1. Evaluating Satellite Rainfall Products for Hydrological Applications
Mekonnen Gebremichael, and Dawit Zeweldi
Civil & Environmental Engineering Department
University of Connecticut

2. Evaluating Satellite Rainfall Products (PERSIANN)
Outline
Introduction
The Approach
The Study Region
Evaluating Satellite Rainfall Products (PERSIANN)
Performance Statistics
Evaluating Utility in Hydrological Applications
A Blueprint
Conclusions

3. Introduction Two validation approaches:
Evaluating against independent rainfall observations
2. Evaluating error propagation in hydrological applications

4. Introduction
High-resolution satellite rainfall products analyzed: PERSIANN, CMORPH (to follow)
PERSIANN
IR-PMW merged algorithm: Neural Network
4-km hourly over the United States
Validation Approach
Evaluation against NEXRAD radar rainfall observations
Evaluation in hydrological applications (to follow)
Study region
Little Washita watershed in Oklahoma, USA
Good quality NEXRAD data; subject of several major experiments (NASA, USDA, etc.)

5. Introduction: The Little Washita Watershed
Area ~ 600 km2

6. Introduction: Inter-annual variability

7. Performance Statistics: 4-km, hourly time scale
Bias

8. Performance Statistics: watershed-averaged, hourly time scale
DJF
JJA

9. Performance Statistics: e-folding distance

10. Performance Statistics of the Largest Storm

11. Performance Statistics of Storms
Storm Duration
Storm rainfall , mm
E-folding distance:
eP / eN
Peak Storm Rate:
qP / qN
RMSE mm
Correlation
Beginning
Ending
NEXRAD
PERSIANN
06/04/02:22
06/05/02:16
37.31
47.62
8.60
0.92
6.11
0.57
08/11/04:07
08/11/04:14
30.07
26.46
8.17
0.62
2.76
0.63
03/03/04:18
03/04/04:23
53.82
58.12
0.89
0.94
8.59
10/07/04:09
10/07/04:18
36.21
28.16
0.70
2.82
0.40
09/19/02:04
09/19/02:13
24.90
14.95
4.26
1.34
8.45
0.23
05/17/02:07
05//17/02:12
36.59
18.78
0.41
0.42
3.20
05/14/03:03
05/14/03:13
21.56
9.30
1.07
0.46
5.74
10/08/02:06
10/09/02:22
72.46
15.84
0.55
0.26
7.16
0.48
10/28/02:19
10/29/02:06
30.73
13.40
0.71
0.79
5.07
0.25
09/08/02:16
09/09/02:19
44.95
5.49
0.09
8.00
0.76

12. Performance Statistics: Scaling with Temporal Scale
DJF
JJA
What is the appropriate space-time scale?

13. Raster-Grid Models (e.g. MIKE SHE)
Hydrological Application
Which is the hydrologic model apt for the satellite rainfall observations?
Numerical simulations of catchment hydrologic processes require a method for representing a basin. Methods can be categorized as lumped versus distributed modeling (contours, grids, polygons, TINs).
Raster-Grid Models (e.g. MIKE SHE)
Predictive performance of hydrologic models as a function of model complexity
and data availability (Grayson and Bloschl 2001).
versus
Basin-Averaged Models (e.g. HEC-HMS)

14. Performance Statistics: A Blueprint for Hydrological Applications
Reference
Rainfall Products
Hydrologic Model
Simulated
Watershed
Response
Observed
Hydrologic
Model
Error
Satellite
Rainfall Products
Simulated
Watershed
Response
Rainfall Error
Propagation
Hydrologic Model

15. Conclusions
Need for rigorous validation of high-resolution satellite rainfall products, at various space-time scales, for different regimes
Need for identifying hydrologic model complexity level apt for satellite-rainfall inputs, and sensitivity to space-time resolutions

16. Thank you