Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Precipitation Detection and Estimation Using Multi-Spectral Remotely Sensed Data
1 Center For Hydrometeorology and Remote Sensing (CHRS), University of California, Irvine2 NASA/GSFC Code 613.1
3 NOAA/NESDIS Center for Satellite Applications and Research (STAR), Camp Springs, MD
Ali Behrangi1
Kuo-lin Hsu1
Bisher Imam1
Soroosh Sorooshian1
George Huffman2
Robert J. Kuligowski3
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
LEO (PMW): LEO (PMW): More accurate estimateMore accurate estimateEven after 3 hour accumulation still we have gapsEven after 3 hour accumulation still we have gaps
GEO (VIS/IR):GEO (VIS/IR):Less accurate estimateLess accurate estimateGlobal coverage is available frequentlyGlobal coverage is available frequently
Introduction: Problem Statement
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Introduction: Solution
1) Interpolating the precipitation intensity obtained from LEO (PMW) Satellites
2) GEO (VIS/IR) satellites provide high-resolution (time and space) images
(Joyce et al., 2004)
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Question:
Can Multi–spectral images help us to improve GEO-based precipitation estimation ?
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
MULTI-SPECTRAL imagesSpinning Enhanced Visible and Infra-red Imager (SEVIRI). 12 different wavelengths once every 15 minutes,
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
The ABI (Advanced Baseline Imager) on
Future GOES-R
Figure courtesy of ITT Industries
(Advanced Baseline Imager )
Multi- Spectral Precipitating Estimation
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Multi- Spectral Precipitating Estimation
Kurino (1997) 6.7, 11 and 12 µm
Rosenfeld and Gutman (1997) 0.65, 3.7, 10.8 and 12 µm
Inoue and Aonashi ( 2000) 0.6, 1.6, 3.8, 11 and 12 µm
Ba and Gruber (2001) 0.65, 3.9, 6.7, 11 and 12 µm
Capacci and Conway (2005)9 MODIS and corresponding SEVIRI
channels
- IR 11µm & 12 µm: => removal of thin cirrus cloud
- IR 11µm & WV 6.7 µm: => sign of deep convective
- NIR 3.7 µm : => sensitive to cloud drop size distribution - VIS : => cloud optical thickness.
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Relative-frequency distributions of different channels under rain and no-rain conditions
No RainRain
(0.65 μm) (3.9 μm) (6.7 μm)
(10.8 μm) (13.3 μm)
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Multi- Spectral Precipitating Estimation
Algorithm Development:
1- Grid-box based :
2- Cloud Patch based :
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Grid-Box Based Approach
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Grid-box Approach
Algorithm Development:
Unsupervised Classification
PCA
A : Thick-Cold cloud (i.e., Convective)B : Thin-Cold cloud (i.e., Cirrus)C : Clear Sky
Rain Probability/Intensity
Multi-spectral Images
Textural information
Clusters (MRR)
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Case Study 1: Florida : August 30 2006
Hit Under Estimation Over Estimation
Ch 1 : 0.6 µm Ch2 : 3.9 µm Ch3 : 6.5 µm Ch4:10.7µm Ch5 : 13.3µm
f) Ch3+Ch5
ETS=25POD=74FAR=45
ETS=29POD=77FAR=42
ETS=27POD=78FAR=44
ETS=36POD=76FAR=35
ETS=30POD=80FAR=42
ETS=30POD=72FAR=39
ETS=35POD=79FAR=37
ETS=37POD=78FAR=35
ETS=37POD=80FAR=36
ETS=48POD=75FAR=22
ETS=49POD=79FAR=24
g) Ch4+Ch5
i) Ch3+Ch4+Ch5
d) Ch5
f) Ch3+Ch5
VIS
IR (10.8 µm)
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Case Study 2:
Over Estimation
d) Ch5
g) Ch4+Ch5
i) Ch3+Ch4+Ch5
f) Ch3+Ch5
Ch 1 : 0.6 µm Ch2 : 3.9 µm Ch3 : 6.5 µm Ch4:10.7µm Ch5 : 13.3µm
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
SEVIRI(MSG)
VIS (0.65 µm) IR (10.8 µm)
Case Study 3: Precipitation Estimation (using SEVIRI)
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Under EstimateUnder Estimate
Over EstimateOver Estimate
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Overall Results
ETS
POD/FAR
BIAS(area)
Scenario
Rain/No-rain Detection
RMSE
BIAS(volume)
CC
Scenario
Rain Rate Estimation
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Multi-spectral & Diurnal Cycle of Precipitation
New York
Florida
Texas
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
BT 10.8 µm
Day:BT (0.65 &10.8) µm
Night:BT (6.7 & 10.8 )µm
Day & Night:BT (6.7 & 10.8 )µm
NEXRAD
Diurnal Cycle over Florida, USA (Summer 2006)
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Patch Based Approach
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
IR-Based Patching
VIS - Based Patching
Multi-spectral - Patch based
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
IR-Based Patching
VIS - Based Patching
Warm thick cloud
Cold Thin cloud
Multi-spectral - Patch based
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Results ofResults of Multi-spectral Cloud ClassificationMulti-spectral Cloud Classification experiment: experiment:
In general results are encouraging !
Detail Statistics will be provided in near future
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
ConclusionsConclusions::- Multi-spectral data are promising for precipitation retrieval,
Particularly for delineation of areal extent of precipitation.
- In addition to 10.8 μm band, VIS channel for day time and WV channel for night time seems to be good candidates.
Future Work- Developing a combined algorithm using multi-
spectral data and PMW estimate, …. is ongoing.
Center for Hydrometeorology and Remote Sensing - University of California, Irvine
Thank You !
Center for Hydrometeorology and Remote Sensing - University of California, Irvine