Satellite remote sensing of

pollution with application to the

Arctic

Chris McLinden Environment Canada

(chris.mclinden@ec.gc.ca)

27 July, 2012

CREATE summer school, Alliston, Ontario

Introduction

Taken from Bohren & Huffman, 1983

The forward problem:

Given the

dragon, what can

be inferred about

its tracks?

The inverse problem:

Given its tracks,

what can be

inferred about

the dragon?

Given the state of

the atmosphere,

what can be

inferred about the

radiation field?

[RT modeling]

Given measurements

of the radiation field,

what can be inferred

about the

atmospheric state?

[remote sensing]

Introduction

Remote sensing is the acquisition of information about an object or phenomenon, without making physical contact with the target.

More simply: measurement at a distance

This is done by collecting electromagnetic radiation and measuring a portion of the spectrum.

Thus remote sensing instruments measure spectra; geophysical properties are only inferred.

Introduction

A remote sensing instrument must:

1. Capture electromagnetic

radiation (EMR) over some well-

defined region

2. Isolate the wavelength interval of

interest

3. Measure the power captured in

this spectral interval, and convert

it to absolute radiance

rE

th

The first meteorological satellites were launched in the

1960s; the first air quality satellites ones in the 1990s

Air quality instruments are in low earth orbit, so 1-2

measurements over a given location per day

air quality = near surface atmospheric composition

The near-surface atmosphere can only

be detected using nadir, or down-

looking, viewing geometry

These instruments measure over a

volume of air, which generally includes

the entire atmospheric column. Nadir geometry

Satellite remote sensing of Air Quality

Satellite remote sensing of Air Quality

Air quality satellite instruments derive the

vertical column density, or VCD

The VCD represents the vertically

integrated number density profile and

has units of molecules per unit area

(e.g., molecules/cm2, or cm-2)

• The primary air quality data product is the

tropospheric VCD – the vertically-integrated number

density between the surface and the tropopause (~10

km)

• This may require the removal of the stratospheric

portion of the total VCD

Satellites measure

vertical column density

or number of

molecules per cm2

Satellite remote sensing of Air Quality

Strengths:

• provide large-scale coverage / an integrated view

• measures over otherwise inaccessible areas

Limitations:

• only a handful of pollutants may be detected

• “moderate” spatial resolution, 10x10 km2 at best

• provides limited (or no) information on where the pollutant is located in the atmospheric column

• only one or two measurements per day; cannot see below cloud tops

Some Air Quality satellite sensors

1996 2000 2004 2008 2012 2016

GOME

SCIAMACHY

OMI

GOME2 / MetOp-A

GOME2 / MetOp-B

TropOMI

MODIS /Terra

MODIS / Aqua

MOPITT

TES

IASI / MetOp-A

IASI / MetOp-B

Year

UV/vis spectral aerosol thermal IR

NO2, SO2

NO2, SO2

NO2, SO2

NO2, SO2

NO2, SO2

Aerosol optical depth

Aerosol optical depth

CO

CO, NH3

CO

CO

Quantity Measured

OMI = Ozone Monitoring Instrument

Ozone Monitoring Experiment (OMI)

~15 km

(2 sec.)

OMI

• Dutch/Finish instrument, launched in 2004 on

the NASA Aura satellite, still operational

• Measures sunlight reflected from Earth’s

surface and atmosphere back out into space

(nadir geometry)

• A spectrometer that measures near-UV and

visible light (280 to 600 nm)

• Horizontal resolution roughly 15 by 30 km2

(best in its class)

• Uses a 2D array detector that simultaneously

measures many wavelengths and across-track

positions

• Air quality gases: NO2 and SO2

Data Inversion

Converting raw data to VCDs (called “inversion”) is a complex process

that requires the use of atmospheric computer models that

simulate the chemical and physical processes

The models supply additional information necessary for the proper

interpretation of the satellite data

Raw

Spectra

(Level 0)

Calibrated,

geolocated

Spectra (L1)

Spectral

Fit

Removal of

stratosphere

Convert Slant

to vertical column Tropospheric

Vertical Column

Density VCD (L2) OMI Processing Sequence

OMI measures spectra – composition

obtained through a careful analysis of

the spectra accounting for all relevant

atmospheric and instrument effects

reflected

Extra-terrestrial

300 350 400 450 50010

-3

10-2

10-1

100

Wavelength [nm]

Op

tical

Dep

th

NO2

SO2

O3

Data Inversion

The high-frequency absorption structure is exploited to determine

amount of absorber in the path.

