Watershed-Scale Hybrid Stochastic-Deterministic Modeling Framework and Diffused Sources Superpositioning
Date of Award
Doctor of Philosophy (PhD)
Civil & Environmental Engineering
Predicting hydrologic system behavior is imperative to planning and management of water resources. The study developed an integrated hybrid stochastic and deterministic framework to improve prediction accuracy for overland flow and diffused sources in a watershed. The methodology includes sampling input parameters at system level and contribution of nonpoint source from hydrologically disconnected areas (heretofore referred to as system-level approach and superpositioning respectively). System-level approach includes the integration of a topography-based sampling grid generalized linear model developed by the study and Monte Carlo methods. The superpositioning method adopts in-stream water quality equation for overland flow pollution estimation.
The system-level approach was applied to the Patuxent watershed to determine runoff, phosphorus and total suspended solids using continuous rainfall. For overland flow, system-level approach (p-value of 0.68) was 0.51% off the observed flow compared with -21.9% for existing method ( p-value of 0.11). Similarly for phosphorus, the model prediction deviated from the observed by 7% compared to that of the existing method which deviated by -32%. The results indicate that the system-level method is a better predictor for overland flow and nonpoint sources. In the superpositioning approach, phosphorus contributions were added to the system-level approach using an event rainfall. The prediction error reduced from 4.82% to -0.29% when the system-level method was superpositioned with nonpoint source. Data from superpositioning analysis showed that including diffused sources contribution from hydrologically disconnected areas further improves the level of accuracy.
The study demonstrates that the framework reduces prediction error and has a high accuracy in reproducing watershed response. The hybrid methodology framework is superior to existing deterministic methods. Ultimately, this dissertation shows the potential of improving prediction accuracy of hydrologic systems by incorporating the strengths of both stochastic and deterministic models. The framework serves as a background for detailed applications for the developed models.
Damalie, Ruby J..
"Watershed-Scale Hybrid Stochastic-Deterministic Modeling Framework and Diffused Sources Superpositioning"
(2013). Doctor of Philosophy (PhD), Dissertation, Civil & Environmental Engineering, Old Dominion University, DOI: 10.25777/sd4a-2358