Modelagem não estacionária GAMLSS aplicada à precipitação no Estado da Paraíba, Brasil.

Abstract

The state of Paraíba makes part of the semiarid region of Brazil, where in recent years, it has been living with severe droughts, resulting in major socioeconomic losses associated with climate variability. Understanding how much and how precipitation may be influenced by sea surface temperature (SST) behavior in the tropical region can assist in mitigating problems like this. Thus, it is necessary to adjust a model that can capture the influence of the SST on the precipitation time series. In this study, the generalized additive models for location, scale, and shape (GAMLSS) was applied to filter the climatic indices with higher predictive efficiency and consequently to perform climatic precipitation predictions. The results show the frequent influence of SST in the State, being the collaboration of the tropical Atlantic Ocean more effective than that of the tropical Pacific Ocean in the distribution of rainfall, highlighting the TNA, TSA, AMO, SOI and PDO indexes, as the main predictors. The GAMLSS model showed predictive ability during the summer and austral fall in Paraíba. This performance is verified during the application of climate forecasts in the years 2016 and 2017, highlighting the trimesters of JFM, FMA, MAM, and AMJ, as those with the highest predictive potential. The methodology demonstrates innovative characteristics by the potential in generating climate prognosis. Allowing different sectors, possibilities for regional and sustainable management of water resources, which can promote, in a practical way, resilience to climate risk.