Improving the accuracy of simple runoff estimates : the role of parameter resolution and data collection

This research project collected, compared, and analyzed rainfall and runoff data from two similar small watersheds within Knoxville's Second Creek drainage area. The primary reason for the study was to determine the most efficient way to increase the accuracy of two well known and simple runoff estimation models: the Rational and SCS Curve Number Methods. This was accomplished by investigating the impact that incorporating different amounts and types of information had on the accuracy of the models. An important component of this investigation was to examine the relative cost-benefit ratios of the different techniques that were attempted.

To optimize the models, the investigation took a three-pronged approach. First, the a priori model parameters and the parameter selection methods were optimized by using increasingly higher-resolution data to characterize the watersheds. The second step of the research was to see by what degree collecting rain and/or runoff data improved model estimations on that watershed. The process began by using the least possible data and incrementally increasing it. The final approach was to investigate whether the measured rainfall-runoff records from one watershed enhanced the estimates on another similar, nearby watershed. Once those steps had been accomplished, a benefit-cost analysis was performed to determine which techniques and data most efficiently improved the models.

For the peak flow estimates, the results showed that fine-tuning the parameters with high-resolution data did not result in better estimates. In fact, for these watersheds the highest resolution parameters produced some of the poorest estimates. Using observed runoff data, however, substantially decreased the estimate errors. Errors were reduced up to 90% using data collected within the watershed. Using data collected from a similar watershed to cross-calibrate the model reduced errors by up to 70%. In addition, the results indicated that more data further improved the estimates. However, the larger amounts of data tended to have a lower benefit to cost ratio.

The results from the volume estimates were not as clear-cut. While all the techniques appeared to work, the evaluation was hampered by the limited observations of storm events. Thus, the first two techniques, the a priori and the calibrated estimates were unreliable. Even so, the third technique, which used data from the similar watershed to calibrate the model, reduced errors by approximately 60 %.

This research provides engineers, hydrologists, and others needing quick and simple runoff estimates with techniques that increase the models' accuracy. This should aid in the sizing of stormwater conveyances, determining mass contaminant loads, making land management decisions, and other actions requiring accurate runoff volumes and peak flows. Because these techniques allow more accurate estimations while maintaining the simplicity and cost effectiveness of the models, it is expected to primarily benefit those in smaller communities, suburban, and rural areas. However, anyone who uses the Rational and SCS Curve Number Methods should find the techniques applicable.

Keywords: Surface Runoff Estimations, Rational Method, SCS Curve Number Method, Representative Watershed, Paired Watershed, Data Collection and Accuracy, Calibration Techniques, A Priori Parameter Resolution, Rainfall and Runoff Data, Calibration Data from a "Similar Watershed", Peak Flow, Volumes, Cross-watershed Calibrations, Time of Concentration, Benefit-Cost Ratio.