Análise de estacionariedade das séries de precipitações da região hidrográfica cearense do rio Poti utilizando testes estatísticos

Abstract

Climate changes over the years have become more and more frequent, we can notice these changes in different contexts, among them, in precipitation. Precipitation is a fundamental component in the composition of a society, moving its economy. Given its importance, it is essential that changes occurring in its cycles are observed and tested, so that its future behavior can be predicted, and problems are avoided. For this, it is possible to analyze through the values of its time series captured in hydrometeorological stations, so that its stationarity is analyzed. With these data in hand, the next step is to apply stochastic tests, returning results that make it possible to detect changes that have occurred over long periods, in addition to measuring the level of these trends, the direction of their slope, their point of change and whether these changes can be considered significant, which will determine whether the series are stationary. For this, in this study, thirteen stations present on the Poti River basin, in Ceará, were selected based on their location and sample size, which vary between the years 1911 and 2018. Then, the normality of the samples was tested by the Shapiro-Wilk test, an important factor for the accuracy of the results of the parametric hypothesis tests applied soon afterwards. In these tests, we evaluate the trend level of the series, to detect possible changes that exceed the adopted significance level, then considering the alternative hypothesis of the tests, which determines that the precipitations are non-stationary. Among the tests available to evaluate the stationarity of historical series, the most used in the literature for hydrometeorological analysis were selected and considering the test power. From parametric tests, the T-Student and F-Snedecor tests were chosen, which respectively evaluate changes in the mean and variance of rainfall. For non-parametric tests, which do not require sample composition, the Wilcoxon, Mann- Kendall, Sen's Slope, Spearman Rho and Pettitt tests were chosen. By applying all the tests, with the aid of the R language, it was possible to compare the results, which showed negative slopes in eleven of the stations and little evidence of non-stationarity, with the Croatá station being the one with the most considerable changes, however still maintaining stability in most tests, concluding that the stations demonstrate strong stationarity arguments.