Metodologia unificada para análise de freqüência de vazões máximas anuais a partir da agregação da informação hidrometeorológica regionalizada

Abstract

Hydrological variables, in general and floodflows, in particular, are considered to be random variables and, consequently, subject to analysis by the theory of probability and mathematical statistics. The estimates of extreme flood quantiles, along with their correspondingexceedance probabilities, are key elements to designing large hydraulic structures and evaluating the associated risks of failure. For less rare floods, the conventional flood frequency analysis provides a number of satisfactory methods of estimation. For extreme floods, however, there is no consensual method to employ, given the large uncertainties inherent to the estimation based on small samples. In this context, the U.S. National Research Council (NRC, 1988) identified three principles tobe followed in order to improve estimation of extreme flood quantiles. These are: substitution of time for space, inclusion of more structure into the employed models, and emphasis on the upper tails of probability distributions. The methodology proposed herein makes use of the three aforementioned principles within a unified context. According to it, the flood peaks that have exceeded an arbitrary threshold and the associated flood volumes are identified and modeled as a marked point stochastic process, using the representation of a compound Poisson process. The essence of the proposed method consists of separatelyestimating the marginal density function of flood volumes, for a duration equal to the watershed time base, and the density function of flood peaks conditioned to volumes. In the sequence, the annual probability distribution of flood peaks can be estimated by integrating the product of both densities. The aggregation of the regional hydrometeorologicalinformation, through a regional TCEV (Two Component Extreme Value) distribution, serves the purpose of guiding the estimation of the flood-volumes density. This aggregation makes use of some premises on the rainfall-runoff transformation, under extreme conditions. The proposed method combines contemporary statistical techniques and aims to extract the maximum information from the available data. In this sense, it is expected that other applications of the proposed method, in addition to the one described herein, may improve the estimation of extreme flood quantiles.