The variability in parameter estimates for a hydrologic model as a function of the number of years of data available for estimating the parameters was investigated. A 25-year record on Spavinaw Creek in Oklahoma and Arkansas was divided into 25 one-year, 12 two-year, and 5 five-year records. Each period was used for parameter estimation. Two objective functions, one based on absolute errors and one based on error sum of squares, were used. Mean parameter values exhibited greater than expected variability, and variances in estimated parameters did not decrease as expected as the number of years used for parameter estimation increased from one to five years. One year was found to dominate one of the estimated model parameters for any record length that included that particular year. Significant differences between parameter values based on the two objective functions were found only for one parameter based on one-year optimizations.

English

Carelia Juarez

Reprints Collection