Data from: Suburban watershed nitrogen retention: estimating the effectiveness of stormwater management structures

Excess nitrogen (N) is a primary driver of freshwater and coastal eutrophication globally, and urban stormwater is a rapidly growing source of N pollution. Stormwater best management practices (BMPs) are used widely to remove excess N from runoff in urban and suburban areas, and are expected to perform under a wide variety of environmental conditions. Yet the capacity of BMPs to retain excess N varies; and both the variation and the drivers thereof are largely unknown, hindering the ability of water resource managers to meet water quality targets in a cost-effective way. Here, we use structured expert judgment (SEJ), a performance-weighted method of expert elicitation, to quantify the uncertainty in BMP performance under a range of site-specific environmental conditions and to estimate the extent to which key environmental factors influence variation in BMP performance. We hypothesized that rain event frequency and magnitude, BMP type and size, and physiographic province would significantly influence the experts' estimates of N retention by BMPs common to suburban Piedmont and Coastal Plain watersheds of the Chesapeake Bay region. Expert knowledge indicated wide uncertainty in BMP performance, with N removal efficiencies ranging from 40%. Experts believed that the amount of rain was the primary identifiable source of variability in BMP efficiency, which is relevant given climate projections of more frequent heavy rain events in the mid-Atlantic. To assess the extent to which those projected changes might alter N export from suburban BMPs and watersheds, we combined downscaled estimates of rainfall with distributions of N loads for different-sized rain events derived from our elicitation. The model predicted higher and more variable N loads under a projected future climate regime, suggesting that current BMP regulations for reducing nutrients may be inadequate in the future.Expert uncertainty estimates for all 60 questions in the SEJ protocol documentThis tab-delimited text file contains raw values provided by all 10 experts for all 60 questions in the structured expert judgment (SEJ) protocol document (Appendix S3). The six columns are defined as follows: "Expert" refers to the expert identification codes used in the article; "Qnum" refers to the number of each question in the SEJ protocol document (Appendix S3); "Qid" is a brief coded description of the key variables in each question (codes are arranged as watershed location – storm event number – site description – influent/effluent load – pre/post restoration condition; see Appendix S3 for the full details of each question); "Percentile.05" is the 5th percentile of the probability distribution envisioned by each expert to describe the total nitrogen load for each question (units are kg total nitrogen); "Percentile.50" and "Percentile.95" refer to the 50th and 95th percentiles of the distribution, respectively (units are kg total nitrogen). Note that expert names are not associated with their estimates; the ordering of expert names in Appendix S2 is different than the ordering of their estimates in the data.Koch_SEJ_all_experts_TN_load_kg.txtRealizations for the 11 calibration variables in the SEJ protocol documentThis tab-delimited text file contains empirically measured values for the 11 calibration variables in the structured expert judgment (SEJ) protocol document (Appendix S3). The three columns are defined as follows: "Qnum" refers to the number of each question in the SEJ protocol document that corresponds to a calibration variable (Appendix S3); "Qid" is a brief coded description of the key variables in each question (codes are arranged as watershed location – storm event number – site description – influent/effluent load – pre/post restoration condition; see Appendix S3 for the full details of each question); "TN.load.kg" is the empirically measured total nitrogen load (units are kg total nitrogen).Koch_SEJ_calibration_variables_TN_load_kg.tx