PERSIANN-CDR based characterization and trend analysis of annual rainfall in Rio De Janeiro State, Brazil

Abstract

Abstract Climate Data analysis has become a fundamental tool for scientists who seek to better evaluate changes in climatic variables worldwide. When it comes to rainfall there are many datasets publicly available, and orbital products have been gradually sharpening its results. The Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks - Climate Data Record (PERSIANN-CDR) annual product is used in this study to characterize rainfall variation over the state of Rio de Janeiro (SRJ), considering the period of 1983 to 2017. A rainfall dataset with 35 year long series for each of the 92 municipalities of the SRJ was created using GIS software. Several statistical tests were then applied to the datasets of each municipality in order to verify normality (Shapiro-Wilk, Anderson-Darling, Lilliefors and Jarque-Bera), homogeneity (Pettitt, SNHT, Buishand and von Neumann), trends (Mann-Kendall) and intensity (Sen) of reduction or increase in annual rainfall. The estimated rainfall datasets were classified mainly as normal and homogenous (non-significant breakpoints), but significant breakpoints were registered by the Buishand's test in the dataset of twenty seven (29.34%) municipalities. Twenty municipalities had their estimated datasets compared to local meteorological stations in order to verify PERSIANN-CDR performance over the SRJ. Municipalities in the Middle Paraiba and Center South regions are the wettest of the state, while locations that presented lower average annual rainfalls in the state are concentrated in the North, Coastal Flats and Northwest regions. Alarming trends of reduction in annual rainfall were identified for all municipalities using the MK test, but to the threshold of 95% reliability, results show fifty four (58.69%) municipalities located in the central and western parts of the SRJ. According to Sen ´s test the intensity of annual rainfall reduction is greater in municipalities of the Middle Paraiba and Green Coast regions, but Center South, Metropolitan and Coastal Flats regions also registered disquieting results. PERSIANN-CDR analysis can be considered an efficient methodology in the characterization of rainfall variability and trend detection for the SRJ, being encouraged for future studies addressing rainfall and drought variability over the state. The analysis of the PERSIANN-CDR products should also be applied in other regions of the country, especially considering the remarkable interannual and intraseasonal variability of rainfall in Brazil.