Using Neural Networks to Create New Indices and Classification SchemesEPA Grant Number: R829784
Title: Using Neural Networks to Create New Indices and Classification Schemes
Investigators: Brion, Gail M. , Lingireddy, Srinivasa
Institution: University of Kentucky
EPA Project Officer: Page, Angela
Project Period: July 1, 2002 through June 30, 2005 (Extended to June 30, 2006)
Project Amount: $523,938
RFA: Microbial Risk in Drinking Water (2001) RFA Text | Recipients Lists
Research Category: Water , Drinking Water , Health Effects
Neural network models can reliably relate shifts in indicator and indigenous bacterial populations to the presence, concentration, age, and source of microbial pathogens. Based on this hypothesis, the proposed project aims at developing a neural network tool that provides early warning of potentially "risky" conditions in source waters for drinking water treatment plants by identifying important combinations of commonly measured indices (physical, chemical, and microbial) related to pathogen presence and concentration.
The proposed project will create a robust, multi-year, multi season, multiparameter database of water quality at a single intake on the Kentucky River that contains new indices relative to the age and source of fecal material. This database will also contain measures of potential pathogens and related or suggested indices. Conventional statistical analysis and advanced neural network analysis will be applied to the resultant database to uncover (a) relationships between pathogen presence and indices, both new and conventional, (b) predictive models that take into account watershed characteristics such as rainfall, seasonality, and flow patterns, (c) classification schemes that indicate the predominant sources, and relative ages, of fecal contamination under varying conditions. In concert with this data collection and modeling effort will be laboratory scale survival studies to elucidate the relationships between pathogens and indices survival in natural waters. The information from the survival studies will provide decay models to be linked with the neural network modeling efforts for river water quality data to accurately track source and age of pollution.
The proposed project is expected to result in a single, long-term, controlled study of the ecology and survival of mixed bacterial populations and their relationship with pathogens. The information (data and data analyses/models) from this study will provide a scientific basis to perform qualitative risk assessments of pathogen inputs into the identified water source. This research will provide data on the relationships between pathogens and other indices that are expected to provide early warning systems for use at all surface water treatment plants and by all watershed managers.