Evaluation of a model-based method for designing water quality observations

An observational design technique supporting water quality modelling was reviewed and evaluated. The practical feasibility and mathematical sensitivity of the method were analysed for Lake Kuortaneenjärvi, Western Finland. A sampling programme to estimate characteristic rate parameters for competition between the nitrogen-fixing cyanobacter Aphanizomenon flos-aquae and eukaryotic phytoplankton was produced and the robustness of the design with respect to parameters and input was studied using the Monte Carlo technique. Optimal timing of observations appeared to be more sensitive to noise in the parameter values than to stochasticity in the model input, which consisted of data on the nutrient loading and climatic conditions of the lake. The perturbated parameter values favoured the weighting of phytoplankton dynamics, whereas the perturbations of the input shifted the design towards cyanobacterial dominance. It was concluded that this was because correct timing of algal succession was the most crucial element in successful estimation. Considerable rationalization of the use of measurement resources can be achieved by adopting the design procedure. This was also demonstrated for Lake Kuortaneenjärvi.