Geolocation Correction of Satellite Precipitation Estimates Using a Radar-Gauge Product

In order to improve hydrologic predictions, accurate precipitation estimates at fine scales (hourly time scale and few kilometers spatial scale) are required. Geostationary Satellites (GOES) Infrared (IR) based precipitation estimates are available at such scale. However, these satellite precipitation estimates (SPEs) are prone to spatial error and can be observed when compared with ground-based measurements. Detecting and applying correction of location error in IR based SPEs have been minimal. The research objective is to observe and assess location error within NOAA / NESDIS / STAR’s satellite precipitation data fields, Hydro-Estimator (HE) and Self Calibrating Multivariate Precipitation Retrieval (SCaMPR), and NOAA / NWS / NCEP / Climate Prediction Center’s QMORPH against Stage-IV ground-based radar-gauge precipitation estimate (ST-IV). A series of MATLAB functions were created to measure SPEs’ shifts in latitude, shifts in longitude, and combinations of them. The satellite data was then shifted in forty-nine different ways over ST-IV to calculate the minimum root-mean-square-error between SPEs and ST-IV estimates. Forty-nine frames were created by shifting the satellite data over calculated longitudinal and latitudinal directions ranging from -3 to +3 grid spaces. The study specifically focuses on tornado cases for the months of May 2010 and 2015. The geographic location of the study area consists of the states of Oklahoma, Kansas, and Nebraska within the region called “Tornado Alley.” After correcting satellite estimations for spatial bias, significant improvements were found in correlation coefficients between satellite and ground radar.