Peakflow prediction using an antecedent precipitation index in small forested watersheds of the Northern California Coast Range

The vast majority of small watersheds in Northwest California lack stream gauge information. Understanding the high flow behavior of these watersheds is crucial for guiding resource managers in project planning. The purpose of this thesis was to develop a predictive relationship between precipitation and peakflow of streams draining small forested watersheds of the Northern California Coast Range. An antecedent precipitation index approach was developed for this purpose. The five selected watersheds are covered by coastal coniferous forests with drainage areas ranging from 0.4 to 34 km2. Streamflow and precipitation data from the South Fork of Caspar Creek was used to create the calibration model. Data from the North Fork of Caspar Creek, Hennington Creek, Little Lost Man Creek, and Freshwater Creek were used for independent model testing. The calibration linear regression model, predicting peakflow as a function of peak antecedent precipitation index, resulted in a r2 of 0.83 and a residual standard error of 1.20 L s-1 ha-1. When peakflow was predicted, using precipitation data from test watersheds, the results were fair to poor with average absolute prediction errors ranging from 28.6 to 66.3 percent. When the ten largest peakflows were predicted separately, the average absolute prediction errors were significantly lower at 10.2 to 44.9 percent. The model was positively biased at all test watersheds except Freshwater Creek. The root mean square error was within 15 percent of the calibration residual standard error at all test watersheds except Little Lost Man Creek. The variability in prediction accuracy could be explained by changing unit-discharge relationships, heterogeneous lithologies, different cumulative land management effects, and spatial variation in precipitation intensity. Prediction errors were the greatest for the smallest peakflows, which may be due to greater variation in interception rates during small rainfall events. The antecedent precipitation index approach outlined in this study is best suited for predicting larger rather than smaller peakflow events that may be influenced more by factors other than short-term rainfall history.