Uncertainty, sensitivity analysis and the role of data based mechanistic modeling in hydrology

In this paper, we discuss the problem of calibration and uncertainty estimation for hydrologic systems from two points of view: a bottom-up, reductionist approach; and a top-down, data-based mechanistic (DBM) approach. The two approaches are applied to the modelling of the River Hodder catchment in North-West England. The bottom-up 5 approach is developed using the TOPMODEL, whose structure is evaluated by global sensitivity analysis (GSA) in order to specify the most sensitive and important parameters; and the subsequent exercises in calibration and validation are carried out in the light of this sensitivity analysis. GSA helps to improve the calibration of hydrological models, making their properties more transparent and highlighting mis-specification 10 problems. The DBM model provides a quick and efficient analysis of the rainfall-flow data, revealing important characteristics of the catchment-scale response, such as the nature of the effective rainfall nonlinearity and the partitioning of the effective rainfall into different flow pathways. TOPMODEL calibration takes more time and it explains the flow data a little less well than the DBM model. The main differences in the mod15 elling results are in the nature of the models and the flow decomposition they suggest. The “quick” (63%) and “slow” (37%) components of the decomposed flow identified in the DBM model show a clear partitioning of the flow, with the quick component apparently accounting for the effects of surface and near surface processes; and the slow component arising from the displacement of groundwater into the river channel (base 20 flow). On the other hand, the two output flow components in TOPMODEL have a different physical interpretation, with a single flow component (95%) accounting for both slow (subsurface) and fast (surface) dynamics, while the other, very small component (5%) is interpreted as an instantaneous surface runoff generated by rainfall falling on areas of saturated soil. The results of the exercise show that the two modelling method25 ologies have good synergy; combining well to produce a complete modelling approach that has the kinds of checks-and-balances required in practical data-based modelling of rainfall-flow systems. Such a combined approach also produces models that are