Hydrological Modelling Improvements Propositions and Applications Results on the Niger River

Abstract:

The modification of lumped models to corresponding semi-distributed ones implies the identification of an optimal weighting factor on different grids. The necessity of such hypothetical weighting factor is shown through two successful proposed protocols, inter-models and inter-basins, using both versions either of GR2M model or SimulHyd model, on sixteen watersheds on the Niger River and its tributaries, over eighteen to forty six years. The assumption of the first protocol is that the difference in performance between the lumped and the semi-distributed versions of a model is due to the spatial variance of soil information in entry of semi-distributed model version. The correlation between the variance coefficient of Water Holding Capacity (WHC of soil) and the relative difference of model versions (between their performances in terms of Nash criteria) decreases in absolute value from 0.76 to 0.21 when a proposed weighting factor is applied in semi-distributed modelling; so, through the weighting factor the variance information of WHC is taken into account, and the correlation between relative difference of model versions (in terms of performance) and this information is decreased: more this de-correlation is, more the weighting factor that improve the performance of a semi-distributed model is better. The proposed weighting factor is built based on the variance coefficient of WHC. Behind the inter-basins protocol is an assumption that: imbedded basins would have influences during the calibration process of a model parameters on a large basin that contains them; it aims to test whether the information from upstream discharge stations could be useful during the modelling process in a downstream discharge station. In the same period, the calibration of a model version on each sixteen basin is performed, following by its validation on fifteens basins using the set of calibrated parameters of the model on the largest basin (Koulikoro). On a basin the lowest the difference of performance is, the highest is the considered influence of this basin to the calibration process of the model parameters on the Koulikoro basin. The coefficient of correlation between the series of relative difference of performances (on a basin) and the series of relative difference of mean WHC values (of each basin against Koulikoro Basin) is -0.33.