Streamflow Calibration of Two Sub-basins in the Lake Whatcom Watershed, Washington, Using a Distributed Hydrology Model
Lake Whatcom provides drinking water to the City of Bellingham and portions of Whatcom County. Therefore, quantifying streamflow into the lake is important to establish the contribution of ground water and surface water runoff in the Lake Whatcom water budget. Runoff is nearly 74% of the total inputs to the lake, thus the runoff provides the most water and nutrients to the lake. The primary goal of this study was to determine the ability of the Distributed Hydrology-Soils-Vegetation Model (DHSVM) to simulate the hydrologic processes in two sub-basins of the Lake Whatcom watershed.
DHSVM is a physically based model that simulates a water and energy balance at the scale of a digital elevation model (DEM). GIS maps of topography (DEM), the watershed boundary, soil texture, soil thickness, vegetation, and a flow network define the characteristics of a watershed. The input meteorological requirements for DHSVM include time-series data representing air temperature, humidity, wind speed, incoming shortwave radiation, incoming longwave radiation and precipitation. Meteorologic data were compiled from recent records of a local weather station, except for longwave radiation, which was estimated. I calibrated and validated DHSVM for water years 2002 - 2003, using streamflow records from Austin and Smith Creeks within the Lake Whatcom watershed. Simulations were performed using one-hour time steps and a 30- meter pixel size. Sensitivity analyses were performed with the model to determine the model’s sensitivities and ability to capture hydrologic processes within the watershed by altering soils, vegetation types, and precipitation inputs.
The calibration simulations for WY 2002 had a calibration error of 1% for Austin Creek and -3% for Smith Creek. Both simulation errors are less than the recommended maximum error of +/-5%. The validation was more problematic because of gaps within the recorded streamflow data. However, for the time frame where the simulated flow and recorded flow did overlap, the validation simulation error was -5% for Austin Creek and 3% for Smith Creek.
The sensitivity analyses provided insight into parameter influences. The soil sensitivity simulations in Smith Creek have high mass balance errors indicating that model calculations were not performing adequately. The high mass balance error suggests that the model is over-estimating either the storage or the output.
The vegetation sensitivity simulations did not affect streamflow other than slightly increasing storm peaks. More realistic simulations that capture vegetation removal through deforestation and urbanization would require the use of a road and storm sewer networks within the model to appropriately simulate decreased infiltration and rerouting of storm water runoff.
Additional precipitation gage data added to the model, illustrated an increase in peaks in Austin Creek. Smith Creek did not have the increase in peaks, primarily due to the distance from the precipitation gage at Brannian Hatchery. The overall streamflow in Austin Creek did not increase with the addition of three precipitation gages to the input file, although the volume of storm event peaks did increase. I also simulated streamflow for Austin Creek and Smith Creek with two other interpolation methods (INVDIST and VARCRESS) using the additional precipitation gages. The INVDIST interpolation method provided the greatest increase in both Austin Creek and Smith Creek, again primarily increasing peak volumes with little change in base flow.
Future efforts should focus on modeling the individual subbasins, rather than attempting to model the entire Lake Whatcom watershed. The heterogeneities between the individual sub-basins are captured by DHSVM which increase the difficulty in modeling the entire watershed.