A DISTRIBUTED RAINFALL-RUNOFF MODEL APPLIED TO FORECASTING PEAK FLOWS IN FAST RESPONDING WATERSHEDS IN THE UPPER DELAWARE RIVER BASIN, CATSKILL MOUNTAINS, NEW YORK

Flash floods pose a significant danger to life and property in the complex terrain of the northeast United States. One effective way to mitigate flood risk lies in implementing a real-time forecast and warning system based on a rainfall-runoff model. The Kinematic Runoff and Erosion Model (KINEROS2 - www.tucson.ars.ag.gov/kineros) is a spatially distributed watershed model driven by high resolution, Doppler radar-derived rainfall input. KINEROS2 runs in a real time mode at the National Weather Service Weather Forecast Office in Binghamton, New York. In this study, KINEROS2 was used to evaluate such a system in several watersheds in New York State over a timeframe of several warm seasons.

KINEROS2 provides a temporal and spatial resolution not currently available with other National Weather Service flash flood forecasting models. The computational time steps in KINEROS2 follow the nominal 5 minute interval from the Digital Hybrid Reflectivity (DHR) radar product which has an average 1-degree by 1-km spatial resolution. KINEROS2 can also run ensembles that allow the forecaster to evaluate various scenarios. Differing scenarios can be produced by changing the reflectivity – rainfall (Z-R) relationship associated with the radar data, thereby varying the precipitation amount input to the model. In addition, model parameters can be altered, resulting in a range of watershed hill-slope and channel characteristics. KINEROS2 can be applied to watersheds with or without stream gauges.

Results will be shown from three watersheds in New York State. Both gauged and ungaged watersheds were evaluated. Model parameter variation, both in saturated hydrologic conductivity and channel length, was observed. Parameter variation was correlated with basin averaged rainfall totals. Model calibration based on these two model parameters with reference to basin average rainfall produced a robust set of streamflow ensembles. The ensembles can be run to predict the timing and magnitude of the peak flow from flood flows in small, fast-responding watersheds in complex terrain.