HR: 0830h
AN: H11F-0903 [PDF]
TI: Precipitation Estimation from Remotely Sensed Information using ANN-Cloud Classification
System
AU: * Hong, Y
EM: yhong@hwr.arizona.edu
AF: Dept. of Hydrology and Water Resources, University of Arizona, Tucson, 1Dept. of Hydrology and Water
Resources, The University of Arizona, Tucson, AZ, 85721, Tucson, AZ 85721
AU: Hsu, K
EM: kuolinh@uci.edu
AF: 2. Dept. of Civil and Environmental Engineering, University of California, Irvine, 2 Dept. of Civil and
Environmental Engineering, University of California, Irvine, CA, 92697, Irvine, CA 92697
AU: Sorooshian, S
EM: soroosh@hwr.arizona.edu
AF: 2. Dept. of Civil and Environmental Engineering, University of California, Irvine, 2 Dept. of Civil and
Environmental Engineering, University of California, Irvine, CA, 92697, Irvine, CA 92697
AB:
Abstract
Artificial Neural Network (ANN) models, which contain flexible architectures and are capable of discerning the underlying
functional relationships from data, are recognized as very useful tools in geophysical applications. In this study, we
demonstrate a hybrid ANN modeling system to estimate surface rainfall from satellite infrared imagery. The proposed network,
Precipitation Estimation from Remotely Sensed Information using ANN-Cloud Classification System (PERSIANN-CCS), includes
several components: (1) cloud image segmentation, (2) cloud patch feature selection, (3) patch feature classification using a
self-organizing feature map network, and (4) patch-based rainfall estimates from a group of multiple nonlinear cloud top
temperature and rainfall functions. The PERSIANN-CCS model was first calibrated using observations from Geostationary
Operational Environmental Satellite (GOES) infrared imagery and the Next Generation Radar (NEXRAD) rainfall network. To
further extend PERSIANN-CCS rainfall estimates over the remote regions, Tropical Rainfall Measurement Mission (TRMM)
microwave rainfall estimates (TMI product 2A12) were used to adjust PERSIANN-CCS model parameters. The calibrated nonlinear
cloud top temperature and rainfall (Tb-R) functions of classified cloud patches show highly variability, reflecting the
complexity of dominant cloud-precipitation processes over various regions. Case studies show that PERSIANN-CCS captures the
variability in rain rate at 12kmx12km grid and 3-hour resolutions, with a standard error of 3.0mm/hr and a correlation
coefficient around 0.65. Additional insights into the cloud evolution and precipitation process from the classified
PERSIANN-CCS cloud patch features and rainfall distributions are discussed.
DE: 1800 HYDROLOGY
DE: 1854 Precipitation (3354)
SC: Hydrology [H]
MN: 2003 Fall Meeting