Implementation of a Unified Methodology to Infer Parameter Distributions and Evaluate the Predictive Uncertainty of a Model Used to Simulate Sediment Transport and Water Quality in a Watershed

摘要

Analysis and prediction of environmental impact from the transport of contaminated sediments in a watershed involves model selection, data collection, model calibration and verification, model application, and evaluation of uncertainty int he results. A unified methodology has been developed and implemented that uses several numerical tools that, when applied together, allow one to infer the distributions of model parameters and assess the uncertainties associated in the predictions of a nonlinear hydrological model. This study starts from the premise that the several sets of paramenter that for particular storms or simulation periods predict the environmental response, should be considered to be composed of parameter values that define the parameter distributions. Inferred parameter distributions are used to perform uncertainty analysis through Monte Carlo sampling. The uncertainty analysis performed here showed that the covariance in the predictions of both, adsorbed pollutant and sediment flux is about 300percent, indicating the limited predictive capability of these components by the model.