were characterized generally by: (1) AEA reporting ‘light’ or ‘barely’ spilling (though not always), and (2) furnace temperature data that was visually indistinguishable from behavior during non-spilling burner cycles, often with the exception of a brief one- or two-minute draft diverter temperature spike at the beginning of the

99

cycle. These events are difficult to categorize as spillage, but they may represent periods of non-robust draft, lacking strong flow of combustion products up the vent. An example is provided in Figure 26. The first burner cycle clearly exhibits backdrafting with diverter temperatures higher than vent temperatures. After the peak around 14:00, there is a period of temperature instability that appears to involve substantial if not complete spillage. Temperatures then stabilize around 14:20. When the next burner cycle starts at around 14:40, there is initially a sharp rise in the diverter temperature that suggests more spillage; but draft is quickly established. Despite this, AEA reported ‘light’ or ‘medium’ spillage up to the point where the vent temperature was 320°F in the third event. More examples of uncertain spillage events are provided in the final task report (Singer et al., 2016b).

Figure 26: Example of Disagreement between Furnace Temperature Sensor Data and Reports of Visually Identified Spillage by AEA during Diagnostic Testing in Apt B_15_1

AEA reported ‘light’ or ‘medium’ spillage well into the second burner cycle. Brief spillage is probable at the start of the second cycle, but strong draft appears to be established within minutes. Light grey highlights period identified as unambiguous spillage.

3.2.7 Identification of Furnace Downdrafting Downdrafting is distinguished from spillage, because it occurs whenever air flows down the vent pipe. This is not necessarily associated with appliance burner operation. It could be caused by weather effects or operation of exhaust devices in the apartment. In Figure 27, an example from Apartment B_15_1 was provided of the wall furnace temperature plots, where vent temperature (lower plot) clearly shows periods of sudden temperature depression characteristic of downdrafting. These periods are characterized by negative dips that drop the vent temperature down to around the outside ambient (roughly 50 or 60°F).

100

Figure 27: Wall Furnace Temperature Measurements Displaying Downdrafting in Apt B_15_1

Downdraft periods are characterized by the vent temperature (blue) dropping suddenly to around outdoor ambient temperature (e.g., 50-70°F).

Our overall approach to identifying these periods was to calculate a 12-hour (720-minute) running average of the vent temperature along with a 12-hour running standard deviation (SD), and classify minutes were classified as downdrafting where the vent temperature was more than three running SD below the running mean. Wall furnace temperature data are affected by burner events, downdrafting, variability in outside temperature, and variability in inside temperatures. So some data cleaning was required prior to calculating the moving average/SD. First, hot vent temperatures (>120°F) or cold vent temperatures (<90°F) were removed. This effectively removed burner events and downdrafting periods themselves from running average/SD calculations. Otherwise during long periods of sustained downdrafting, the running average value would become the temperature under downdraft conditions, and this approach would no longer work. A downdrafting index was created that classified a minute as downdrafting when the vent temperature was three running SD below the running mean. Example downdrafting events are pictured in Figure 28 and Figure 29.

101

Figure 28: Example of a Downdrafting Events Identified in Apt B_15_1

One sustained 4-hour downdraft is pictured in the center. Note how an additional downdrafting event occurred prior to the first furnace burner cycle at 20:00, and the beginning of this event was missed, but then it was identified as the vent temperature dropped a second time.

Figure 29: Example of Downdrafting Events Identified in Apt B_15_1

The first downdraft event occurred prior to the furnace firing, but it did not stop upward draft from being established.

102

As was common in this work, the only way of assessing these results was through a careful visual review of plotted data. The approach described above accurately classified most periods identified as downdrafting by visual review. During visual review, it was noted noted some instances where a long, extended downdrafting event was broken up into two or more smaller events, due to an intermediate minute that rose above the 3 SD threshold. As a result, estimates of the duration of downdrafting events may be negatively skewed.

Some interesting patterns were noted in the data during the visual review. Some periods were observed of unusual vent and draft diverter temperatures that were usually coincident with exhaust fan operation. Examples are provided in Figure 30, Figure 31, and Figure 32. In the first example (Figure 30), both the vent and draft diverter temperatures increased during bathroom exhaust fan operation (highlighted in grey) and the draft diverter temperature started to fluctuate. The team hypothesize that the temperatures increased because of reduced dilution airflow through the draft diverter caused by depressurization from the bath fan operating. The reduced outflow of air through the draft diverter and temperature fluctuations suggest an increased risk of spilling combustion products from the standing pilot burner (which is the cause of the elevated vent temperature). The next example, provided in Figure 31, shows a comparison of the effect occurring with a low airflow bathroom exhaust (~30 cfm) and a higher airflow kitchen exhaust fan (~80-100 cfm). Bathroom fan operation caused the vent temperature to increase with no clear change in draft diverter temperature. When the kitchen exhaust fan operated, both temperatures increased and the draft diverter temperature had large fluctuations. In the final example, in Figure 32, vent temperature jumped rapidly by almost 10°F, but only midway through the bath fan cycle. A window being closed could explain the delay in this example.

103

Figure 30: An Example of Bathroom Operation (Gray Bands) Impacting Draft Diverter and Vent Temperatures in Apt A_11_1, with Possible Implications for Spillage of Pilot Exhaust Gases

Figure 31: An example of Kitchen (Turquoise Shading) and Bathroom (Grey Shading) Exhaust Fan

Operation Impacting Vent and Draft Diverter Temperatures in Apt A_7_2

The higher airflow of the kitchen exhaust fan (roughly 80 vs. 28 cfm) produces greater depressurization which could reduce the flow of dilution air through the draft diverter, leading to higher temperatures and potentially increasing the chance of partial spillage of pilot burner pollutants.

104

Figure 32: Example of Increasing Vent Temperature during Bath Fan Operation in Apt A_12_1

The increase is hypothesized to correspond to a window being closed with consequent depressurization, causing a reduction in flow through the draft diverter.

3.3 Results 3.3.1 Description of Apartment Units Summary descriptions of apartments included in the study are provided in Table 28.

Apartment units at Building A were mostly identical studio apartments with left and right hand orientations. There were two 1-bedroom units that had more square footage and an additional interior doorway creating more conditioned floor area. Many were interior units with only two exterior facing walls. Two participating units were located on the second floor of a two-story building. In addition to the roof adding more exterior surface area, this characteristic has an effect on drafting ability and is therefore noted above. Orientation information was not recorded for any of the units. Stovetops at Building A were electric. Apartments A_1_2 and A_10_1 used medical oxygen tanks. Apartment A_1_2 had an inoperable range hood through the midpoint data collection visit for that unit.

Apartments in Building B were all identical studio units. They had continuously exhausting bathroom fans that ramp up to a higher CFM when an occupant is detected in the space. There were also passive, tenant-adjustable passive air vents near the rear sliding glass doors. Three of the four participating units were located on the second floor of two story buildings. All stovetops at Building B were natural gas powered.

105

Table 28: Characteristics of Tested Apartment Units

Apt ID1 Floor area (ft2) Description A_1_2 470 1st floor, Studio A_2_2 560 3 exterior walls, 1st floor, 1 bedroom unit A_3_2 470 2.5 exterior walls, 1st floor, Studio A_4_1 470 2 exterior walls, 1st floor Studio A_5_1 470 2.5 exterior walls, 1st floor, Studio A_6_1 470 1st floor, Studio A_7_2 470 1st floor, Studio A_8_1 470 1st floor, Studio A_9_1 470 3 exterior walls, 1st floor, Studio A_10_1 560 3 exterior walls, 1st floor, 1 bedroom A_11_1 470 2nd floor, 1st floor, Studio A_12_1 470 3 exterior walls, 2nd floor, Studio B_13_1 510 2nd floor, studio, Gas Range B_14_2 510 1st floor, Studio, Gas Range B_15_1 510 2nd floor, studio, Gas Range B_16_1 510 2nd floor, studio, Gas Range

1ID includes building (A,B), apt code, and number of occupants.

3.3.2 Diagnostic Testing Envelope Airtightness The airtightness of each apartment unit was tested upon initial inspection. The airflows at 50 Pascals of depressurization varied from roughly 400 to 700 cubic feet per minute (CFM) with a mean and median roughly at the midpoint of these values. These values represent total apartment leakage, which includes leakage areas between the unit tested and all adjacent parts of the building. These leakage areas are relevant for depressurization-induced spillage assessments, but are not relevant for energy calculations or outdoor air exchange.

Exhaust Device Airflow Testing Calibrated fan flow meters were used to measure the airflow of exhaust devices in each apartment unit. The individual and maximum total airflows are reported in Table 29. Kitchen fans all had two settings. While the higher flow setting is used in worst-case depressurization testing, it is worth noting that the higher speed added only 21 CFM of exhaust flow on average and the highest increment was 47 CFM.

In Building B, bathroom fans8 constantly operated on low speed and had motion sensors to ramp up airflow from low to high when the room was occupied. In the first unit tested (B_14_2) any sign of the fan transitioning between low and high settings was undetectable. The transition was noticeable in other units, but “low” flow rates could not be tested since any movement to measure the flow rate caused the fan speed to increase, and there was no way to permanently set the fans to the low speed. In addition, the HOBO motor loggers used to assess bathroom fan

8 Panasonic Whispergreen

106

operation were unable to pick up the bathroom fans, likely due to the DC motors. Bathroom fan airflows were much higher in Building B than in Building A (98 vs. 49 CFM on average).

Table 29: Summary of Exhaust Fan Airflow Measurements and Maximum Installed Exhaust Capacity in Each Apartment

AptID

Kitchen Fan Bathroom Fan

(CFM)

Maximum Exhaust Capacity

(CFM) Low (CFM) High (CFM) A_1_2 73 88 42 130 A_2_2 126 154 32 186 A_3_2 105 120 55 175 A_4_1 62 77 54 131 A_5_1 106 120 57 177 A_6_1 79 92 NA 92 A_7_2 80 98 28 126 A_8_1 68 115 56 171 A_9_1 87 108 49 157 A_10_1 118 140 81 221 A_11_1 107 131 37 168 A_12_1 107 131 38 169 B_13_1 28 48 97 145 B_14_2 40 62 39 101 B_15_1 31 63 100 163 B_16_1 33 49 98 147 Median 80 103 54 160

Step-Wise Depressurization and Spillage Testing Combustion spillage diagnostics were performed for the wall furnace in each apartment and results are presented in Table 30. The testing included a step-wise assessment of draft/spillage at conditions varying from natural to worst-case depressurization. Table 30 reports the maximum combustion appliance zone depressurization with the installed exhaust fans. In some cases, this was not sufficient to cause spillage, additional exhaust airflow was introduced using the blower door fan. Table 30 reports the exhaust fan and door configuration and lowest depressurization that caused spillage.

The ability of the installed exhaust fan capacity to depressurize the combustion appliance zone in each apartment was highly variable, from roughly -2 to -15 Pascal. In five apartments, the installed exhaust devices did not cause spillage of the wall furnaces even when operated under worst-case conditions. In these cases, a blower door fan was used to induce spillage with 150 cfm of exhaust airflow (labeled as ‘Induced 150CFM’). The lowest level of depressurization leading to spillage in any apartment (via installed capacity or blower door) was -5 Pa. In some units, furnaces did not spill until depressurization reached 11 Pa. In the ten apartments with enough installed exhaust capacity to cause spillage, three could be spilled with only a single

107

exhaust device operating. Furnaces in units A_2_2 and A_11_1 spilled with only the kitchen fan on low speed (at 126 and 107 CFM, respectively). The furnace in B_16_1 spilled with only the bathroom fan operating (at 98 CFM). All other apartments required coincident operation of two exhaust fans for the furnace to spill.

In general, the impact of opening or closing the bathroom door was small on CAZ depressurization, with median absolute change in CAS depressurization of 0.2 Pa for matched fan conditions. The maximum change in any case was 1 Pa (kitchen fan high, bath fan on in A_5_1); yet the same apartment at another fan combination condition (kitchen fan low, bath fan on) showed no change in depressurization with bathroom door position. These results suggest that most changes resulting from the bathroom door position are within the noise of the instrument and may not be worth test effort.

Notably, all apartments in Building B had passive ventilation air openings that could be set to high (mostly open) or low (mostly closed) settings. Tenants were encouraged to keep these in the high setting. At unit B_16_1 the tenant had previously blocked the vent opening by stuffing a T-shirt into the entrance and covering it with duct tape. This was done to impede the entry of cigarette smoke from downstairs neighbors. New rules have been enacted disallowing cigarette smoking in that area, so the tenant agreed to have the passive vent opened and operating as intended. It should also be noted that, as described in Methods Section 4.2.4, the wall furnaces in Building B had spill switches that were repeatedly tripped during this spillage testing.

Combustion Appliance Flue Carbon Monoxide Levels Flue gases from each apartment’s wall furnace were measured for air-free carbon monoxide concentrations in the flue during spillage testing. Only one apartment had problematic CO during inspection, apartment A_2_2 initially tested flue CO above 1,000 ppm, but with repairs this was reduced to 0 ppm. The next highest value was 15 ppm, which is extremely low for air-free CO in the exhaust. Low appliance CO reduces some of the risks associated with combustion appliance spillage, but other pollutants are still of concern (e.g., NOx, particles, water vapor).

108

Table 30: Summary of Wall Furnace Spillage Testing in Each Apartment Unit

AptID

CAZ pressure referenced to outside (Pa)

Least challenging condition that produced spillage

Maximum depressurization with installed

fans

Minimum depressurizatio

n to induce spillage

Kitchen fan Bath fan Bath door Vents

A_1_2 -2.9 -8 Induced 150 cfm NA NA NA

A_2_2 -11.7 -7.2 Low Off Open NA A_3_2 -9.9 No Fail No Fail No Fail No Fail No Fail

A_4_1 -8.5 -9.6 Induced 150 cfm NA NA NA

A_5_1 -12 -10 Low On Closed NA

A_6_1 -8.2 -8.1 Induced 150 cfm NA NA NA

A_7_2 -9.5 -8.4 Low On Closed NA A_8_1 -11 -11 High On Open NA A_9_1 -11 -10.5 High On Closed NA A_10_1 -8.6 -7 Low On Closed NA A_11_1 -14.9 -8.5 Low Off Open NA A_12_1 -6.9 -5.3 Low On Closed NA

B_13_1 -4.2 -8.9 Induced 150 cfm NA NA NA

B_14_2 -2.3 (Result not recorded)

Induced 150 cfm NA NA NA

B_15_1 -7.5 -5 Low On Closed Open B_16_1 -4.1 None Off On Open Closed

3.3.3 Field Monitoring Outside Conditions The monitoring in the apartment units took place over the course of roughly two months—February and March of 2016—and outside weather varied during these periods. Daily average outside temperatures were retrieved from the Weather Underground website for the two building locations, using the Hayward Airport (KHWD) for Building A and the Moffett Federal Air Field (KNUQ) for Building B. The start and end of the monitoring periods for each apartment unit are listed in Table 31, along with the calculated base 65°F heating degree days per day during that time period. The coldest monitoring period (A_5_1) was 30% colder than the mildest period (A_2_2). This variability is expected to have a modest impact on heating system run times.

109

Table 31: Monitoring Start and End Dates (Year-Month-Day) for Each Apartment Unit, along with Mean Heating Degree-Days per Day for the Period using a 65°F Base Temperature

AptID Start End HDD65 per day A_1_2 2016-02-05 2016-02-26 6.7 A_2_2 2016-02-05 2016-03-07 6.3 A_3_2 2016-03-08 2016-03-25 7.4 A_4_1 2016-03-07 2016-03-25 7.7 A_5_1 2016-02-01 2016-02-25 8.1 A_6_1 2016-02-04 2016-02-26 7.0 A_7_2 2016-02-02 2016-02-25 7.8 A_8_1 2016-02-01 2016-02-25 8.1 A_9_1 2016-02-02 2016-02-25 7.8 A_10_1 2016-02-04 2016-02-29 6.6 A_11_1 2016-02-03 2016-02-26 7.4 A_12_1 2016-02-03 2016-02-26 7.4 B_13_1 2016-03-03 2016-03-24 7.2 B_14_2 2016-02-29 2016-03-25 6.7 B_15_1 2016-03-02 2016-03-30 7.3 B_16_1 2016-03-03 2016-03-25 7.2

Living Space Conditions Measurements of temperature, relative humidity, carbon monoxide and carbon dioxide were made in the main living space of each apartment. Below are basic statistical summaries for these parameters across the participating apartment units.

Temperature and Relative Humidity Indoor temperature and relative humidity was measured in the central living space of each apartment unit at five-minute intervals during the study period. Summary statistics are provided in Table 32 and Table 33. Indoor temperatures are in the expected range for occupied residences, with some staying stable throughout the day and others varying substantially over the course of the day. None of the apartments experienced any period below 60°F. Some units had maximum temperatures in the upper 80s and lower 90s; it is unclear if these were by preference (as Building A occupants were all seniors) or other reasons. Indoor relative humidity was in the expected range for occupied dwellings, with study period minima of 30-40% and maxima of 50-70% RH.

