Abstract
We modify a framework to model time series of discharges, on watersheds of the Niger River and its tributaries, over 1901-2020. This framework is built from a French hydrological model -Genie Rural with 2 parameters at Monthly time step -GR2M. The first parameter, X1, concerns the input variables. This parameter is applied to the rainfall and the evapotranspiration data; its reciprocal, 1/X1, is applied to the Soil's Water Holding Capacity data (assimilated to reservoir height). The second parameter, X2, is applied to the time series of discharges, which is the output variable in modelling. Modifications lead to a new model version called SimulHyd, which stands for Simulation of Hydrological systems. A weighting rule on grids is inserted into the framework to create a gridded variant of semi-distributed modelling from a lumped model. Following a World Meteorological Organization's goal, we develop a scoring technique for intercomparisons of model versions on both one single and several catchments. Over the 16 watersheds, a three-model intercomparison shows the preponderance of SimulHyd semi distributed (with 117.83 out of 192 total scoring-points, equivalent to 61.37%) over both GR2M nondistributed (with 61.33 scoring-points, or 31.94%) and GR2M semi-distributed (with 12.83 scoring-points, or 6.68%).