Satellite-based precipitation estimates can potentially benefit many critical hydrological applications, such as flood and landslide warning, especially due to their high spatial and temporal resolutions and near real-time availability. However, evaluated against gauge-based measurements, these remotely-sensed measurements exhibit considerable systematic errors. Existing efforts to reduce these errors, such as TMPA 3B42, use the most recent, long-term (e.g., monthly) average of gauge observations to adjust satellite-based estimates. However, the latency in the availability of the gauge data prohibits the use of the corrected satellite data in real-time applications.

In this work, we present a pilot study of using gauge data for real-time error reduction in satellite-based estimates over contiguous U.S. Our previous studies show that the systematic errors in most satellite-based precipitation products, including 3B42RT, CMORPH, PERSIANN and NRL-blended, are highly consistent. This consistency allows us to establish an empirical relationship from historical data between satellite estimates and gauge measurements, without the need of real-time gauge data. Then we apply this relationship to real-time satellite-based estimates when gauge data are not yet available.

Evaluation of the corrected satellite data with this approach shows that the systematic errors are reduced at least by 50%. In addition, it is demonstrated that a much sparser network of rain gauges (300 gauges over U. S.) does not significantly degrade the performance of this correction scheme.