Abstract:
The organic layer of evergreen tropical mountain forests plays a constitutive role concerning the water balance in these ecosystems. Knowledge of the hydraulic conductivity function and the water
retention function permits the description of this water balance by application of the Richards equation (Richards, 1931) for liquid flow in unsaturated porous media. For this purpose ten samples of
the organic layer, taken in a mature tropical mountain forest in South Ecuador, were collected and treated in laboratory multi-step irrigation experiments. By applying inverse modelling techniques on the observed drainage data, the soil hydraulic Mualem-van-Genuchten parameters were derived, using three different model approaches (van Genuchten, 1980; Durner et al., 1999; ?imůnek et al., 2003). Calculations of the Richards equation based on these parametersets were congruent with the observed drainage hydrographs. The estimated parameters showed only low variation in case of the
Mualem-van-Genuchten model (van Genuchten, 1980), whereas for the two dual-porosity models (Durner et al., 1999; ?imůnek et al., 2003) some parameters scattered considerably. Problems concerning
nonunique solutions were assessed by dint of response-surface analysis as well as a sensitivity study of variations in the initial parametersets. It became clear that the observed drainage hydrograph may be explained by multiple different parametersets. Thus, a physical interpretation of the optimized parameters must be neglected. Simulations run both on optimized parametersets and
parametersets averaged over all samples were cross-validated against alternative irrigation experiments. Each model was able to predict the observed drainage hydrograph well. Shifts with increasing
irrigation rates were problematic, which can be explained by hysteresis effects. In this range best predictions were achieved by the dual-porosity model proposed by ?imůnek et al. (2003). However, this model lacks accuracy in regions of high irrigation rates and large steps between irrigation rates, where the other models presented in this work perform better. For the purpose of simulating water flow at this site it is proposed to use the averaged parameterset of the Mualem van Genuchten model, since it is distinguished by an acceptable simulation power but low parametrization.
