Print

O'Neil-Dunne, Jarlath P. M.

Rubenstein School of Environment and Natural Resources

2004

M.S.

Nonpoint source pollution from agriculture continues to be a burgeoning problem in the United States. Best management practices such as the use of riparian buffers, can significantly reduce the amount of agricultural runoff and thus the amount of nitrogen, phosphorous, and pathogens reaching aquatic systems. In Vermont, the Natural Resource Conservation Service (NRCS) established a program to provide financial incentives to farmers to develop riparian buffers adjacent to their fields in the Mad River Valley. In order to estimate the cost of the program and to determine prospective fields for buffer establishment, NRCS personnel manually analyzed hardcopy aerial photographs. This approach, however, does not account for geophysical factors such as soil type, slope, and land cover that also significantly and directly influence surface drainage. Furthermore, manual interpretation of panchromatic aerial photographs is far more time consuming when compared to semi-automated feature extraction of multispectral imagery. To aid NRCS in their efforts, a GIS-based method was developed centered on the Revised Universal Soil Loss Equation (RUSLE) that incorporates key factors to identify and prioritize candidate agricultural fields in the Mad River Valley for riparian buffer development. Development and implementation of the RUSLE model within a GIS framework also provides a methodology that can be extrapolated to surrounding regions in a cost effective manner. The model is limited by the resolution and accuracy of the digital elevation model (DEM) and the flow modeling algorithm, both of which impact the ability to accurately predict overland flow. Due to the importance of imagery in determining the presence of existing buffer zones, an assessment comparing the utility of panchromatic digital orthophotography and high resolution multispectral IKONOS satellite data to map land use and land cover was conducted. Object-oriented classification of IKONOS imagery saves time and yields similar results (overall accuracy = 79%) when compared to the manual interpretation of panchromatic digital orthophotographs (overall accuracy = 80%). However, the most accurate method of generating land use and land cover information was through the manual interpretation of multispectral IKONOS imagery (overall accuracy = 91%).