Reza Khanbilvardi (CREST Director)

Reza Khanbilvardi (CREST Director)

NOAA Eastern Region Flash Flood Conference, Wilkes-Barre/PA, June 3, 2010

Flash Flood & CREST(NOAA-Cooperative Remote Sensing Science and Technology Center) (NOAA-Cooperative Remote Sensing Science and Technology Center)

(Hydro-Climate & Land Hydrology)

Hydrologic Models

Operation

Flash Flood Forecasting

In Situ Data

Hydrograph

Stream FlowSnowmelt

Reservoir Release

Watershed & Stream Characterization:

Topography, roughnessSoil Moisture

Vegetation density, Infiltration, Interception

INPUT Information

Observation or Estimation of

Hydrologic Variable:

Precipitation(Rainfall & Snowfall)

from: Rain Gauge, Radar, Satellite

Calibration Data

Products

Surface Runoff

CREST Related Actvities/Projects:

Precipitation Enhancement:

To Improve Flash Flood Forecasting:

• Remotely Sensed Precipitation Estimation & Improvement• Precipitation Prediction using Satellite-based Information

Soil Moisture Enhancement

• CREST L-band and High Frequency Microwave Radiometer• using Passive & Active Microwave for Soil Moisture Estimation• Soil Moisture & Hydrological Modeling• Application of SMAP for Flash Flood Forecasting Test-bed

Snow Characteristics & Flash Flood Guidance (FFG) Syatem

Sea Ice Monitoring

Improve Coastal and Estuarine Flood Monitoring

Development of a sub-watershed hydrologic model in Puerto Rico

global retrieval of microwave land surface emissivity Improvement

Precipitation Estimation & NowcastPrecipitation Estimation & Nowcast

ImprovementImprovement

• MW based Snowfall Detection & Estimation,

• Multi Sources Rainfall Estimation to Generate Rainfall for Radar Gap Areas,

• Validation and Improvement of Satellite based Rainfall Products

• Satellite-based Thunderstorm Nowcasting,

Methodology: An Artificial Neural Networks System (ANN)

• Data Selection - 75% of data as Input &

25% of data for Testing

Data Used: AMSU-B channels: 89-, 150-, 183±1-, 183±3-, 183±7 – GHz & Ground-based snowfall

MW-based Snowfall Detection & Estimation

• Model Features: - Channel Combinations,

- Number of Nodes

- Transfer Functions

- Number of Runs

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

10

14

15

20

23

27

28

30

31

38

40

42

48

50

54

68

71

72

73

74

98

103

104

108

112

113

114

117

121

173

174

175

176

177

183

187

188

191

193

216

R (

av

era

ge

)

Calibration Validation

25 cal 25 val

# of RUNS with P-value < 0.5

150, 183 ± 1, and 183 ± 7 GHz

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0 5 10 15 20 25 30 35

Nodes in HL

R c

orre

cte

d

150, 183 ± 1, 183 ± 7 150, 183 ± 1, 183 ± 7

Cali Vali

# of Nodes in Hidden Layer

- Data Filtering

Precipitation > 0.0 inches

Step # 2

Only snow? Tsurf<273°K

Precipitation type?

Surface type = 1 (land)

