Inferring the Temporal Sampling Uncertainty of Satellite Precipitation in Data-Sparse Regions

The global distribution of precipitation is an important key to understanding the earth's water and energy budgets. While developed countries tend to have relatively good precipitation observation networks, our understanding of the distribution of precipitation in data sparse regions relies on a few rain gauges and the information gathered by space borne sensors. Several recent datasets (CMORPH, PERSIANN, TMPA) attempt to represent the global distribution of precipitation at relatively high spatial and temporal scales by combining observations from multiple passive microwave and infrared satellite platforms and ground-based observations. The highly variable nature of precipitation, combined with the low number of globally available sources of ground validation make it difficult to estimate uncertainty for these multi-satellite precipitation datasets. By relating the precipitation uncertainty calculated in data-rich regions to the characteristics of the observed precipitation, it may be possible to infer the uncertainty of precipitation estimates in data-sparse regions that share similar characteristics. A methodology is developed to infer the uncertainty of multi-satellite precipitation measurements based on the monthly probability distribution of 3-hourly accumulated rainfall and is tested over the contiguous United States using the TRMM Multisatellite Precipitation Analysis (TMPA) validated against the NCEP Stage IV radar product.