The Water Framework Directive: Setting the phosphorus loading target for a deep lake in Denmark using the 1D lake ecosystem model DYRESM-CAEDYM

We used the one-dimensional DYRESM-CAEDYM model to elucidate and quantify the influence of the external phosphorus loading on ecosystem dynamics in moderately deep Lake Ravn, which is situated in an agricultural landscape in Denmark. Model simulations were used to quantify the extent to which the external phosphorus loading must be reduced to meet upcoming lake ecological quality requirements according to the European Union Water Framework Directive (WFD). The model generally showed good agreement with observed data for temperature and oxygen from the epilimnion and hypolimnion during the calibration period (7 years) as well as the validation period (5 years); although peaks of oxygen concentrations in epilimnion during late spring often were underestimated. Phosphate and total phosphorus (TP) concentrations were generally well reproduced in both the epilimnion and hypolimnion, though hypolimnetic phosphorus was occasionally underestimated in late summer. There was also good agreement between monitored data and modelled biomass of diatoms and dinoflagellates as well as the zooplankton biomass of cladocerans and calanoid copepods, although the timing of biomass peaks occasionally deviated from observations. Root-mean-square-errors (RMSE), used to quantify the model error, were overall similar for the calibration and the validation period. Simulations of scenarios with a reduced external TP loading suggest that a substantial reduction (40-50%) of the TP loading is required if phytoplankton biomass is to drop to a level sufficiently low to meet the proposed WFD requirements (summer average <6.5 μg chlorophyll a l−1). The predicted outcomes of considerable loading reductions should, however, be treated with some caution, as the conceptual model in this study could not fully account for the changes in trophic structure occurring at radically reduced TP loading. This particularly applies to changes in the fish stock, which may have extensive cascading effects via increased zooplankton grazing on the phytoplankton when external TP loading is reduced. This would most likely lead to higher transparency than that predicted by the model. To further improve the reliability of the predicted outcomes of various model scenarios, future work should include a series of test simulations also including fish predation. Today, fish predation is available in the DYRESM-CAEDYM model as well as other dynamic lake models; however, only sparse data is available on fish stock dynamics in Lake Ravn (as for most other lakes). An applicable test of the fish algorithms in today's dynamic models will require reliable estimates of fish stock biomass in the study lakes during both the model calibration and validation period.