Intercomparison of land-surface parameterization schemes: sensitivity of surface energy and water fluxes to model parameters

In this study, sensitivities of 10 land-surface schemes (LSS) to five prescribed model parameters (i.e. the maximum soil moisture content (MSMC), effective available water content (EAWC), Clapp-Hornberger B parameter, leaf area index (LAI), and minimum stomatal resistance) are investigated based on the fractional factorial analysis method. The sensitivities of four model responses (i.e. evapotranspiration, total runoff, sensible heat flux, and soil moisture in the total zone) are evaluated as functions of these five parameters considering both individual and parameter interaction effects. To facilitate these sensitivity analyses, which are conducted for three hydroclimatic scenarios, two indices are introduced along with a criterion for measuring relative parameter effects. The two new indices are single response effect index and multiple response effect index. Results show that for the majority of LSS, the four model responses are generally most sensitive to the MSMC parameter, followed by the Clapp-Hornberger B parameter, under the three different hydroclimatic scenarios. The effects of MSMC, the Clapp-Hornberger B parameter, and EAWC on the model responses are generally much larger than those of LAI and minimum stomatal resistance among most of the 10 schemes. This implies that the variations associated with the soil properties possibly due to the measurement uncertainties and/or spatial heterogeneity may play a more significant role in partitioning water and energy budgets than those associated with vegetation properties in the current generation of land-surface model parameterizations. Results also show that large sensitivities of model responses exist in relation to the choice of LSS when using the same parameter values, and in relation to the hydroclimatic scenario when using the same parameter and LSS. The differences can be sometimes quite large. In addition, the effects of parameter interactions are generally weaker than those of single parameters. The preliminary conclusions obtained from this study offer some insight on why large response differences between schemes occurred every time in the Project for Intercomparison of Land-surface Parameterization Schemes (PILPS) phases 1, 2(a), 2(b), and 2(c) intercomparison studies, and perhaps on why each scheme performs better at its own testing site(s) than at the PILPS sites. (C) 2003 Elsevier B.V. All rights reserved.