Spectral fit: a multi-linear regression is performed using laboratory

measured absorption spectra of all relevant gases

310 315 320 325 3300

0.005

0.01

0.015420 440 460

0

0.005

0.01

0.015

SO2

NO2

Strong ozone

absorption interferes

with SO2 signal

SO2 window NO2 window

NO2 over the GTA

0

100

400

900

1600

2500

3600

4900

6400

10

20

30

40

50

60 NO

2 Tro

po

sp

he

ric V

CD

[10

14 c

m-2]

Popu

latio

n D

ensity

OMI 2005-2007 summertime average

The Nanticoke Generating Station is the largest coal-fired power plant in North

America delivering 4000 MW at peak capacity.

Ontario attempting to phase out coal burning by 2014; four of its units have been

retired.

Nanticoke Generating Station

Nanticoke power plant

2005 2006 2007 2008 2009 2010 20110

2

4

6

x 1015

Year

VC

D [

cm

-2]

or

em

issio

ns

VCD

Annual NOx emissions (scaled)

Power Generated (scaled)

Reported to

Gov’t database

Weekends Weekdays

40% increase

NO2 over the GTA

10 20 30 40 50 60

NO2 Tropospheric VCD [1014 cm-2]

2005-2011, summer, all wind directions

NO2 over the GTA

10 20 30 40 50 60

NO2 Tropospheric VCD [1014 cm-2]

Windspeed and direction from ECMWF reanalysis tied to OMI observations

2005-2011, summer, Southerly winds

Suwanose-jima Kikai

Aso Sakura-jima Miyake-jima

Global SO2 emission source catalogue (~200 sources)

Example: Volcanoes in Japan

Asama

SO2 Pollution Controls Bring Results December 2, 2011

See NASA Earth Observatory, http://earthobservatory.nasa.gov/IOTD/view.php?id=76571#

Scientists using the Ozone Monitoring Instrument (OMI) on NASA’s Aura satellite observed

major reductions in sulfur dioxide (SO2) between 2005 and 2010 in Alabama, Georgia, Indiana,

Kentucky, North Carolina, Ohio, Pennsylvania, and West Virginia. Led by Vitali Fioletov of

Environment Canada, the research team found that sulfur dioxide levels near the region’s coal-

fired power plants fell by nearly half since 2005.

Fioletov et al., GRL., 2011

The largest SO2 source in the Arctic: Norilsk, Russia, 70N.

0.0-0.3 0.3 0.6 DU

1% of Russia’s GDP

2% of Russia’s industrial production

3% of Russia’s export

… and 2,400 kT of SO2 per year

(Canada <2,000 kT/yr)

Norilsk

-1.0 0.0 1.0 2.0 3.0 DU

70N

The largest SO2 source in the Arctic: Norilsk, Russia, 70N.

Copper, nickel smelting

Application to oil sands monitoring

• “Oil sands”, or “tar sands”, refer to a type of petroleum

deposit in which the oil is very thick and sticky (called

“bitumen”) and mixed with sand, water, and clay

• Only in recent years has it been profitable to extract and

refine oil from these deposits

• Canada has a proven reserve of ~170 billion barrels

surface mining region

Province

of Alberta

• Bitumen found close to the

surface may be mined;

deeper deposits need to be

heated and then pumped to

surface from Energy Resources Conservation Board, 2011

Surface mining & upgrading processes emit NOx and SO2

into the atmosphere

OMI well suited to study these pollutants

data products - NO2: Dutch TEMIS version 2

SO2: NASA OMSO2 V003

Application to oil sands monitoring

Mining & Transport Extraction Separation Primary Upgrading Secondary Upgrading

Steps in Surface Mining

NOX

SO2

from The Oil Sands Process, CNRL

a

Vertical Column Density (x1015 cm-2)

0 1 2 3 4 5 6

Alberta

Toronto

Edmonton Oil Sands

Surface Mining

Region

OMI NO2 2005-2010

tropospheric VCD [0.25 0.25 grid]

