Modelling the Spatio-temporal Process Variability of Soil Water Balance in a Forest

In terrestrial forest ecosystems the partitioning of precipitation into evapotranspiration, runoff (recharge) and soil moisture change is influenced by the forest canopy. At the scale of a forest stand, factors such as tree species composition, crown architecture, and different canopy interception results in spatial variable evapotranspiration and soil moisture patterns. Both processes directly affect the recharge and runoff in space and time. To accurately predict the runoff from forested ecosystems, we need to know how much of this spatial process variability needs to be represented in our models. In this study, we model runoff and recharge from a forest stand based on spatial uniform values of precipitation, evapotranspiration and compare it to model results that include spatial variation and variation plus spatial correlation. In a mixed European beech-Norway spruce stand spatio-temporal variation of volumetric water content (VWC) was measured by TDR (time domain reflectometry) at 194 locations across 0.5 ha plots with permanently installed 30 cm vertical wave guides. Measurements were repeated 28 times before and after rewetting periods during the vegetation seasons in 2000 and 2001. Additionally, the locations of all trees within the plots were recorded. Geostatistics was used to describe spatial correlation between VWC measurements stochastically and to interpolate recharge patterns in space. Spatial patterns of recharge were analyzed according to antecedent soil moisture content, tree species locations, and canopy throughfall. We used measured surface and subsurface runoff from sprinkling experiments (50 m) at the same site to calibrate and validate hydrological parameters in our model. Runoff for the entire site was modeled in three different ways: first by spatially uniform precipitation, soil water content and evapotranspiration, second by incorporating the spatial variation and third by also considering the spatial correlation in the input variables. The model performance was tested by comparing model results with measured spatio-temporal change in soil water storage and evapotranspiration. Results of the geostatistical analysis show for example that spatial correlation and variation (expressed by the coefficient of variation) is not constant in time but shows a hysteresis that depends on the wetting and drying history of the site. Spatial patterns of transpiration are reflecting the spatial distribution of tree species (beech and spruce have different strategies in water consumption). Including these spatial patterns in the soil water content as a measure to describe the connectivity of wetter areas events improves the model performance (in respect to runoff) in particular during periods of high precipitation. Consideration of spatial process variability not only improves the overall model performance but also helps to better understand the interaction of soil science, plants and hydrology.