Validation of two SVAT models for different periods during the West African monsoon

Abstract

In climate and weather forecast models, multi-layer soil-vegetation-atmosphere transfer schemes (SVAT) simulate the soil-atmosphere exchange. Often, these SVAT schemes are used without a profound knowledge of the essential input parameters. They rely instead on a standard setup for known soil textures. Comprehensive measurements of soil moisture, soil temperature, and energy balance components during the 2006 African Monsoon Multidisciplinary Analysis (AMMA) campaign offered a unique opportunity for validation of two SVAT schemes, TERRA-ML and VEG3D, with standard setups. Both schemes are coupled by default or experimentally to the frequently used COSMO model. For the soil moisture content in the top 0.1 m of the soil Veg3D (soil texture: loamy sand) and TML (soil texture: sand) were able to reproduce observations, in lower layers performance was unsatisfactory. Soil temperature observations were represented rather well by the SVAT model results for the top 0.1 m of the soil, whereas at depths below 0.1 m, the performance of the models was poor. This could be traced back to the fact that different horizons were observed but not represented by the models. The decay of soil moisture anomalies in the different model runs was faster than the one observed. This affects the time span in model simulations favourable for the triggering of convection and the generation of thermally forced circulations. The turbulent fluxes simulated by VEG3D (soil texture: loamy sand) and TERRA-ML with the soil texture of sand were in good agreement with observations, while TERRA-ML with a soil texture of sandy loam completely missed the right transformation of the available energy into sensible and latent heat. This proves that an appropriate soil texture classification with appropriate parameter settings is as important as soil moisture.