Regionalization of a rainfall-runoff model for flood prediction over the Island of Tahiti, French Polynesia

Résumé fourni par la source

Study region The Island of Tahiti, prone to heavy rainfall and flash floods that regularly cause severe damage. Study focus The aim of this paper is to build a rainfall-runoff model to estimate floods in any catchment of the island. A distributed conceptual and parsimonious event-based model was designed to fit the observed hydrological spatial variability. It combines the SCS-SMA (Soil Conservation Service - Soil Moisture Accounting) as a runoff production function with a Lag-and-Route routing function, both operating over a mesh of cells. The complete model was calibrated in 11 catchments across the island, and their parameters were related to catchment descriptors related to soil moisture, geology and level of urbanization. New Hydrological Insights for the Region Two parameters were chosen to handle the hydrological variability between catchments: the initial level in the reservoir of the runoff model, H 0 , and an index linked to the flow velocity in the routing model, µ. For natural catchments, an effective relationship was established between the median values of H 0 and the index of regionalized antecedent precipitation. The proportion of urbanization was a secondary factor in explaining H 0 . µ was related to the geological characteristics of the catchment. This rainfall-runoff model can be used to calculate design floods for the whole island of Tahiti, but needs to be improved for use in flood forecasting for real events.