Impacto da Incerteza na Representatividade da Repartição Espacial da Precipitação sobre os Hidrogramas e as Curvas de Permanência Obtidos de um Modelo Chuva-Vazão Distribuído

Abstract

Great progress has been made in the last decades in the development of hydrologic rainfall-runoff models. Currently, scientists are focusing on how to assess and treat uncertainties related to hydrologic modelling due to the model structure or to the input data errors. This study aims to assess the uncertainty of monthly flows and their duration frequency curves, due to the weak representation of the spatial variability of the rainfall field with the existing rainfall gauges. The flows are generated by the distributed hydrologic model ACUMOD. The Gramame basin in the northeastern coastal region of Brazil is used to carry out this study because in the past it was equipped with a good rainfall gauge network. This basin is representative of others in the region where the rainfall gauge density is generally much lower. In the simulations, the rainfall gauge density is diminished by subtracting one to seven rainfall gauges sequentially and randomly from the original network. Three complementary analyses are done to assess uncertainty on the frequency duration curve: i) studying the impact of the rainfall gauge density on model output; ii) studying the intradensity impact of the spatial variability of the rainfall gauges field, and, iii) estimating uncertainty on some points of the frequency duration curves. It is shown that the distance between the outputs of the model and the reference flows increases as the rainfall gauge density decreases from the density threshold of one rainfall gauge in 65 km². This situation is generalized in the coastal north-eastern region of Brazil. It is also shown that the sensitivity of the model and of the frequency duration curves varies in space, depending on how the rainfall gauges are spatially distributed. But the results show a different sensitivity for both the outputs of the model and the frequency duration curves. In the Mamuaba and Mumbaba rivers, the uncertaint