Spatial scale variability in model development and parameterization

Jeffrey George Arnold, Purdue University

Abstract

A distributed parameter, continuous time model was developed to assist water resource managers in assessing water supplies and nonpoint source pollution on watersheds and large river basins. The model operates on a daily time step and can simulate several years of output. The model also allows a basin to be subdivided into hundreds of subbasins. The basin can be discretized into grid cells or natural subwatersheds. A command language was developed to route and add flows down through a watershed. Major components of the hydrologic balance are simulated including surface runoff, lateral flow in the soil profile, groundwater flow, evapotranspiration, channel routing, and pond and reservoir storage. The model that was developed was applied and validated on three different spatial scales; the field, the small watershed, and the river basin. At the field scale, model input data was developed and simulation runs were made using generated weather data for the 329 ha Animal Science watershed. Validation was not attempted since measured flows and precipitation were not yet available. At the small watershed scale, ARS station G (17.7 $km\sp2$) at Riesel, Texas was used for validation of water and sediment yields. The Lower Colorado River basin was simulated and compared to measured USGS streamflow data to test the model on a relatively large river basin (9000 $km\sp2$). Generally, simulated means and standard deviations compare well with measured values. Also, at Station G at Riesel, the Nash-Sutcliffe coefficients of efficiency are between 0.70 and 0.80, indicating a reasonable goodness-of-fit. For the Lower Colorado River Basin, the coefficient of efficiency was 0.65 for annual flows and 0.60 for monthly flow comparisons. These are reasonable values for an uncalibrated simulation. A modeling approach was taken to determine the impact of spatial variability in soils and land use on simulated water and soluble chemical yields. Model users often select a dominant soil and land use for each subwatershed. The impacts of this assumption on the simulated hydrologic response are relatively unknown. The modeling results from this study show that using only the dominant soil to represent an entire subwatershed can result in relatively large errors. However, simulating the dominant five or six soils (50% of the area) consistently gave results within five percent of simulating the entire subwatershed. Lumping land use created errors in simulated water and soluble nutrients even when 70-80% of the area was simulated.

Degree

Ph.D.

Advisors

Engel, Purdue University.

Subject Area

Agricultural engineering|Hydrology|Geography

COinS