Step # 2 Station data

AMSU data

discard data

discard data

no no

yes yes

discard data

discard data

yes

no no

yes

discard data

no

To model

non-precipitating pixels

Snowy pixelsModel Estimate (Snowfall Rate) Model

Data Filtering

BR - 25-25 - LZA - 13

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

89 G

Hz

150

GH

z

183

± 1

GH

z

183

± 3

GH

z

183

± 7

GH

z

89 a

nd 1

50 G

Hz

89 a

nd 1

83 ±

1 G

Hz

89 a

nd 1

83 ±

3 G

Hz

89 a

nd 1

83 ±

7 G

Hz

150

and

183

± 1

GH

z

150

and

183

± 3

GH

z

150

and

183

± 7

GH

z

183

± 1

and

183

± 3

GH

z

183

± 1

and

183

± 7

GH

z

183

± 3

and

183

± 7

GH

z

89,

150

, an

d 18

3 ±

1 G

Hz

89,

150

, an

d 18

3 ±

3 G

Hz

89,

150

, an

d 18

3 ±

7 G

Hz

89,

183

± 1,

and

183

± 3

GH

z

89,

183

± 1,

and

183

± 7

GH

z

89,

183

± 3,

and

183

± 7

GH

z

150

, 18

3 ±

1, a

nd 1

83 ±

3 G

Hz

150

, 18

3 ±

1, a

nd 1

83 ±

7 G

Hz

150

, 18

3 ±

3, a

nd 1

83 ±

7 G

Hz

183

± 1,

183

± 3

and

183

± 7

GH

z

89,

150

, 18

3 ±

1, a

nd 1

83 ±

3 G

Hz

89,

150

, 18

3 ±

1, a

nd 1

83 ±

7 G

Hz

89,

150

, 18

3 ±

3 an

d 18

3 ±

7 G

Hz

89,

183

± 1,

183

± 3

and

183

± 7

GH

z

150

, 18

3 ±

1, 1

83 ±

3 a

nd 1

83 ±

7 G

Hz

89,

150

, 18

3 ±

1, 1

83 ±

3 a

nd 1

83 ±

7 G

Hz

Channel Combination

R (a

vera

ge)

Calibration

Validation

Channel Combination

0 0.5 1 1.5 2 2.5 30

0.5

1

1.5

2

2.5

3

Snowfall Obs. (mm)

Sn

ow

fall

Estim

ate

s (m

m)

Snowfall Observation (mm) S

now

fall

Est

imat

es (

mm

)

0 0.5 1 1.5 2 2.5 3 3.5 40

0.5

1

1.5

2

2.5

3

3.5

4

Snowfall Obs. (mm)

Sn

ow

fall

Estim

ate

s (

mm

)

Sno

wfa

ll E

stim

ates

(m

m)

150-, 183±3- and 183±7 - GHz

R=0.869

RMSE=0.22

Model Performance

Validation

Calibration

R=0.600

RMSE=0.34

Multi-Sensors Precipitation EstimationObjective: To Develop a Multi-Sensor Rainfall Retrieval Algorithm to Generate more accurate Rainfall for Radar Gap Areas. In this project, NESDIS Hydro-Estimator & NEXRAD Stage-IV at Hourly 4km x 4km

Procedures (3 steps):

1) Spatial Error Correction: Apply Method of Least Squares (Brogan 1985): the method of Hills Climb to cluster Rainy pixels because the corresponding clusters are to pick up.

2) Bias CorrectionBias ratio field using Inverse Distance method provides a more radar like output both spatially and intensity wise.

3) Merge Rain gauge-Radar-Satellite Rainfall to Generate Rainfall for Radar Gap areas

Pixel by pixel based Successive Correction Method (SCM) has been tried for a selected gap area in the radar rainfall.

Validation & Improvement of Rainfall Products Validation of satellite-based rainfall retrieval algorithms for hurricane storms

Satellite–based Hydro-estimator, GMSRA from NESDIS, PERSIANN and TRMM-3B42 RT rainfall retrieval algorithms have been evaluated at hourly and daily basis for Five very strong hurricanes: Charley, Frances, and Jeanne from 2004 and Wilma and Rita from 2005 against NEXRAD Stage-IV.

Hydro-Estimator GMSRA PERSIANN TRMM 3B42 Stage IV RadarHurricane Frances (5 September 2004, 0-23 UTC)

A new algorithm is been developed to detect rainy cloud pixels using visible and infrared GOSE data. This algorithm improves the performance of the Hydro-Estimator.

The plots show the preliminary comparison between the Hydro-Estimator and new algorithm.

Minimizing the index number implies minimizing the average false alarm rate (FAR), maximizing the average of probability of detection (POD), and maximizing the average hit rate (HR)

Validation & Improvement of NESDIS Hydro-Estimator

Satellite Thunderstorm Nowcasting(Transitioning GOES-based Nowcasting Capability into the GOES-R Era)Joint Project: CREST-CUNY, NWS-MDL, NESDIS, OAR-NSSL, & CIMMS

RDT (Rapid Developing Thunderstorms) Model, developed by Météo-France in the framework of EUMETSAT-SAF Nowcating.

- Single IR channel statistics used throughout cell lifecycle to evaluate convective activity,

• Cells are detected by whether they form towers higher than

a given BT threshold (6 degrees). Cells are tracked, and all

properties, such as: contours, areas, growth rates, BT min,

BT avg, and BT gradients around the periphery are stored.