110

Table 32: Summary Statistics of Measured Indoor Temperatures

AptID Indoor Temperature (°F)

Min 25th Median Mean 75th Max A_1_2 70.0 72.7 73.4 73.6 74.3 80.0 A_2_2 62.8 73.1 74.5 74.3 75.5 79.5 A_3_2 66.9 75.3 76.4 76.4 77.8 82.2 A_4_1 67.2 72.0 72.6 72.7 73.2 81.1 A_5_1 69.1 72.2 73.1 72.9 73.6 77.4 A_6_1 71.0 73.6 74.8 74.6 75.5 83.9 A_7_2 NA NA NA NA NA NA A_8_1 72.4 75.4 76.7 76.6 77.6 84.1 A_9_1 68.4 75.8 76.3 75.7 76.7 78.8 A_10_1 65.2 71.5 72.3 72.3 72.9 86.3 A_11_1 71.2 74.1 75.7 75.7 77.1 83.9 A_12_1 72.2 76.5 78.3 78.2 80.2 82.5 B_13_1 66.0 70.0 71.9 72.1 74.3 84.1 B_14_2 65.9 74.5 75.3 75.4 76.4 90.3 B_15_1 65.2 69.8 71.7 72.1 74.1 90.5 B_16_1 NA NA NA NA NA NA

Table 33: Summary Statistics of Measured Indoor Relative Humidity

AptID Indoor Relative Humidity (%)

Min 25th Median Mean 75th Max A_1_2 30.7 48.0 51.1 51.5 55.0 68.1 A_2_2 36.2 55.8 59.0 58.3 61.7 76.4 A_3_2 37.0 45.8 49.6 49.5 53.5 68.3 A_4_1 42.5 49.4 51.6 51.4 53.9 59.6 A_5_1 37.7 46.3 49.5 49.3 52.5 59.7 A_6_1 36.8 47.7 49.7 49.7 51.4 70.7 A_7_2 NA NA NA NA NA NA A_8_1 34.0 42.8 45.2 45.6 48.5 60.4 A_9_1 33.1 43.8 46.3 46.5 49.4 67.9 A_10_1 34.4 44.9 47.9 47.4 49.7 58.1 A_11_1 36.8 45.3 46.7 46.3 47.6 51.5 A_12_1 26.4 36.4 39.5 38.8 41.8 51.8 B_13_1 34.4 44.3 46.7 46.3 48.5 56.8 B_14_2 29.9 42.2 44.3 44.6 46.9 58.6 B_15_1 32.9 42.6 45.4 45.9 48.4 71.2 B_16_1 NA NA NA NA NA NA

111

Carbon Dioxide CO2 concentrations were measured on a five-minute time-step in the main living room of each apartment unit. The distribution of living room CO2 is presented in Table 34 for each apartment. Measured levels were in-line with the expected range in occupied residences. Maxima were in the 1,500 to 2,000 ppm range, and median concentrations were just below 800 ppm.

Table 34: Summary Statistics of Measured 5-Minute CO2 Concentrations in the Main Living Room of Each Apartment

AptID

Room CO2 (ppm) Min 25th Median Mean 75th Max

A_1_2 518 619 759 772 866 1423 A_2_2 3131 903 1337 1317 1861 1892 A_3_2 405 655 805 793 947 1240 A_4_1 435 603 649 638 684 900 A_5_1 396 797 911 885 1010 1314 A_6_1 494 894 996 965 1065 1280 A_7_2 455 956 1100 1089 1242 1783 A_8_1 424 571 627 626 677 1076 A_9_1 424 557 618 625 685 1003 A_10_1 536 821 898 896 963 1318 A_11_1 411 725 957 897 1081 1318 A_12_1 410 518 797 793 1036 1481 B_13_1 410 488 569 578 655 974 B_14_2 400 610 704 709 786 1917 B_15_1 402 490 612 675 822 1941 B_16_1 NA NA NA NA NA NA 1 This value is sufficiently below the outdoor background that it is clearly erroneous. The cause of the error was not determined.

Carbon Monoxide Carbon monoxide is the most commonly cited pollutant of concern that CAS testing is meant to address in residences. Accordingly, CO levels were measured on a one-minute time step in the main living area of each apartment unit throughout the study. There was only one instance of a CO measurement of concern, in Apartment A_1_2. In this event, CO increased from a low background to about 60 ppm over a two-hour period. The event is of uncertain cause since the study participant in this apartment did not complete daily log sheets. The increase is not explained by burner use as the CO increase did not coincide with wall furnace cycles and the furnace was confirmed to have very low CO (15 ppm air-free) during diagnostic testing. And the cooktop burners were electric. It is also not plausible that the CO could have come from outdoors, as this would have been observed in other apartments.

Summary Statistics for Equipment Use, Spillage, and Downdrafting

112

Developing an appropriate reduced challenge condition requires some knowledge of how apartments are operated over time; starting in this sub-section data for exhaust fan and cooking appliance operation that are relevant to this objective is presented. However, it is recall that households participating in this study were asked to operate their homes in a way that may not represent their typical behavior patterns. If during AEA diagnostic testing, the apartment’s wall furnace showed any signs of spillage under natural conditions, the participant was instructed to open their windows when operating their range hood and/or cooking. If the tested apartment was not spilling under natural conditions, the participant was instructed to operate their windows as they normally do. Participants were also instructed to use their kitchen exhaust fans and bathroom fans whenever cooking or bathing.

Overall summary results for equipment use, spillage and downdrafting are provided in Table 35. Detailed results are provided for each apartment, along with discussion and illustrative plots in the following sections.

Table 35: Overall Summary Statistics for Activities Recorded across All Apartments

Metric

Fraction of Minutes Events

Mean Max

Total Count

(#)

Average per

Apartment (#)

Average Duration (Minutes)

Max Duration (Minutes

) Bath Fan 3.88% 13.45% 839 76 17 540 Kitchen Fan 2.42% 10.19% 368 23 28 267 Cooking Burners 3.32% 11.35% 583 36 32 302 Cooking (Burners or Kitchen Fan) 4.63% 13.27% 657 41 34 324

Heating 2.76% 8.89% 743 46 14 162 Spillage 0.03% 0.07% 25 6.3 1.4 5 Downdrafting 3.11% 18.47% 307 24 34 750

Burner Cycles The heating cycles are summarized for each apartment unit in Table 36. The average heating system runtime was 2.76% of the monitoring period, and 743 discrete burner cycles were identified. Across apartments, the average number of heating burner cycles was 46 with an average length of 14 minutes. Cumulative heating system run time was highly variable between units, varying between four minutes and 2,126 minutes. Cycle lengths were also highly variable, even between units with similar amounts of total furnace operation. For example, B_15_1 and A_6_1 both had approximately 1,000 minutes of heating operation, but with 30 versus 120 heating cycles (average cycle lengths of 36 versus 8 minutes). The longest heating cycles were on the order of one to two hours.

113

Table 36: Summary Statistics for All Heating System Cycles in All Apartment Units

AptID

Total runtime

(min)

Runtime fraction

(%)

Cycle coun

t

Furnace Cycle Length (minutes)

Min 25th Median Mean 75th Max A_1_2 26936 3.86% 75 5 8 10 14 14 74 A_2_2 13749 0.63% 8 2 7 13 11 14 19 A_3_2 13586 8.89% 81 3 8 13 15 20 59 A_4_1 23836 0.20% 7 5 7 7 7 7 7 A_5_1 27259 2.64% 64 5 10 11 11 11 27 A_6_1 27361 3.66% 118 2 6 6 8 8 58 A_7_2 26571 0.07% 2 6 8 9 9 11 12 A_8_1 27303 2.48% 45 2 8 16 15 21 35 A_9_1 26626 4.81% 149 1 8 9 9 9 25 A_10_1 27303 2.17% 10 5 37 41 59 82 162 A_11_1 27344 1.50% 10 31 36 39 41 47 52 A_12_1 26984 7.88% 115 3 9 11 18 19 95 B_13_1 25727 1.14% 26 3 5 11 11 16 28 B_14_2 27238 0.25% 5 10 10 10 11 11 14 B_15_1 26466 4.05% 28 5 17 29 38 47 130 B_16_1 27251 0.00% 0 NA NA NA NA NA NA Average 2.76% 46 1 7 9 14 15 162

Downdrafting Downdrafting occurs when air flows down the vent pipe due to weather or exhaust fan effects, and is not limited to furnace operation. Summary statistics for downdrafting events occurring during the measurement periods are provided in Table 37. These downdrafting events likely occurred whenever exhaust fans were operated in the apartment and windows were closed. Variability in downdrafting was high, with some units experiencing no downdraft, and others downdrafting up to 14% of the monitoring period. A total of 307 downdraft events were identified across all apartments, with the averages of 24 downdraft events and events lasting 34 minutes each. The longest individual event lasted 750 minutes in Apt B_15_1 (roughly half of one day). The vast majority of downdraft events lasted less than 100 minutes. As discussed earlier, these results only include full downdrafting conditions; other reduced draft conditions are not included (e.g., when an exhaust fan causes a slowing of the exhaust vent gases or even partial spillage of pilot burner pollutants).

Downdrafting was more common in Building B, possibly because of the continuously operating bathroom exhaust fans. AEA was not able to measure the bath fan airflows on the continuous, low setting, but their airflows on high (activated by local sensor) were roughly 100 cfm. The team hypothesized that in these apartment units, the wall furnaces may have been under downdraft conditions whenever all windows were closed. In apartments where wall furnaces spilled, residents were instructed to keep windows open during cooking. As a result, the data

114

presented here may be underestimating downdrafting and spillage events that would occur under unguided circumstances (presumably with windows closed more often). The only exception may be Apt B_14_2 where no downdrafting was recorded. This apartment had much lower bathroom exhaust airflow (40 cfm at high, compared with 100 cfm in other building 2 apartments), and a low flow kitchen exhaust. As such, this apartment had the second lowest of all installed fan capacities (~100 cfm total). Downdrafting was also high in one Building A apartment (Apt A_8_1), which had by far the most bathroom and kitchen exhaust fan operation (13% and 10% of the monitoring period, respectively).

Table 37: Summary Statistics for Downdrafting Cycles in All Apartment Units

AptID

Total minutes operatio

n

Total Downdraft Period (%) Cycle

count

Length of Downdraft Cycles (minutes)

Min 25th Median Mean 75th Max A_1_2 26936 0.11% 6 1 1 2 5 2 23 A_2_2 13749 0.56% 10 2 4 7 8 8 17 A_3_2 NA NA NA NA NA NA NA NA NA A_4_1 NA NA NA NA NA NA NA NA NA A_5_1 27259 1.21% 15 2 9 13 22 31 74 A_6_1 27361 0.09% 4 1 1 4 6 9 17 A_7_2 26571 0.68% 12 1 3 7 15 17 75 A_8_1 27303 6.58% 87 1 2 14 21 29 154 A_9_1 26626 0.27% 5 4 12 13 15 14 30 A_10_1 27303 0.93% 10 1 2 8 25 35 112 A_11_1 27344 0.59% 9 1 2 6 18 29 51 A_12_1 26984 0.16% 1 43 43 43 43 43 43 B_13_1 25727 8.73% 80 1 4 16 27 33 175 B_14_2 NA NA NA NA NA NA NA NA NA B_15_1 26466 18.47% 56 1 4 12 87 76 750 B_16_1 27251 2.04% 12 1 6 50 46 71 128

Average 3.11% 24 1 3 13 34 33 750

Spillage Spillage occurs when some fraction of the combustion products produced by burner operation pass into the living space. Spillage can encompass a fraction of the burner combustion products or all of them. If the burner fires when the vent is in a downdraft condition, proper venting may still occur, as the buoyancy of the hot combustion gases may be sufficient to reverse flow. And even if the flow is not reversed immediately upon burner firing, the exhaust may over time warm the air above to create the intended updraft venting. Spillage events are of direct concern as they could present a health or material risk (depending on the contents of the combustion products emitted), whereas downdrafting is of secondary concern only, as it increases the likelihood of spillage. Many appliances may spill on start-up yet still establish good draft within a few minutes. The data presented here are for unambiguous spillage events—those where

115

spillage is strong and clear, with dramatic growth in the draft diverter temperature sensor while the vent temperature remains stagnant. This may dramatically under-predict partial or unclear spillage, which may or may not be hazardous.

Spillage statistics are reported in Table 38 for the four apartments that had any spillage. Most (12 of 16) had no unambiguous spillage occur outside of the AEA diagnostic test period, which was removed from these analyses. A total of 25 unambiguous spillage events were identified, with an average length of one minute and maximum length of five minutes. All of these unambiguous spillage events were shorter than the 5-minute cold burner start-up period during which spillage is allowed by CAS test protocols. In almost all identified cases of unambiguous spillage, the same pattern occurred. First, the wall furnace was in downdraft mode, then the burner turned on, the appliance spilled for one to five minutes, and then what appeared to be normal draft conditions were established.

It is interesting to note that these four apartments all required two fans to induce spillage during AEA diagnostic testing (see Table 30), and in the case of B_13_1, a duct blaster fan was required to induce spillage, because the installed fans were not able to do so. The installed fan exhaust airflows were above average, but not necessarily the highest of those measured. These apartments (with the exception of A_9_1) had high levels of downdrafting (see Table 37). Downdrafting was also high in A_8_1 and B_16_1, but they showed no unambiguous spillage.

Table 38: Summary Statistics for Wall Furnaces Spillage in Apartments with Any Spillage Identified

AptID

Total minutes

of operation

Spillage Fraction

(%) Cycle count

Furnace spillage cycle duration (minutes)

Min 25th Median Mean 75th Max A_5_1 27259 0.02% 1 5 5 5 5 5 5 A_9_1 26626 0.00% 1 1 1 1 1 1 1 B_13_1 25727 0.07% 14 1 1 1 1.4 2 2 B_15_1 26466 0.04% 9 1 1 1 1.2 1 2

Average 0.03% 6.3 1 1 1 1.4 2 5

As noted previously, in apartments where wall furnaces spilled under natural test conditions, occupants were instructed to open their windows during kitchen exhaust fan use. This may or may not have affected the frequency of measured unambiguous spillage. If anything, it is likely these reports of spillage frequency and duration are biased low, relative to ‘typical’ operation, where the window may have remained closed during kitchen fan use.

Bathroom Exhaust Usage Bathroom exhaust fan usage was monitored in 11 of 16 apartment units using fan motor fraction sensors. Motor sensors did not work in five of the units, likely due to DC fan motors. Notably, these included the four Building B apartments, in which continuous bathroom fans were operated. Total bathroom fan operation and summaries of events are reported in Table 39; ‘NA’ results for Apt A_1_2 indicate that the sensor failed to log operation. Bath fan use varied

116

substantially across units from 0.5 to 13.4% of the monitoring period, averaging 3.9% of monitored minutes. A total of 839 bath exhaust use events were identified, with an average number of 76 per apartment at 17 minutes in duration. The longest continuous bathroom exhaust use period was 540 minutes, likely the result of it being left in the on position.

Table 39: Summary of Bathroom Exhaust Fan Use Building A Apartments

AptID Total (min)

Total Bath Fan Use

(%)

Count of Bath Fan Events

Bath Fan Event Durations (min)

Min 25th Median Mean 75th Max A_1_2 NA NA NA NA NA NA NA NA NA A_2_2 36013 0.54% 24 1 2 7 8 14 17 A_3_2 24289 1.02% 24 1 3 5 10 11 48 A_4_1 26000 0.85% 80 1 2 2 3 3 14 A_5_1 34558 1.22% 124 1 2 2 3 3 17 A_6_1 31647 1.35% 36 1 3 8 12 14 90 A_7_2 32940 2.62% 72 1 3 7 12 17 99 A_8_1 34414 13.45% 234 1 4 7 20 32 136 A_9_1 33119 5.17% 75 1 3 11 23 26 227 A_10_1 35674 0.83% 60 1 2 3 5 5 23 A_11_1 33099 6.26% 55 2 8 22 38 48 172 A_12_1 33333 9.34% 55 1 3 8 57 83 540 Average 3.88% 76 1 2 5 17 16 540

Cooking and Kitchen Exhaust Fan Usage Cooking activities were assessed in each apartment unit using stovetop temperature measurements in all burner locations, as well as by monitoring of the operation of the kitchen exhaust fan. Residents of each apartment unit agreed as part of their participation to operate the kitchen exhaust fan whenever they were cooking. Kitchen exhaust fan usage therefore may be much higher than would typically occur in these apartments or in other residences.

Statistics for kitchen fan operation are presented in Table 40. Some kitchen exhaust fan use was logged in most apartment units, with usage periods varying between 0 and 10% of monitored minutes, averaging 2.4% of minutes. A total of 368 kitchen exhaust use events were identified, with an average count of 23 per apartment and average duration of 28 minutes. The longest single kitchen fan use event lasted 267 minutes.

117

Table 40: Summary of Kitchen Exhaust Fan Use in Each Apartment Unit

AptID Total (min)

Total Kitchen Fan (%)

Count of Kitchen Fan

Events

Kitchen Fan Event Durations (min)

Min 25th Media

n Mean 75th Max A_1_2 30391 2.31% 14 3 22 43 50 63 132 A_2_2 14771 4.54% 39 2 6 10 17 18 200 A_3_2 24300 0.45% 11 1 4 10 10 14 32 A_4_1 26020 0.37% 17 1 3 5 6 9 12 A_5_1 30184 3.25% 29 5 17 21 34 36 144 A_6_1 30478 0.34% 10 1 5 9 10 17 22 A_7_2 28818 1.72% 22 2 7 15 23 27 84 A_8_1 30055 10.19% 83 3 18 33 37 47 154 A_9_1 28880 2.75% 53 1 5 10 15 18 66 A_10_1 31853 1.00% 8 6 11 42 40 55 89 A_11_1 30433 3.15% 21 4 17 29 46 63 207 A_12_1 15817 0 0 NA NA NA NA NA NA B_13_1 14419 0 0 NA NA NA NA NA NA B_14_2 30132 4.11% 38 1 10 13 33 42 267 B_15_1 15806 3.59% 15 1 13 29 38 46 127 B_16_1 31630 0.94% 8 1 19 48 37 52 60 Average 2.42% 23 1 9 18 28 38 267

Statistics for cooking burner use are presented in Table 41. Some cooking was logged in every apartment in this study, with use times varying between 0.7 and 11% of monitored minutes, averaging 3.3% of minutes. A total of 583 cooking events were identified, with an average count of 36 events in each apartment, lasting an average of 32 minutes. The longest recorded cooking event was just over 300 minutes.

