A physically-based parsimonious approach for spatial disaggregation of NEXRAD precipitation data in mountainous terrains

Accurate characterization of the spatial variability of precipitation is critical for understanding and prediction of both natural and human-influenced hydrologic responses (e.g., recharge, runoff, reservoir releases). However, this information is often not available in mountainous terrains because of sparse precipitation observation networks. NEXRAD provides essentially continuous temporally and spatially precipitation data with coarse spatial resolution. With ~4 km spatial resolution, NEXRAD may not capture the spatial variability of mountain precipitation, which is critical to modeling hydrologic response. Recent studies [Andreassian et al., 2004; Bindlish and Barros, 2000] demonstrate significant improvement of hydrologic model performance using higher resolution precipitation input. The objective of this study is to disaggregate NEXRAD precipitation data in mountainous terrain, using a newly-developed physically-based statistical approach [Guan et al., 2005]. Using a multivariate linear regression, conditioned by original NEXRAD data, the approach auto-searches topographic and atmospheric effects on mountain precipitation distribution. The resulting regression function is used to disaggregate original NEXRAD data to a spatial resolution of 1 km. This algorithm is tested in the mountainous terrain of northern New Mexico.