Use of Artificial Neural Networks for Agricultural Chemical Assessment of Rural Private Wells

by Chittaranjan Ray,
Kristopher K. Klindworth,

Document Type: Proceeding Paper

Part of: North American Water and Environment Congress & Destructive Water


The detection of agricultural chemicals in rural private wells has prompted for numerous studies at regional, state, and national levels. A principal concern arises from the fact that private wells do not come under any regulatory compliance for periodic testing of the pumped water. Numerous factors contribute to the occurrence of these chemicals in ground water: pesticide use, distance of the well to the nearest crop land, depth to uppermost aquifer material from land surface and the overall well depth, distance to other potential sources of contamination, timing and intensity of rainfall with respect to pesticide application, the location of possible accidental spills and improper disposal of containers, and soil hydrologic, geochemical, and biological factors. The complex interaction of these factors makes it difficult to model the system for predicting the vulnerability of a well for contamination. An approach, using artificial Neural Networks (NN), has been taken to identify factors that contribute to contamination of the well. At present, data from 192 drilled and driven wells and 123 dug and bored wells are available for training and testing purposes. The NN has a Feed Forward architecture with hidden layer(s). For training purposes we will separate the well types (drilled/driven versus dug/bored) and the pesticide types (herbicides and insecticides versus nitrate). The data are currently being worked upon.

Subject Headings: Pesticides | Soil pollution | Groundwater pollution | Chemicals | Wells (water) | Neural networks | Training

Services: Buy this book/Buy this article


Return to search