Fuzzy-logic based approach for E. coli load prediction in cascading dams

Authors: ABIMBOLA, OLUFEMI - University Of Nebraska; MITTELSTET, AARON - University Of Nebraska; MESSER, TIFFANY - University Of Nebraska; Berry, Elaine

Submitted to: ASABE Annual International Meeting
Publication Type: Abstract Only
Publication Acceptance Date: 2/25/2019
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: In recent times, Escherichia coli has attracted special attention in environmental research due to its microbiological impairment of water. E. coli in surface water is governed by several physical, chemical, and biological factors. Several models have been developed to predict E. coli loads in surface water, but they tend to be site or system specific or too complex with too many unknown input parameters. As a result, modeling the transport of E. coli in surface water is challenging with predictions yielding high uncertainty. Machine-learning algorithms use computational methods to learn information directly from data without relying on a predetermined equation as a model. In order to better predict E. coli loads in two cascading dams, this study aimed to examine the use of machine learning with physical, chemical, and biological input variables along with their interactions. Sampling was done at the US Meat and Animal Research Center (USMARC), Nebraska. Comparisons of a traditional model (LOADEST, USGS) and an adaptive neuro-fuzzy inference system (ANFIS) with different clustering methods were conducted. Precipitation and pasture management were also included as features in addition to water temperature, phosphate, nitrate and ammonia to improve the prediction accuracy. The model results are compared to estimated E. coli loads based on available water-quality data at USMARC, Nebraska. Cross-validation and model performance measures indicated that overall E. coli load predictions by the ANFIS models are better than the E. coli load predictions by the LOADEST model. Hence, the ANFIS models can be used to effectively identify targets for intervention plans and monitoring programs and implement regulatory standards through Total Maximum Daily Load programs.