Landsurface characterization

Soil texture (Figure 20.2(a)) and soil hydraulic properties were extracted from the database developed by the National Cooperative Soil Survey (USDA 1995). The soil column in the land surface model was divided into three, four, or five layers of various depths across the basin according to local geology and type of vegetation cover. The first layer is a thin superficial layer of 8 cm, which functions as the interface between the ground and the atmosphere for the energy balance. The second layer is an intermediate layer expanding throughout the root zone with a varying depth across the river basin as a function of land cover type. The depth of this layer is generally 50 cm for short vegetation (grass and shrubs) and 100 cm (two 50-cm layers) in forested areas, respectively. The bottom layer is a deeper layer extending between the root zone layer and the local water table, or impermeable boundary, the thickness of which varies in the watershed depending on local geology: typically, 50 cm at high altitudes and 100 cm (two 50 cm layers) at low elevations.

Land cover consists predominantly of deciduous trees in the uplands, and short grass and crops at low elevations (Figure 20.2(b)). A small area in the southeastern region of the basin is populated by coniferous, while a narrow band of nearly bare ground can be found along the northeast-southwest direction. Note that the areas identified as bare ground refer actually to agricultural fields that are intermittently cultivated. Although a vegetation dynamics model has been recently incorporated into the LSHM, in the simulations described here changes in vegetation cover are represented dynamically via direct assimilation of indices derived from remotely sensed data. Specifically, leaf area index (LAI) and fractional vegetation cover were estimated by parameterizations using NDVI (Normalized Difference Vegetation Index) data from the Advanced High Very High Resolution Radiometer (AVHRR; Tucker 1979; Jackson etal. 1983).

Figure 20.2 (a) Soil texture distribution at 1-km spatial resolution; (b) Landcover distribution

Figure 20.2 (a) Soil texture distribution at 1-km spatial resolution; (b) Landcover distribution

NDVI data were available at 8-km spatial resolution every 10 days. For use in this case study, the data were downscaled to 1- and 5-km and linearly interpolated in time down to one-hour time intervals. For grass and shrubs, LAI was derived from NDVI according to the parameterization proposed by Choudhury etal. (1994) [LAI = -1.81 ln(1.36 - 2 x NDVI)]. For forested areas, Spanner et al. (1990) was used as reference [LAI = 0.438(NDVI/0.43 8)3 77]. Fractional vegetation cover (F) was estimated using the relationship from Carlson and Ripley (1997) [F = (NDVI)2]. During 1988, LAI ranged from about 2 to 8 in spring and from about 4 to a maximum value of 14 at some locations in summer (Figure 20.3). LAI decreases immediately after the peak in June because of the effect of water stress on vegetation, with occasional increases of LAI at high elevations of the basin during July in response to localized thunderstorms. During 1993, LAI ranged from 2 to 10 in spring and from 4 to 16 in summer (not shown). Starting from the end of May until the end of July, LAI remains relatively high consistent with the annual cycle of deciduous trees in the basin and with the peak of the crop growing season. The space-time variability of LAI reflects the differences in the physical controls of the water cycle between the drought of 1988 and the wet summer of 1993. This

Table 20.2 Surface albedo, roughness height for the atmospheric boundary layer, and Manning's roughness coefficient for overland and channel flow routing






height (m)










Deciduous trees




Coniferous trees




Bare ground






indicates that LAI can be viewed as a proxy of water stress and drought conditions in the Monongahela.

Given the soil and vegetation information, the model parameters for which there was no ancillary data (surface albedo, roughness height, and Manning's roughness coefficient) were extracted from the literature (Table 20.2). Minimum stomatal resistance, a limiting factor in the canopy resistance parameterization during the daytime, is 150 s/m for grass and shrubs and 200 s/m for deciduous and coniferous trees (Devonec and Barros 2002). Note that the parameters had the same values at both 1- and 5-km resolutions. The difference between

11-20 July 21-31 July 01-10 August 11-20 August 21-31 August

20 km

10 12 14 16

Figure 20.3 Leaf area index (LAI) distribution April through August of 1988

the model set-ups stems from the spatial resolution at which the physical processes are resolved, and from the representation of only the terrain.


A total of 48 model simulations (2 climate regimes x 4 rainfall interpolation methods x 2 spatial resolutions x 3 stream networks) were conducted. The hydrologic model was initialized for a period of one month (spin up period) at the beginning of each run in order to allow the state variables to reach internal equilibrium. Each model simulation was performed at one- and five-km spatial resolutions with an hourly time step for a five-month period from April through August. Following Devonec and Barros (2002), the hydrologic model was not calibrated, or specific parameters optimized. Bindlish and Barros (2000) showed that the calibration of model parameters is particularly sensitive to the underlying hydroclimatic regime and spatial resolution. Thus, any optimal set of model parameters cannot be dissociated from the conditions under which calibration took place, which goes against the structural stability requirements enunciated earlier.

Simulations of streamflow hydrographs at the outlets of subcatchments indicated in Figure 20.1 (Table 20.1) are compared against daily streamflow observations. Note that, in the ensuing discussion, the simulated streamflow

refers to total streamflow Q that comprises both overland and subsurface flows, while the simulated subsurface flow Qs includes both interflow and baseflow.

0 0

Post a comment