• Lookup tables of cell lifecycles are used to determine if the

cell may be convective. Growth rates and roundness of the

top are important parameters.

t1 t2 t3 t4

Cloud Tracking => Storm Detection

Borrow ideas from existing algorithms that do each step best.

=> Extrapolation

Steps in Thunderstorm Nowcasting

Applying Extrapolation

Red: previousYellow: currentGreen: extrapolated

Investigation is needed to stabilize extrapolation

Extrapolation is based on RDT cloud lifecycles study.

The RDT model in New YorkData from direct broadcast, 15 min refresh rate

http://air.ccny.cuny.edu/

Land Group: Soil - Snow - Vegetation

Projects: • Snow cover and snow water equivalent (SWE) retrieval from active and passive microwave sensors.

• Development of merging algorithms that combines microwave and thermal infrared observations for soil moisture observations.

• Improve soil moisture and snow retrieval algorithm reducing vegetation effect using NOAA-CREST Microwave Radiometers.

• To improve flash flood forecasting system using satellite based gridded soil moisture data (SMAP Testbbed soil moisture data).

• Sea ice monitoring using geostationary satellite data

• Integration of the produced soil moisture and snow cover characteristics maps into hydrological models.

Vegetation

Snow CoverSoil Moisture

• Active Microwave

• Passive Microwave

• Optical Sensors

Soil Moisture Estimation & ImprovementSoil Moisture Estimation & Improvement

• CREST L-Band & High Frequency Microwave Radiometer

• Soil Moisture Estimation using Passive and Active Microwave & Optical Sensors,

• Soil Moisture Retrieval & Hydrological Modeling,

• Application of SMAP Test-bed for Flash Flood Forecasting

NOAA-CREST L-band Microwave Radiometer

L-Band Radiometer Frequency: 1.40 to 1.55 GHz (SMAP mission Frequency) Dual polarization (H, V) Antenna system: 1.5 x 0.7 meters Manufacturer: Radiometrics Corporation, Boulder CO.

Research Objectives:• L-band soil moisture remote sensing field experiments for

calibration and validation of soil moisture retrieval algorithms.

• Temporal analysis of brightness temperature variation with respect to measured soil moisture and vegetation characteristics (NDVI and NDWI) to develop (or strengthen) vegetation component of Radiative Transfer Model.

• Study the land emissivity variation under a controlled environment (roughness and vegetation).

• Investigate the impact of inter rainfall time interval on the retrieval of soil moisture particularly over vegetated areas.

Sensitivity Analysis of b-factor in Microwave Emission Model for Soil Moisture Retrieval

Calibration and validation of radiative transfer model for soil moisture retrieval at low frequency (1.4 GHz) using better vegetation component.

Neural network and fuzzy logic modeling for soil moisture retrieval. Evaluate the impact of land cover heterogeneity on soil moisture

retrieval. Evaluate the vegetation impact on soil moisture retrieval for

different land cover type.

Soil Moisture Research

Lakhankar et al (2009 b)

SAR Image

350 km x 300 km(Res. 25 m)

Study Area (A and B)

A: 26.4 km x 96 kmB: 31.2 km x 103.2 km

A

B

9900’W 9800’W 9700’W 9600’W 9500’W 9400’W

3800’ N

Soil Moisture Data

165 km x 495 km(Res. 800 m)

3700’ N

3600’ N

3500’ N

Truth SM Simulated SMLakhankar et al (2009 a)

Neural network and fuzzy logic modeling for soil moisture retrieval

Impact of land cover heterogeneity on soil moisture retrieval

Seo et al (2010)

Analysis of an Adaptive NRCS Curve Number

Intraseasonal variation of the CN over a selected watershed in NJ. MOPEX data has been used. discharge and precipitation observations since 1927

Qualitative vs quantitative estimate of soil moisture

Soil moisture experiment (SMEX) campaigns

0 0.2 0.4 0.6 0.8 10

0.05

0.1

0.15

0.2

0.25

ASMR-E 6.9 GHz Soil Wetness Index (SWI)Vol

umet

ric S

oil M

oist

ure

(VS

M),(

m3 /m

3 )

Arizona(SMEX04)

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

ASMR-E 6.9 GHz Soil Wetness Index (SWI)

Alabama(SMEX03)

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5


Georgia(SMEX03)

