Análise de consistência e regionalização das precipitações máximas ocorridas no Rio Grande do Sul entre 1912-2014

Abstract

The knowledge of intense precipitation is primordial in hydraulic structures design. However, the determination of this variable stumbles on its monitoring deficiency. Thus, the use of statistical methods, such as regional frequency analysis, aims to support the determination or caracterization of these precipitation. The purpose of this study was to identify behavioral patterns of daily intense precipitation that allows the estimative of design storm in ungauged basins. The study was directed to the southernmost state of Brazil, Rio Grande do Sul (RS), a region with one of the world highest intense precipitation events. The hydrometeorological monitoring of the region is accomplished mainly by the National Water Agency (ANA). In this study were analyzed 1070 rain gauges, containing records between 1912 and 2014. Firstly, the consistency procedure was performed, by pre-analyzing and classifying the precipitation events (adapted Z score), selection of annual maximum series of continuous periods, and punctual validation. In the second stage, the maximum events were submitted to a regional frequency analysis, based on the regionalization method of Hoskins e Wallis. Finally, was demonstrated a precipitation estimative method for ungauged basins. The results of the pre-analysis characterized the periods of greatest temporal and spatial continuity from the available records, which occurred in the decade of 1970. On the other hand, this period has the greatest number of monitoring inconsistencies. The classification of the precipitation records indicated possible inconsistent data. Parallel to this, the selection process allowed the elaboration of maximum series of 15 and 30 years, uninterrupted, which occurred between 1963 and 1977, and between 1971 and 2000, with 91 and 40 gauges, respectively. Among the series, 163 events were indicated as possibly inconsistent data, which then were analyzed by the punctual validation, resulting in more trustworthy maximum series. The results of the maximum series regionalization indicated that RS has from two to three regions of homogenous behavior. The estimative from the 15 years series generated the most accurate values for Return Periods (T) lower than ten years. In opposite, the 30 years series resulted in better approximations for T greater than 20 years. Nevertheless, the quality of the estimates is prejudiced in the regions with a low density of gauges. The processes applied to the consistency analysis were effective, and the suggested methodology demonstrated to be appropriate and flexible.