Deriving Nutrient Concentration Criteria for Rivers: The Model‐Then‐Classify Method Incorporating Under‐Protection Risk

ABSTRACT The model‐then‐classify method allows resource managers to set nutrient concentration criteria to achieve biological targets, with a specified level of risk that the target will not be achieved. Because modeling precedes classification, the approach produces comprehensive criteria that apply to all waterbodies in a domain of interest. We demonstrate the approach for New Zealand rivers and periphyton biomass targets. We fitted generalized linear models expressing peak periphyton biomass at 268 monitoring sites as a function of nutrient concentrations (TN or TP) and other variables including site shade, temperature, and hydrology. We used the models to predict biomass along gradients in nutrient concentrations for all rivers nationally. The predictions were inverted and averaged within environmental classes to derive criteria for each nutrient, class, three biomass targets, and shaded or unshaded conditions. The criteria include choices of under‐protection risk (UPR), which is the probability the target will be exceeded despite compliance with the criteria. The UPR provides transparency about the societal cost of reducing the risk of not achieving the target. A validation procedure demonstrated that, when concentrations comply with the criteria, the proportion of sites not achieving the target aligns with the UPR. The approach could be applied to other stressors and types of aquatic ecosystems.