Spurious correlation in the USEPA rating curve method for estimating pollutant loads

Water quality models are important tools used for developing total maximum daily loads (TMDLs), which are used to implement water quality standards. The writers have developed a tool which spatially interpolates output from the Chesapeake Bay Program Watershed Model (WSM) and requires little time, data, or training. In contrast, the WSM requires extensive time, data, and training to run. This study examines if the writers' tool provides accurate estimates of pollutant loads as compared to observed pollutant loads calculated using two different rating curve methods. The accuracy of the tool was assessed based on Nash-Sutcliffe coefficients. This study found that the tool provided moderate estimates of total nitrogen, poor estimates of total phosphorus, and that the rating curve method used to calculate observed loads affected the conclusion of the accuracy of the writers' tool for predicting total sediment. A key finding of this study is that the widely used load-derived method for calculating observed loads exhibits spurious correlation. The writers discuss the imperative need to use sound statistical methods when estimating pollutant loads and TMDLs, and suggest an alternative approach for developing rating curves which is not subject to the problem of spurious correlation.