Large-scale land surface hydrology models have been used to make routine hydrological forecasts and to predict future climate changes. For us to have any degree of confidence in model forecasts and predictions, the model structures must be physically sound and, equally important, model parameters must be realistic and reasonable. But how to determine model parameters still remains as one of the unresolved scientific issues faced by land surface modelers. The issue of parameter estimation becomes more imminent when models are applied to locations where there are scarce observational data for model validation purposes.
This study investigates a strategy to improve parameter estimation procedures. Since parameter estimation can be model dependent, Eta Land Surface Sub-system (ETA-LSS) is chosen for this study. The focus of this study is on the estimation of model parameters mainly related to runoff generation. These parameters are generally not directly measurable. They may, however, relate indirectly to local climate, soil and vegetation properties. To uncover these relationships require a detailed scrutiny of how runoff generating processes are formulated in the models and how model behaviors are influenced by individual model parameters. This paper will present some findings on the "proper" values for model parameter at given locations. Insights into sound representations of sub-processes such as surface runoff, interflow, and baseflow are given. This study will take advantage of the data assembled from Model Parameter Estimation Experiment (MOPEX), which include about 40 watersheds in Arkansas/Red River basin. The findings from this study should facilitate further examination of issues related to transferability of model parameters to ungauged basins.