Formulação e avaliação de índices preditores em área semiárida do Brasil.

Abstract

Stability indexes are numeric parameters obtained from upper-air reports in many operational stations and are widely used as auxiliary tools in weather forecasts, given that they state the relationship between atmospheric thermodynamics and precipitation. In this study, we analyze a range of stability indexes at play in the region of Petrolina (9 24′S, 40 30′W), with a special focus on January months, a pre-rainy season month that is locally determined by diverse meteorological systems that shape the weather conditions in the area. Our data base is formed by daily upper-air reports throughout the period of 15 January months, from 1999 to 2016, alongside with precipitation data. Weather forecasting techniques are used as to assess local thermodynamic conditions, by focusing on both the days prior to rain events and rain days catalogued by Petrolina rain-gauge station. Yeo-Johnson transformation was applied to the stability indexes, aiming at the normalization of data. The transformed indexes were then analyzed with the aid of statistical tools, such as Quantis techniques and multivariate analysis, as a way to evaluate whether such indexes are useful to determine conditions that precede rainfall days. By means of the quantile orders of 20, 40, 60 and 80%, our results indicate that both quantis analysis and multivariate analysis pave the way to a better understanding of such indexes as sound indicators for the assessment of instability levels and the prediction of rain occurrence in the Petrolina area. The perfomance analysis of stability indexes ranging from rain days to days prior to rain events indicates that K and SWEAT indexes are the best stability indicators. Yeo-Johnson transformation was efficient enough to approximate indexes distribution to that of normal distribution and have shown that the transformed variable amplify the instability indicator values of K, SWEAT and IC indexes. TMED* has satisfactorily addressed the temporary effects of thermal inversion as a favorable condition to the rain occurrence (by means of a disruption of the stable layer). The discriminant analysis of the preliminary factor allowed us to conclude that the groups we have obtained from groupment analysis have been efficiently discriminated and that SWEAT*, K* TNCL* and IC* were the mot important indexes for the determination of standardized discriminant function.