Design of water quality monitoring using surrogate approach in central Germany

High-frequency monitoring of wide range of parameters is still costly and laborious. Consequently, techniques that use surrogate data for estimating, grab-sampled parameters using continuously monitored descriptors are more efficient to inform water resources management. The aim of this study is to develop a set of surrogates to continuously estimate in-stream sediment, carbon, nitrogen and phosphorus concentrations as well as ions contributions, based on commonly available high-frequency measurements (i.e., turbidity (Turb), discharge (Q), electrical conductivity (EC), nitrate-N concentration (NO₃-N) and specific adsorption coefficient (SAC)). To this end, high-frequency data (15 min interval) were collected during three years period at the UFZ-TERENO platform in Bode catchment, Germany. Surrogate models were analyzed for suspended sediment concentration (SSC), particulate organic carbon (POC), total organic carbon (TOC), dissolved organic carbon (DOC), particulate nitrogen (PN), total nitrogen (TN), particulate phosphorus (PP) and different ions contributions (Ca²⁺, SO₄²⁺, Cl^-, Na⁺, Mg²⁺, K⁺) using linear regression models. They were developed in four distinct (in terms of size, soil type, dominant land use, and topography) sub-catchments.

Results showed that Turb, Q, EC and NO₃-N were the most predictive variables. SSC could be predicted uniformly by Turb in all sub-catchments with a lowest correlation coefficient of 0.93, while different models were obtained for each sub-catchment. The non-uniqueness of those models can be explained by complexity of soil erosion processes and the sensitivity of Turb to soil roughness, topography and soil type. Best predictions of POC, TOC, PP, TP and PN were achieved using multi-parameter regressions (including Turb, Q, EC and NO₃-N). The ion contributions were reconstructed reasonably well when the Turb and EC were considered (lowest regression coefficient is 0.74). Results showed that the SAC together with the Turb and Q are good surrogates for the DOC concentration (lowest correlation coefficient is 0.60).

Merging high-frequency and surrogate techniques, in appropriate manner, seems to be very efficient design of water quality monitoring for better characterization of water body in hotspots and during hot moments periods.