118

Table 41: Summary of All Cooking Burner Activities in Each Apartment Unit, Where at Least One Cooking Burner Was Being Used

AptID Total (min)

Total Cooking

Burner (%)

Count of Cooking Burner Events

Cooking Burner Event Durations (min)

Min 25th Media

n Mean 75th Max A_1_2 30372 4.89% 53 3 12 20 28 39 127 A_2_2 14741 2.38% 54 4 12 14 20 17 130 A_3_2 24360 1.95% 16 6 15 24 30 35 87 A_4_1 26078 0.31% 8 2 9 11 10 11 16 A_5_1 34558 1.58% 16 2 14 26 34 41 105 A_6_1 31662 0.93% 13 2 13 15 23 39 60 A_7_2 32940 1.74% 23 4 10 18 25 30 83 A_8_1 30432 6.06% 87 1 13 20 24 30 89 A_9_1 33123 9.42% 91 4 9 18 34 51 159 A_10_1 33541 0.89% 7 10 18 39 45 56 119 A_11_1 33144 3.23% 23 4 13 33 47 58 271 A_12_1 33329 1.67% 25 4 10 13 22 22 148 B_13_1 31319 0.66% 4 8 10 15 52 57 169 B_14_2 30527 3.25% 32 4 12 16 36 45 267 B_15_1 34693 11.35% 100 4 16 37 46 54 302 B_16_1 32069 2.84% 31 2 9 22 29 43 97 Average 3.32% 36 1 12 19 32 41 302

A cooking index that included either burner use or kitchen exhaust fan use was assessed, with summary statistics provided in Table 42. The purpose of this index is to indicate the maximum amount of time that a kitchen exhaust fan would be used if it were used for all cooking events in addition to actual use. This index varied from 0.6 to 13% of the monitored period, averaging 4.6% of minutes. A total of 657 such events were identified, with an average count of 41 events per apartment and duration of 34 minutes. The longest single event lasted 324 minutes. This combined index is only used to assess coincident operation with the bathroom exhaust fan and wall furnace. This provides an estimate of spillage or downdrafting that would occur if the kitchen fan were always used during cooking burner activity.

119

Table 42: Summary of All Cooking Activities in Each Apartment Unit, as Characterized by Either Cooking Burner Operation or Kitchen Fan Operation

AptID Total (min)

Total Cooking

(%)

Count of Cooking Events

Cooking Event Durations (min)

Min 25th Median Mean 75th Max A_1_2 30636 6.30% 58 3 12 21 33 41 142 A_2_2 14771 8.42% 53 2 12 16 23 21 209 A_3_2 24523 2.25% 24 1 7 17 23 32 87 A_4_1 26278 0.63% 22 1 3 7 8 11 26 A_5_1 34851 2.98% 31 2 17 23 33 40 144 A_6_1 31800 0.97% 15 1 9 15 21 30 60 A_7_2 33111 2.30% 29 2 9 18 26 31 96 A_8_1 30660 11.10% 88 1 20 35 39 49 155 A_9_1 33404 9.71% 95 2 10 18 34 50 159 A_10_1 33738 1.42% 9 6 24 47 53 71 119 A_11_1 33359 3.81% 24 4 17 35 53 62 287 A_12_1 33569 1.66% 25 4 10 13 22 22 148 B_13_1 31974 0.64% 4 8 10 15 52 57 169 B_14_2 30956 5.54% 49 4 10 14 35 41 324 B_15_1 34933 13.27% 100 4 16 38 46 55 302 B_16_1 32318 3.01% 31 2 9 22 31 46 97

Average 4.63% 41 1 12 21 34 44 324

Statistics for coincident use of the kitchen exhaust fan and cooking burner operation are summarized in Table 43 and displayed visually in Figure 33. As expected, coincident operation was less than either kitchen exhaust fan or cooking use statistics, because sometimes cooking occurred with no exhaust use and the exhaust fan may have been left on after cooking finished or possibly used to remove other odors from the kitchen. Coincident use varied between 0 and 5.6% of monitored minutes, averaging 1.5% of minutes. 292 events were identified, with an average count of 18 per apartment at 22 minutes duration. The longest single coincident use event lasted 210 minutes.

120

Table 43: Summary of Coincident Cooking Burner and Kitchen Fan Operation

AptID Total (min)

Total Cooking

(%)

Count of Cooking Events

Cooking Event Durations (min)

Min 25th Median Mean 75th Max A_1_2 30372 0.85% 9 9 17 25 29 35 59 A_2_2 14741 3.27% 40 2 5 8 12 12 121 A_3_2 24300 0.13% 3 4 9 13 10 14 14 A_4_1 26018 0.04% 3 2 3 3 3 4 5 A_5_1 30184 1.63% 14 5 18 27 35 38 105 A_6_1 30478 0.30% 8 4 6 10 11 16 21 A_7_2 28818 1.07% 16 2 9 14 19 20 71 A_8_1 30055 5.80% 82 3 11 17 21 27 80 A_9_1 28880 2.32% 49 1 4 9 14 17 57 A_10_1 31853 0.49% 6 6 8 24 26 43 49 A_11_1 30418 2.49% 20 4 10 21 38 59 191 A_12_1 15817 <0.005% 0 NA NA NA NA NA NA B_13_1 14419 <0.005% 0 NA NA NA NA NA NA B_14_2 30132 2.27% 19 1 12 19 36 46 210 B_15_1 15806 3.07% 15 1 13 25 32 40 127 B_16_1 31630 0.74% 8 1 19 26 29 44 60 Average 1.53% 18 1 8 14 22 27 210

Figure 33: Use of Kitchen Exhaust Fan, Cooking and Coincident Use for Each Apartment

121

Despite the commitment by participants to operate exhaust fan whenever cooking, the cooking and exhaust fan use figures diverge quite sharply in some cases; and in some cases where the values are roughly similar, only moderate fractions are coincident. In apartment B_15_1, which did the most cooking of any unit, the exhaust fan was only operated roughly one-third of the time that cooking was logged, and nearly all kitchen exhaust use was coincident with cooking. Apartments A_2_2, A_5_1, and A_8_1 had exhaust fan usages double or triple monitored cooking rates. Apartments A_7_2, A_10_1 and A_11_1 had similar cooking and exhaust fan usage fractions, but their coincident usage was somewhat lower.

Many factors may be contributing to the apparent divergence between cooking and kitchen exhaust fan use. First, participants were not given specific instructions about when to start and stop the exhaust fan. Occupants may have sometimes failed to operate the exhaust fan while cooking, and at other times the exhaust fan may have been left on after cooking was completed. Algorithms for detecting cooking may have missed some events, particularly during oven usage. It is also possible that timestamps on the various sensors used were not in perfect alignment, such that coincident cooking and fan usage were artificially separated. Given this variability and the uncertainties, all assessments of ‘cooking’ coincident with heating system operation include both cooking and exhaust fan usage. The combined cooking index was not used to inspect wall furnace temperature data, which wsa not expected to respond in any way to cooking burner operation with no kitchen exhaust fan operation.

Coincident Operation Assessments As reported on heating system operation, cooking activity, and kitchen and bathroom exhaust fan use. These data are combined to assess the frequency that each apartment unit was in a challenge condition. Summary results are provided in Table 44. We assessed operation of the heating appliance coincident with cooking (combined cooking burner and kitchen exhaust fan index, identified as ‘Heating+Cooking’) and bathroom exhaust fan use, both individually and combined together (‘Heating+BathFan’, ‘Heating+Cooking+BathFan’). The team also assessed the coincident operation of the bathroom exhaust and cooking, irrespective of heating system operation (‘Cooking+BathFan’). These coincident operation assessments give an additional estimate of the potential frequency of downdrafting and spillage conditions. Spillage or downdrafting likely did not occur during all coincident exhaust fan usage, but these results should put an upper bound on the frequency of occurrence in these apartments.

We conservatively consider any combination that includes Cooking+BathFan as possibly being the worst-case depressurization condition, and describe them as such for the remainder of this report. This is a conservative or upper-bound estimate of worst case conditions because (a) the Cooking metric includes all the time that the kitchen exhaust fan was used and any other time the main cooking appliance was used); (b) this considers that all kitchen exhaust fan use could be on the highest setting (i.e. the setting used for worst-case conditions), even though some use is undoubtedly on the lower speed, quieter setting; and (c) times don’t exclude when windows were open (which is appropriate since some participants were advised to open windows when using the kitchen exhaust fan, and would have in any case been impossible to analyze since we didn’t monitor window opening).

122

The results indicate correlations between activities in some apartments but not others. Activities are considered to be correlated if there is more frequent overlap than would be predicted as the simple product of the individual activity frequencies. The team expected activities to be correlated, as cooking and bathing only occur during waking, occupied time periods, and heating hopefully occurs more frequently during occupied times. Yet, in many cases, the frequency of coincident operation is roughly the same as predicted by simply multiplying the use fractions for each element together, e.g. 1% bath fan use and 1% kitchen fan use would be coincident 0.01 x 0.01 = 0.0001 or 0.01% of the time if not correlated. There were exceptions, for example in Apt A_9_1, there was much more coincident heating, bath fan use and cooking than would be predicted if the activities were unrelated (actual coincidence was 0.075% of minutes versus 0.025% calculated as the product of the three independent frequencies). Similarly, in Apt A_7_2, there was more frequent Cooking + Bath Fan use than would have occurred if the two activities were uncorrelated (0.17% vs. 0.06% calculated from the product of Cooking and Bath Fan use. Setting aside these cases, when events were only very weakly associated, the coincident operation values were generally one or more orders of magnitude lower than the individual activities. Notably, the correlation between activities may be lower in continuously occupied low-income senior housing than in a home occupied by a family, where activities are more likely to be condensed into shorter periods before and after work.

Table 44: Summary of Coincident Operation of the Heating System, Cooking (Either Cooking Burners or Kitchen Exhaust Fan), and Bathroom Exhaust Fan

AptID Heating

(%)

Cooking (Burners or

Kitchen Fan) (%)

Bath Fan (%)

Heating+ Cooking

(%)

Heating + Bath Fan (%)

Heating+ Cooking + Bath Fan (%)

Cooking + Bath Fan (%)

A_1_2 3.86% 6.30% NA 0.69% NA NA NA A_2_2 0.63% 8.42% 0.54% 0.01% <0.005% <0.0005% <0.005% A_3_2 8.89% 2.25% 1.02% 0.33% 0.16% <0.0005% 0.10% A_4_1 0.20% 0.63% 0.85% 0.02% <0.005% <0.0005% 0.01% A_5_1 2.64% 2.98% 1.22% 0.09% 0.11% <0.0005% 0.02% A_6_1 3.66% 0.97% 1.35% 0.04% 0.08% <0.0005% 0.03% A_7_2 0.07% 2.30% 2.62% 0.05% <0.005% <0.0005% 0.17% A_8_1 2.48% 11.10% 13.45% 0.55% 0.74% 0.033% 1.98% A_9_1 4.81% 9.71% 5.17% 0.35% 0.31% 0.075% 1.09% A_10_1 2.17% 1.42% 0.83% <0.005% 0.06% <0.0005% 0.01% A_11_1 1.50% 3.81% 6.26% <0.005% 0.58% <0.0005% 0.22% A_12_1 7.88% 1.66% 9.34% 0.09% 0.50% 0.015% 0.09% B_13_1 1.14% 0.64% NA <0.005% NA NA NA B_14_2 0.25% 5.54% NA <0.005% NA NA NA B_15_1 4.05% 13.27% NA 2.34% NA NA NA B_16_1 0.00% 3.01% NA 0.00% NA NA NA Average 2.76% 4.63% 3.88% 0.29% 0.23% 0.011% 0.34%

123

Coincident operation was generally very low, but not always. For example, in Apt B_15_1 2.3% of monitored minutes were spent with the heating system operating coincident with cooking. In this apartment, the minimum condition required for spillage during diagnostic testing was the kitchen fan on low and bathroom fan on high (i.e., not continuous low flow) (see Table 30). This was another example of correlated operation leading to higher than random coincidence. The coincidence of Heating+Cooking and Heating+BathFan occurred 0.29% and 0.23% on average, with maxima of 2.34% and 0.74% of monitored minutes across apartments. Heating+Cooking+BathFan was an order of magnitude lower (0.011% of minutes, maximum of 0.075%). Coincident Cooking+BathFan was similar to Heating+Cooking and Heating+BathFan, with an average of 0.34% of monitored minutes.

Complete statistics on coincident Cooking+BathFan events are reported in Table 45, representing the conservative estimate of worst-case depressurization frequency. A total of 93 such events were recorded, with average coincident operation of 0.34% of monitored minutes (maximum of 1.98%). An average of six events for each apartment unit were recorded, with an average duration of 13 minutes (maximum duration of 92 minutes). As noted above, these values do not include the apartments in Building B, where continuously low-speed bathroom fans operated and the sensors did not record fan data.

Table 45: Summary Statistics for Coincident Cooking+Bathfan Events in Each Apartment Unit (i.e., Worst-Case Depressurization Conditions)

AptID Total (min)

Total Worst-

Case (%)

Count of Worst-Case

Events

Worst-Case Event Durations (min)

Min 25th Median Mean 75th Max A_1_2 NA NA NA NA NA NA NA NA NA A_2_2 36013 0.00% 0 0 0 0 0 0 0 A_3_2 24289 0.10% 2 10 11 12 12 13 14 A_4_1 26000 0.01% 1 2 2 2 2 2 2 A_5_1 34558 0.02% 3 2 2 2 2 2 2 A_6_1 31647 0.03% 2 5 5 6 6 6 6 A_7_2 32940 0.17% 7 1 3 5 8 12 21 A_8_1 34414 1.98% 47 1 3 8 14 23 61 A_9_1 33119 1.09% 23 1 3 9 16 17 92 A_10_1 35674 0.01% 1 2 2 2 2 2 2 A_11_1 33099 0.22% 4 6 7 8 19 19 53 A_12_1 33333 0.09% 3 4 8 12 10 13 13

Average 0.34% 6 1 3 7 13 17 92 No results are shown for Building B because monitoring data were not obtained for bath fans.

The risk of spillage is greatest when the wall furnace burner operates during worst-case depressurization (Heating+Cooking+BathFan). Most apartments experienced no minutes at this condition. In fact, this only occurred in three of the 11 apartments that had heating, cooking and

124

bath fan data. A total of five such events were identified with an average length of seven minutes and maximum length of 12 minutes. Notably, if bath fan sensors had recorded data from the continuously operated two-speed fans in the apartments in Building B, the team expected the more worst-case events would have been identified, almost certainly with longer durations. If, in fact, the bathroom exhaust operated continuously in Apt B_15_1, then 2.43% of monitored minutes would have been logged at Heating+Cooking+BathFan, though the bath fan airflow rate would have been at the ‘low’ rate (i.e., not the ‘worst-case’ condition).

All of these coincident events are dependent on the operation of the heating system, which can vary with some building features, namely the heating system output relative to the load it is serving. In general, heating system output capacities scale with heating demand, and it is expected the system runtimes to be roughly independent of climate severity. This is because smaller systems are paired with lower loads, and larger systems are paired with larger loads, resulting in similar runtimes. But a system with excess heating output capacity might short-cycle (less runtime) or generally have very different operating times than a ‘right-sized’ or under-sized heating system (longer runtime). Thermostat set points and other occupancy patterns can also have substantial effects on system runtimes. These factors could substantially impact the assessments of coincident exhaust fan and heating system operation.

3.4 Discussion This study sought to improve understanding of the risk of wall furnace exhaust spillage in apartments that failed the worst-case depressurization testing common in CAS protocols. Ten of the 16 apartments recruited to the study failed worst-case testing with installed exhaust fan capacities, and seven of these passed a one-fan challenge condition. The study initially sought to quantify the frequency of spillage when kitchen and bath exhaust fans are used as intended, i.e. whenever cooking or bathing occurs. However, since many of the apartments did not pass the reduced challenge test of kitchen on low and bath fan, and some did not even pass with just the kitchen fan on high speed, participants of most apartments were advised to open a window when using the exhaust fans. As a result of this, the frequency of spillage and downdrafting under typical operating conditions was not accessed. The study nevertheless produced valuable data to inform the frequency of coincident exhaust fan use and wall furnace operation when occupants have been directed to follow the best practice of using available venting whenever cooking or bathing. These coincidence data provide an estimate of potential downdrafting and spillage frequency that are independent from window operation.

Study results support the suggestion by (Rapp et al., 2015) and others that many combustion appliance hazards in residences can be identified by test procedures that focus on visual inspection methods and/or non-worst-case testing. In these apartments, the greatest hazards were: (1) high appliance CO (>1,000 ppm in Apt A_2_2), which was identified and corrected prior to monitoring, (2) appliances that were spilling and tripping spill switches under ‘natural’ test conditions (low-flow continuous bathroom exhaust, in this case), and (3) occupant efforts to defeat the engineered safety features included in their systems (in this case, the intentional blocking of combustion air vents using t-shirts and repeated manual resetting of spill switches).

125

These hazards were identified by visual inspection, testing of appliances under natural conditions, and brief discussions with the occupant(s).

Results also support the idea that the worst-case test condition commonly used in CAS testing might be irrelevant to occupied residences even when exhaust fans are used as recommended (i.e. with all cooking and bathing) and certainly with how they are actually used in most residences. In the apartments measured in this study, it was quite rare for the both kitchen and bath fans to be operated together, at the worst-case depressurization condition used in CAS testing. Coincident usage averaging 0.34% of monitored minutes across apartments, and was 2% in the apartment with the highest coincident usage. The longest continuous period of coincident use of the bath fan with any Cooking (including both cooking burner activity and kitchen exhaust fan use) was 92 minutes and the average event duration was 13 minutes. In those two apartments (with sufficient data) where only one exhaust fan was required to spill the appliance during diagnostic testing, time spent at this ‘Challenge’ condition was greater. In these cases, the Challenge condition occurred for up to 8% of monitored minutes. A wall furnace was operated during the worst-case depressurization condition in only 3 of the 11 units with all the data needed to make this determination. At the very most, a unit spent 0.075% of monitored minutes at this condition, with the longest continuous event lasting 12 minutes. The team was not able to differentiate low- vs. high-speed kitchen fan operation, so time spent at the worst-case condition (kitchen fan on high and bath fan on) may be even lower than reported here. Worst-case depressurization conditions in residences with a greater number of exhaust fans may be even more rare, since coincident operation of all fans to be even lower than reported in this work were expected. But in some of the units most prone to spillage (i.e., in Building B), bathroom fan data failed to log, and these units had continuous low-flow bathroom exhaust fans, which almost certainly increased time spent at or near worst-case or in a reduced challenge condition.

