Estimating Iowa's riverine phosphorus concentrations via water quality surrogacy

Phosphorus (P) is a widespread waterborne pollutant that impairs many waterbodies. However, it is challenging to measure directly, and much research has been dedicated to developing surrogacy models that can repeatedly predict its concentration. Optimal approaches for modeling strategies are often unclear and depend upon local P dynamics and the availability of financial and technical resources. This study presents a schema for developing P surrogacy models at a statewide scale (16 major rivers in Iowa, USA). Specifically, we examined the relationship between particulate phosphorus (Part P) and orthophosphate (OP) and explored the viability of eight potential surrogates in predicting their concentrations using multiple linear regression and power regression methods. We also investigated similarities between surrogate models for Part P and total suspended solids (TSS). At all sites, OP and Part P were not strongly correlated (mean R = 0.20 ± 0.17). Many instances were observed where samples had high concentrations of one form but not the other. Modeling results demonstrated that turbidity was consistently the best predictor (t-statistics >10) of Part P, and adding other surrogates alongside turbidity did little to improve model performance. No surrogates proved useful in estimating OP. Viable power regression models were created using turbidity to predict Part P (mean R² = 0.69 ± 0.12). These models had a nonlinear form where Part P concentrations leveled off as waters became exceptionally turbid. This contrasted with TSS, which maintained a strong linear relationship across all turbidity levels. Turbidity-based models show promise in quantifying statewide P levels, as they enable high-resolution and real-time Part P estimates.