Prediction of boundary and Stormwater E. Coli concentrations using river flows and baseflow index

E. coli (EC) concentrations of the upstream boundary, tributaries, and stormwater in the lower Passaic River at Paterson, New Jersey, were modeled using multivariate polynomial regression (MPR). Baseflow indexes (BFIs) and river flows from upstream and downstream boundaries of the study area were used as predictors. The MPR models were developed by stepwise addition of the candidate terms. The candidate terms were selected based on their t-statistics and the final term was selected based on the Nash-Sutcliffe efficiency (NSE) of the overall model. The NSE values of the models ranged from 0.61 to 0.88. The boundary concentrations were earlier modeled using symbolic regression without BFI as a predictor, resulting in a set of highly complex models for the same data. This study demonstrates the suitability of BFI as a water quality predictor and the importance of identifying suitable predictors to develop defensible empirical water quality models. Further, the relation between EC concentrations and BFI could be used to infer whether the predominant pollutant source at a location is independent of rainfall or is rainfall driven.