Investigating Satellite Precipitation Uncertainty Over Complex Terrain

The estimation of heavy precipitation events is a particularly difficult task, especially over high mountainous terrain typically associated with scant availability of in situ observations. Therefore, quantification of precipitation variability in such data‐limited regions relies on remote sensing estimates, due to their global coverage and near real‐time availability. However, strong underestimation of precipitation associated with low‐level orographic enhancement often limits the quantitative use of these data in applications. This study utilizes state‐of‐the‐art numerical weather prediction simulations, toward the reduction of quantitative errors in satellite precipitation estimates and an insight on the nature of detection limitations. Satellite precipitation products based on different retrieval algorithms (Climate Prediction Center morphing method, Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks‐Cloud Classification System, and Global Satellite Mapping of Precipitation) are evaluated for their performance in a number of storm events over mountainous areas with distinct storm characteristics: Upper Blue Nile in Ethiopia and Alto Adige in NE Italy. High‐resolution (1 and 2 km) simulations from the Regional Atmospheric Modeling System/Integrated Community Limited Area Modeling System are used to derive adjustments to the magnitude of satellite estimates. Finally, a microphysical investigation is presented for occurrences of erroneous precipitation detection from the satellite instruments. Statistical indexes showcase improvement in numerical weather prediction‐adjusted satellite products and microphysical commodities among cases of no detection are discussed. Key Points Satellite precipitation estimates over two challenging mountainous regions are associated with underestimation during HPEs NWP estimates proficiently accommodate the adjustment of satellite precipitation magnitude when in situ observations are scant Low ice‐mixing ratio values and confined vertical cloud structures are the common characteristics in cases of inefficient satellite detection