| dc.description.abstract | Models require as inputs weather data like rain and temperature, and other parameters which must be estimated for the model to function. In hydrologic/water quality models, input parameters are generally represented by average values. However, there is a great deal of uncertainty as to the accuracy of the average values assigned to the parameters used. Naturally if the parameters of a model are uncertain, the output produced by the model will be uncertain as well. Because of the randomness of hydrologic events, one would naturally think about dealing with the random variability of the various model parameters so that their uncertain behavior may be characterized through probability distributions (Ben Salem, 1986). These distributions express the modelers degree of belief that parameter values will be in certain intervals in parameter space. There are two primary means for quantifying this uncertainty. Monte Carlo Simulation and First Order Analysis. One of the problems encountered in Monte Carlo Simulation is the form of the input parameter distribution. In this study the focus will be on the effect of input parameter uncertainty on the uncertainty of model outputs. In fact this study is a continuation of a study conducted by Prabhu (1995). In his study, Prabhu used the AGNPS (Agricultural Non Point Source) model to illustrate a statistical model evaluation protocol. Prabhu used one set of probability distributions for the input parameters; however, he was not certain that the correct distributions had been selected. The uncertainty that results in model outputs is logically dependent on the amount and form of the uncertainty in the input parameters as reflected in the probability distributions of these parameters. In this study the same input parameters used by Prabhu will be employed. To assess the importance of the input parameter distributions, different combinations of input parameter distributions and variances will be used. The impact of changing the distributions and reduction of variances ofinput parameters on the uncertainty of model outputs will be studied. | |
