cuny_finaln

1
Improving Satellite-based Precipitation Estimates
using Multi-Sources Data
Shayesteh Mahani
City College (CCNY) at the City University of New
York (CUNY),
Sixth Annual NOAA-CREST Symposium, Mayagüez, PR
Feb. 20-22, 2008
National Oceanic and Atmospheric Administration
Cooperative Remote Sensing Science and Technology
Center
2
Generating Accurate Rainfall over the Radar Gap
Areas
Radar Coverage over the U.S. (Radar Measured
Frequency of Rainfall, for 1998 2000)
3
Introduction
Merging Satellite-based Precipitation Estimates
(SPE) from with ground-based radar and rain gauge
measurements for reducing uncertainties of SPE
estimates over radar gap regions.

• Following two steps need to be performed for
  generating more accurate
• rainfall products for radar gap areas
• 1) Merging SPE rainfall estimates with NEXRAD
  Stage-IV measurements of the neighboring pixels
  of radar gap areas
• 2) Merging combined SPE and NEXRAD Stage-IV
  rainfall estimates with rain gauge observations.

4
1) Merging Satellite-based Rainfall
Estimates with Radar-based Rainfall Measurements

• Bias Correction of Satellite-based Rainfall
  Estimates
• Merging using the Successive Correction Method
  (SCM)
• Results

5
Merging SPE with Radar Rainfall

• Data Used
• NEXRAD Stage IV, calibrated radar-based rainfall
  measurements with rain gauge observations at
  hourly 4 km x 4 km resolutions.
• High resolution satellite-based precipitation
  estimates (SPE) , NESDIS operational
  Hydro-Estimator (HE).

• Pixel-based IR-Tb from GOES-Ch4 with 10.7 µm
  wavelength,
• Moisture availability (precipitable water),
• Relative humidity to reduce precipitation in arid
  regions,
• Corrections are made for subcloud RH, orography,
  and convective EL (warm clouds),
• Resolution 4 - 5 km resolution generally every
  15-30 minutes, spatial
• and temporal resolution of the available
  geostationary satellite data.

6
Merging HE with Radar Rainfall
Study Sites Different size areas 0.8 x 0.8,
1.5 x 1.5, and 2.0 x 2.0, selected as radar
gap sites, over an area with available NEXRAD
data.
Methodology 1. Bias Correction of HE, Bias
corrected HE is the original HE multiply by
average ratios of mean and maximum of
NEXRAD by the mean and maximum of HE estimates.
Original HE HE after Bias Correction
Radar Rainfall
7
Successive Correction Method (SCM)
2. Merging SPE with NEXRAD
where i the observation pixel p the center
pixel, wi weight factor for pixel (i) ri
pixel (i) distance from center R Maximum
distance in a window Er(p) Error for pixel (p)
RR(i) NEXRAD at pixel (i) HE HE rainfall
estimate PRM Merged rainfall at pixel, p
Hydro-Estimator (HE)
i
p
i
ri
p
wi (ri) wi (R2 ri2) / (R2 ri2)
Radar Rainfall (RR)
Er(p) ?wi (RR(i) HE(i)) / (? wi)
PRM(p) HE(p) Er(p)
8
ResultsMerged Radar- Satellite- based
Rainfall over Radar Gap Areas
Satellite Rainfall NEXRAD
Generated Rainfall
August 15, 2004, hour 1600 UTC, for a 0.8?
x 0.8? gap area
Satellite vs. NEXRAD Merged vs. NEXRAD
9
Results (Cont.)
Satellite Rainfall NEXRAD
Generated Rainfall
July 14, 2004, hour 0300 UTC, for a 1.5? x
1.5? gap area
Satellite vs. NEXRAD Merged vs. NEXRAD
10
Results (Cont.)
Satellite Rainfall NEXRAD
Generated Rainfall
July 14, 2004, hour 0300 UTC, for a 2? x 2?
gap area
Satellite vs. NEXRAD Merged vs. NEXRAD
11
Results (Cont.)
Correlation Coefficient RMSE
July and August 2004, for a 1.5? x 1.5? gap area
July and August 2004, for a 0.8? x 0.8? gap area
12
Results (Cont.)
Correlation Coefficient RMSE
July and August 2003, for a 2? x 2? gap area
July and August 2004, for a 2? x 2? gap area
13
2) Adjustment of SPE NEXRAD using Rain Gauge
Measurements

