Freshwater Colloquium: Data-Efficient Stream Nutrient Modeling: An Application of Data-driven Sparse Sensing

Regular estimation or monitoring of nutrient (i.e., nitrogen or phosphorus) concentrations and loads is necessary for assessing stream health and evaluating nutrient pollution mitigation programs. The state of science concludes that at least daily monitoring is necessary. However, current nutrient monitoring programs involve either infrequent grab samples or costly, maintenance-intensive in-situ sensors, resulting in low sampling frequencies and high uncertainties. To address this, improved stream nutrient monitoring and modeling methods are needed. Data-driven Sparse Sensing (DSS) offers an efficient solution by reliably estimating high-frequency time series from fewer samples, leading to more efficient sampling designs. This study applies the DSS technique to estimate stream nutrient concentrations and loads (nitrate-nitrite as NOx and phosphorous as P) using sparse measurements of nutrient concentrations and flows. The overall aim is to evaluate how DSS can identify the informative sampling times and locations and reconstruct concentration (NOx and TP) time series with minimal measurements. Furthermore, this study examines the model’s sensitivity to the size of training data, time span of monitored data, and hydrologic regions. This method has a high potential to reduce the cost and effort of nutrient monitoring programs and assessments of catchment material balances.