Validation of the CMORPH satellite-rainfall product using an experimental rain gauge network and radar rainfall product

We evaluated the National Oceanographic and Atmospheric Administration Climate Prediction Center (NOAA-CPC) satellite rainfall product which is based on the CPC morphing technique (CMORPH). The evaluation is done at the highest resolution of the product (30-minute and ~8 km) using two sources of independent rainfall data: a dense rain gauge network in South Louisiana and the radar-based NWS MPE product over the same area. In terms of the probability of exceedance of the hourly rainfall, the distributions of the CMORPH, radar and gauge data show a very good agreement for rainfall thresholds between ~3 mm – 8 mm which correspond to the 2 % - 8 % exceedance probabilities. However, there appears to be a large difference for extreme low and high rainfall intensities as derived from the three data sources. The radar product has small total bias (-1.91% of the annual rainfall) that is comparable to the magnitude of its components. The total bias of the CMORPH product is also small (0.02%) but the missed rain and the false rain reached 25 % and 30 % of the annual rainfall amount. The result indicates the need to interpret the total bias jointly with its components. The use of radar data instead of gauge data as a reference for the CMORPH product significantly changed the magnitudes of the hit bias and the missed rain but it only slightly affected the total bias and the false rain. CMOPRH successfully detected occurrences of rainfall intensities that exceeded zero only 30 % of the time missing 30 % of the rainfall depth. However, it detected rainfall intensities that exceeded 1.75 mm/h about 71% of the time. The false detection of rain occurred only <4 % of the time but the falsely detected rainfall depth reached up to 30%. The use of the radar estimates instead of the gauge estimates as a reference did not change the probability of detection but significantly changed the probability of false detection. The CMORPH estimates better agreed to the gauge and the radar estimates at the event scale than at the hourly scale. The satellite product overestimated the rainfall depth at the start of an event but significantly underestimated the rainfall during most parts of the event duration. Overall, this study showed that the poor performance of CMORPH is mainly due to its poor detection capability of light rainfall. We also showed how the density of gauges that is used to estimate the reference rainfall affects the evaluation result.