Comparative analysis of IMERG satellite rainfall and elevation as covariates for regionalizing average and extreme rainfall patterns in Greece by means of Bilinear Surface Smoothing

Remotely sensed data, including rainfall estimates and digital elevation models (DEMs), are increasingly available at various temporal and spatial scales, offering new opportunities for rainfall regionalization in regions with limited ground-based observations. We evaluate the efficacy of NASA’s Integrated Multi-satellitE Retrievals for GPM (IMERG) rainfall estimates and SRTM-derived elevation data as alternative spatial covariates for regionalizing average and extreme rainfall patterns across Greece. Using the Bilinear Surface Smoothing (BSS) framework, we assess and compare the regionalization of average daily rainfall and average annual maximum rainfall across multiple timescales (0.5 h to 48 h) by leveraging both IMERG-derived estimates and the elevation data as covariates. Additionally, the BSS framework is herein extended to provide Bayesian credible intervals for the final estimates, using the posterior variance estimate and the equivalent degrees of freedom determined through the Generalized Cross Validation error minimization procedure. Elevation-based models outperformed IMERG, particularly for indices of extreme rainfall, capturing the differential effects of orography. The exploration of the orographic effect based on the BSS framework revealed that the average annual rainfall maxima at small timescales exhibit a negative relation to elevation, which becomes positive and more significant with increasing timescale. However, IMERG proved valuable for regionalizing average daily rainfall, demonstrating its utility as a complementary tool. The results also underscore the role of temporal scale in regionalization efficiency of extreme rainfall, with higher accuracy observed at longer timescales (24 h and 48 h) and greater uncertainty at finer scales.