Modeling Stream Flow Using GIS

Abstract

The Delaware Water Gap National Recreation Area (DEWA) would like to utilize hydrologic modeling coupled with GIS to help with the prediction of water quality changes in the watersheds entering the upper Delaware River. The first step towards completing this goal is to create a model that can accurately predict flow. The hydrologic model SWAT was used to model flow in the Broadhead watershed for DEWA by the Center for Earth Observation (CEO) at North Carolina State University (NCSU). The Broadhead watershed is located in North Eastern Pennsylvania and flows through DEWA on its way to the Delaware River.

Based on limited data and the criteria set forth by DEWA, SWAT was chosen from 11 different models as best suited to meet DEWA's needs. The data used to run the SWAT model included a 30-meter DEM, STATSGO soils data, a USGS landuse/ landcover map, and daily weather data from January 1, 1993 through October 20, 1999. The data used to calibrate the model consisted of flow data from two USGS gage stations, Minisink Hills and Anamolink, which are located within the Broadhead basin. The flow data from the two USGS gage stations were separated into surface flow and base flow using the USGS model HYSEP.

The Broadhead basin was separated into ten sub-watersheds. Two sub-watersheds contributed to the Anamolink catchment and eight sub-watersheds contributed to the Minisink catchment (the Minisink catchment contains the Anamolink catchment). The Hydrologic Response Units (HRUs) were set to include soil types and landuse/ landcover types greater than or equal to 5% of the sub-watershed area. The calibration period of the model was run from January 1, 1993 to December 31, 1995, and the validation period was run from January 1, 1996 through October 20, 1999.

The model was run on both an annual and monthly time step. For the monthly time step the model was tested for both winter and non-winter months. The model predicted total flow on an annual time step within 16% of observed flow for the Anamolink basin, and within 18% of observed flow for the Minisink basin. However, more data and calibration is required to achieve the goal of predicting flow on a monthly time step.