Assessments of unambiguous spillage and downdrafting frequencies are additional outcomes of this work. Downdrafting was quite common, whereas unambiguous spillage was extremely rare. The presence of continuous bathroom exhaust fans in the Building B apartments led to wall furnaces being in a downdraft condition for extended periods. The longest downdraft event was roughly 12-hours. Nearly all of the recorded spillage events occurred when the appliance was already in downdraft; however, draft was established relatively quickly in all cases. Also, due to low rates of CO production by the wall furnaces, the team could not associate spillage with any substantial CO events in the living space. And even with higher CO generation rates, the exhaust fans that caused the downdrafting and spillage would have provided substantial dilution for any CO that was released.

Overall, the results of this study indicate that caution is warranted about the robustness of venting of natural draft wall furnaces in small residences with airtight shells. This finding derives from the observation of frequent downdrafting in such apartments, corresponding to the operation of either the kitchen or bathroom exhaust fan at highest settings. If a small residence with an airtight shell has continuous mechanical exhaust – as several apartments in this study had – then any operation of the kitchen exhaust fan will produce greater depressurization than it would otherwise and raise the risk of downdrafting and spillage.

126

The observed tendency of some apartments in this study to frequently reach a downdraft condition is particularly troubling, because engineered safety features (e.g., spill switches and CVA openings) were disabled in some apartments. Had the one apartment with an intentionally clogged combustion air inlet not been opened prior to monitoring, the team hypothesized higher frequencies of downdrafting and spillage in that unit would have been seen. Similarly, had occupants not been instructed to open windows during cooking, downdrafting and spillage likely would have increased (at least in apartments with spillage under ‘natural’ conditions).

It is difficult to draw firm conclusions about CAS test methods in the larger population of residences in California, due to the small number of residences assessed in this study. It is believed a larger sample set would be required to justify revisions to CAS test protocols. Any future work should address, at least in part, the following additional limitations. First, these apartment units did not have vented clothes dryers, which can contribute to house depressurization and may have long runtimes in some cases. Second, this study also did not cover a range of heating system sizes and thermostat behaviors that might dramatically affect heating system runtimes. For example, apartment units with right-sized heating systems might have substantially more heating burner runtime, such that coincident operation of exhaust devices and heating burners could be substantially higher. Outdoor temperature conditions were especially mild during this study, so heating runtimes may be underestimated relative to a typical heating season. That being said, maximum total operation of any exhaust device was generally in the 10-15% range (see Table 44), so we expect the theoretical maximum period spent at a Challenge condition to still be substantially below 10-15%. Third, this study did not assess residences containing multiple natural draft gas appliances (e.g., tank water heater and a wall furnace). While coincident operation of all appliances would go down, coincident operation of any single combustion appliance with exhaust fans would increase. This is due to the additional gas burner runtime on the second appliance. A caveat to this point is that the larger burner sizes in water heaters and especially central furnaces will produce more heat and stronger draft. Similarly, larger residences tend to have more leakage area, which lessens the depressurization effect of exhaust equipment. Finally, the demography of the study population undoubtedly influenced the results. Eleven of the sixteen apartments in the study had only one resident and the other five had only two residents. Homes with more residents are expected to have higher coincident fan usage rates, though not strictly proportional to the number of people. The fact that many of the apartments had only senior occupants may also have impacted activities such as cooking and bathroom exhaust fan use. Lastly, participants in this study likely used exhaust fans more frequently than is ‘typical’, because they were instructed to use them during all cooking and bathing (which is atypical). Occupants also likely opened windows during cooking more than typical, as these were specific instructions provided to some participants at the start of the study.

3.5 Recommendations Based on the findings of this study the following are recommended:

127

• Additional research should be conducted to assess the suitability of the worst-case challenge conditions that are currently used in combustion appliance safety protocols, with the specific goal of identifying challenge conditions that occur for long enough to result in an air pollutant hazard should a spillage event occur.

• The research into a suitable challenge condition should be conducted in residences with higher than average occupancy and various combinations of exhaust fans in particular including residences with dryers, which were not present in the apartments in this study.

• Future research into coincident exhaust fan usage and potential impacts on venting should not ask residents to modify their normal window use. Advising residents to use kitchen exhaust ventilation when cooking is a reasonable condition because such use is generally recommended as a healthy homes measure.

• Research may also be warranted to assess the indoor air quality impacts of frequent downdrafting causing pilot burner pollutants to enter the living space of the home.

• Consideration should be given to modifying CAS protocols to use the alternative challenge condition that is more commonly encountered in normal use or when residents use kitchen and bath exhaust fans as recommended.

• Combustion safety testing should continue to emphasize carbon monoxide measurements in the flue or vent to identify improperly functioning burners, and careful visual inspection to identify hazards such as blocked combustion ventilation air openings, both of which were observed in this study.

• A potentially highly cost-effective approach to improving knowledge about combustion safety hazards would be to capture information already being collected during CAS tests by developing digital CAS data collection forms that can be readily uploaded to a statewide database of CAS test results.

• Special attention should be paid to the venting of natural draft wall when including envelope air sealing in energy efficiency retrofits of small homes.

• Residents should be educated about the importance of not blocking trickle vents and other combustion ventilation air openings, and inspections of CVA opening by building managers should be considered if they are deemed essential to safety.

• Though not included in this study, the suitability of technology options including furnaces that draw combustion air directly from outside and supply or balanced mechanical ventilation systems (in place of exhaust ventilation) can improve protections in residences that are found to have depressurization-induced combustion safety hazards.

• In light of the finding of frequent downdrafting, there should be an investigation of the impacts to occupant exposures to combustion pollutants from pilot burners and the potential benefits of replacing pilot burners with spark igniters.

128

CHAPTER 4: Capture Efficiency of Cooking-Related Fine and Ultrafine Particles by Residential Exhaust Hoods 4.1 Introduction Pollutant emissions from cooking burners and the cooking of food can substantially and adversely impact air quality in homes. Natural gas burners commonly emit nitrogen dioxide (NO2) and under some conditions emit substantial quantities of carbon monoxide (CO), formaldehyde (HCHO) and ultrafine particles (UFP) (Dennekamp et al., 2001; Moschandreas and Relwani, 1989; Singer et al., 2010a; Wallace et al., 2004). Electric coil resistance burners produce UFP (Dennekamp et al., 2001). Cooking activities produce fine and ultrafine particles and a wide range of irritant and other potentially harmful gases including acrolein (Abdullahi et al., 2013; Buonanno et al., 2009; Fortmann et al., 2001; Fullana et al., 2004; Seaman et al., 2009; Zhang et al., 2010). Gas burners and cooking also release substantial quantities of water vapor that can contribute to moisture related indoor air quality problems (Parrott et al., 2003). Individual cooking events can produce short-term PM2.5 concentrations exceeding 300 µg m-3 and UFP concentrations exceeding 105 cm-3 in homes (Abdullahi et al., 2013; Afshari et al., 2005; Buonanno et al., 2009; He et al., 2004; Wallace et al., 2004; Zhang et al., 2010). A recent study estimated that among Southern California homes that cook with natural gas on a weekly basis, >5% have acute CO and >50% have 1-h NO2 concentrations that exceed the corresponding concentration thresholds for health-based ambient air quality standards (Logue et al., 2013).

Indoor concentrations of pollutants generated during cooking can be reduced through use of an exhaust hood positioned above the cooktop or an exhaust fan in the kitchen. Exhaust hoods remove some fraction of the emitted pollutants before they mix into the general air volume of the kitchen with additional removal as air from the kitchen is exhausted outdoors. Exhaust fan effectiveness can be described as a capture efficiency (Li et al., 1997; Li and Delsante, 1996; Singer et al., 2012a) that quantifies the fraction of generated pollutants removed either directly or over the duration of exhaust fan operation. Alternately, effectiveness can be framed as the reduction in pollutant concentrations in the kitchen or other location in the home (Rim et al., 2012; Zhang et al., 2010). Direct removal without mixing into the kitchen is termed first-pass CE; including removal from the exhaust provided by the fan gives total CE.

Published studies conducted in laboratories, test homes and in the field have reported CE or other metrics of exhaust hood removal effectiveness for gases, particles, or moisture produced by natural gas burners (Delp and Singer, 2012; Farnsworth et al., 1989; Rim et al., 2012; Singer et al., 2012a). Performance has been shown to vary with airflow; hood geometry and height with respect to the cooktop; whether front, back or oven burners were used; and hood design, which is most prominently differentiated by the collection volume offered by the hood. A few studies have reported on the effectiveness of range hoods at reducing concentrations of particles generated by cooking activities (Sjaastad and Svendsen, 2010; Zhang et al., 2010).

129

Currently in the US there is no direct information available to consumers about the pollutant removal effectiveness of cooking exhaust hoods. The Home Ventilating Institute (HVI) provides third party certification of airflows and sound levels determined using standard test methods (HVI, 2013a; HVI, 2013b). Rated airflows and sound levels are listed in a catalog that is updated monthly by HVI (http://www.hvi.org/proddirectory/index.cfm). Standard IEC-61591 (IEC, 2005) quantifies grease removal effectiveness for an oil-drop heating event by pre- and post-weighing of the hood and filters and includes a test for odor removal; but this test is not commonly used to rate products sold in the US.

A relatively simple test method was used recently by Delp and Singer (2012) to characterize CE of seven exhaust hoods in a controlled laboratory study and by Singer et al. (2012) to quantify CE for 13 exhaust hoods and two downdraft systems installed in residences. The method calculates CE as the ratio of incremental CO2 mass removal through the exhaust hood to CO2 generation by the burners. Incremental CO2 exhaust flow is calculated as the product of the measured airflow rate through the exhaust hood and the measured increase in CO2 concentration in the exhaust flow during burner use. The method looks at incremental CO2 flow because there is a baseline flow of CO2 from ambient air and from additional CO2 exhaled by the cook and other building occupants. The method uses CO2 as a surrogate for all combustion products emitted by a gas burner. The test method incorporates pots of water on the cooktop burners to simulate the impact of cooking vessels on the dynamics of the exhaust plumes.

The work described in this paper was initiated with the dual objectives of (1) quantifying CE for particles generated during typical cooking events and (2) conducting a preliminary assessment of the applicability of the simple, CO2-based test method as an indicator of capture efficiency for cooking-related particles.

4.2 Materials and Methods 4.2.1 Overview Capture efficiencies (CE) were determined for four under-cabinet exhaust hoods under carefully controlled conditions in an experimental room. CE for burner pollutants was determined directly by comparing the CO2 mass flow through the exhaust hood to the CO2 produced at the burner. CE for cooking particles was determined indirectly by comparing particle concentrations in the room when a hood was in use to concentrations measured during the same cooking activity with no hood installed. The indirect approach was required because particle losses in the hood and ductwork would bias the concentrations seen in the hood exhaust stream and consequently bias the calculated particle CE. The room was supplied with particle free air through HEPA filters and the room was maintained at a positive pressure of 1 Pa relative to the outdoors. The flow rate of the main exhaust air pathway – through a blower door – was modulated to maintain a constant total exhaust flow as airflow through the cooking exhaust hoods was varied.

4.2.2 Range Hoods Particle CEs were determined for four hoods previously investigated by Delp and Singer (2012). The tested hoods included a low cost model (L1), an Energy Star qualified model (E2), a

130

premium hood (P1), and a combined microwave exhaust hood (M1) that represent common geometries and ranges in airflow rates. The microwave was mounted at the same height as the other hoods but the bottom of the unit was closer to the stovetop, as occurs in typical installations. The microwave drew air from both the bottom and top front of the unit.

4.2.3 Experimental Setup A schematic of the laboratory layout is shown in Figure 34. The volume of the room was 121 m3 (5.8 m by 7.2 m by 2.9 m high). A simulated residential cooking area was affixed to a 2.4 m by 2.4 m section of wall located approximately 1/3 of the length of the room closest to the supply air. Each range hood was mounted on this wall with the top of the hood positioned 76 cm above the cooktop of a 76 cm wide cooking range (Delp and Singer, 2012). The hoods were mounted between drywall boxes installed to simulate a kitchen with wall cabinets adjoining the hood. The cooking range was installed between drywall boxes topped with steel sheeting to simulate side cabinets and countertops. The top of the room consisted of structural 22.9 cm I-beams and the bottom of the steel roof deck with fiberglass batt insulation. Semi-rigid plastic sheeting was attached across the bottoms of the I-beams over the cooking area to keep the rising exhaust plume out of the shallow channels formed by the I-beams. Baseline experiments were conducted with no hood in place above the stove.

The cooktop had one nominal 12,000 BTU h-1 (12.7 MJ h-1) burner at the front right position and three nominal 9500 BTU h-1 (10.0 MJ h-1) burners. The range was supplied 99.97% methane from certified cylinders (Airgas). Fuel flow was measured using a mass flow meter (Alicat, Model MLD-20SLPM-D/5M), factory calibrated for methane with an accuracy of 1%. Flow was reported at a reference condition of 1 atm and 25 C and logged at 1 Hz. Fuel flow was controlled using the burner adjustment knobs on the appliance.

Upon installation, a range hood was connected to a 0.6 m long section of 15.2 cm smooth galvanized ducting followed by 6 m of 15.2 cm diameter aluminum flexible ducting to vent to the outside. An Energy Conservatory Minneapolis Duct Blaster flow measurement device comprising a calibrated fan and throttling ring was placed inline approximately 2.4 m downstream of the hood. The Duct Blaster quantifies airflow based on the pressure difference measured across the throttling ring; it reports flow at a reference air density of 1.2014 kg m-3. An Energy Conservatory Automated Performance Testing (APT) measurement and control unit measured and recorded airflows and the Teclog software provided with the APT maintained a constant airflow rate. The flow rate through the hood was provided using both the hood fan and the Duct Blaster. An additional inline fan (Soler & Palau, PM-150X) was installed in the ducting to boost airflow when the exhaust hood and Duct Blaster together were not able to achieve the nominal airflow rates of the exhaust hood. Pressures and airflows were logged at 2 s intervals using Teclog software. The APT also measured temperature and relative humidity in the duct.

131

Figure 34: Experimental Configuration of the Room Used to Measure Gas and Particle Phase Capture Efficiency

The room is 5.8 m wide, 7.2 m in length, and 2.9 m high, and is drawn to scale.

The room was operated at the same overall ventilation airflow rate for all experiments. Particle-free air was supplied to the room through HEPA filters (Airhandler, Terminal module 3EJY1) connected to inline blowers (Dayton, model 5TCK9) using 30 cm diameter aluminum flexible ducting. Room air was exhausted using an Energy Conservatory Minneapolis Blower Door. The pressure difference across the blower door was monitored using an Energy Conservatory DG-700 pressure and flow gauge with Teclog software to control the fan speed. The combined airflow supplied to the room was 765 L s-1 (1620 cfm) as measured by constant injection tracer gas and confirmed by the measured SF6 tracer decay rate of 23 hr-1. Two household axial fans

132

were added to improve mixing and reduce directional airflows around the range. Estimated airflows induced by Fans A and B were 292 L s-1 and 552 L s-1. The pressure across the building shell was monitored and recorded. When a range hood was used, the flow out of the blower door was lowered by the amount of flow through the hood while maintaining a positive pressure of approximately 1 Pa in the room.

Preliminary analysis of experimental results raised concerns about the accuracy of the recorded hood flow rates when the inline fan was used. The exhaust airflow rate for each range hood that was operated with the inline fan was checked using a second Duct Blaster connected to the inlet of the hood with a custom-fabricated connector. This is a routine field method, often called powered capture hood, used to measure flows for non-standard air system registers as described in the Duct Blaster manual (Energy Conservatory, 2007). For two of the three hoods used with the inline fan, the actual flow rate was higher than the set point. Adjustment factors were used to obtain accurate airflows for the hoods operated with the inline fan.

4.2.4 Airflow and Mixing Verification Experiments During the set-up phase, tracer release experiments were conducted to investigate mixing and airflow patterns in the room. In one experiment, SF6 was released at the supply air inlet and concentrations were measured at the exit and at other locations within the room using an infrared gas analyzer (MIRAN SapphIRe Model 205B-XL, Thermo Scientific). The flow in the region in front of the stove was visualized using a smoke machine. The number, placement, and settings of mixing fans were adjusted to achieve generally consistent concentrations around the room and to minimize short-circuiting from the stove area or from the supply to the outlet.

The supplemental information in Lunden et al. (2014, 2015) provides details of experiments conducted to explore two key mixing questions: (1) whether there was any short-circuiting recirculation from the area above the hood – where pollutants would rise after not being captured on first pass – to the room air being drawn into the hood from the nearby surroundings, and (2) whether concentrations measured at the sampling location used throughout the CE tests accurately reflect concentrations in the air leaving the room through the blower door. Another experiment was conducted to compare CE values calculated using the direct method (based on mass flow of CO2 through the exhaust hood) and the indirect method (based on particle flow out of the room). During a stir-fry cooking procedure, SF6 was released into the wok through copper tubing formed into a Y and SF6 concentrations were measured in both the hood (E2) and the room exhaust. Another experiment examined the impact of ventilation and mixing fans on CE of burner pollutants: the blower door, supply and mixing fans were all turned off and CO2 CEs were measured for hood E2. Results for these experiments are provided in the Supplement to Lunden et al. (2015).

4.2.5 Pollutant Measurements Carbon dioxide concentrations in the range hood exhaust were measured each 2 s using a PP Systems EGM-4 infrared analyzer. The flow in the exhaust was turbulent. The sampling location was 2 m downstream of the hood connection, corresponding to 12 duct diameters. Radial uniformity at this location was verified by moving the sampling probe through a full transverse during pilot experiments.

