Project ID: 2004FL59B

Title: Ground Water Vulnerability Delineation Using Integrated GIS and Neuro-Fuzzy Methods

Project Type: Research

Focus Categories: Models, Nitrate Contamination, Groundwater

Keywords: Ground Water Modeling

Start Date: 03/01/2004

End Date: 04/28/2005

Federal Funds: $38,817

Non-Federal Matching Funds: $113,420

Congressional District: 10th

Principal Investigator:
Barnali Dixon

Abstract

Assessment of ground water (GW) vulnerability or delineation of vulnerable areas for monitoring purposes is difficult since GW contamination depends upon numerous, complex interacting parameters. Statistical correlation between a set of causal factors that potentially contribute to the contamination often does not produce acceptable results. More accurate GW vulnerability estimates can be delineated from site-specific studies; however, these studies are quite expensive and not feasible at all well sites. Therefore, there is a need to develop techniques that (i) will provide reliable GW vulnerability estimates at varying watershed scales, (ii) require less extensive site-specific data, and at the same time, be robust when data are uncertain and incomplete, and (iii) can be updated easily.
The objectives of this research are to innovatively extend the mapping of GW vulnerability at a large watershed scale by developing and adapting a hybrid method of Neural Networks (NN) and Fuzzy Logic (FL) known as a Neuro-fuzzy model in a GIS platform. This project also will characterize spatial variabilities of parameters critical to GW vulnerability using geostatistical approach. NN is a multi-input and multi-output model that develops a non-linear relation between input and observed output parameters. FL is a fuzzy rule-based system that minimizes error propagation and incorporates expert opinion. Recently, industrial applications show that NN and FL complement each other. Therefore, we propose to develop an integrated tool that incorporates a hybrid system of NN and FL called a Neuro-fuzzy model to predict GW vulnerability to contamination. We have access (through the Florida Aquifer Vulnerability Assessment (FAVA project) to data sets for 100 wells in Hillsborough and adjacent counties of the Southwest Florida Water Management District. A statistical comparison of the outputs from the models and kriged surfaces will be conducted with the field data for model validation. This research should contribute to the development of a robust and economically feasible tool for mapping GW vulnerability. Once the mapping tool is developed, it will be available via the Internet, and the methodology can be extended to other watersheds. City and county planning councils may incorporate resultant vulnerability maps into future urban growth planning and simulation to facilitate urban growth and development more effectively while protecting our water. Watershed managers and county extension agents may use this map to implement Best Management Practices (BMPs).

Progress/Completion Report PDF