Methodology and comparative study of monthly water balance models in Belgium, China and Burma

Abstract

A set of new monthly rainfall runoff models (water balance models) is defined, for use in river catchments smaller than about 4000 km 2, without appreciable frost or natural or artificial lakes. The input series are areal precipitation and potential evapotranspiration. The output is riverflow. The number of parameters, used in the description of the hydrological phenomena in the catchment, is in most cases three, sometimes four. The statistical methodology use for calibrating the models of given catchment is described; it reduces essentially to regression analysis, including residual analysis, sensitivity to calibration period and extrapolation test. In particular, automatic calibration is used, excluding subjective elements. The models are applied to 79 river basins in Belgium, China and Burma. The results are compared with four similar models taken from the literature. The results of applying the new models are satisfactory from a statistical point of view and are much better than those quoted in the literature; a greater part of the observed runoff is explained and there is no residual seasonality. This results from the different mathematical structure of the models, and especially from the use, in the published models, of several storages with maximum ‘capacities’, with no distinction between slow and fast runoff corresponding to baseflow and direct runoff, respectively.