133

Particle number concentrations in room air were measured using two instruments. A condensation particle counter (CPC, Model 3781, TSI Incorporated) measured all particles ≥6 nm. An optical particle counter (OPC, Model BT-637S, MetOne Instruments) measured particles in 6 size bins: 0.3 to 0.5 µm, 0.5 to 0.7 µm, 0.7 to 1.0 µm, 1.0 to 2.0 µm, 2.0 to 5.0 µm, and above 5.0 µm. Measurements were recorded every 10 s.

Particle instruments sampled approximately 1.2 m away from the blower door fan 0.6 m off of the floor. This location was selected as the exhaust flow through the blower door ensured that all cooking emissions passed through this general location. The particles were sampled through a 1m length of 0.64 cm ID conductive tubing that was divided using flow splitters within a few centimeters of each instrument.

4.2.6 Cooking Procedures Measuring CE for cooking-related particles requires cooking procedures with suitably repeatable particle generation rates; these depend on both the food being cooked and the cooking procedure. Carefully scripted cooking procedures were developed with the aim of repeatability.

We sought to develop at least two cooktop procedures and one oven procedure with distinct plume characteristics that vary in the challenge they present for capture. A procedure for pan-frying a hamburger on medium heat on the back burner was developed because the lower pan height, lower plume velocity, and the location on the back burner fully underneath all of the range hoods presents a geometry that facilitates high capture efficiency for exhaust gases. A stir-fry procedure using high heat with a wok on the front burner was developed as a more challenging, lower capture efficiency condition due to the wider and higher pan, more disruption of the plume by the activity of the cook (Huang et al., 2010), faster plume rise from higher energy input (Kosonen et al., 2006) and the location of the pan on the front burner. Efforts to develop an oven cooking activity are described in the Supporting Information.

Hamburgers were cooked on medium heat in vegetable oil in a nominal 11-inch (28 cm) stainless steel frying pan (Winco SSFT-11 Master Cook). The hamburgers were 85% lean / 15% fat, pre-formed, pure ground beef patties purchased at a local store of a nationwide specialty grocer. Individual patties were separated, wrapped in foil, and stored in a freezer. Patties were fully thawed and weighed before cooking. The pan was weighed and 3.5 ± 0.05 g of canola oil was added. The burner was lit and adjusted to a gas flow rate of 1.7 ± 0.1 lpm using the appliance burner control. The pan with oil was placed on the burner and the hamburger was added after 2 min. The burger was lightly pressed with a spatula for 5 s at each 1 min interval, flipped at 3 min, then pressed again at 1 min intervals until the burner was turned off after 6 min. The pan was allowed to cool for 30 s, after which it was covered, weighed, and removed from the room.

Green beans were selected for the stir-fry because they maintained structural integrity for the desired duration of cooking. Green beans were stir-fried using peanut oil in a nominal 15-inch (38 cm) carbon steel wok (Thunder Group Inc.). Green beans were purchased frozen in 680 g bags at the same store as the burgers. Twenty bags were purchased together and mixed to

134

produce a homogeneous supply for experiments. Beans were measured into 150 ± 0.5 g portions, sealed in new plastic bags, and returned to the freezer. For each experiment, the wok was weighed and 10 ± 0.1 g of peanut oil added. The burner was lit and adjusted to a fuel flow rate of 4.4 ± 0.1 lpm. The wok was placed on the burner, and the beans added after 90 s. The beans were stirred continuously using a silicone spatula. After 3 min of cooking the burner was turned off and the wok was allowed to cool for 30 s; it then was covered, weighed, and removed from the room. Stir-fry experiments were conducted on the front burner for all hoods. The stir-fry procedure was conducted on the back burners to evaluate the effect of cooking burner position, using Hood E2 operating on low and high. A nominal 12-inch (30 cm) wok was used for these experiments because the 38 cm wok did not fit over the middle of the back burner.

4.2.7 CO2-Based Capture Efficiency Capture efficiency for burner combustion products was measured using the CO2-based method outlined in Delp and Singer (2012). Two burner configurations were used: 1) both back burners and 2) both front burners. Covered 5L stainless steel pots filled with approximately 3L of water were placed on the cooktop burners to simulate use. The pots were placed on the stovetop, the burners were ignited and operated for 3 min, then turned off. The researcher moved away from the range after placing the pots of water to minimize activity-based air currents that can affect CE. This approach will be referred to as the POW (pots of water) CE test. Fuel flow rates were 8.5 ± 0.3 lpm for the two front burners and 7.6 ± 0.6 lpm for the two back burners.

The CO2-based method was also used to calculate burner combustion product CE during the cooking activities; results of these calculations are identified in the figures as “Cook”.

There were four important differences between the POW procedure and the scripted cooking activities: (1) the POW procedure used two burners whereas the cooking activities each used a single burner, resulting in different fuel flows and heat generation rates, (2) the water in the POW procedure provides a heat sink that can impact the burner plume; (3) the pans used in the cooking protocols have a different shape than those used in the POW procedure; and (4) a technician stood in front of the range to execute each cooking activity but avoided this area during the POW protocol. All four are expected to affect CE. Differences in vessel geometry, heat generation and removal rates, and the use of two vs. one burner all should impact both the geometry and fluid dynamic properties of the plume rising from the cooking burner(s). The activity of the cooking technician may intermittently disrupt the plume.

To explore the extent to which these factors affect CO2-based CE, the CE of the frying pan and wok each were measured using a modified version of the POW test. The wok or pan was half-filled with water and covered with foil. Fuel flow rates and burner position were the same during the pan tests as during the cooking activities and CE was quantified similarly as the POW test.

4.2.8 Experimental Schedule One of the supply fans providing HEPA filtered air operated continuously to maintain room air particle concentrations below ambient levels. The second supply fan and the exhaust fan(s)

135

were started at the beginning of each day. Cooking experiments did not begin until the total particle concentrations in the room were stable and low, typically around 500 cm-3.

Experiments were conducted June through October 2013. Five experiments each were conducted for most combinations of range hood, fan speed, and cooking procedure. Four experiments each were conducted for the hamburger and green beans on hood L1 at low fan speed and six were conducted with the hamburger on hood L1 at high speed. Experiments without a hood were conducted between the series of tests on the first and second hoods and toward the end of the experiments for a total of 18 experiments for both the pan and stir-frying. The range top was cleaned between experiments.

4.2.9 Capture Efficiency Calculations As noted in the Introduction, total capture efficiency (CE) or pollutant removal effectiveness by a range hood is simply the mass exhausted through the hood divided by the mass emitted from the source:

1

Total CE includes pollutants captured by the hood before they mix into the room as well as those that escape into the room and are then removed with the room air that is exhausted from the hood,

2

The mass that is captured directly from cooking and does not escape into the room can be used to calculate a “First Pass” capture efficiency CEFP,

3

In theory, the CEFP offers a more useful measure of the performance of the hood as it can be combined with whatever general exhaust benefit provided by the extant space and ventilation conditions. In practice, the airflow dynamics of the room can impact first pass capture efficiency, so neither metric is uniformly applicable across installations.

CE for the exhaust of a combustion-based cooking burner can be determined by relating the mass flow of exhaust gases through the hood to the generation rate at the cooking surface. Carbon dioxide (CO2) can be used as a surrogate for the exhaust gas mixture. For the typical case of complete (or nearly complete) combustion, the mass generation rate of CO2, S, can be calculated (estimated) from the fuel rate of the burner and knowledge of the carbon content of the fuel. The total mass flow of CO2 through the hood can be calculated from the measured concentration of CO2 in the exhaust air stream and an independent measurement of the exhaust airflow rate. To focus on the direct capture of CO2 from the burner, the concentration of CO2 from the room, Croom, is subtracted, as shown in Equation 4:

136

4

In calculating CE, the team accounted for a density difference of 1.5% in the reference conditions used by the Duct Blaster compared to those of the Alicat fuel flow sensor.

If a person is present in the immediate vicinity of the cooking surface and range hood, the contribution of exhaled CO2 must be considered in the analysis. Exhaled CO2 from occupants of the room is included in the room air CO2 measurement and requires no special attention when quantifying first-pass CE. For these calculations, a CO2 exhalation rate of 0.33 ± 0.15 LPM (Emmerich and Persily, 2001) was assumed. This value is based on the assumption of a single female cook of 163 cm height with a metabolic activity factor of 1.3 (ASHRAE, 2013). The ± 0.15 LPM incorporates uncertainty in the rate for an individual female, the potential contribution of a cooking assistant coming close enough to contribute to CO2 seen in the hood and some fraction of the cook’s exhaled CO2 not being drawn directly into the hood.

Determining the CE – either total or first-pass – for pollutants generated by an actual cooking process is more challenging because the generation rate is typically not calculable. This challenge may be addressed by generating a reproducible quantity of pollutants. In principle, one could volatilize or release an inert tracer into a cooking vessel and release a suitable quantity of heat at the cooking surface to simulate the buoyant plume that is characteristic of the represented cooking activity. The inert tracer generation approach was deemed unsuitable for the current study, which evaluated CE for particles relevant to actual cooking activities.

The approach employed in this study was to attempt to generate a reproducible aerosol through execution of the tightly controlled cooking protocols described above. Instead of tracking particle mass, particles were tracked by number concentration in specified size ranges. The number of particles generated from the cooking event was quantified by measuring the airflow rate and concentration of particles leaving the chamber through the blower door, which was the route of all air exiting the room when no range hood was used. Since particle loss in the room is expected to be low due to the low residence time (high air exchange rate), the number of particles being removed by ventilation when no hood was operating is a good estimate of the emission rate and thus a valid reference point for calculating CE when hoods were operating:

5

This equation uses the symbol M for consistency with prior equations even though the team switched from talking about mass to number concentration. The equations could be used for either quantity. Similar to the CO2-based calculation above, there is a need to account for particles in the room air that are not associated with the cooking activity, Cbkg.

The total CE associated with range hood use is calculated as the difference in the number of particles exhausted through the blower door when no hood was used relative to the number exhausted through the blower door when a hood was used:

6

137

The number of cooking-associated particles not captured by a range hood – i.e. those exiting through the blower door when a hood was used – is calculated as shown in Equation 7:

7

The particle concentration in the room in Equation 7 (Croom@exh,with-hood) reflects the effects of particles being captured directly from cooking as well as some particles being removed with the room air exhausted through the hood during cooking. To calculate CEFP, the room air particles removed by the hood must be added to the count of particles not captured on the first pass:

8

Substitution of Equations 5 and 8 into Equation 6 results in the first pass capture by the hood

9

The resulting equation for CEFP for particles is

10

The difference in particle concentrations in the numerator and denominator of Equation 10 are calculated by integrating the particle concentration measured from the time when cooking begins, t0, to when the particle concentration returns to background, tb,

11

Ci,room@exh is the particle concentration measured at the room exhaust for particle size bin i and Ci,bkg is the background when no cooking is occurring. Particle concentrations vary between individual cooking events. As a result, the CEFP is calculated using the average emissions over multiple cooking events

12

The calculated CE represents the average CE over the measurement period.

The formulas for first-pass capture efficiency for the direct and indirect methods use the concentrations measured in the room, but in two different places: near the hood and near the exhaust. The two values may not be equivalent depending on the mixing conditions in the room. For a well-mixed room, the two concentrations will be equal. Any difference between these two values will be important when comparing resulting CEs.

All CEs reported in the Results section of this paper are first-pass CEs.

138

4.2.10 Uncertainty in Calculated Capture Efficiencies Uncertainty in the calculated CE for burner pollutants is a function of uncertainties in (a) the measured hood airflow rate, (b) the measured CO2 concentrations in the hood exhaust and background room air, and (c) the calculated mass emission rate of CO2. Uncertainty in the hood airflow rate is estimated to be 5% based on verification tests with a second duct blaster. The uncertainty in CO2 concentration measured in the hood exhaust was calculated to be 3% based on the standard deviation of the EGM4 calibrations; uncertainty in the background concentration was estimated as 2%. The mass emission rate of CO2 depends on both gas flow and breathing rate. Uncertainty in fuel flow rate was 1%. Uncertainty in the breathing rate is estimated to be 45%, as noted earlier. Because these uncertainties are combined in quadrature, the fractional uncertainty of the breathing rate increases as the gas flow rate to the burner decreases and the contribution of breathing to the emitted CO2 become greater. Total uncertainty in the CO2-based CE ranged from 6.6% to 14.4%.

Uncertainty in particle CE was estimated using the standard error of CEs from replicate experiments combined in quadrature with the instrument measurement uncertainty. Day-to-day instrument repeatability was estimated to be 3%. In general, the standard error across replicate experiments was larger than instrument accuracy and dominated uncertainty in particle CE.

4.3 Results and Discussion 4.3.1 Measured Concentrations Figure 2 shows typical experimental results for pan-frying and stir-frying. Each figure shows time resolved measurements of CO2 (ppm) in the hood exhaust, fuel flow (lpm), and size-resolved number concentrations of particles measured in the room: ≥6 nm (# cm-3) as measured by the CPC and 5 of 6 particle ranges measured by the OPC (# L-1). There was no substantial response for the largest particle size (>5 µm). Grey vertical lines on the figures mark the following experimental events: pan with oil placed on the stovetop, food added to the pan, fuel turned off, and pan covered and removed from the room. The top two plots show CO2 in the hood and fuel flow increasing sharply just after the burners are ignited and remaining roughly constant throughout the experiment.

Stir-frying emitted more particles than pan-frying, and the difference was more pronounced for particles larger than 0.5 µm. Particle emissions from pan-frying were predominately in the 0.3 to 0.5 µm size range and below, as indicated by the OPC and CPC responses. The stir-fry plot shows particle concentrations (≥6 nm) starting to increase after the pan and oil were added and before the beans were added, whereas the other particle size channels increased sharply only after the beans were added. Adding the pan and oil to the medium flame did not produce such a clear increase in particles as adding the wok and oil to the larger flame did. This reflects a more rapid heating of the oil in the 1 mm steel wok over high heat compared to the pan with 5 mm aluminum bottom over medium heat. For the stir-fry, the OPC response shows that after the quick initial increase in all particle sizes when beans were added, particles larger than 0.3 µm increased only gradually through the remainder of the cooking event. The continuing, steep increase in total particle concentration (CPC) throughout the experiment indicates ongoing

139

emissions of particles smaller than 0.3 µm, including ultrafine particles. Concentrations of all particle sizes followed an exponential decay (at the room AER) after the fuel was turned off and the pan or wok was covered and removed from the room.

Figure 35: Concentrations of CO2 in the Hood Exhaust and Particles in Room Measured by the CPC and OPC Resulting from (a) Pan-Fry and (b) Stir-Fry Cooking Activities

An example of the variability in measured particle concentrations across replicate cooking experiments is displayed in Figure 36, which shows concentrations of 6 nm and larger particles for pan-frying hamburgers without a hood and for E2 operating on low and high speed. These data demonstrate a core challenge in conducting performance assessments using cooking-generated particles: despite precisely defined cooking protocols there was substantial variability in particle concentrations across replicate implementations for many of the conditions. The relative standard deviation (RSD) of the time-integrated concentration for the no-hood condition was 23%, while those measured with the hood were 10% and 50% for the low and high fan speeds. Particle concentrations returned to the same range of background concentrations for both hood and no hood conditions. Plots analogous to Figure 36 are shown for each hood and cooking condition in Supplemental Figures S6 and S7 of Lunden et al. (2015).

140

Figure 36: Concentration or Particles ≥6 nm as Measured by the CPC from a Sampling Point Near the Exhaust Outlet of Room

Data shown for pan-frying experiments with no hood present (black dashed line) and with hood E2 operating on low (red) and high (blue) fan speeds. Note that particle concentrations are graphed using a log scale. The time scale denotes the seconds since the pan was put on the burner.

The calculated number of particles leaving the room through the blower door and RSDs for each hood and flow condition is provided in Table 46 for pan-frying and Table 47 for stir-frying. RSDs for specific combinations of hood and airflow rates ranged from approximately 10% to 90%. The mean RSD for the CPC and the five OPC channels with no hood operating was 48% for stir-frying and 35% for pan-frying. The mean RSDs across all combinations of hood and fan speed were 36% for stir-frying and 36% for pan-frying. Variability in emissions reflected by the RSDs is similar for the two cooking activities both with and without hood use, indicating that the variability in emissions is primarily from the cooking rather than airflow. A previous study that used a scripted procedure for pan-frying of steak (Sjaastad and Svendsen, 2010) reported similar variability in particle emissions in the 0.3 to 0.5 µm size range: RSDs ranged from 12% go 62% with an average value of 37%.

141

Table 46: Summary Results for Hamburger Pan-Frying Experiments1

CPC OPC (particle size range) Flow2 N Total

Particles 0.3–0.5 µm 0.5–0.7 µm 0.7–1.0 µm

1.0 −2.0 µm

2.0–5.0 µm

Set- up

lps Mn (1012)

RSD (%)

Mn (109)

RSD (%)

Mn (108)

RSD (%)

Mn (107)

RSD (%)

Mn (107)

RSD (%)

Mn (107)

RSD (%)

No Hood

18 36.1 23 31.1 91 29.7 33 124 22

76.4 22 32.4 21

L1 51 4

3.94 21 1.79 39 3.73 31 18.7 29

11.5 28 4.93 29

L1 81 5

0.94 42 0.36 21 0.89 27

4.35 30. 2.72 36 1.15 36

E2 52 5

3.22 10 6.28 55 5.40 24

23.9 20 14.5 19 6.06 18

E2 109 5

0.21 50 0.88 55 0.83 34 2.33 24

1.20 18 0.42 21

M1 68 5

0.74 25 1.18 80 0.90 27

3.48 18

1.95 20 0.68 18

M1 137 5

1.83 30 1.54 19 1.67 17

7.57 25 4.62 24 1.65 30

P1 138 5

0.30 73 0.14 69 0.34 87

1.24 70

0.70 67 0.23 87

1 Mean (Mn) total number of particles removed through the blower door and relative standard deviation (RSD) of measurements by the condensation particle counter (CPC) and optical particle counter (OPC). 2

Measured values.

