A Bayesian approach to adjusting precipitation areal reduction factors through rainfall-runoff modeling in flood frequency analysis

Código

I-EBHE0012

Autores

Pablo Paskosky, Rafael Banega, Rafael Navas

Tema

WG 1.10: Hydrologic Design - Solutions & Communication

Resumo

The frequency of extreme floods for hydraulic engineering design is usually determined based on assumptions that rely on the availability of rainfall and streamflow data. Where there is a long record of floods, probability distributions can be adjusted to estimate the frequency of particular flood events. In cases where flood data is absent, but precipitation data is available, rainfall-runoff models come into play. In such situations, a design storm with a specified frequency is used to simulate a flood event that matches the frequency of the design storm (although this is a strong assumption). In more complex scenarios, where flood records are insufficient to develop robust probability distributions and comprehensive rainfall data are available; the question arises of how can these two types of data be effectively combined to estimate the frequency of extreme floods more accurately. In this context, the study aims to optimize a rainfall-runoff model and the areal reduction factor of precipitation of design storms using Bayesian inference. The optimisation routine is designed to reduce uncertainties in the probability distribution of extreme floods. The research focuses on the Tacuarembó Chico at Tacuarembó city (660 km², Uruguay), an area prone to frequent flooding and material losses. The runoff simulations are carried out using the Mesoscale Hydrological Computation Tool (MHCT), which relies on the curve number and the unit hydrograph. The design storm follows the national hydraulic design guidelines but with a variable areal reduction factor (ARF). The Bayesian inference approach allows for the joint optimisation of MHCT parameters, design storm ARF, and quantification of flood frequency uncertainties. The study reveals that the ARF depends on the return period of the storm, indicating that extreme floods occur during frontal and convective storms which have different ARF. These findings contribute to better flood estimation and enhance the design of effective mitigation strategies.