Effective SVAT-model parameters through inverse stochastic modelling and second-order first moment propagation

Inverse stochastic modelling and direct moment propagation is applied for deriving effective land surface parameters of a Soil-Vegetation-Atmosphere-Transfer (SVAT) model. Special attention is given to the proper representation of latent heat fluxes at grid scale when an underlying subgrid-scale heterogeneity is assumed. Derived effective land surface and vegetation parameters yield the same total latent heat fluxes on the grid scale as the averaged heat fluxes on the subgrid-scale. Upscaling relations are derived that relate mean and variance (first and second moment) of subgrid-scale heterogeneity to a corresponding effective parameter at grid-scale. Explicit upscaling relations are exemplary derived for (a) land surface property parameters roughness length, albedo, emissivity and (b) vegetation parameters wilting point soil moisture, field capacity, leaf area index, vegetation fraction, and minimal and maximal stomata resistance. It is shown that the value of effective parameters depends on the sign of the first and second derivatives of aggregated fluxes with respect to the parameter. It is demonstrated that the Second-Order-First-Moment (SOFM) Method yields results that are comparable to those obtained with inverse stochastic Monte Carlo simulations. Effective parameters were found to be independent of meteorological forcing conditions and initial conditions for a 7-day time period.