An adaptive methodology to enhance infrared satellite precipitation estimates using radar observations

We have developed a methodology to enhance an infrared –based high resolution rainfall retrieval algorithm by intelligently calibrating the rainfall estimates using both ground and space-based radar observations. Our approach involves the following four steps: 1) segmentation of infrared cloud images into patches; 2) feature extraction using a wavelet-based method; 3) clustering and classification of cloud patches; and 4) dynamic application of brightness temperature (T_b) and rain rate relationships, derived using ground and space-based based radar observations. Cloud-top brightness temperature measurements from geostationary satellite (GOES-12), in conjunction with cloud-to-ground lightning data from the National Lightning Detection Networks (NLDN), are used for cloud classification and rain fall estimation. The study area covers parts of New Mexico, Texas, Oklahoma, Kansas, Colorado, Arkansas, and Missouri. Our methodology for satellite-based rainfall estimation is similar to the PERSIANN approach. However, our algorithms are enriched with lightning data and further enhanced with a wavelet-based technique for feature extraction. Wavelet transform is a powerful tool for texture analysis. So it is applied in our methodology to extract information from features of cloud texture. In addition, past studies have shown that lightning is correlated to rainfall amounts and cloud top temperatures. So we expect to get better results by incorporating lightning information which is the number of flashes that occur in the respective cloud patch areas during -15 min to +15 min window of observed infrared data from geostationary platforms. Further, an artificial neural networks (ANN) with Self Organizing Map (SOM), an unsupervised technique, for cloud clustering. This algorithm has been used to extract cloud features from GEOS-12 (Channel 4) to estimate rain rates at 0.04° x 0.04° spatial resolutions every 30 min.

Some preliminary results, based on an data from June 27, 2007, are shown in Figure 1. The cloud top brightness temperature from the GOES-12 infrared is shown in top-left; the corresponding cloud patches, obtained by segmentation and morphological image processing, are shown in the top-right panel with an overlay of lightning flashes in 15 minute widow around the time of nominal scan. The results (bottom-right) are qualitatively validated against an estimate of rain rates (bottom-left) from the TRMM Microwave Imager (TMI). In this initial iteration of algorithm development, for comparison purposes the PMW-based rainfall estimates from TMI, from a 30-day window prior to this testing period, were also used to train and calibrate the algorithm. The initial results are promising; but the algorithm currently overestimates the area of rainfall. We are presently using TRMM-PR and the ground-based dual-polarimetric NSSTC ARMOR radar estimates to further enhance and test the algorithm. Then, in-situ measurements will be used for bias correction. In addition, an optimum nonlinear fitting function will be applied to improve T_b rain rate relationships. This high resolution precipitation product will have potential applications in near-realtime hydro-meteorological applications as well as in climate studies.