Heightened nitrogen (N) inputs to watersheds threaten both human communities and natural ecosystems. Contaminated drinking water, eutrophication, and toxic algal blooms can all result from enhanced N loading to streams and nearshore marine systems. Better understanding of nutrient sources and dynamics will help establish more productive monitoring and mitigation. Common practice uses monthly grab sample measurements, often combined with continuous flow monitoring and LOADEST modeling, to estimate nutrient fluxes. However, new sensors are available for real-time nitrate monitoring. Do real-time estimates differ enough from grab samples and modeling to justify the cost? The Nooksack River and two of its lowland tributaries, Kamm and Fishtrap Creeks, Whatcom County, WA, provided opportunities to compare these methods in different stream environments. The Nooksack encompasses a large watershed, while Kamm and Fishtrap have varying land use and groundwater influence. We sampled stream flow and nitrate concentrations once per month, using those measurements to estimate daily and monthly N flux by applying two techniques: extrapolation of grab sample data, or grab sample concentrations paired with continuous flow and LOADEST modeling. We tested these modeled estimates of nitrate flux against real-time measurements from nitrate sensors: Sea-Bird SUNA in Fishtrap; OTT Hydromet ecoN in Kamm and Nooksack.
Results/Conclusions
Dynamic flows in the Nooksack River led to significant differences between grab sample, LOADEST and real-time estimates of nitrate flux. Nooksack daily discharge followed predictable patterns, with base flows as low as 11 m3/s in summer and peak flows as high as 570 m3/s in winter. Not surprisingly, the rainy season (Oct. to March) had more extreme and less predictable oscillations than summer. Based on the real-time sensor, nitrate concentrations (range: 0.02-0.75 mg-N/L) correlated positively with discharge across seasons, but winter high flow events often decreased nitrate concentrations. Nooksack nitrate fluxes largely correlated with discharge, following similar annual and intraseasonal patterns. Grab samples were necessary to validate sensor measurements, however their low temporal resolution was a major limitation for calculating cumulative nutrient loads, as they missed flux variability between measurements. Continuous flow monitoring combined with grab samples and LOADEST modeling provided better resolution. Still, in several months, real-time flux estimates differed significantly from LOADEST calculations (95% CI), despite having 20 years of data for LOADEST calibration. Differences were greatest in winter, when flow variability was highest. We are exploring similar relationships in Kamm and Fishtrap creeks to see if they hold in different stream environments.