56.5

57

57.5

0 0.1 0.2 0.3 0.4

Surface Mining Area

• NO2 and SO2 both show area of enhancement over surface mining;

some differences in distribution evident

• NO2 also shows secondary maximum further to the north

• Primary source of SO2 is thought to be upgrading, and the only on-

site upgraders are at the location indicated

Fort McMurray Fort McMurrayFort McMurray

LandSat OMI NO2 (2005-2010) OMI SO2 (2005-2010)

90 km

Surface Mining Operations

with on-site Upgraders

56.5

57

57.5

0 1 2 3

1015 molecules/cm2 Dobson Units

Evolution

12

A

BC

D

E

b 2005-2007

12

A

BC

D

E

c 2008-2010

Vertic

al C

olu

mn

Den

sity

(x10

15 c

m-2)

0

0.5

1

1.5

2

2.5

2003-2006 2007-2010

12

b 2005-2007

12

c 2008-2010

Vertic

al C

olu

mn

Den

sity

(DU

)

-0.1

0

0.1

0.2

0.3

0.4

OMI NO2

OMI SO2

SCIAMACHY NO2

- SO2 in 2008-2010

appears to be larger, but

area of enhancement

slightly smaller

- uncertainties too large

to conclude an increase

- NO2 in 2008-2010

clearly larger, and also

area of enhancement

also appears larger

- SCIAMACHY data is

consistent with OMI

A to E = location of in-situ NO2 measurements

Evolution of NO2 over Oil Sands

Examine NO2 from a seasonal perspective – less spatial

information

Use fit of 2D Gaussian to characterize seasonal NO2 VCD

data (DJF, MMA, JJA, SON); calculate trends

Total mass [t(NO2)]

of enhancement

Maximum VCD

Widths of distribution [km]

“Background” VCD

WBEA in-situ NO2 (average over sites A-D)

Production [millions of barrels per day]

Max VCD

Background VCD

Widths

Air Mass Factors

VCDtrop = (SCD – VCDstrat AMFstrat) / AMFtrop

Air mass factor (AMF) describe the sensitivity of the satellite sensor to

absorbing layer. They are computed using a multiple-scattering radiative transfer model and their accuracy relies in large part on the validity of input parameters, including:

1. Shape of the absorbing profile

2. Surface reflectivity or albedo

Raw

Spectra

(Level 0)

Calibrated,

geolocated

Spectra (L1)

Spectral

Fit

Removal of

stratosphere

Convert Slant

to vertical column Tropospheric

Vertical Column

Density (Level 2) UV/vis Processing Sequence

measured modelled

Surface Albedo

Landsat 1993

Landsat 2005

Landsat 2010

Complications: surface albedo

• AMFs are sensitive to the reflectivity of the underlying surface

– measured light that is reflected from the surface will have

passed through the entire atmosphere twice

No light from surface

Some light from surface

Bright

Dark

Currently, a surface reflectivity “climatology” is used

and so does not take into account changes in land

use/cover.

AMF sensitivity studies suggest this would impact the

calculated trend in NO2 by 1%/year.

0.025

0.03

0.035

0.04

Refl

ecti

vit

y [

-]

a

0.1

0.15

0.2

AO

D a

t 550 n

m [

-] b

0.8

1

1.2

AM

F [

-]

c

2005 2006 2007 2008 2009 2010 20110.8

1

1.2

VC

D c

orr

ecti

on

, a

m [

-]

d -1.9 0.3%/yrd -1.9 0.3%/yrd -1.9 0.3%/yr

Year

d -1.9 0.3%/yrd -1.9 0.3%/yr

MODIS OMI (471 nm) OMI (442 nm)

0.025

0.03

0.035

0.04

Refl

ecti

vit

y [

-]

a

0.1

0.15

0.2

AO

D a

t 550 n

m [

-] b

0.8

1

1.2

AM

F [

-]

c

2005 2006 2007 2008 2009 2010 20110.8

1

1.2

VC

D c

orr

ecti

on

, a

m [

-]

d -1.9 0.3%/yrd -1.9 0.3%/yrd -1.9 0.3%/yr

Year

d -1.9 0.3%/yrd -1.9 0.3%/yr

MODIS OMI (471 nm) OMI (442 nm)

Oil Sands

NO2

SO2

Oil production:

1,600,000 bpd

Reported SO2 emissions:

about 115 kT/y

OMI-estimated SO2 emissions:

about 85 kT/y

Context

The enhancements in NO2 and SO2 are comparable to what OMI observes over a “large” coal-burning power plant

SO2 emissions from the Oil Sands are about 100 kT/year. There are many other (>50) industrial sources with the same or larger level of emissions in the world. The largest industrial source produces >2000 kT/year.

