Author(s): Vo Ngoc Duong; Philippe Gourbesville
Linked Author(s): Philippe Gourbesville
Keywords: Rainfall distribution; Hydrological modelling; Large catchment; Mike she
Abstract: In deterministic hydrological modelling, rainfall can be defined as a major input data. Unfortunately, in many regions, rainfall records are often incomplete and not dense enough to accurately represent the reality of rainfall spatial distribution over large catchments. In most cases, the spatial re-distribution of collected data over the catchment according to availability leads to introduce a significant uncertainty within the model before any calibration or validation step. In order to assess these consequences on hydrological modelling applied to the Vu Gia Thu Bon catchment in central Vietnam, the rainfall of 15 local meteorological stations over an area 10350 km2 have been redistributed spatially with several different interpolation methods such as Thiessen polygons, Inverse-distance weight, Spline, Natural neighbor, Ordinary Kriging, Geographically weighted regression. The effect of each method on hydrologic simulation was estimated via Mike She model from DHI-a distributed deterministic hydrological model. This study has demonstrated the added value of each method and clearly identified the uncertainty bias introduced by the rainfall hypothesis in hydrological modelling. The analysis has shown the uncertainty of spatial distributed rainfall and their effects on runoff process simulation within the model. The result confirmed again the advantage of geostatistical techniques in rainfall distribution, particularly, the Kriging methods and suggested to apply this interpolation for Vu Gia Thu Bon catchment.