Table 47: Summary Results for Green Bean Stir-Frying Experiments1

CPC OPC Flow2 N Total

Particles 0.3–0.5 µm 0.5–0.7 µm 0.7–1.0 µm

1.0 −2.0 µm

2.0–5.0 µm

Set- up

lps Mn (1013)

RSD (%)

Mn (1010)

RSD (%)

Mn (1010)

RSD (%)

Mn (1010)

RSD (%)

Mn (1010)

RSD (%)

Mn (109)

RSD (%)

No Hood

18 4.84 26 8.89 37 4.64 54 2.70 57 1.78 58

8.05 56

L1 51 4 4.25 35 8.59 45 4.44 57 2.49 59 1.59 59

6.78 58

L1 81 6 3.12 40 7.43 25 3.88 35 2.20 35 1.43 35

6.12 33

E2 52 5 2.94 33 8.39 12 4.46 18 2.56 21 1.63 22

7.12 24

E2 109 5 3.02 70 7.53 18 3.80 23 2.07 25 1.27 26

5.35 25

M1 68 5 3.18 24 8.22 30 4.39 39 2.54 41 1.63 42

7.28 42

M1 137 5 1.68 52 3.69 29 1.53 37 0.82 38 0.53 37

2.27 36

P1 138 5 2.11 54 3.77 34 1.65 45 0.89 46 0.57 47

2.46 47

1 Mean (Mn) total number of particles removed through the blower door and relative standard deviation (RSD) of measurements by the condensation particle counter (CPC) and optical particle counter (OPC). 2

Measured values.

142

4.3.2 Capture Efficiency Results Figure 37 shows the burner pollutant (CO2) CEs calculated for pots of water (POW), pan with water (Pan), and cooking (Cook), along with cooking particle CEs calculated for hamburger pan-frying on a back burner. These CEs and all others presented in this section are first-pass CEs. CEs are shown for particles ≥6 nm (CPC) and for five size ranges measured by the OPC. CEs for pan-frying on the back burner were mostly above 90% for high fan speeds and mostly above 80% for low fan speeds. CO2-based CEs measured during the cooking events were slightly lower than those measured by POW and pan with water for most combinations of hood and fan speed. Possible contributions to this result are the cook disrupting flow near the burners and the POW and pan with water drawing more heat than the pan with burger, producing less energetic plumes. With the exception of the POW test for E2 on low speed (62% CE), CEs were 75% or better for both CO2 and particles for all the hoods and fan settings.

Figure 38 displays CO2-based CEs calculated for pots of water (POW), wok with water, and cooking, as well as particle CEs for stir-frying of green beans on front burner(s). CO2-based CEs were similar for POW (2 pots on 2 burners) and wok with water (1 burner), suggesting that neither cooking vessel shape nor the number of burners had a large effect (unless the effects happened to balance).

CO2-based CEs measured during cooking were substantially lower than those measured during POW and wok with water experiments for most combinations of hood and fan setting.

In stir-frying experiments, CEs for particles >0.3 µm were lower than CEs determined for CO2 for all but one combination of hood and setting (M1 on high). The difference between particle CE and CO2 CE during cooking varied across hoods. For example, P1 (only one fan speed) had particle CEs of 56–69% across the particle size bins and a cooking CO2 CE of 72%. By contrast, Hood E2 on high speed had size-resolved particle CEs of 15–38% and a CO2 CE of 54% during cooking. When hoods L1, E2 and M1 were operated at low speed, particle CE values for the stir-fry were extremely low for the particles in the OPC size range: CEs for 0.3 µm and larger particles were 3–16%. CEs for all particles ≥6 nm were 34% for M1 on low, 39% for E2 and 12% for L1 on low fan speed. CO2-based CEs measured during the stir-fry experiments with hoods on low speed were all in the range of 35–38%.

Figure 39 compares capture efficiencies for stir-frying on either the front or back burners with hood E2. This plot further illustrates the sharp contrast between particle CEs at the two locations. While the particle capture efficiencies measured on the front burner are low – varying between 4% and 39% by hood, fan setting and particle size – those on the back burner were much higher, varying between 70% and 99%. The results for cooking on the back burner are similar to those measured during pan-frying for the same hood. In addition, like the pan-frying results, CEs for the different particle size bins were similar to the CO2 CE measured using POW. At the high flow rate, CEs for both cooking activities on the back burner for hood E2 are close to 100%. At the low fan setting, particle CEs averaged 83% for pan-frying and 78% for stir-frying. It is clear that the burner location has a much more significant effect on capture efficiency than

143

the cooking activity. Cooking on the rear burner means that the hood fully covers the area over which the emissions occur, resulting in higher capture efficiency.

Figure 37: Capture Efficiencies (CE) Calculated Using CO2 Measured in Hood Exhaust and Particles Measured in Room for Pan-Frying Hamburger

CEs calculated using CO2 in the shaded, left side of the graph show results for two pots of water on rear burners operating on highest setting (POW) using the method described in Delp and Singer (2012); for a single frying pan half-filled with water, covered with foil, and placed on a single back burner operating on medium heat (Pan); and during the cooking experiments (Cook). CE calculated using particle concentrations measured in the room are shown on the right side of the graph. Values for CE are shown for all particles ≥6 nm, as measured by the CPC, and for five size bins measured by the OPC.

144

Figure 38: Capture Efficiencies (CE) Calculated Using CO2 Measured in Hood Exhaust And Particles Measured in Room for Stir-Frying Green Beans

CEs calculated using CO2 in the shaded, left side of the graph show results for two pots of water on front burners operating on highest setting (POW) using the method described in Delp and Singer (2012); for a single wok half-filled with water, covered with foil, and placed on a single front burner operating on high heat (Wok); and during the cooking experiments (Cook). CE calculated using particle concentrations measured in the room are shown on the right side of the graph. Values for CE are shown for all particles ≥6 nm, as measured by the CPC, and for five size bins measured by the OPC.

145

Figure 39: Capture Efficiencies (CE) Calculated Using CO2 and Particle Measurements for Stir Frying on the Front or Back Burners

CEs calculated using CO2 in the shaded, left side of the graph show results for two pots of water on rear or front burners operating on highest setting (POW) using the method described in Delp and Singer (2012); for a single wok half-filled with water, covered with foil, and placed on a single front or back burner operating on high heat (Wok); and during cooking experiments (Cook) on front or back burners. CE calculated using particle concentrations in the room are shown on the right side of the graph. Values for CE are shown for particles ≥6 nm, as measured by the CPC, and for size bins measured by the OPC.

4.3.3 Airflow and Mixing Verification Experiments Results of the airflow and mixing experiments are presented in the supplemental information to Lunden et al. (2015); only the summary findings are noted. One set of experiments indicated a short-circuit loop that increased the apparent first-pass capture efficiency relative to the ideal pattern of pollutants mixing evenly throughout the room if they are not captured on the first pass. The same experiments found variations of roughly 15% in SF6 concentration in the vicinity of the blower door during an experiment. These variations are much smaller than the variance in room air particle concentrations across replicate experiments. These two features are displayed in Supplemental Figures S2 and S3 of Lunden et al. (2015). In a set of experiments in which SF6 was released into the wok during stir-fry procedures, capture efficiencies calculated based on SF6 measurements in the hood exhaust were roughly 5% higher than CEs calculated from measurements in the room. This result, displayed in Supplemental Figure S4, is consistent with short-circuiting causing a bias in the apparent first-pass CE. Supplemental Figure S5 (Lunden et al., 2015) shows that CO2-based CEs measured with the blower door, supply and

146

room air mixing fans off were indistinguishable from CEs measured under the standard mixing conditions that applied through the experiments measuring CE.

4.3.4 Comparisons to Published Studies on Cooking Hood Effectiveness The new results for burner exhaust pollutants, obtained using the CO2-based POW, pan with water, and cooking tests, in some cases indicate higher CEs than previously reported for the same hoods executing the same POW procedures in a different laboratory (Delp and Singer, 2012). At this time, the differences cannot be explained. Experiments exploring the potential effects of airflow patterns were conducted with the configuration reported in this new study but not in the earlier study.

The particle CEs presented in Figure 37 to Figure 39 expand on the published record of range hood effectiveness for particles. Rim et al. (2012) quantified exhaust hood effectiveness for ultrafine particle concentrations measured in the bedroom of a test house, examining burner-generated particles in the size range of 2-20 nm. That study found a strong size dependence on effectiveness over the range studied with higher effectiveness for higher hood airflow rates and back burners compared with oven or front burners. Yet even for the lowest airflow hood with oven or front burner use, effectiveness exceeded 70% for the largest particles (14-20 nm) observed. Sjaastad and Svendsen (2010) measured concentrations of 0.3–0.5 µm particles 1.3 m to the side of the burner during and just after pan-frying a beefsteak in a 56-m3 chamber. The scripted cooking procedure was repeated for 9–13 replicates of each condition, examining cooking area location (wall, corner, or middle of room), variations in exhaust airflow, and two installation heights for three hoods. The study did not report effectiveness per se and hood airflow variations had a large effect on overall chamber air exchange. These factors and the single measurement location to the source limit the generalizability of the reported results.

4.4 Conclusions Capture efficiencies (CE) were measured for cooking-generated particles during scripted cooking procedures and for burner produced CO2 during the same cooking activities and separately for pots and pans containing water. CEs were determined for four exhaust hoods including a basic low-volume hood, an energy-efficient device with a flat bottom, a microwave exhaust hood and a high performance hood with large capture volume. For pan-frying a hamburger over medium heat on the back burner, CEs for particles were similar to those for burner combustion products, as indicated by CO2, and mostly above 80%. For stir-frying green beans in a wok on the front burner over high heat, CEs for burner produced CO2 during cooking varied by hood and airflow: CEs were 34–38% for low (51–68 L s-1) and 54–72% for high (109–138 L s-1) settings. CEs for 0.3–2.0 µm particles during front burner stir-frying were 3–11% on low and 16–70% on high settings. High CE was obtained when stir-frying on a back burner. Results indicate that CO2-based CEs measured for combustion pollutants are not predictive of CEs for cooking-generated particles under all conditions; but they may be suitable to identify devices with CEs above 80% for both burner exhaust gases and cooking-related particles.

147

CHAPTER 5: Effect of Fuel Wobbe Number on Pollutant Emissions from Advanced Technology Water Heaters 5.1 Introduction The research reported in this chapter is part of a larger effort to explore the potential air quality impacts resulting from increasing use of liquefied natural gas (LNG) in California. The reported research was developed to focus on pollutant emissions from advanced technology residential water heaters when these devices are operated with fuels having the higher Wobbe numbers associated with many LNG supplies. Wobbe number (WN) is a relative measure of the energy delivery rate of a gas through a fixed orifice to a burner and is an indicator of the interchangeability of fuel gases. This research expands on an earlier study by Lawrence Berkeley National Lab (LBNL) that reported on the sensitivity of pollutant emission factors to fuel Wobbe number for conventional technology storage and on-demand (tankless) water heaters (Singer et al., 2010a). That study measured emissions of carbon monoxide (CO), nitrogen oxides (NOX), nitric oxide (NO), nitrogen dioxide (NO2, estimated as the different between NOX and NO), formaldehyde, and number of particles (PN). The storage water heaters evaluated in that study had low emissions of all pollutants other than NOX and there was only a slight sensitivity of NOX to fuel Wobbe number for a subset of the storage water heaters. Several of the on-demand water heaters had moderate or higher emissions of CO, NO2, and formaldehyde. NOX was generally lower among the on-demand water heaters but more sensitive to fuel WN.

After the earlier research was completed, a variety of new technologies were incorporated into residential water heating products. The focus of the study reported here was to examine the effect of fuel WN on pollutant emissions from these advanced technology water heaters. Performance of “ultra low-NOX” water heaters that comply with South Coast Air Quality District Rules 1121 and 1146.2 was of particular interest.

The South Coast Air Quality Management District (SCAQMD) has two certifications for water heaters. Rule 1121 (amended 2004) requires residential type natural gas water heaters manufactured and sold on or after January 1, 2008 with heat input rates less than 75,000 Btu/hr be certified to a NOX emission level less than or equal to 10 ng (calculated as NO2) per joule of heat output (23 lb per billion Btu of heat output) or 15 ppmv NOX at 3% O2, dry (17.5 lb per billion Btu of heat input). Rule 1146.2 (amended 2006) requires water heaters manufactured and sold on or after January 1, 2013 with heat input rates between 75,000 Btu/hr and 400,000 Btu/hr (Type I units) be certified to a NOX emission level less than or equal to 14 ng (calculated as NO2) per joule of heat output or 20 ppmv NOX at 3% O2, dry. These water heaters are referred to as ultra low-NOX. Water heaters manufactured and sold prior to and through December 31, 2012 were certified to 40 ng (calculated as NO2) per joule of heat output or 55 ppmv NOX at 3% O2, dry. These water heaters are referred to as low-NOX.

148

5.2 Methods Water heaters utilizing specific technology combinations were identified as targets for testing. These target technologies were communicated to manufacturers who expressed willingness to donate production units for the purposes of advancing the research effort. Of the twelve technology combinations targeted, ten were donated and two were purchased. Eight of the water heaters were storage type, three were on-demand and one was a hybrid. The following technologies were represented:

• 2 Ultra-low NOX, natural draft storage water heaters

• 2 Ultra-low NOX, power vent storage water heaters

• 2 Energy Star qualified, induced draft, low NOX storage water heaters

• 2 Condensing storage water heaters, one ultra-low NOX and one power-vent low-NOX

• 2 Ultra-low NOX, non-condensing on-demand water heaters

• 1 Ultra-low NOX, condensing on-demand water heater

• 1 Ultra-low NOX, hybrid on-demand water heater

Experiments were conducted on new water heaters installed in a combustion laboratory at LBNL. Appliances were operated through test cycles designed to capture key variations and transient features of actual use patterns. Appliances were operated with line-supplied natural gas from Pacific Gas & Electric (the local gas utility) as a baseline fuel and with simulated LNG blends. Blends were formulated to represent imported LNG and diluted with nitrogen to achieve Wobbe numbers of roughly 1420, 1395, and 1365, as calculated from fuel heating value in British thermal units [Btu] per standard cubic foot. PG&E line gas was in the range of 1344–1365 Wobbe across experiments. Simulated LNG blends were delivered in premixed cylinders. Premixed blends were 12% ethane, 1.6% propane, and 86.4% methane (1420 Wobbe, blend 3C); 8.0% ethane and 92% methane (1390 Wobbe, blend 1C); and 7.9% ethane, 1.9% nitrogen and 90.2% methane (1360 Wobbe, blend 1N). Each water heater was tested on two separate days with PG&E line gas and also tested with the simulated LNG fuels.

With each fuel, storage water heaters were operated through two 15 minute [min] burns with at least 10 min of non-operation before the first burn and 8 min between burns; on-demand water heaters were operated through water draws of 1, 2, and 4 gallons per min [gpm], each lasting for 8 min with 10 min between the burns.

The exhaust stream for each water heater was monitored to determine time-resolved CO, NOX, NO, CO2 and O2 concentrations, and integrated air samples were collected to determine formaldehyde and acetaldehyde concentrations through both burns. These data were used to calculate air-free exhaust concentrations, which were then combined with fuel flow information to calculate emission factors in units of nanograms of pollutant emitted per Joule of fuel energy consumed (ng/J). These units are equivalent to micrograms of pollutant per kilojoule of fuel energy (µg/kJ).

149

Measurement results were analyzed to assess whether emission factors for CO, NOX, NO2, formaldehyde and acetaldehyde followed any trend with fuel Wobbe number. For each appliance and each pollutant, the analysis identified the best-fit linear relationship between emission factors and Wobbe number. A t-test analysis was then performed to determine if the best-fit line is statistically discernible from a line with slope equal to 0, i.e. to determine the likelihood that there is in fact no trend with Wobbe number. The threshold used for this analysis was p<0.10; when this condition is met, there is less than a 10% chance that there is in fact no relationship between emission factor and Wobbe number.

5.3 Results Table 48 and Table 49 provide summary results in terms of pollutant emission factors for each water heater using the average results from the PG&E line gas experiments. The tables also present the statistically discernible trends of emission factors changing with increasing fuel Wobbe number. The impact of Wobbe number is provided for a change of 50 Wobbe number.

Table 48 shows that the three water heaters that were not rated as ultra low-NOX – AW04, AW06 and AW07 – had NOX emissions of 36–38 ng/J with PG&E fuel. Those emission factors are higher than the NOX emissions measured for conventional technology devices in the recent LBNL study (Singer et al. 2010). These three water heaters use conventional burner technologies but utilize enhancements in venting and heat transfer to extract more energy from the combustion process to improve energy efficiency. Increasing Wobbe number caused small increases in NOX emissions of these devices, estimated at 2-5% for a 50 Wobbe number change.

Consistent with design intent, the storage water heaters with ultra-low NOX burner technologies had much lower emissions than those with conventional technology burners. Four of the models had NOX under 10 ng/J and the fifth had NOX emissions of 12 ng/J. It is noted these results are for the test conditions used in this study and should not be considered as applicable to standards based on regulatory test procedures. Consistent with results obtained by Singer et al. (2010) for the lower-NOX on-demand water heaters, these ultra-low NOX water heaters showed substantial and statistically discernible (p<0.10) sensitivity to fuel Wobbe number. The storage water heaters with ultra-low NOX burners had NOX increases of 15 to 51% with a 50 Wobbe number increase in fuel.

The majority of NOX was in the form of NO; estimated NO2 (NOX-NO) was 2-15% of total NOX for conventional burners, 12-22% for 4 of 5 ultra-low NOX devices and 40% for the last device.

