Transcript Satellite Remote Sensing, Rainfall, and Ground Validation

Ling Tang and Caitlin Moffitt
CEE 6900

Introduction
◦ Flooding in Southern Texas
◦ Satellite Rainfall Data
 GPCP and TRMM
◦ Dartmouth Flood Observatory




Objectives
Methodology
◦
◦
◦
◦
Study Region and Time Period
Datasets
Statistical Analysis
Qualitative Analysis
Results
Conclusions






Torrential rains totaled as
much as 2-3 feet
River levels reached record
heights with crests as high as
30-40 feet above flood stage
9 fatalities
48,000 homes damaged5,000 people evacuated
$1 billion in damage
Extensive impact to livestock
and agriculture in the region
Flooding not just a localized issue- since1970 more than 7,000
major flooding and drought events have caused $2 trillion in
damage and 2.5 million casualties world-wide. (World Water
Assessment Programme, 2009)



Satellite-based flood data could be solution to early flood
warning and disaster management
Important component for flood analysis is rainfall
Two satellite rainfall products considered in this study
GPCP
◦ Sensor Packages:
 Special Sensor
Microwave/Imager (SSMI)
 GPCP Version 2.1 SatelliteGauge (SG) combination
 Atmospheric Infrared
Sounder (AIRS)
 low-orbit IR (leo-IR) GOES
Precipitation Index (GPI)
data from NOAA
 Television Infrared
Observation Satellite
Program (TIROS)
Operational Vertical
Sounder (TOVS)

TRMM
◦ Sensor Packages:
 TRMM Microwave Imager
(TMI)
 Precipitation Radar
 Visible Infrared Scanner
(VIRS)

Uses satellite observations from MODIS to
monitor flooding as it occurs
◦ MODIS
 Visible and infrared bands
 Used to determine properties of Earth’s surface and
atmosphere

MODIS observations confirmed by flooding
reports


To understand the level of agreement of two
satellite-based rainfall products
To understand which satellite-based rainfall
and flood product would be more appropriate
for early flood warning and disaster
management
Study Region: Texas
Latitude:
25.5N - 36.5N
Longitude:
93.5W - 107.5W
Time period: one month
June 09 - July 09, 2002
Three river gauge stations in southern Texas are
selected for the flooding event
1. Frio River
28˚28'02“ N
98˚32'50"W
2. Nueces River
28˚18'31“N
98˚33'25“W
3. San Antonio River
28˚57'05“N
98˚03'50“W

Two Satellite Products:
TRMM 3B42RT and Global Precipitation Climatology Project (GPCP)

Ground Radar Data :
Next Generation Radar (NEXRAD) Stage IV
Data
Area
Spatial
Resolution
Temporal
Resolution
TRMM
3B42RT
60˚N~60˚S
0.25˚
3-hourly
GPCP
global
1˚
daily
NEXRAD
Stage IV
Mainly in US 0.04˚
1˚ and daily
hourly
-- All datasets were uniformed to 1˚ and daily resolution
and cropped at the Texas region for statistical analysis.
1. Estimate the statistical properties of the datasets
- Calculate the mean and standard deviation of the
datasets in each day in the study time period.
2. Compare the level of agreement between the two
satellite datasets
- Estimate the correlation between the satellite
products, and also with the ground data.
3. Estimate the uncertainty of satellite products based
on the truth data (ground radar)
This includes the estimation of four error metrics:
1). Bias
2). Root Mean Square Error (RMSE)
3). Probability of Detection (POD)
4). False Alarm Ratio (FAR)
Error assessment
1. Bias the average of difference between the
study data and the truth of the days being
estimated
2. RMSE the second moment of error, for an
unbiased estimator, RMSE is the standard deviation
3. Probability of Detection (POD) Rain
The fraction of observed events that were correctly
forecast
4. False Alarm Ratio (FAR)
The fraction of forecast events that were observed to
be non-events
(source from : Ebert E. et. al 2007)


Compare hydrographs for point locations
along satellite-claimed “flooded” rivers to
determine if flooding occurred
Side-by-side comparison of DFO flood maps
with 3B42RT and GPCP to determine which
satellite product would be better for early
flood detection and disaster management
3B42RT overestimates
rainfall
GPCP underestimates
rainfall
3B42RT is more
variable than GPCP
3B42RT
Mean
STD
NEXRAD
GPCP
3
2.5
Overall, 3B42RT has
higher bias than GPCP

2
1.5
Mean bias
1
STD bias
0.5
0
-0.5
-1
3B42RT
GPCP
10
•
3B42RT has higher
RMSE than GPCP
8
6
Mean RMSE
4
STD RMSE
2
0
3B42RT
GPCP
0.6
0.5

POD is higher for
3B42RT
0.4
Mean POD
0.3
STD POD
0.2
0.1
0
3B42RT
GPCP
0.4
0.35
0.3

FAR is higher for
3B42RT
0.25
Mean FAR
0.2
0.15
STD FAR
0.1
0.05
0
3B42RT
GPCP


Flood indicated by large jump
in hydrographs
Occurs immediately after
rainfall begins

3B42RT shows stronger
indication of high rainfall
upstream from flood points




Overall, both satellite-based rainfall products
indicated areas of high accumulation
upstream of flooding points
3B42RT had higher probability of detecting
rainfall and flooding, and a higher correlation
with ground measurement
However, 3B42RT also has higher uncertainty
compared to GPCP
GPCP is more appropriate for climatologic
analysis or application








World Water Assessment Programme (2009). The United Nations World Water
Development Report 3: Water in a Changing World. Paris:UNESCO, and
London: Earthscan
Adler, R.F., G.J. Huffman, A. Chang, R. Ferraro, P. Xie, J. Janowiak, B. Rudolf,
U. Schneider, S. Curtis, D. Bolvin, A. Gruber, J. Susskind, P. Arkin, E. Nelkin
2003: The Version 2 Global Precipitation Climatology Project (GPCP) Monthly
Precipitation Analysis (1979-Present). J. Hydrometeor., 4,1147-1167.
Huffman, G.J., R.F. Adler, M. Morrissey, D.T. Bolvin, S. Curtis, R. Joyce, B
McGavock, J. Susskind, 2001: Global Precipitation at One-Degree Daily
Resolution from Multi-Satellite Observations. J. Hydrometeor., 2, 36-50.
Kummerow, Christian et al. “The Tropical Rainfall Measurement Mission
(TRMM) Sensor Package.” Journal of Atmospheric and Oceanic Technology.
Volume 15 (June 1998). 809-817
http://trmm.gsfc.nasa.gov/ (TRMM)
http://www.ncdc.noaa.gov/ (NEXRAD)
http://precip.gsfc.nasa.gov/ (GPCP)
http://www.dartmouth.edu/~floods/ (Dartmouth Flood Observatory)