A trend analysis and model comparison of total phosphorus concentrations and loads in the Boise River near Parma, southwestern Idaho, water years 2003–21

Total phosphorus (TP) concentrations and loads in the Boise River near Parma, Idaho, were examined to identify changes by month over a 19-year period from water year 2003 through water year 2021 and to evaluate the performance of three common water-quality models. Mean annual TP concentrations and loads were estimated to have reduced by approximately 60 percent over the study period. Mean annual TP concentrations were reduced from 0.42 milligrams per liter in 2003 to 0.18 milligrams per liter in 2021. Mean annual TP loads were reduced from 816 kilograms per day in 2003 to 302 kilograms per day in 2021. Mean annual concentrations and loads reduced by approximately 3 percent per year with the largest changes occurring in the non-irrigation season of October through April. The TP load remained highest in May across the model period while peak concentration shifted from January to March.

High-frequency TP data collected with an automated sampler every 49 hours enabled detailed model performance evaluation of the Load Estimator (LOADEST), Weighted Regressions on Time, Discharge, and Season (WRTDS), and WRTDS method with Kalman filtering (WRTDS_K) water-quality models generated with near-monthly data. All three models were generally able to reproduce the observed concentrations, with the largest errors occurring in the spring when observed concentrations were most variable. Annual TP loads varied by up to 27 percent, or approximately 128,000 kilograms, between the three models calibrated on monthly data. In this system with highly variable concentrations, we note that performance metrics for WRTDS_K based on monthly calibration data masked serious errors that were only revealed by comparing results against higher frequency (49-hour) autosampler data. This emphasizes the value of high frequency validation data to quantify uncertainty in water-quality models when applied to systems where concentrations change rapidly. Lastly, we identify that hydraulic routing may be a valuable addition to discharge, season, and time in water-quality modeling for systems with significant human intervention in natural hydro-biogeochemical processes.