5B.2
Multi-sources precipitation estimates
Kibrewossen Tesfagiorgis, NOAA/CREST, New York, NY; and S. Mahani, R. Khanbilvardi, and D. Kitzmiller
Development of a multi-source merging approach, for improving Satellite-based Precipitation Estimates (SPE) over the radar gap coverage by merging SPE with ground-based radar (RR) and rain gauge (RG) rainfall measurements is the objective of the present study. High resolutions, hourly 4km × 4km, SPE from NESDIS Hydro-Estimator (HE) algorithm and NEXRAD Stage-IV are selected to be merged and generate a multi-sensor product. The merging algorithm should be capable of extending rainfall information from neighboring pixels inside the radar gap area. Successive Correction Method (SCM) is viable of doing this job. In the present project SCM in combination with Kriging-Bayesian technique is planned to be implemented for merging multi-sensor SPE, RR, and RG information. The experiment indicates that there are considerable intensity and distribution differences between SPE-HE and RR. Therefore, the first step should be bias correction of SPE with respect to RR and/or RG data to reduce the rainfall intensity discrepancy. In this study, we examined bias correction of satellite-based HE rainfall with respect to NEXRAD Stage IV rainfall measurements using ratio of mean, median and maximum values for corresponding rainy pixels as bias estimators. Based on our results the average of mean-ratio and max-ratio could give good bias corrected HE for almost all of the study data. The next step before implementing the merging algorithm is to reduce the spatial discrepancy between SPE-HE and RR. To solve the problem of the dislocation between the radar and satellite rainfall estimates, one of them has to be geometrically corrected with reference to the other. For this particular study the RR was taken as a reference or template. The dislocation measures the spatial error between the two rainfall products. Several corresponding pair points were selected between the radar and satellite estimations for parameter estimations. A set of pair points is selected to apply linear registration using least square technique as the first trial to minimize the dislocation error. This process will continue until we find the best match between the satellite and radar rainfall estimates.
Session 5B, Remote Sensing and Satellites V
Friday, 13 November 2009, 1:45 PM-3:35 PM
Previous paper Next paper