150

Table 48: Summary of Experiments and Results for Storage Water Heaters1, 2

ID Description Rating (Btu/h)

PG&E3

WN CO

(ng/J) NO2

(ng/J) NOX

(ng/J) HCHO (ng/J)

Acetald. (ng/J)

AW01 Ultra-low NOX; natural draft; “round” burner; 38 gal 38,000

1350 0.0 1.2 7.5 0.06 0.04

1348 (ns) (ns) 2.0 (ns) (ns) AW02 Ultra-low NOX; natural draft;

“cake-pan” burner; 40 gal 40,000 1347 -0.94 1.4 6.3 0.04 0.04

1346 (ns) (ns) 1.3 (ns) (ns) AW03 Ultra-low NOX; power vent;

“cake-pan” burner; 40 gal 42,000 13485 2.1 2.5 6.2 0.15 0.11

(ns) (ns) 1.4 (ns) (ns) AW04 Condensing; “pancake”

burner; 50 gal 76,000 1347 11.3 5.2 35.2 0.02 0.07

1353 (ns) (ns) (ns) (ns) (ns) AW05 Ultra-low NOX; power vent;

“round” burner; 40 gal

36,000 1348 1.2 1.2 9.9 0.12 0.1

1349 (ns) (ns) 1.5 (ns) (ns) AW06 Energy Star, induced draft;

“pancake” burner; 29 gal 60,000 1352 1.2 0.9 38.1 0.02 0.05

1348 1.4 (ns) 1.4 (ns) (ns) AW07 Energy Star, induced draft;

“pancake” burner; 40 gal 40,000 1348 1.5 4.5 36.1 0.07 0.08

1349 -0.3 (ns) 1.7 (ns) (ns) AW08 Condensing; ultra-low NOX;

“cylinder” burner; 50 gal 100,000 1350 0.6 1.8 11.6 0.01 0.03

1348 0.3 (ns) 5.9 (ns) (ns) 1 For CO, NO2, NOX, HCHO, and Acetaldehyde, the first value in each cell is the average full burn emission rate for the PG&E line gas tests. The emissions change for a 50 Btu/scf increase in fuel WN, below the average full burn emission rate, is only shown for a p value ≤0.10. Results with p<0.05 are italicized. Low p-values indicate statistically significant results 2 NOX = nitrogen oxide, CO = carbon monoxide, NO2 = nitrogen dioxide, HCHO = formaldehyde, Acetald = acetaldehyde; ns = not significant 3 PG&E Line Gas Wobbe Number is listed for each experiment 4 Negative value indicates near zero CO and is within instrument linearity (±2% of Full Scale) 5 Only one experiment was conducted with PG&E line gas.

The results in Table 48 reveal that, similar to the prior results for conventional technology storage water heaters, advanced technology (ultra-low NOX, induced draft, and energy efficient condensing) designs have very low emission rates of CO, formaldehyde and acetaldehyde. Only one device – the condensing water heater AW04 – had CO above 10 ng/J. None of the storage water heaters had emissions of either aldehyde exceed 0.2 ng/J.

Results for the on-demand devices, including the hybrid AT04, are shown in Table 49.

With PG&E line gas, three of the four had NOX emissions in range of 8.7 to 10.6 ng/J; AT03 had substantially higher NOX emissions of 19.2 ng/J with line gas. Only two of the four devices (AT01 and AT04) had NOX increase with Wobbe number with statistical discernibility; the

151

increase was 44-45% with a 50 Wobbe number fuel change. Estimated NO2 accounted for 34-52% of total NOX for these devices.

All four of the on-demand water heaters had CO emissions with PG&E line gas exceed 10 ng/J and NO2 exceed 3 ng/J. Three had formaldehyde exceed 0.2 ng/J. CO emissions for the four on-demand devices operating with PG&E fuel were 11 to 65 ng/J as compared to a range of CO emissions of 19 to 87 ng/J for the tankless water heaters reported by Singer et al. (2010). One interesting difference in the new data is that the device with highest CO – AT03 – also shows the greatest sensitivity to fuel Wobbe number. This result is driven primarily by emissions for the 2 GPM water flow.

Table 49: Summary of Experiments And Results For On-Demand Water Heaters, Adjusted for Effect of Water Flow Rate1, 2

ID Description Rating (Btu/h)

PG&E3 WN

CO (ng/J)

NO2 (ng/J)

NOX (ng/J)

HCHO (ng/J)

Acetald. (ng/J)

AT01 Ultra-low NOX, non-condensing

11,000-180,00

0

1348 36.0 4.7 10.6 0.6 0.23

1349 7.2 1.7 4.8 (ns) (ns)

AT02 Ultra-low NOX, non-condensing

11,000-150,00

0

1362 32.1 4.9 9.5 0.50 0.14

1365 -1.5 (ns) (ns) -0.1 (ns)

AT03 Ultra-low NOX, condensing

15,000-150,00

0

1344 64.5 6.6 19.2 0.26 0.08

1347 34.5 (ns) (ns) (ns) (ns)

AT04 Ultra-low NOX, hybrid

16,000-100,00

0

1349 10.9 3.1 8.7 0.05 0.11

1346 7.1 0.7 3.8 (ns) (ns) 1 For CO, NO2, NOX, HCHO, and Acetaldehyde, the first value in each cell is the average full burn emission rate of the PG&E line gas tests. This value averages emissions from the 1, 2, and 4 gallon per minute flow rate burns. The emissions change for a 50 Btu/scf increase in fuel WN, below the average full burn emission rate, is only shown for a p value ≤0.10. Emission changes in italics indicate p-values ≤ 0.05. Low p-values indicate statistically significant results 2 NOX = nitrogen oxide, CO = carbon monoxide, NO2 = nitrogen dioxide, HCHO = formaldehyde, Acetald = acetaldehyde; ns = not significant 3 PG&E Line Gas Wobbe Number is listed for each experiment

5.4 Conclusions The controlled experiments conducted in this study indicate that for some advanced technology water heaters, emissions of some pollutants are sensitive to the Wobbe number of the fuel. The most prominent sensitivity was an increase in NOX emissions with increasing Wobbe number. All five of the ultra low- NOX storage water heaters and two of the four ultra low- NOX on-demand water heaters had statistically discernible increases in NOX with fuel Wobbe number. The largest percentage increases occurred for the ultra low-NOX water heaters. Another prominent result was a statistically discernible change in CO emissions with Wobbe number for all four of the on-demand devices tested. The device with highest CO emissions with line gas also had the largest CO increase (53%) with increasing fuel Wobbe number.

152

5.5 Chapter 7 Recommendations The research described above leads to the following recommendations.

1. The hazard of acute exposures to combustion pollutants from natural gas cooking burners should be addressed through initiatives in multiple areas inlcluding the following: • Expand education and awareness among the public and building industry professionals

(builders, retrofitters, realtors) that kitchen exhaust is a necessary building service that should be employed regularly when cooking, particularly with gas burners, and that cooking on back burners improves pollutant capture for almost all range hoods.

• When the test method for range hood becomes formalized as an ASTM standard, the Title 24 code should be revised to require a minimum capture efficiency – ideally of 90% or greater – rather than a minimum airflow as it currently requires.

• Research should be conducted to determine if vented over the range microwaves – which are installed in many new California homes despite not strictly meeting the Title 24 code – provide adequate protection when used.

• California should look for opportunities to support and incentivize the development and demonstration of automatic range hoods.

2. Efforts should be made to accelerate replacement of existing appliances with pilot burners, especially cooking appliances, ideally installing direct vent, sealed combustion to eliminate any possibility for spillage. Energy efficiency programs may offer an avenue for these efforts.

3. Combustion appliance safety protocols should identify and utilize challenge conditions suitable to the homes under study, and not defer to “worst case” conditions of all exhaust fans on highest settings. Protocols should also emphasize visual assessment of vent sizing.

4. Regulatory testing of low-NOX water heaters should include high Wobbe fuels.

153

GLOSSARY

Term Definition AEA Association for Energy Affordability AHU Air handling unit AM Arithmetic mean ASTM American Society for Testing and Materials CAAQS Californial Ambient Air Quality Standard CAS Combustion appliance safety CASIF Combustion Appliance Safety Inspection Form CAZ Combustion appliance zone

CE Capture efficiency, the fraction of pollutants emitted during cooking or burner use that are removed by the exhaust fan or range hood

cfm Cubic feet per minute CH2O Formaldehyde CO Carbon monoxide CO2 Carbon dioxide CPC Condensation particle counter, measures UFP CREL Chronic reference exposure level CSD California Department of Community Services and Development CVA Combustion ventilation air DOE Department of Energy EPA Environmental Protection Agency ESA Energy Savings Assistance EUC Energy Upgrade California FAU Forced air unit (heating or cooling system that circulate air in the home) ft Foot or feet (length) g Grams (Mass) GM Geometric mean GSD Geometric standard deviation GTI Gas Technology Institute h Hour HCHO Formaldehyde HONO Nitrous acid HVI Home Ventilating Institute IAQ Indoor air quality

154

Term Definition IOU Investor owned utilty kJ KiloJoules (103 Joules) L/s, lps Liters per second LBNL Lawrence Berkeley National Laboratory LIHEAP Low-Income Heating Energy Assistance Program LIWP Low-Income Weatherization Program LNG Liquefied natural gas LOQ Limit of quantitation m Meter MERV Minimum efficiecy rating value, relevant to filter quality mg Micrograms (10-6 grams) min Minutes mm Micrometer (10-6 meter) MOR Microwave (exhaust hood) over range NAAQS National Ambient Air Quality Standard NGAT Natural Gas Appliance Test NGCB Natural gas cooking burners nm Nanometer (10-9 meter) NO Nitric oxide NO2 Nitrogen dioxide NOX Nitrogen oxides ºC Degrees Celsius OEHHA Office of Environmental Health Hazard Assessment (California) ºF Degrees Farenheit Pa Pascals (pressure measurement) PG&E Pacific Gas & Electric company PIER Public Interest Energy Research

PM2.5 Fine particulate matter; the mass of airborne particles with diameters <2.5 micrometers

PN Particle number (typically used to express the concentration in air) POW Pot of water ppb Parts per billion ppm Parts per million RD Relative deviation

155

Term Definition REL Reference exposure level RH Relative humidity RHA Richard Heath and Associates SCAQMD South Coast Air Quality Management District SD Standard deviation T Temperature Title 24 California's building code U.S. / US United States of America U.S. EPA United States Environmental Protection Agency UFP Ultrafine particles, aerosol particles with diameters of <100 nm WCD Worst-case depressurization WN Wobbe number Wx Weatherization

156

REFERENCES

Abdullahi, K.L., Delgado-Saborit, J.M. and Harrison, R.M. (2013) Emissions and indoor concentrations of particulate matter and its specific chemical components from cooking: A review, Atmos. Environ., 71, 260-294.

Afshari, A., Matson, U. and Ekberg, L.E. (2005) Characterization of indoor sources of fine and ultrafine particles: A study conducted in a full-scale chamber, Indoor Air, 15, 141-150.

ASHRAE (2013) Chapter 9. Thermal Comfort, ASHRAE Handbook: Fundamentals, Atlanta GA, ASHRAE, 9.6.

Belanger, K., Gent, J.F., Triche, E.W., Bracken, M.B. and Leaderer, B.P. (2006) Association of indoor nitrogen dioxide exposure with respiratory symptoms in children with asthma, American Journal of Respiratory and Critical Care Medicine, 173, 297-303.

Bhangar, S., Mullen, N.A., Hering, S.V., Kreisberg, N.M. and Nazaroff, W.W. (2011) Ultrafine particle concentrations and exposures in seven residences in northern California, Indoor Air, 21, 132-144.

Brown, S.K., Mahoney, K.J. and Cheng, M. (2004) Room chamber assessment of the pollutant emission properties of (nominally) low-emission unflued gas heaters, Indoor Air, 14, 84-91.

Buonanno, G., Morawska, L. and Stabile, L. (2009) Particle emission factors during cooking activities, Atmos. Environ., 43, 3235-3242.

Delp, W.W. and Singer, B.C. (2012) Performance Assessment of U.S. Residential Cooking Exhaust Hoods, Environ. Sci. Technol., 46, 6167-6173.

Dennekamp, M., Howarth, S., Dick, C.a.J., Cherrie, J.W., Donaldson, K. and Seaton, A. (2001) Ultrafine particles and nitrogen oxides generated by gas and electric cooking, Occupational and Environmental Medicine, 58, 511-516.

Dutton, S.J., Hannigan, M.P. and Miller, S.L. (2001) Indoor pollutant levels from the use of unvented natural gas fireplaces in Boulder, Colorado, Journal of the Air & Waste Management Association, 51, 1654-1661.

Emmerich, S.J. and Persily, A.K. (2001) State-of-the-Art Review of CO2 Demand Controlled Ventilation Technology and Application, Springfield VA, Technology, N. I. O. S. A., National Institute of Standards and Technology, NISTIR 6727.

Energy Conservatory (2007) Minneapolis Duct Blaster Operation Manual (Series B Systems), Minneapolis MN, The Energy Conservatory.

Farnsworth, C., Waters, A., Kelso, R.M. and Fritsche, D. (1989) Development of a fully vented gas range, ASHRAE Trans., 95, 759-768.

157

Fortmann, R., Kariher, P. and Clayton, R. (2001) Indoor air quality: Residential cooking exposures, Sacramento, CA, Prepared for California Air Resources Board, ARB Contract Number 97-330.

Francisco, P.W., Gordon, J.R. and Rose, B. (2010) Measured concentrations of combustion gases from the use of unvented gas fireplaces, Indoor Air, 20, 370-379.

Franklin, P., Dingle, P. and Stick, S. (2000) Raised exhaled nitric oxide in healthy children is associated with domestic formaldehyde levels, American Journal of Respiratory and Critical Care Medicine, 161, 1757-1759.

Franklin, P., Runnion, T., Farrar, D. and Dingle, P. (2006) Comparison of peak and average nitrogen dioxide concentrations inside homes, Atmos. Environ., 40, 7449-7454.

Franklin, P.J., Dingle, P.W. and Stick, S.M. (1999) The effect of formaldehyde levels in homes on lung function and exhaled nitric oxide in healthy children, American Journal of Respiratory and Critical Care Medicine, 159, A775-A775.

Fullana, A., Carbonell-Barrachina, A.A. and Sidhu, S. (2004) Volatile aldehyde emissions from heated cooking oils, J. Sci. Food Agric., 84, 2015-2021.

Garrett, M.H., Hooper, M.A. and Hooper, B.M. (1999) Nitrogen dioxide in Australian homes: Levels and sources, Journal of the Air & Waste Management Association, 49, 76-81.

Garrett, M.H., Hooper, M.A., Hooper, B.M. and Abramson, M.J. (1998) Respiratory symptoms in children and indoor exposure to nitrogen dioxide and gas stoves, American Journal of Respiratory and Critical Care Medicine, 158, 891-895.

Hansel, N.N., Breysse, P.N., Mccormack, M.C., Matsui, E.C., Curtin-Brosnan, J., Williams, D.L., Moore, J.L., Cuhran, J.L. and Diette, G.B. (2008) A longitudinal study of indoor nitrogen dioxide levels and respiratory symptoms in inner-city children with asthma, Environ. Health Perspect., 116, 1428-1432.

He, C.R., Morawska, L.D., Hitchins, J. and Gilbert, D. (2004) Contribution from indoor sources to particle number and mass concentrations in residential houses, Atmos. Environ., 38, 3405-3415.

Hei Review Panel on Ultrafine Particles (2013) Understanding the Health Effects of Ambient Ultrafine Particles. HEI Perspectives 3. , Boston, MA., Health Effects Institute.

Huang, R.F., Dai, G.Z. and Chen, J.K. (2010) Effects of mannequin and walk-by motion on flow and spillage characteristics of wall-mounted and jet-isolated range hoods, Ann. Occup. Hyg., 54, 625-639.

HVI (2013a) HVI Airflow Test Procedure, Wauconda, IL, Home Ventilating Institute, HVI Publication 916-2013.

HVI (2013b) HVI Loudness Testing and Rating Procedure, Wauconda IL, Home Ventilating Institute, HVI Publication 915-2013.

158

Iec (2005) Household range hoods - Methods for measuring performance, Standard IEC 61591, Ed. 1.1, International Electrotechnical Commission.

International Agency for Research on Cancer (2006) IARC Monographs on the Evaluation of Carcinogenic Risks to Humans Volume 88 (2006): Formaldehyde, 2-Butoxyethanol and 1-tert-Butoxypropan-2-ol. .

Kaden, D.A., Mandin, C., Nielsen, G.D. and Wolkoff, P. (2010) 3. Formaldehyde, WHO Guidelines for Indoor Air Quality: Selected Pollutants., Geneva, World Health Organization.

Kile, M.L., Coker, E.S., Smit, E., Sudakin, D., Molitor, J. and Harding, A.K. (2014) A cross-sectional study of the association between ventilation of gas stoves and chronic respiratory illness in U.S. children enrolled in NHANESIII, Environ Health, 13, 71.

Klug, V.L., Lobscheid, A.B. and Singer, B.C. (2011) Cooking Appliance Use in California Homes – Data Collected from a Web-Based Survey, Berkeley CA, Lawrence Berkeley National Laboratory, LBNL-5028E.

Knol, A.B., De Hartog, J.J., Boogaard, H., Slottje, P., Van Der Sluijs, J.P., Lebret, E., Cassee, F.R., Wardekker, A., Ayres, J.G., Borm, P.J., Brunekreef, B., Donaldson, K., Forastiere, F., Holgate, S.T., Kreyling, W.G., Nemery, B., Pekkanen, J., Stone, V., Wichmann, H.E. and Hoek, G. (2009) Expert elicitation on ultrafine particles: likelihood of health effects and causal pathways, Particle and Fibre Toxicology, 6.

Kosonen, R., Koskela, H. and Saarinen, P. (2006) Thermal plumes of kitchen appliances: Cooking mode, Energy Build., 38, 1141-1148.

Less, B.D. (2012) Indoor air quality in 24 California residences designed as high performance green homes. Masters Masters Thesis, University of California at Berkeley, Available via eScholarship: http://www.escholarship.org/uc/item/25x5j8w6.

