Pollutant Load Estimates Using Regression Models with In-Stream Measurements

Abstract

A continuous in-stream water quality measurement (CWQ) study was carried out in two Lincoln, Nebraska urban watersheds. Discrete stormwater samples were collected at the study sites during 17 storm runoff events over a three-year period. In-stream flow and water quality (e.g., turbidity) measurements were combined with climatic data to develop multiple-linear regression (MLR) models for the estimation of six stormwater pollutant concentrations [i.e., total suspended solids (TSS), soluble reactive phosphorus (SRP), total phosphorus (TP), nitrate plus nitrite-nitrogen (N+N), total Kjeldahl nitrogen (TKN), and Escherichia coli (E. coli)]. MLR concentration models based on in-stream measurements (CWQ-C) were developed to estimate pollutant concentrations at any time during a storm. Three additional MLR models were developed to estimate event mass loads based on (1) climatic data only, (2) both CWQ and climatic data (CWQ-L), and (3) use of literature event mean concentrations (simple mass load). The comparison suggests that for small, urban watersheds, using correlated in-stream water quality and flow measurements along with climatic data (e.g., CWQ-L models) best captures variability, especially for TSS, SRP, and TP. The study also showed that nitrate atmospheric deposition data improved the N+N and TKN Climatic and CWQ-L load models.