14
GPCC Rain-Gauge Data
Global GPCC Gauge Observation
Monthly, 1º x 1º gridded area-mean
precipitation from land surface rain gauge
observations,
15
Merging-Adjustment Algorithm (SCM)
Rain Gauge
where wi weight factor related to
pixel (i) ni of gauges in pixel i, di
distance between pixel (i) and center
pixel (p), Er(p) Error for pixel p G(i)
Rain-gauge observations SR SPE RR MRA
Adjusted SPERR Merged SPE RR G
i
p
SR SPE RR
wi (ni , di)
Er(p) ?(wi ? (G(i) SR(i))) / (? wi)
MRA(p) SR(p) Er(p)
16
Merging-Adjustment Algorithm (SCM)
SPE RR Estimates
Number of Gages
MRA(p) G(p)
G Rain-gauge MPA Merged/Adjusted Rainfall
17
Merging-Adjustment Algorithm (SCM)
SPE RR Estimates
Number of Gages
18
Merging-Adjustment Algorithm (SCM)
SPE RR Estimates
Number of Gages
19
Merging-Adjustment Algorithm (SCM)
SPE RR Estimates
Number of Gages
of gauges ? num-cells x 4
20
Merging-Adjustment Algorithm (SCM)
SPE RR Estimates
Number of Gages
21
Validation of Adjusted SPE vs. Rain Gauge
GPCC Gauge, July 2001
Original PERSIANN, July 2001
Adjusted PERSIANN, July 2001
22
Validation of Adjusted SPE vs. NEXRAD
Original PERSIANN, July 2001
NCEP Radar, July 2001
Adjusted PERSIANN, July 2001
Original Adjusted
Corr. 0.75 Corr. 0.83
23
Conclusions

• The developed merging algorithm is viable for
  extending the patterns and intensity of the
  radar-based rainfall to the gap area from the
  surrounding pixels.
• Generated rainfall is more similar to radar-based
  rainfall, for smaller gap areas.
• Generated rainfall is more correlated with
  radar-based rainfall, if it was available, than
  the original satellite-based rainfall estimates
  with RR.
• Merging monthly SPE with rain gauge observations
  reduces rainfall uncertainties.

24
Hydro-Climate ProjectsThe City University of
New York (CUNY),

• Improving Precipitation Forecast
• (NOAA-NWS, NESDID, MDL, OAR CIMMS)
• Satellite-based Precipitation Nowcasting
  Capability for New York City Metropolitan Area
• Exploring the Relationships between Aerosols
  Hydrological Variables.

• Precipitation Estimation
• (NOAA-NWS -NESDIS)
• Development of MW based Multi-Spectral Remotely
  Sensed Detecting Classifying Light, Moderate
  Heavy Snowfall
• Development of a Multi-Spectral MW based Snowfall
  Rate Estimation
• Development of a Satellite-based Rainfall
  Retrieval Algorithm using Multi-Sensor IR
  Lightning Data
• Multi-Sensor Precipitation Estimation (QPE) over
  the Radar Gap Areas.

Satellite-based Hydro-Climate Projects

• Validation of Rainfall Products
• (NOAA-NWS -NESDIS)
• Validation of Satellite-based NESDIS Rainfall
  Products
• Validation of Satellite-based Rainfall Products
  for hurricane.

25
Time-Line of Hydro-Climate Projects
Developing a Multi-Spectral Detecting
Classifying Snowfall Model
Validation of Satellite-based NESDIS Rainfall
Algorithms
Validation of Satellite-based Rainfall Retrieval
Algorithms for Hurricane
2003 2004 2005 2006 2007 20
08 2009 2010 2011
Improve IR-based Rainfall Retrieval Algorithm for
Thunderstorms using IR Lightening
Developing a Multi-Spectral MW Snowfall Retrieval
Algorithm
Impacts of Aerosols on Hydrological Variables
Satellite-based Nowcasting Capability over the
New York Metropolitan Area
Multi-Sensor Precipitation Estimates (QPE).
YEARS
26
CREST Hydro-Climate Participants
NOAA-CREST Scientists
CCNY- CUNY CCNY- CUNY CCNY-CUNY CCNY-CUNY
Reza Khanbilvardi Shayesteh E. Mahani Arnold
Gruber Brian Vant Hull
NOAA Collaborators
NESDIS NESDIS NWS MDL OAR NESDIS MDL NWS
Ralph Ferraro Bob Kuligowski Pedro
Restrepo Mamoudou Ba Robert Rabin Cezar Kongoli
Stephan Smith David Kitzmiller
27
Hydro-Climate Post-doc Students
Ali Amirrezvani PhD Student, CUNY Start Fall
2004 Expected Graduation Date December 2008

Brian Vant Hull Post Doctor, CUNY Joint CREST
since January 2007

• Awards
• Runner up Prize, Competition Poster
  Presentation,10th Annual CUNY Conference in
  Science and Engineering, The Graduate Center of
  The CUNY,  February 23, 2007.

Yajaira Mejia PhD Student, CUNY Start Fall
2004 Expected Graduation Date December
2007 (Defended on Oct. 3rd)
Nasim Nourozi PhD Student, CUNY Start Spring
2007 Expected Graduation Date December 2010
Completed her MS in December 2007

Cecilia Hernandez PhD Student, CUNY Start
Spring 2005 Expected Graduation Date December
2008
Bernard Mhando PhD Student, CUNY Start Fall
2006 Expected Graduation Date December 2009

Heather Glickman PhD Student, CUNY Start Fall
2005 Expected Graduation Date June 2009