A fecal score approximation model for analysis of real-time quantitative PCR fecal source identification measurements

Numerous qPCR-based methods are available to estimate the concentration of fecal pollution sources in surface waters. However, qPCR fecal source identification data sets often include a high proportion of non-detections (reactions failing to attain a prespecified minimal signal intensity for detection) and measurements below the assay lower limit of quantification (minimal signal intensity required to estimate target concentration), making it challenging to interpret results in a quantitative manner while accounting for error. In response, a Bayesian statistic based Fecal Score (FS) approach was developed that estimates the weighted average concentration of a fecal source identification genetic marker across a defined group of samples, mathematically incorporating qPCR measurements from all samples. Yet, implementation is technically demanding and computationally intensive requiring specialized training, the use of expert software, and access to high performance computing. To address these limitations, this study reports a novel approximation model for FS determination based on a frequentist approach. The performance of the Bayesian and Frequentist models are compared using fecal source identification qPCR data representative of different 'censored' data scenarios from a recently published study focusing on the impact of stormwater discharge in urban streams. In addition, data set eligibility recommendations for the responsible use of these models are presented. Findings indicate that the Frequentist model can generate similar average concentrations and uncertainty estimates for FS, compared to the original Bayesian approach. The Frequentist model should make calculations less computationally and technically intensive, allowing for the development of easier to use data analysis tools for fecal source identification applications.