Scale of analysis drives the observed ratio of spatial to non-spatial variance in microbial water quality: insights from two decades of citizen science data

Abstract

While fecal indicator bacteria (FIB) testing is used to monitor surface waters for potential health hazards, recent studies report substantial variation in FIB levels and that observed variation appeared dependent on scale of analysis (SOA). Citizen science data and random effects models were used to quantify variance in FIB levels attributable to spatial versus temporal factors. Separately, Bayesian models were used to quantify the ratio of spatial to nonspatial variance in FIB levels, and identify associations between environmental factors and FIB levels. Separate analyses were performed for each stream, lake, and watershed as well as at the state-level (three SOAs). As SOA increased (from waterway to watershed to statewide models), variance attributable to spatial sources generally increased and variance attributable to temporal sources generally decreased. While relationships between FIB levels and environmental factors, such as flow conditions, were constant across SOA, the effect of land cover was both highly dependent on SOA and consistently smaller than the effect of outfalls. This study demonstrates the importance of scale when designing monitoring programs or using FIB monitoring to inform management. Moreover, these data represent a comprehensive survey of water quality in Upstate NY, and the study highlights a valuable publicly available database.

Methodological differences between studies confound one-size-fits-all approaches to managing surface waterways for food and water safety

“…A 2020 review of agricultural water in the Southeastern United States noted that the geographical location of a water source played an important role in the prevalence and survival of foodborne pathogens ( 35 ). This conclusion is supported by other studies that compared microbial water quality between growing regions nationally ( 16 , 37 ) and locally ( 12 , 17 , 38 ), and between water types ( 32 , 39 – 41 ). Indeed, multiple studies have shown that this variability in water quality limits the efficacy of one-size-fits-all approaches to monitoring and mitigating microbial hazards in aquatic environments.…”

Structural equation models suggest that on-farm non-crop vegetation removal is not associated with improved food safety outcomes but is linked to impaired water quality

“…Datasets. This study used historical data from 16 peer-reviewed papers and two citizen science databases where the original source was willing to share the complete dataset (12,(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45)(46). Specifically, we compiled data from the Northeastern USA that reported (i) the presence or absence of foodborne pathogens in soil, water, vegetation, or wildlife feces, or (ii) dissolved oxygen, fecal indicator bacteria, chloride, conductivity, nutrient, salinity, or sediment levels in streams.…”

“…Data were available on water quality outcomes from published studies (32)(33)(34) and citizen science datasets, including outcomes with food safety implications [e.g., fecal indicator bacteria (FIB) levels]. Several of the water quality datasets [e.g., (35)(36)(37)] overlap spatially and temporally with surveys that tested farm and other environmental samples for foodborne pathogens [e.g., (12,32,(38)(39)(40)]. Since we compiled datasets from multiple sources, it was important to determine if methodological differences between datasets would swamp signals of interest.…”