It is also useful to contrast these results with other oil-industry sources

Ufa, Russia (oil refineries, power plants, etc.)

(same latitude as oil sands, ~same obs. conditions)

Three oil refineries located in Ufa with a

combined capacity of >1,000,000 BPD

Ufa Population: ~1,000,000

Oil Sands

NO2

SO2

Oil production:

1,600,000 bpd

Reported SO2 emissions:

about 115 kT/y

OMI-estimated SO2 emissions:

about 85 kT/y

OMI-estimated SO2 emissions:

about 100 kT/y

NO2

SO2

Cantarell and Ku-Maloob-Zaap Oil Fields, Mexico

(Large North American source, growing rapidly) Oil Sands

NO2

SO2

Oil production:

1,600,000 bpd

Reported SO2 emissions:

about 115 kT/y

OMI estimated SO2 emissions:

about 85 kT/y

NO2

SO2

SO2

2005-2007

2007-2011

Oil production:

800,000+500,000 BPD

OMI estimated SO2 emissions:

about 200 kT/y in 2005-2007

about 330 kT/y in 2008-2011

2005-2011

Oil refineries in Aruba and Venezuela

(near vacation site; SO2 source comparable to oil sands)

The Aruba refinery processes lower-cost heavy sour crude

oil and produces a high yield of finished distillate products.

Total capacity of 235,000 bpd

Paraguaná Refinery Complex, Venezuela, one of the world

largest refinery complexes (940,000 bpd)

Oil Sands

NO2 NO2

SO2 SO2

Oil production:

1,600,000 bpd

Reported SO2 emissions:

about 115 kT/y

OMI estimated SO2 emissions:

about 85 kT/y

Future of Space-based

Monitoring TROPOMI (2015, Europe): OMI-like but 6+ times better

spatial resolution, better sensitivity, 10+ times more data points

OMI (15 x 30 km2) TropOMI (7 x 7 km2)

Future of Space-based

Monitoring • GEO-CAPE (2018+, USA):

Geostationary platform, should observe

up to 60N, target resolution 4 x 4 km2;

hourly repeat

PCW concept

• PCW (2018+, Canada, Polar Communications

and Weather): A pair of satellites in highly-

ellipitical orbits that together provide near-

geostationary coverage of Arctic/sub-Arctic;

target 8 x 8 km2 resolution, hourly repeat

References

Texts:

Remote Sensing of the Lower Atmosphere: An Introduction, G. L. Stephens,

Oxford University Press, 1994.

The Remote Sensing of Tropospheric Composition from Space, John P.

Burrows, Ulrich Platt, Peter Borrell (editors), Spring, 2011. *

Papers:

Martin, R. V., Satellite remote sensing of surface air quality, Atmospheric

Environment, 42, 7823–7843, 2008. *

McLinden, C. A., V. Fioletov, K. F. Boersma, N. Krotkov, C. E. Sioris, J. P.

Veefkind, and K. Yang, Air quality over the Canadian oil sands: A first

assessment using satellite observations, Geophys. Res. Lett., 39, L04804,

doi:10.1029/2011GL050273, 2012. *

* pdf available from ftp://exp-studies.tor.ec.gc.ca/pub/ftpcm/CREATE/

Data Inversion

• Stratosphere removed using simulations from a global chemical-

transport model

• There are many paths that involve reflection and/or one or more

scattering events; to interpret the measurements computer models

are used that simulate multiple-scattering and absorption

• Computer models are also used to provide an estimate of the profile

shape

0 0.5 1 1.5 2 2.5 3 3.50

5

10

15

20

25

30

35

40

Alt

itu

de [

km

]

Air Mass Factor

NO2 (440 nm)

SO2 (313 nm)

Lower probability of reaching surface

Higher probability of reaching surface 0 0.1 0.2 0.30

1

2

3

VMR [ppb]

z [

km

]

Profile shape

(from model)

sensitivity

visible

UV

Pixel-averaging method to better resolve

features in satellite data:

* need to use a large amount of data

The value assigned to a grid-box is the

average of all data within radius r

Mapping

LandSat 2009

25 km

320 km

x

y

r

r

e.g.: x=y=1 km, r=8 km

Surface Mining Area

Approximate size

of OMI footprint