GIS-based spatial precipitation estimation using next generation radar and raingauge data
Additional Document Info
Precipitation is one important input variable for land surface hydrologic and ecological models. Next Generation Radar (NEXRAD) can provide precipitation products that cover most of the conterminous United States at high resolution (approximately 4 km 4 km). There are two major issues concerning the application of NEXRAD data: 1) the lack of a NEXRAD geo-processing and geo-referring program and 2) bias correction of NEXRAD estimates. However, in public domain, there is no Geographic Information System (GIS) software that can use geostatistical approaches to calibrate NEXRAD data using raingauge data, and automatically process NEXRAD data for hydrologic and ecological models. In this study, we developed new GIS software for NEXRAD validation and calibration (NEXRAD-VC) using raingauge data. NEXRAD-VC can automatically read in NEXRAD data in NetCDF or XMRG format, transform projection of NEXRAD data to match with raingauge data, apply different geostatistical approaches to calibrate NEXRAD data using raingauge data, evaluate performance of different calibration methods using leave-one-out cross-validation scheme, output spatial precipitation maps in ArcGIS grid format, calculate spatial average precipitation for each spatial modeling unit used by hydrologic and ecological models. The major functions of NEXRAD-VC are illustrated in the Little River Experimental Watershed (LREW) in Georgia using daily precipitation records of fifteen raingauges and NEXRAD products of five years. The validation results show that NEXRAD has a high success rate for detecting rain and no-rain events: 82.8% and 95.6%, respectively. NEXRAD estimates have high correlation with raingauge observations (correlation coefficient of 0.91), but relatively larger relative mean absolute error value of 36%. It is also worth noting that the performance of NEXRAD increases with the decreasing of rainfall variability. Three methods (Bias Adjustment method (BA), Regressing Kriging (RK), and Simple Kriging with varying local means (SKlm)) were employed to calibrate NEXRAD using raingauge data. Overall, SKlm performed the best among these methods. Compared with NEXRAD, SKlm improved the correlation coefficient to 0.96 and the relative mean absolute error to 22.8%, respectively. SKlm also increased the success rate of detection of rain and no-rain events to 87.47% and 96.05%, respectively. Further analysis of the performance of the three calibration methods and NEXRAD for daily spatial precipitation estimation shows that no one method can consistently provide better results than the other methods for each evaluation coefficient for each day. It is suggested that multiple methods be implemented to predict spatial precipitation. The NEXRAD-VC developed in this study can serve as an effective and efficient tool to batch process large amounts of NEXRAD data for hydrologic and ecological modeling. 2010 Elsevier Ltd.