Effect of land surface representation on forest water and carbon budgets

Forested landscapes often show very well-pronounced heterogeneity in the factors that control evapotranspiration, runoff production and carbon assimilation at a variety of length scales. In hilly or mountainous environments, strong contrasts in net radiation, available soil water, soil structure and stand characteristics can produce a large variance in both the meteorological drivers and surface resistance to carbon and water exchange with the atmosphere over distances measured in tens of metres. Because of the strong nonlinearities characterizing the influence of the environmental variables on surface resistance (particularly available soil water), the parametrization of surface process models with mean values of the environmental variables and no distribution often leads to significant bias in areal average carbon and water flux. However, it is often not feasible to incorporate directly the full distribution and patterns of the landscape for regional-scale models. Continental- and subcontinental-scale vegetation data sets currently being collected by synoptic-level satellites (e.g. the Advanced Very High Resolution Radiometer, AVHRR) do not capture the large proportion of landscape variability that exists below the resolution of the sensors. This paper explores the impacts of various landscape representation schemes that retain a range of detail in the description of land surface form and processes on simulated areal average evapotranspiration, runoff production and net carbon exchange with the atmosphere. Specific comparison is made of schemes that attempt to incorporate the topographic structure, soil and vegetation distributions of a region with schemes that sample the surface at levels similar to current coarse-resolution satellites. For strongly heterogeneous basins (mountainous topography), it is found that spatial variations in available soil water can have significant effects on areal averaged carbon and water flux rates, particularly under drying conditions, whereas the spatial variations in radiation, temperature and humidity over the terrain appear to have a lesser impact.