Less, B.D., Mullen, N.A., Singer, B.C. and Walker, I.S. (2014) Indoor air quality in 24 California residences designed as high performance homes, HVAC&R, Invited revision submitted Aug, 2014.

Li, Y., Delsante, A. and Symons, J. (1997) Residential kitchen range hoods - Buoyancy-capture principle and capture efficiency revisited, Indoor Air, 7, 151-157.

Li, Y.G. and Delsante, A. (1996) Derivation of capture efficiency of kitchen range hoods in a confined space, Build. Environ., 31, 461-468.

Logue, J.M., Klepeis, N.E., Lobscheid, A.B. and Singer, B.C. (2013) Pollutant exposures from unvented gas cooking burners: A simulation-based assessment for Southern California, Environ. Health Perspect., Published online 05-Nov-2013.

Logue, J.M., Klepeis, N.E., Lobscheid, A.B. and Singer, B.C. (2014) Pollutant Exposures from Natural Gas Cooking Burners: A Simulation-Based Assessment for Southern California, Environ. Health Perspect., 122, 43-50.

159

Lunden, M.M., Delp, W.W. and Singer, B.C. (2014) Capture efficiency of cooking-related fine and ultrafine particles by residential exhaust hoods, Berkeley CA, Lawrence Berkeley National Lab, LBNL-6664E.

Lunden, M.M., Delp, W.W. and Singer, B.C. (2015) Capture efficiency of cooking-related fine and ultrafine particles by residential exhaust hoods, Indoor Air, 25, 45-58.

Morales, E., Julvez, J., Torrent, M., De Cid, R., Guxens, M., Bustamante, M., Kunzli, N. and Sunyer, J. (2009) Association of Early-life Exposure to Household Gas Appliances and Indoor Nitrogen Dioxide With Cognition and Attention Behavior in Preschoolers, American Journal of Epidemiology, 169, 1327-1336.

Moschandreas, D., Relwani, S., Johnson, D. and Billick, I. (1986) Emission rates from unvented gas appliances, Environ. Int., 12, 247-254.

Moschandreas, D.J. and Relwani, S.M. (1989) Field-measurements of NO2 gas range-top burner emission rates, Environ. Int., 15, 489-492.

Moschandreas, D.J. and Zabransky Jr, J. (1982) Spatial variation of carbon monoxide and oxides of nitrogen concentrations inside residences, Environ. Int., 8, 177-183.

Mullen, N.A., Li, J., Russell, M.L., Spears, M., Less, B.D. and Singer, B.C. (2016) Results of the California Healthy Homes Indoor Air Quality Study of 2011-2013: impact of natural gas appliances on air pollutant concentrations, Indoor Air, 26, 231-245.

Mullen, N.A., Li, J. and Singer, B.C. (2013a) Participant-assisted data collection in the California Healthy Homes Indoor Air Quality Study of 2011-2013, Berkeley, CA, Laboratory, L. B. N., Lawrence Berkeley National Laboratory.

Mullen, N.A., Russell, M.L., Lunden, M.M. and Singer, B.C. (2013b) Investigation of formaldehyde and acetaldehyde sampling rate and ozone interference for passive deployment of Waters Sep-Pak XPoSure samplers, Atmos. Environ., 80, 184-189.

Neas, L.M., Dockery, D.W., Ware, J.H., Spengler, J.D., Speizer, F.E. and Ferris, B.G. (1991) Association of Indoor Nitrogen-Dioxide with Respiratory Symptoms and Pulmonary-Function in Children, American Journal of Epidemiology, 134, 204-219.

Nitschke, M., Smith, B.J., Pilotto, L.S., Pisaniello, D.L., Abramson, M.J. and Ruffin, R.E. (1999) Respiratory health effects of nitrogen dioxide exposure and current guidelines, International Journal of Environmental Health Research, 9, 39-53.

Noris, F., Adamkiewicz, G., Delp, W.W., Hotchi, T., Russell, M., Singer, B.C., Spears, M., Vermeer, K. and Fisk, W.J. (2013) Indoor environmental quality benefits of apartment energy retrofits, Build. Environ., 68, 170-178.

Parrott, K., Emmel, J. and Beamish, J. (2003) Use of kitchen ventilation: Impact on indoor air quality, The Forum for Family and Consumer Issues, 8.

160

Piazza, T., Lee, R.H., Sherman, M. and P., P. (2007) Study of Ventilation Practices and Household Characteristics in New California Homes, Sacramento, CA, California Energy Commission and California Air Resources Board, CEC-500-2007-033.

Pilotto, L.S., Douglas, R.M., Attewell, R.G. and Wilson, S.R. (1997) Respiratory effects associated with indoor nitrogen dioxide exposure in children, International Journal of Epidemiology, 26, 788-796.

Rass (2009) Residential Appliance Saturation Survey, Sacramento CA, California Energy Commission. http://www.energy.ca.gov/appliances/rass/.

Rim, D., Wallace, L., Nabinger, S. and Persily, A. (2012) Reduction of exposure to ultrafine particles by kitchen exhaust hoods: The effects of exhaust flow rates, particle size, and burner position, Sci. Total Environ., 432, 350-356.

Rim, D., Wallace, L. and Persily, A. (2011) Reduction of exposure to ultrafine particles by kitchen exhaust fans of varying flow rates. In: Proceedings of Indoor Air 2011, Vol. Paper 806, pp. Pape 806.

Roda, C., Kousignian, I., Guihenneuc-Jouyaux, C., Dassonville, C., Nicolis, I., Just, J. and Momas, I. (2011) Formaldehyde exposure and lower respiratory infections in infants: Findings from the PARIS Cohort Study, Environ. Health Perspect., 119, 1653-1658.

Ryan, P.B., Hemphill, C.P., Billick, I.H., Nagda, N.L., Koontz, M.D. and Fortmann, R.C. (1989) Estimation of nitrogen dioxide concentrations in homes equipped with unvented gas space heaters, Environ. Int., 15, 551-556.

Ryan, P.B., Soczek, M.L., Treitman, R.D. and Spengler, J.D. (1988) The Boston Residential NO2 Characterization Study: 2. Survey methodology and population concentration estimates, Atmos. Environ., 22, 2115-2125.

Salthammer, T., Mentese, S. and Marutzky, R. (2010) Formaldehyde in the Indoor Environment, Chem. Rev., 110, 2536-2572.

Schwab, M., Mcdermott, A., Spengler, J.D., Samet, J.M. and Lambert, W.E. (1994) Seasonal and yearly patterns of indoor nitrogen dioxide levels - data from Albuquerque, New Mexico, Indoor Air-International Journal of Indoor Air Quality and Climate, 4, 8-22.

Seaman, V.Y., Bennett, D.H. and Cahill, T.M. (2009) Indoor acrolein emission and decay rates resulting from domestic cooking events, Atmos. Environ., 43, 6199-6204.

Singer, B.C., Apte, M.G., Black, D.R., Hotchi, T., Lucas, D., Lunden, M.M., Mirer, A.G., Spears, M. and Sullivan, D.P. (2010a) Natural Gas Variability in California: Environmental Impacts and Device Performance: Experimental Evaluation of Pollutant Emissions from Residential Appliances, Sacramento CA, California Energy Commission, CEC-500-2009-099.

161

Singer, B.C., Delp, W.W. and Apte, M.G. (2010b) Natural Gas Variability in California: Environmental Impacts and Device Performance: Experimental Evaluation of Installed Cooking Exhaust Fan Performance, Lawrence Berkeley National Laboratory.

Singer, B.C., Delp, W.W., Lorenzetti, D.M. and Maddalena, R.L. (2016a) Pollutant Concentrations and Emission Rates from Scripted Natural Gas Cooking Burner Use in Nine Northern California Homes, Berkeley CA, Laboratory, L. B. N.

Singer, B.C., Delp, W.W., Price, P.N. and Apte, M.G. (2012a) Performance of installed cooking exhaust devices, Indoor Air, 22, 222-234.

Singer, B.C., Delp, W.W., Price, P.N. and Apte, M.G. (2012b) Performance of installed cooking exhaust devices, Indoor Air, 22, 224-234.

Singer, B.C., Hodgson, A.T., Hotchi, T. and Kim, J.J. (2004) Passive measurement of nitrogen oxides to assess traffic-related pollutant exposure for the East Bay Children's Respiratory Health Study, Atmos. Environ., 38, 393-403.

Singer, B.C., Less, B.D., Delp, W.W., Brooks, A., Cohn, S. and Finn, B. (2016b) A field study of wall furnace venting and coincident exhaust fan usage in 16 northern California apartments, Berkeley CA, Laboratory, L. B. N., Lawrence Berkeley National Laboratory.

Sjaastad, A.K. and Svendsen, K. (2010) Different types and settings of kitchen canopy hoods and particulate exposure conditions during pan-frying of beefsteak, Indoor and Built Environment, 19, 267-274.

Spengler, J., Schwab, M., Ryan, P.B., Colome, S., Wilson, A.L., Billick, I. and Becker, E. (1994) Personal exposure to nitrogen dioxide in the Los Angeles Basin, Journal of the Air & Waste Management Association, 44, 39-47.

Spengler, J.D., Duffy, C.P., Letz, R., Tibbitts, T.W. and Ferris, B.G. (1983) Nitrogen dioxide inside and outside 137 homes and implications for ambient air quality standards and health effects research Environ. Sci. Technol., 17, 164-168.

Spicer, C.W., Coutant, R.W., Ward, G.F., Joseph, D.W., Gaynor, A.J. and Billick, I.H. (1989) Rates and mechanisms of NO2 removal from indoor air by residential materials, Environ. Int., 15, 643-654.

Spicer, C.W., Kenny, D.V., Ward, G.F. and Billick, I.H. (1993) Transformations, lifetimes, and sources of NO-2, HONO, and HNO3 in indoor environments, Air and Waste, 43, 1479-1485.

Spicer, C.W., Kenny, D.V., Ward, G.F., Billick, I.H. and Leslie, N.P. (1994) Evaluation of NO2 measurement methods for indoor air quality applications, Journal of the Air & Waste Management Association, 44, 163-168.

Stratton, J.C. and Singer, B.C. (2014) Addressing Kitchen Contaminants for Healthy, Low-Energy Homes, Berkeley CA, Laboratory, L. B. N., LBNL-6547E.

162

Traynor, G.W., Apte, M.G. and Chang, G.M. (1996) Pollutant emission factors from residential natural gas appliances: A literature review, Berkeley, CA, Lawrence Berkeley National Laboratory, LBNL-38123.

Traynor, G.W., Girman, J.R., Apte, M.G., Dillworth, J.F. and White, P.D. (1985) Indoor Air-Pollution Due to Emissions from Unvented Gas-Fired Space Heaters, Journal of the Air Pollution Control Association, 35, 231-237.

Us Epa (2010) Integrated Science Assessment for Carbon Monoxide (Final Report), Washington, DC, U.S. Environmental Protection Agency.

Us Epa (2016) Integrated Science Assessment for Oxides of Nitrogen - Health Criteria (2016 Final Report), Washington DC, U.S. Environmental Protection Agency.

Van Strien, R.T., Gent, D.F., Belanger, K., Triche, E., Bracken, M.B. and Leaderer, B.P. (2004) Exposure to NO2 and nitrous acid and respiratory symptoms in the first year of life, Epidemiology, 15, 471-478.

Walker, I.S., Wray, C.P., Dickerhoff, D.J. and Sherman, M.H. (2001) Evaluation of flow hood measurements for residential register flows, Berkeley CA, Lawrence Berkeley National Laboratory, LBNL-47382.

Wallace, L. (2006) Indoor sources of ultrafine and accumulation mode particles: Size distributions, size-resolved concentrations, and source strengths, Aerosol Sci. Technol., 40, 348-360.

Wallace, L., Wang, F., Howard-Reed, C. and Persily, A. (2008) Contribution of gas and electric stoves to residential ultrafine particle concentrations between 2 and 64 nm: size distributions and emission and coagulation remission and coagulation rates, Environ. Sci. Technol., 42, 8641-8647.

Wallace, L.A., Emmerich, S.J. and Howard-Reed, C. (2004) Source strengths of ultrafine and fine particles due to cooking with a gas stove, Environ. Sci. Technol., 38, 2304-2311.

Wallace, L.A., Ott, W.R. and Weschler, C.J. (2015) Ultrafine particles from electric appliances and cooking pans: experiments suggesting desorption/nucleation of sorbed organics as the primary source, Indoor Air, 25, 536-546.

Wilson, A.L., Colome, S.D., Baker, P.E. and Becker, E.W. (1986) Residential Indoor Air Quality Characterization of Nitrogen Dioxide - Phase I Final Report, Prepared for Southern California Gas Co., Los Angeles CA, Multiple Volumes and Appendices.

Wilson, A.L., Colome, S.D. and Tian, Y. (1993) California Residential Indoor Air Quality Study, Sponsored by Gas Research Institute, P. G. E., San Diego Gas & Electric, Southern California Gas, Integrated Environmental Services, Irvine CA, Multiple Volumes and Appendices.

163

Wilson, A.L., Colome, S.D., Tian, Y., Becker, E.W., Baker, P.E., Behrens, D.W., Billick, I.H. and Garrison, C.A. (1996) California residential air exchange rates and residence volumes, Journal of Exposure Analysis and Environmental Epidemiology, 6, 311-326.

Yamamoto, N., Shendell, D.G., Winer, A.M. and Zhang, J. (2010) Residential air exchange rates in three major US metropolitan areas: results from the Relationship Among Indoor, Outdoor, and Personal Air Study 1999-2001, Indoor Air, 20, 85-90.

Yang, W., Lee, K. and Chung, M. (2004) Characterization of indoor air quality using multiple measurements of nitrogen dioxide, Indoor Air, 14, 105-111.

Zhang, Q.F., Gangupomu, R.H., Ramirez, D. and Zhu, Y.F. (2010) Measurement of ultrafine particles and other air pollutants emitted by cooking activities, International Journal of Environmental Research and Public Health, 7, 1744-1759.

164

A-1

APPENDIX A Listed are the scientific papers and technical reports from this project. Publications prepared as deliverables or otherwise resulting directly from this project are marked with an asterisk. The others are publications to which the project contributed. Reports with an LBNL number are available at eetd.lbl.gov/publications.

Francisco PW, Brand L, Cautley D, Singer BC, Gloss S. 2016. The latest developments in residential combustion safety testing. IAQ 2016: Defining Indoor Air Quality: Policies, Standards, and Best Practices. Washington DC, Sep 12-14.

*Less BD, Singer BC, Walker IS, Mullen NA. 2015. Indoor air quality in 24 California residences designed as high performance homes. Science and Technology for the Built Environment 21(1): 14-24. LBNL-6937E.

Logue JM, Klepeis NE, Lobscheid AB, Singer BC. Pollutant exposures from unvented gas cooking burners: A simulation-based assessment for Southern California. 2014. Environmental Health Perspectives. 122(1): 43-50. LBNL-6712E.

Logue JM and Singer BC. 2014. Energy impacts of effective range hood use for all U.S. residential cooking. HVAC&R Research 20(2): 264-275. LBNL-6683E.

*Lunden MM, Delp WW, Singer BC. 2015. Capture efficiency of cooking-related fine and ultrafine particles by residential exhaust hoods. Indoor Air 25(1): 45-58. LBNL-6664E.

*Mullen NA, Li J, Singer BC. 2012. Impact of Natural Gas Appliances on Pollutant Levels in California Homes. Lawrence Berkeley National Laboratory, Berkeley, CA. December 2012. LBNL-5970E.

*Mullen NA, Li J, Singer BC. 2013a. Participant Assisted Data Collection Methods in the California Healthy Homes Indoor Air Quality Study of 2011-13. Lawrence Berkeley National Laboratory, Berkeley, CA. August 2013. LBNL-6374E.

*Mullen NA, Russell ML, Lunden MM, Singer BC. 2013. Investigation of formaldehyde and acetaldehyde sampling rate and ozone interference for passive deployment of Waters Sep-Pak XPoSure samplers. Atmospheric Environment 80: 184-189. LBNL-6386E.

*Mullen NA, Li J, Russell ML, Hotchi T, Singer BC. 2016. Results of the California Healthy Homes Indoor Air Quality Study of 2011-13: Impact of natural gas appliances on air pollutant concentrations. Indoor Air 26(2): 231-245. LBNL-185629

Rapp VH, Singer BC, Stratton JC, Wray CP. 2015. Assessment of Literature Related to Combustion Appliance Venting Systems. Lawrence Berkeley National Laboratory, Berkeley, CA. Feb 2015. LBNL-176805.

*Rapp VL, Singer BC. Effect of Fuel Wobbe Number on Pollutant Emissions from Advanced Technology Residential Water Heaters: Results of Controlled Experiments. Lawrence Berkeley National Laboratory, Berkeley, CA. May 2014. LBNL-6626E.

A-2

*Singer BC, Less BD, Delp WW, Brooks A, Cohn S, Finn B. A Field Study of Wall Furnace Venting and Coincident Exhaust Fan Usage in 16 Northern California Apartments. Lawrence Berkeley National Laboratory, Berkeley, CA. September 2016. LBNL-1006274.

*Singer BC, Delp WW, Lorenzetti DM, Maddalena RL. 2016. Pollutant concentrations and emission factors from scripted natural gas cooking burner use in nine Northern California homes. Lawrence Berkeley National Laboratory, Berkeley, CA. LBNL-1006385.

Stratton JC and Singer BC. Addressing Kitchen Contaminants for Healthy, Low-Energy Homes. Lawrence Berkeley National Laboratory, Berkeley, CA. January 2014. LBNL-6547E.

Walker IS, Sherman MH, Singer BC, Delp WW. 2016. Development of a tracer gas capture efficiency test method for residential kitchen ventilation. IAQ 2016: Defining Indoor Air Quality: Policies, Standards, and Best Practices. Washington DC, Sep 12-14.