Lessons learned from hydrological models for the improvement of climate model

Abstract

Hydrological models are designed to represent the interactions between the
physical process and the water storage in short-term or long-term forecasting. The hydrological models in climate models (also know as Land Surface Models) aim to represent hydrological processes and interactions at a global scale. In this research, one Land Surface Model, HTESSEL is introduced. Previous studies have shown that HTESSEL is not reproducing hydrological fluxes well at a catchment scale.
In this study, the problems in representing discharge in HTESSEL can be summarized in three aspects: mismatch in peaks, slower recession,and the monthly delay. Thus, the aim of this research is to investigate the reasons of the poor performance in HTESSEL, and provide possible suggestions for a better fit. Therefore, the hydrological models HBV and GR4J (that operate on catchment scales) are introduced to identify the problems in HTESSEL by model comparison.
HTESSEL, HBV and GR4J models are used to simulate river discharge in 15 catchments are compared in terms of structure and parameterization. HTESSEL use tabulated parameter values, while HBV and GR4J are calibrated to match observations firstly. In order to investigate the influence of different model processes and parameters, a second calibration of the HBV and GR4J parameters is applied. Here, the model parameters are calibrated to the HTESSEL model output. Comparing the two calibration results, the parameter differences can be identified.
The results show that the soil column in HTESSEL is a key factor that influences the surface and subsurface runoff. On the one hand, HBV and GR4J can reproduce the slower falling limb in humid region by increasing their slow reservoirs. On the other hand, the top 50 cm of soil column is the effective depth that influences maximum infiltration rate. Thus, the changing of effective depth and the parameterization of orography variable b, which influences the fast runoff in HTESSEL, are necessary in temperate and mediterranean catchments.
According to this study, to solve the problem of mismatch in peaks this 50cm should be a spatial variable firstly. The increase of effective depth could overcome the overestimation in some places and the decrease of effective depth could overcome the underestimation of peaks in other places. In addition to effective depth, optimizing parameter b is also necessary, because it influences the fast runoff. Moreover, for the problem of slower recession and monthly delay, decreasing the size of soil column in HTESSEL is one way to get a better fit. Thus, in future, more study could focus on the interplay of the soil infiltration capacity and the fast runoff parameters. It might be helpful to improve the simulation.