Evaluation of E. coli in sediment for assessing irrigation water quality using machine learning

Area of Science:

• Environmental microbiology
• Water quality assessment
• Machine learning applications

Background:

• Contaminated irrigation water poses a significant risk to human health via fresh produce.
• Accurate quantification of E. coli in irrigation water is essential for public health protection.

Purpose of the Study:

• To evaluate the impact of sediment data on machine learning model performance for E. coli quantification.
• To identify key sediment-related factors influencing E. coli levels in irrigation water.

Main Methods:

• Field samples collected from Southwest U.S. irrigation canals.
• Analysis included meteorological, chemical, physical water quality, and sediment properties (E. coli concentration, median size, bed shear stress).
• Machine learning models (Support Vector Machine, Logistic Regression, Ridge Classifier) were trained with and without sediment features using multi-variant filter feature selection.

Main Results:

• Including sediment features improved the correlation with E. coli standards compared to models without them.
• The Support Vector Machine model demonstrated the best performance for both E. coli standards (1 and 126 CFU/100 ml).
• All tested machine learning models showed improved performance when sediment features were incorporated.

Conclusions:

• Sediment properties, specifically E. coli concentration in sediment and bed shear stress, are major determinants of E. coli levels in irrigation water.
• Integrating sediment data into machine learning models is a viable strategy for enhancing the accuracy of E. coli quantification in irrigation water.
• This approach can lead to more effective risk assessments and management strategies for agricultural water safety.