Wednesday, 17 January 2007
Distributed hydrological modeling in a micro-scale rainforest catchment in Central Amazonia, Brazil: DHSVM calibration
Exhibit Hall C (Henry B. Gonzalez Convention Center)
The Asu catchment is the first and best instrumented rainforest watershed in the Amazon basin. It has a second-order blackwater stream draining an area of 6.5 km2 entirely covered by pristine rainforest. Practically all components of the land surface hydrological cycle have been monitored in real time since 2002. The site includes a fluxtower whose footprint coincides with the Asu catchment area. The DHSVM (Distributed Hydrology Soil Vegetation Model) offers a dynamic representation of evapotranspiration, soil moisture and runoff generation, in the same spatial scale of the digital elevation model (DEM). We validated the Shuttle Radar Topographic Mission DEM data of the area (SRTM 90 m x 1 m horizontal and vertical resolutions respectively) with field topographic data. The validation was necessary because the SRTM data refers to the topography of the vegetation canopy, and the forest can mask some ground topographic features of significance to hydrology. We then resampled this data to increase horizontal resolution to 30 m, using it thereafter as the basis for all DHSVM runs. The DHSVM has not yet been calibrated to tropical rainforest conditions, therefore we used our hydrological field data for developing a first calibration. Because our data is spatially discrete and not evenly distributed on the catchment, we have carried out a careful data interpolation and field verification. For doing that we've developed a new algorithm that uses the same DEM to generate equipotential surfaces with hydrological meaning, and then proceeded to verify its validity through griding the area and checking ground water along the grid. The robustness of this interpolated data gave us confidence to use it in the calibration of the DHSVM. This detailed study of the Asu catchment using a micro-scale hydrological model will generate essential parameters to integrate into larger scale models. Scale aggregation has been a difficult issue to tackle, especially because thus far there have been not much detailed field data in finer scales. Better represented hydrological processes will certainly create an impact in numerical models of weather and climate, which today rely solely on surface schemes like SSib, where lateral water transfers among grid cells are not represented. We believe that substituting surface parameterized schemes by fully functional surface hydrological models will increase the capacity of climate models to predict soil water and its connections to the land-atmosphere interactions.