Intercomparison of H SAF and IMERG heavy rainfall retrievals over a Mediterranean coastal region

Abstract

Satellite-based precipitation products play a crucial role in providing global, continuous, and reliable estimates of rainfall, essential for understanding and managing Earth’s water cycle. This study evaluates the accuracy of three H SAF satellite products (H61B, H64, and H68) and compares their performance with IMERG V06B (Early and Late Runs) products in detecting and estimating extreme precipitation events in the western Mediterranean region. The analysis is based on hourly and daily rainfall data collected from 186 rain gauges in Catalonia (NE Iberian Peninsula), using a point-to-pixel approach. The results show that satellite estimates tend to overestimate low precipitation accumulations (less than 2 mm and 10 mm in one hour and one day respectively), with this overestimation being more evident in the H68 product at the hourly scale and in IMERG Late at the daily scale. However, all products show a substantial decline in accuracy for higher precipitation amounts, particularly when they exceed 10 mm in one hour and 30 mm in one day (relative errors up to −60% and critical success index less than 33% in all cases). Despite its biases, IMERG Late product emerged as the most reliable for detecting substantial rainfall accumulations. Additionally, this analysis examined the relationship between the microphysical properties of the precipitating cloud top and estimated or observed surface precipitation. Accurate precipitation estimates from the satellite products were consistently associated with high values of Cloud Optical Thickness (COT) and Ice Water Path (IWP), while false alarms were often linked to low values of these variables. As expected, the results indicated a poorer performance in estimating precipitation associated with warm clouds. Additionally, these findings highlight the importance of integrating cloud optical and microphysical properties into retrieval algorithms to improve satellite precipitation estimates.