Análise do desempenho das estimativas da versão 7 do produto IMERG Final Run baseado em dados de satélites para identificar eventos extremos de chuva em escala diária

Abstract

Extreme rainfall events significantly contribute to the occurrence of environmental disasters, affecting populations and causing substantial damage, as these events are associated with floods, inundations, and other natural disasters. Therefore, the accuracy in detecting and quantifying these events is crucial for mitigating these environmental disasters and formulating public policies. Thus, this study aims to evaluate the performance of the estimates of the new version of IMERG (Integrated Multi-satellite Retrievals for GPM) Final Run V07 in detecting daily extreme rainfall events. The determination of the threshold of extreme rainfall events was based on the analysis of percentiles, using the 99th percentile. The evaluation was based on point-pixel comparisons with daily precipitation records, using as statistical metrics the coefficient of determination (R²), the concordance correlation coefficient (ρc), the Nash-Sutcliffe efficiency coefficient (NSE), the mean absolute relative error (MRAE), the root mean square error, and the percent bias (PBIAS). This analysis was performed for Brazil, using observed data from 3,119 high-quality rain gauges operated by the National Center for Monitoring and Early Warning of Natural Disasters (CEMADEN), divided into regional groups defined based on the similarity of the rainfall regime, and considering estimated data from IMERG Final Run V07, for the period from 2014 to 2022. The analysis of the distribution of extreme rainfall events revealed different patterns for the homogeneous precipitation groups. The Southeastern group (SE) has the highest number of recorded extreme precipitations (14,517), while the Central group (CEN) has the lowest number of records (3,356), and the Northern and Southeast Coast group (NCSE) is the one that presented the highest average rainfall per event (64.68 mm), while the Semi-arid group (SM) presented the lowest average (31.43 mm). The CNE group presented the best results, in relation to the other groups, in most of the metrics (i.e., R², NSE, and ρc). While the NCSE region presented the worst results in four of the six metrics applied (i.e., NSE, ρc, PBIAS, and MRAE). Finally, it is worth noting that the IMERG Final Run V07 did not present satisfactory results when estimating extreme rainfall events in any of the homogeneous precipitation groups analyzed, which means that there was no significant improvement in relation to this type of event when compared to previous versions of the product