Modelación no estacionaria de la magnitud y frecuencia de las crecidas en el Alto Cauca mediante índices climáticos y de operación de embalse

Abstract

In this study, changes in the frequency and magnitude of annual floods in the Cauca River (Southwest Colombia) are modeled using a nonstationary framework by means of the Generalized Additive Models of Localization, Scale and Shape. Non-stationary flood frequency analysis incorporates two climatic indices and an anthropic index that allows us to assume that changes in the water reservoir and the percentage of regulated tributary area are factors which disturb the capacity of the dam to withhold the flood. The results highlight the role of the El Niño Southern Oscillation (ENSO) phenomenon and the proposed Reservoir Index, as significant covariates in the parameters of the selected distributions. The dependence of model parameters on covariates improves the model's capacity for representing temporal variability of the flood regime. Nonstationary models indicate significant differences in the flow associated with a specific return period, and in the failure risk of flow design, depending on the working life, in contrast to classical stationary models. The main conclusion is that since 1986 in the gauging station Juanchito, flooding has shown a gradual increase in magnitude, which is unambiguously associated with the cold phase of the ENSO phenomenon. Therefore, the non-stationary models provide valuable information on the reservoir, its regulation strategy for high flow, and for flood risk management in the Cauca river basin.