“…Ninety-five percent confidence intervals around estimated mean annual loads and FWCs quantify uncertainty around estimates and generally were larger for constituents with greater variability in measured loads (tables 4 and 5). Recent research has shown that LOADEST models can produce biased estimates of long-term average constituent loads (Stenback and others, 2011;Garrett, 2012;Hirsch, 2014), and LOADEST has since been modified to facilitate residual analysis and bias identification (Runkel, 2013). Long-term bias in model estimates was less of a concern because the modeling period was only 3 years and predictor variables describing trends in time were not used.…”