0 0.2 0.4 0.6 0.8 10

0.05

0.1

0.15

0.2

0.25

0.3

0.35


umet

ric S

oil M

oist

ure

(VS

M),(

m3 /m

3 )

North Oklahoma(SMEX03)

0 0.2 0.4 0.6 0.8 10

0.05

0.1

0.15

0.2

0.25

0.3

0.35


South Oklahoma(SMEX03)

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5


Iowa(SMEX02)

SM = 0.53*SWIR = 0.565, RMS = 0.09

SM = 0.486*SWIR = 0.728, RMSE = 0.12

SM = 0.572*SWI+0.014R = 0.453, RMSE = 0.10

SM = 0.318*SWI+0.005R = 0.753, RMSE = 0.06

SM = 0.486*SWIR = 0.462, RMS = 0.07

SM = 0.599*SWI+0.005R = 0.161, RMS = 0.06 0 0.2 0.4 0.6 0.8 1

0

0.05

0.1

0.15

0.2

0.25


umet

ric S

oil M

oist

ure

(VS

M),(

m3 /m

3 )

Arizona(SMEX04)

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5


Alabama(SMEX03)

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5


Georgia(SMEX03)

0 0.2 0.4 0.6 0.8 10

0.05

0.1

0.15

0.2

0.25

0.3

0.35


umet

ric S

oil M

oist

ure

(VS

M),(

m3 /m

3 )

North Oklahoma(SMEX03)

0 0.2 0.4 0.6 0.8 10

0.05

0.1

0.15

0.2

0.25

0.3

0.35


South Oklahoma(SMEX03)

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5


Iowa(SMEX02)

SM = 0.318*SWI+0.005R = 0.821, RMS = 0.05

SM = 0.572*SWI+0.014R = 0.561, RMS = 0.09

SM = 0.486*SWIR = 0.761, RMS = 0.12

SM = 0.53*SWIR = 0.546, RMS = 0.14

SM = 0.599*SWI+0.005R = 0.417, RMS = 0.05

SM = 0.321*SWI+0.006R = 0.488, RMS = 0.07

Regression analysis between soil wetness indexes [TB (H)

6.9 GHz] using the end members derived at local and In-situ soil moisture observed

Regression analysis between soil wetness indexes [TB (H) 10.7 GHz]

using the end members derived at local and In-situ soil moisture observed

LULC influences CN and is closely related to its changeable behavior. SM affects the CN values and also contribute to its variation due to the amount of water infiltrated. LULC and SM, therefore, are key factors for understanding CN ‘s behavior.

Soil moisture retrieval and hydrological modeling

Soil moisture accounting model

3. SMAP testbed Soil moisture HL-RDHM

Flash Flood Forecasting

River Flow or Stage, Gauge data

1. Sacramento based Soil moisture

Precipitation

Evapo-transpiration

Snow Model (snow17)

SMAP Test-bed data for Flash Flood Forecasting

Runoff

• To improve flash flood forecasting system using satellite based gridded soil moisture data (SMAP Testbbed soil moisture data).

• Hydrology Laboratory-Research Distributed Hydrologic Model (HL-RDHM) developed by NOAA-NWS is currently used for Flash Flood Forecasting.

2. Continuous API based Soil moisture

Comparison and Evaluation

Flow

Snow melt

Flash Flood Forecasting EnhancementFlash Flood Forecasting Enhancement

• Snow Characteristics Flash Flood Guidance (FFG) System,

• Towards a better global retrieval of microwave land surface emissivity ,

• Sea ice monitoring over the Caspian Sea using geostationary satellite data

• Improve Coastal and Estuarine Flood Monitoring

• Development of a sub-watershed hydrologic model within the western PR study basin

Snow Characteristics /Flash Flood Guidance (FFG) System

CREST High Frequency Microwave Radiometers

C-Band Radiometer Frequency: 37 and 89 GHz Dual polarization (H, V) Manufacturer: Radiometrics Corporation, Boulder CO.

Research Objectives:

• Use high frequency microwave radiometer field experiments for calibration and validation of Snow cover and SWE retrieval algorithms.

• Temporal analysis of brightness temperature variation with respect to snow depth, SWE, and Snow Grain Size.

