Comparison of Methods for Identifying Persistent Spatial Patterns in the Microbial Water Quality of Irrigation Ponds

The microbial quality of irrigation water is a key factor in food safety and security. Farm ponds are common sources of irrigation water and are monitored to ensure the concentrations of E. coli do not exceed threshold values. Results of monitoring may vary substantially depending on where samples are collected. Determining persistent spatial patterns in the microbial quality of pond water appears necessary to reduce the sampling burden while adequately estimating the food safety risks. The objective of this work was to 1) evaluate the persistence in the spatial variability of E. coli concentrations in irrigation ponds, and 2) compare two different methods for characterizing spatial patterns in E. coli concentrations and water quality. Two irrigation ponds in Maryland were each sampled on permanent spatial grids six or seven times in the 2016, 2017, and 2018 growing seasons. In situ sensing of water quality parameters was performed concurrently with sampling. The mean relative difference (MRD) and the Empirical Orthogonal Function (EOF) analyses were applied to deviations of local measurements from the averages across the ponds to identify patterns in the concentrations of E. coli and water parameters. The two analyses showed similarities in ranking locations in the ponds by the deviations of E. coli concentrations from averages. The identified patterns persisted over all three years of observations. The location rankings from the MRD and EOF analyses were highly correlated (Spearman correlation coefficient >0.92). Both analyses revealed significant correlations between patterns of E. coli concentrations and water quality variables. Results of this work show that both methods are suitable for identifying areas in ponds that contain consistently low, average, and high E. coli concentrations with respect to the average concentrations. The MRD method produces more readily interpretable results while the EOF method could reveal more information about the variability.