H11G-0979:
Event-Based Hydrologic Model Calibration using NEXRAD Data in a Southern New Jersey Watershed

Monday, 15 December 2014
Christopher Seigel and Joseph A Daraio, Rowan University, Glassboro, NJ, United States
Abstract:
Proper calibration of hydrologic models requires both reliable observed stream flow and precipitation data. Southern New Jersey has a notable lack of observed precipitation data, in particular, at the event scale; therefore model calibration represents a significant challenge. From a design standpoint this has not been a major concern as hydrologic models have been driven using design storms and are not calibrated. However, the need for research and impact analyses in the face of climate change (changes in extreme precipitation and sea-level rise in particular) and expected population growth in the region requires calibrated models, and reliable observed precipitation data are necessary for model calibration. Several HEC-HMS watershed models were developed and calibrated using NEXRAD data for the upper Maurice River watershed in southern New Jersey, which contains the two growing urban areas of Vineland and Glassboro, to meet three objectives: (1) to validate that NEXRAD data can be effectively utilized in this region, (2) to better understand the roles that spatial variability and scaling play in the use of NEXRAD data within a hydrologic model, and (3) to determine the possible effects of urbanization on stream flow within this watershed. Gridded NEXRAD data were converted to virtual rain gages using the Watershed Modeling System and used as input for HEC-HMS models. Preliminary results indicate that models forced by a spatially coarse distribution (one rain gage for the watershed) of precipitation perform better than models driven by higher resolution precipitation data (14 rain gages for the watershed) for large or extreme events (e.g. hurricanes), and higher resolution precipitation driven models performed better than lower resolution models for smaller events indicating that models should be calibrated from events of the same scale for valid model application.