Simple Solutions for Improving Urban Peak Flow Estimates at the Watershed Scale

Peak flow rates are a critical hydrologic variable. They are used to size urban drainage infrastructure, predict flooding, serve as inputs to erosion and sediment transport models, size and gauge the effectiveness of treatment systems such as stormwater control measures, and estimate daily runoff simulations in support of regulations. Therefore, use of peak flow models that produce large errors can result in severe consequences, regardless of whether peak flows are over- or under- estimated. The two most popular peak flow models in the United States are the rational method (RM) and the Natural Resources Conservation Service (NRCS) Technical Release 55 (TR-55) method. In order to reduce errors from modified versions of these models across storm sizes and intensities that are more frequently observed, multiple low-cost, simple adjustment and calibration approaches were incorporated into the models, and their outputs were compared to observed peak flow rates in two adjacent watersheds with similar attributes. Tested approaches included (1) adjustments by iteratively increasing the resolution of soil, land use, and topographical data to better calculate model input variables; (2) direct calibration using experimentally determined ratios (ratio of sum of modeled peak flows to sum of observed peak flows) from rainfall-runoff data collected within the watershed of interest from a variety of events that were randomly subsampled; and (3) indirect calibration using experimentally determined ratios from field collected rainfall-runoff data from a similar neighboring watershed for a variety of events that were randomly subsampled. Peak flow estimates from modified versions of the RM (QRAT-MOD) and NRCS TR-55 method (QNRCS-MOD) parametrized with higher resolution data did not necessarily result in better model performance and did not produce any suitable model fits as defined by a series of goodness-of-fit metrics. Both the QRAT-MOD and QNRCS-MOD models were substantially improved after applying direct calibration ratios, with the resulting calibrated models providing suitable accuracy in most scenarios. Furthermore, many QNRCS-MOD models were also considered suitable after applying indirect calibration ratios calculated from the adjacent watershed. However, this was not the case for QRAT-MOD models when applying indirect calibration ratios, though all goodness-of-fit metrics improved substantially. Results suggest local peak flow estimations resulting from more frequent events can be greatly improved by implementing a single flow monitoring station to collect peak flow calibration data for a relatively small number of events. Though not explored in this analysis, it is likely that these approaches can also improve peak flow estimations for larger, design-sized events.