Assessment of satellite products for filling rainfall data gaps in the Amazon region

Abstract

AbstractRainfall data series with adequate quality and length are often incomplete or nonexistent. Thus, filling in rainfall gaps becomes necessary to complete databases. This article proposes the use of satellite products (TRMM—Tropical Rainfall Measuring Mission, CHIRPS—Climate Hazards Group InfraRed Precipitation with Stations and CMORPH—CPC Morphing Technique) to fill gaps in the rainfall historical series. The simple regression method, using satellite rainfall estimates, was tested to fill the missing data from 164 rainfall gauge stations in the Amazon region. Large dispersions were observed between rainfall data, with R2 ranging from 0.383 to 0.844, the best results were found in areas with less rainfall. As well, the greatest performance of the products was verified in the dry period, with r and d higher than 0.899 and 0.950, respectively. The product with the best representation in the region was CHIRPS, which had the lowest monthly values of mean absolute error (0.979 mm) and root mean square error (3.656 mm). The results confirm that the satellite estimates satisfactorily represent the seasonal variation of rainfall in the region, despite presenting cases of overestimation and underestimation of data. The higher performance of CHIRPS can be explained by the higher spatial resolution (0.05°), allowing for more accurate weather forecasts. In fact, CHIRPS has the CHPclim model, which adds other factors to the good product performance. These characteristics justify the better performance of the CHIRPS product for filling gaps in daily rainfall data in the Amazon region, favoring the best monthly rainfall estimates for each region state analyzed.Recommendations for Resource Managers Satellite products have been increasingly used for estimating rainfall data in regions with a low number of installed rainfall gauge stations. Thus, the assessment and selection of these products needs to be elaborated for the best decision making of water resource managers. Rainfall data are important to recognize the occurrence patterns for prediction of the climatic behavior of a region. Sectors such as agriculture and disaster prevention (droughts, floods, erosion of watersheds, and river silting) need knowledge of rainfall for planning, management, and mitigation. Knowledge of rainfall behavior is very important in the Amazon region. In this case, the dry season and temperatures have been increasing due to global climate change. These changes establish conditions for more intense fires, which increases the deforestation of the region.