Comparing 19 GHz in a) and 37 GHz in b) V and H polarization emissivities

Towards a better global retrieval of microwave land surface emissivity

Where

a)a)

b)b)

Atmospheric correction according to Liebe’s model

Upwelling (a) and downwelling (b) brightness temperature as an atmospheric contribution to the satellite observation on July 14th 2003 at 37 GHz

a)a) b)b)

Products

Model

SEVIRI-based sea ice map over the northern part of the Caspian Sea on 28 February 2007 at 11h15 AM UTC (right) and the MODIS true-colour image for the same day (left)

MSG SEVIRI full disk false color composited image and the portion of the image over Caspian Sea reprojected to latitude-longitude grid on 23 January 2007 at 10:15 AM UTC.

Instantaneous ice maps (left column) and original MSG SEVIRI images on 23 January 2007. False color images in the right column are constructed with Ch.3 reflectance (red), HRV reflectance (green) and inverted infrared

brightness temperature (blue)

The obtained correlation coefficients with IMS charts for 2007 and 2008 were 0.92 and 0.83 respectively. The technique has been proposed as one of candidate ice mapping techniques for the future GOES-R ABI instrument.

The average percentage of cloud reduction because of the daily compositing ranged from 22% to 25%. Daily maps of ice distribution and concentration with minimal cloud coverage were produced.

TEMIMI, M., ROMANOV, P., GHEDIRA, H., KHANBILVARDI, R. & SMITH, K. (2009) Sea ice monitoring over the Caspian Sea using geostationary satellite data. International Journal of Remote Sensing, Accepted.

Sea ice monitoring over the Caspian Sea using geostationary satellite data

Improve Coastal and Estuarine Flood Monitoring(Establishing the Application of High Resolution Satellite Imagery)

Hurricane Charley 2004 track and hurricane eye location on the 08/14/2004

8/15/2004 (right after hurricane

Charley 2004)

11/14/2005 (low tide conditions)

Use of Radarsat 1 images

Example of inland Example of inland flooded area in redflooded area in red

Additional Additional flooded area (in flooded area (in red) can be seen red) can be seen inland and along inland and along the coastthe coast

Development of a sub-watershed hydrologic model within the western PR study basin

Example of Daily Reference Evapotranspiration (mm) in Puerto Rico, May 27, 2009

A GOES product has been developed for PR and Hispañola to estimate ground-based solar radiation and evapotranspiration. An algorithm is being developed to perform a pixel-by-pixel daily water balance. The algorithm will provide soil moisture content which is an initial condition for the flood Nowcast model.

*

Example of Daily Reference

Evapotranspiration (mm) in Haiti and

the Dominican Republic, March 10,

2010

Specific topics being investigated:

•A Flood Forecast Alarm System for Western Puerto Rico•Evaluation of Upscaling Parameters and their Influence on Hydrologic Predictability in Upland Tropical Areas•Calibration and Validation of High Resolution Radar Rainfall Estimation

Remote Sensing of Evapotranspiration in the Caribbean Region (UPRM E. Harmsen)

Testbed Subwatershed Basin-Scale Model

Upscaling Procedure

CCNY/CUNYCCNY/CUNYCCNY?CUNYCCNY/CUNYUPRMUPRMUPRM

NOAA-CREST Scientists: Reza KhanbilvardiShayesteh E. MahaniArnold GruberBrian Vant HullEric HarmsenNazario D. RamirezRamon Vasquez

CREST Hydro-Climate Participants:

NOAA Collaborators: Ralph FerraroBob KuligowskiPedro RestrepoMamoudou BaRobert RabinCezar Kongoli Stephan Smith David Kitzmiller Daniel Lindsey John R. Mecikalski

NESDISNESDISNWSMDLOARNESDISMDL NWSCIMMSCIMMS

NOAA-CREST Scientists: Reza KhanbilvardiMarouane TemimiAlvaro GonzalesPradipat SumakalNaira ChaouchLenny RoytmanAtiq RahmanTarendra LakhankarAmir Azar

CCNY- CUNYCCNY- CUNYCCNY-CUNYCCNY-CUNYCCNY-CUNYCCNY-CUNYCCNY-CUNYCCNY-CUNYCCNY-CUNY

CREST Land Participants:

NOAA Collaborators: Peter Romanov

Fuzhong Weng

Sid Boukabara

Jerry Zhan

Felix Kogan

Mitch Goldberg

NESDIS

NESDIS

NESDIS

STAR

STAR

NESDIS

Question?

Documents

Reza Khanbilvardi (CREST Director)