Hybrid and non-hybrid interpolators applied in the spatial distribution of rains in the coastal mountain region of Rio de Janeiro state

Abstract

Despite advances in space-time monitoring of rainfalls, this information in regions of complex topography is scarce. Space interpol based on orbital data might surprise such a solution. Therefore, the study evaluated the performance of hybrid and non-hybrid interpolators in estimating rainfalls in Ilha Grande Bay Hydrographic region (RHBIG), located in Serra do Mar (the Sea Mountain, in Portuguese), in the states of Rio de Janeiro and São Paulo. Rainfall data from surface variable seasons and virtual seasons, derived from the CHIRPS product, the years 2004 and 2013, were used for Ordinary Kriging (KO) and Regression Kriging (KR) interpolator models, having as explanatory the Digital Elevation Model (DEM). The results showed a correlation (r = 0.68) between the observed and CHIRPS data, with the underestimated rainfall rate on the coast, average difference (di) of -10%, and overestimated on the plateau (di = 9%), which gives rise to a smoothed spatial distribution. Regarding the rainfall models, the KR, linear and logarithmic, tended to extrapolate the minimum and maximum values, increasing the rainfall rate from the coast to the plateau, with an average of -28% of the coastal stations. While, in the KO models, the volume tends to decrease from the coast to the plateau (di = -1.4), which is corroborated by studies already carried out at RHBIG with observed data. Finally, the models of both interpolators, with the virtual stations corrected for the observed data, show an increase in the rainfall amplitude, which reduces the smoothing of the CHIRPS model. In this way, the KO models derived from a dense and regular network of virtual stations, corrected for observed data, can be an alternative for using CHIRPS data in complex topography locations without a dense network of surface stations, which results in models with higher spatial resolution and rainfall amplitude.