On evaluating the validity of continuous, distributed hydrologic model predictions in spatially heterogeneous Hortonian watersheds
Date of Completion
To improve the usefulness of distributed hydrologic models as effective prediction tools, a detailed evaluation of the validity of distributed hydrologic model predictions in spatially heterogeneous watersheds is necessary. In this study, the distributed-parameter, two-dimensional, hydrologic model, CASCC2D is converted from an event-based model into a continuous one by adding a new parameterization for the simulation of the evolution of soil moisture. This model is calibrated using a 40-day flow record from the outlet of the Goodwin Creek watershed in Mississippi. Areas of the watershed with similar land-use and soil texture properties are changed uniformly during the calibration process. Four validation tests are conducted to test the applicability of continuous CASCC2D for the simulation of catchment dynamics at ungaged internal locations. Results of these tests reveal that the simulations of continuous CASC2D are statistically comparable to the runoff observations at Goodwin Creek. To compare the effects of parameter uncertainty of distributed hydrologic model predictions with those of a lumped model, two other single-event validation tests and a Monte Carlo uncertainty analysis are conducted by using HEC-1 and CASCC2D Results of this comparative study reveal that, in general, CASC2D model predictions are less uncertain than those of HEC-1. These results also show that Monte Carlo uncertainty analysis causes significant biases in expected runoff volume. A more detailed Monte Carlo uncertainty analysis is conducted using continuous CASC2D to investigate the effect of increased spatial heterogeneity of saturated hydraulic conductivity (Ks) on model predictions. Results of this investigation confirm that increased spatial heterogeneity of Ks leads to an over-prediction in expected runoff volume and peak discharge. A hypothetical catchment with an open-book type of configuration is used to investigate the effect of this bias on model predictions under different combinations of rainfall and initial soil-moisture conditions, and coefficients of variation of Ks. Results of this study show that intense, short-duration rainfall-events in watersheds with highly spatially-varied Ks-values lead to significant increases in sampling biases. These findings also show that the coefficient of variation of Ks has the most significant impact on sampling biases. ^
Senarath, Sharika Upendra Sedara, "On evaluating the validity of continuous, distributed hydrologic model predictions in spatially heterogeneous Hortonian watersheds" (2000). Doctoral Dissertations. AAI9997206.