Canonical Variable Selection for Ecological Modeling of Fecal Indicators

Abstract

More than 270,000 km of rivers and streams are impaired due to fecal pathogens, creating an economic and public health burden. Fecal indicator organisms such as Escherichia coli are used to determine if surface waters are pathogen impaired, but they fail to identify human health risks, provide source information, or have unique fate and transport processes. Statistical and machine learning models can be used to overcome some of these weaknesses, including identifying ecological mechanisms influencing fecal pollution. In this study, canonical correlation analysis (CCorA) was performed to select parameters for the machine learning model, Maxent, to identify how chemical and microbial parameters can predict E. coli impairment and F+‐somatic bacteriophage detections. Models were validated using a bootstrapping cross‐validation. Three suites of models were developed; initial models using all parameters, models using parameters identified in CCorA, and optimized models after further sensitivity analysis. Canonical correlation analysis reduced the number of parameters needed to achieve the same degree of accuracy in the initial E. coli model (84.7%), and sensitivity analysis improved accuracy to 86.1%. Bacteriophage model accuracies were 79.2, 70.8, and 69.4% for the initial, CCorA, and optimized models, respectively; this suggests complex ecological interactions of bacteriophages are not captured by CCorA. Results indicate distinct ecological drivers of impairment depending on the fecal indicator organism used. Escherichia coli impairment is driven by increased hardness and microbial activity, whereas bacteriophage detection is inhibited by high levels of coliforms in sediment. Both indicators were influenced by organic pollution and phosphorus limitation. Core Ideas Maxent models were developed to infer source and transport of two fecal indicators. Canonical correlation analysis was applied to select parameters for Maxent models. Canonical selection and sensitivity analysis improved E. coli models. Elevated concentrations of E. coli are associated with hardness and heterotrophic activity. Bacteriophage detection is inhibited by sediment coliforms and enzyme activity.

Storm sampling to assess inclement weather impacts on water quality in a karst watershed: Sinking Creek, Watauga watershed, East Tennessee

“…Furthermore, the presence of OBs at the spring site suggests an anthropogenic wastewater (septic tank effluent) source for some of the spring water. Although only one of two samples at the spring site was positive for OBs, these preliminary results support the findings of prior research at Sinking Creek that proposed failing septic systems as one source of E. coli in the creek (Dulaney, 2003). To further evaluate residential wastewater as a source of recharge for the spring, it is recommended that OB sampling be repeated for more storm events to determine the conditions under which OB is present in the stream and spring.…”

“…It has a humid subtropical climate (Köppen Cfa climate) (Kottek et al, 2006), with average annual rainfall of 104 cm (41 in) (National Weather Service, 2020 2003identified locations with high concentration of fecal coliform bacteria. The greatest concentration was measured at sites with the largest amount of agricultural land cover, with a geometric mean range of 271-1,767 cfu 100 ml −1 ; mitigation strategies such as creating vegetated riparian buffers zones to limit transport of fecal coliform to streams were suggested (Dulaney, 2003). Floresguerra 2003measured E. coli, nitrates, and OBs to locate nonpoint sources of pollution.…”

“…Luffman 2016 Models predicting the presence of fecal indicators using ecological drivers in Sinking Creek were developed in Max-Ent (maximum entropy modeling) to understand whether chemical and microbial variables could predetermine E. coli impairment (Gilfillan, Hall, et al, 2018;Gilfillan, Joyner, et al, 2018). Hall et al (2014) attempted to confirm the presence of E. coli O157:H7 at Sinking Creek but were unsuccessful, likely due to interference of the soil water chemistry.…”

“…Water flow paths or origin may be determined using dye tracing, which has been used in several groundwater dye tracing studies in northeast Tennessee (Doyka, 2017;Wilson et al, 2016). Combined with geologic mapping and geospatial analysis similar to the methods of Burnham et al (2015), dye tracing may be a suitable method to identify the water source for the spring at Sinking Creek.…”

Microbial Water Quality: Monitoring and Modeling

“…There is often a need to select and compare statistical techniques that can reduce the number of predictors and create more‐robust predictive models (Singh et al, 2004; Truchado et al, 2018). One such technique—canonical correlation analysis—was applied as the input‐selection procedure for a machine learning model in the work of Gilfillan et al (2018), aimed at identifying potential drivers of the impairment of microbial water quality using E. coli and F+ and somatic bacteriophages as indicators in the mixed‐use stream in East Tennessee. The input selection led the authors to the conclusion that E. coli impairment is driven by run‐off and erosion, whereas bacteriophage detection is inhibited by high levels of coliforms in sediment.…”