In situ high-frequency measured turbidity can potentially be used as a surrogate for riverine phosphorus (P) concentrations to better justify the effectiveness of nutrient loss mitigation measures at agricultural sites. We explore the possibilities of using turbidity as a surrogate for total phosphorus (TP) and particulate phosphorus (PP) in four snowmelt-driven rivers draining agricultural clayey catchments. Our results suggest slightly stronger relationship between in situ measured turbidity and PP than between turbidity and TP. Overall, linear TP and PP regressions showed better error statistics in the larger catchments compared with their sub-catchments. Local calibration of the in situ sensors was sensitive to the number of high P concentration discrete water samples. Two optional calibration curves, one with and one without influential data, resulted in a 17% difference in the estimated mean TP concentrations of a snowmelt storm contributing 18% of the annual discharge volume. Accordingly, the error related to monthly mean TP estimates was the largest in spring months at all sites. The addition of total dissolved phosphorus (TDP) improved the model performance, especially for sites where the TDP/TP ratio is large and highly variable over time. We demonstrate how long-term discrete samples beyond sensor deployment can be utilized in the evaluation of the applicability range of the local calibration. We recommend analysing the validity of P concentration estimates, especially during high discharge episodes that contribute substantially to annual riverine nutrient fluxes, since the use of surrogates may introduce large differences into the P concentration estimates based on selected local calibration curves.
Keywords: In situ monitoring; Linear regression; Local calibration; Proxy relations; River water quality.