Evaluating modifications of the soil module TERRA Felix Ament, MeteoSchweiz MeteoSwiss has just recently introduced the multi-layer version of the soil module TERRA operationally. One reason for this delay was a too strong drying of the soil during spring and summer which was revealed by test integrations. Operationally, this problem is now fixed by a relaxation of soil moisture in deep layers towards analysis data of the forcing ECMWF model. However, this is just a pragmatic approach and should be replaced by physically based improvements of the model itself. A direct analysis of the dry-out problem is very difficult since many components of the water budget are hardly measurable only observations at super-sites like Lindenberg (see talk by Gerd Vogel) might give some insight. However, it remains unclear whether soil moisture is too low due to too strong evaporation, excessive surface runoff, ground runoff created by oversaturated layers or runoff at the lower boundary. Therefore it is reasonable to address the dry-out problem indirectly by testing various model modifications which have the potential to reduce the effect. For this purpose, a set of 13 sensitivity integrations of TERRA in a stand-alone version have been performed. Atmospheric forcing is prescribed from aLMo7 analysis data. Each integrations covers a domain including Switzerland, parts of southern Germany, northern Italy and France with 7km grid spacing. A long integration period of 13 month allows the model to reach at least partly a balanced state with respect to the modified model formulation. Atmospheric forcing is prescribed from aLMo7 analysis data. The set of tested model modifications comprises e.g. alternative lower boundary conditions, other parameterizations of hydraulic conductivity and diffusivity, inclusion of marco-pore effects, testing of external parameters derived from ECOCLIMAP or bare soil evaporation according to Noilhan and Platon (1989). We will present the result of this sensitivity experiment by an intercomparision of means of all components of the soil water budget in order to demonstrate the effects of individual changes. This assessment indicates promising directions being worthwhile a more detailed analysis and tests in the fully coupled COSMO model system.