The role of eutrophication models in water management

In this thesis the role of eutrophication models in water management is analysed. The thesis consists of an extended introduction followed by five Appendices with papers describing different mathematical models dealing with eutrophication in surface waters. At first systems analysis is described as a general framework for eutrophication modelling. Emphasis is put on the uncertainties inherent in predictions. Next, a set of criteria is derived from this. The criteria may be useful in evaluating previous work and in guiding new initiatives on eutrophication modelling. Accordingly, the criteria are applied to review studies described in three of the five papers and to evaluate a number of projects of different lake managers in The Netherlands.Systems analysis, an approach to solve problems, is utilised as the framework to study problems related to eutrophication of surface waters. The following stages are distinguished in this approach: problem analysis and the formulation of objectives, the set up of a conceptual framework, analysis of available data, specification of the mathematical formulations, sensitivity analysis, parameter estimation, validation and finally application of the model. Application of this approach generally results in repeatedly going through several stages, so systems analysis is a cyclic rather than a one-way procedure. Several examples from the field of eutrophication illustrate the approach.Contrary to common practice of systems analysis the proper technical development and/or application of the model, referred to as _credibility_, is explicitly distinguished in this thesis from the acceptance of the results, referred to as _acceptability_. This is in agreement with actual pratice in water management, where the modeller, the user of the model and the user of the results of the model often are two or even more different persons. Credibility and acceptability are thus specified in a set of criteria which may be useful for both modellers and users of the model (results). The criteria addressing credibility are: the objectives of the model and the ensuing choice of the state variables, the dimensions of the modelled system and aggregations in time and space, the utilisation of the available data, the appropriateness of the model structure, the determination of parameter values, the validation of the model and finally the assessment of the uncertainties in the structure, the parameters and the results of the model. Criteria addressing acceptability are: attention to the motivation to undertake a modelling project, specification of the constraints in time and money, explication of the arguments to accept or reject (parts of) the model (results) and consideration of the consequences of the use of the model (results).Next, the criteria are used to review three studies described in the Appendices. A statistical model aiming at predicting the phosphorus concentration in lakes is described in the first study. All available information was used to determine the structure and parameters of the model, so validation on another data set could not be carried out. However, recent studies indicated that the general conclusions from the study were correct. On the other hand, accurate prediction of the phosphorus concentration appeared to be unattainable.A conceptual model for long term prediction of the phosphorus content in the top layer of the sediment is described in the second study. The paramete values were derived from previous work. Consequently, the model could be validated on observations of the sediments of Lake Veluwe, The Netherlands. The results of the validation were sufficient to qualitatively interpret and compare model predictions for a number of management alternatives.In the third study a relatively complex model is presented. The model was validated on observations of Lake Veluwe as well. In this thesis the assumption of complete mixing of the water was confirmed. Nevertheless, spatial homogeneity of the sediment was questioned. Furthermore, the previous qualitative validation was quantified. A deviation of 10% - 40% between model results and observations was obtained for the main model variables and this was acceptable taking into account the inherent errors of the observations. Neither the parameters nor the model structure could unambiguously be identified. Furthermore, the validation was biased in so far some parameters were adjusted. Therefore, the model was rightly used for hypothesising and for speculating and not for comparing detailed alternative courses of action.The assessment of uncertainties related to the use of these models did not always receive sufficient attention in the three studies discussed above. This thesis partly redeems the shortcomings and this contributes to the credibility of the studies. Next, subjects related to acceptability were highlighted. For that purpose lake managers at the operational level of three districts in The Netherlands were interviewed. Acceptability of a model or of the results of a model appeared to be only partly related to the credibility of a modelling project. Informal and personal relationships and accidental factors contributed significantly to the managers_ choice of the model(ler). The acceptation of the results of a modelling project was also affected by the motivation of a manager to undertake a project and by the constraints in time and money imposed by the manager. In their turn, managers were willing to adjust their objectives and to pay little attention to the reliability of the results.The constraints in time and money imposed by managers were the main reason why insufficient attention was paid to uncertainties in the development and the application of models. The assessment of uncertainties is laborious and complex, while completeness and objectivity are probably not feasible for larger models. That is why it is so important that the modellers clarify what is addressed and what is omitted. This, and not a high reliability in itself, is the essence of credibility. The managers are then better equipped to make verifiable choices to accept or reject (part of) the model (results). Managers are not always aware of the possible advantages of addressing uncertainties. Doing this may guide further actions, save time and money in certain situations, enhance public support and contribute to the confidence in the authority of the manager. Analogously to modellers with respect to credibility, managers may be more explicit in their arguments for accepting or rejecting (part of) the results. This may alleviate the present shortcomings and may contribute to a more transparent policy making and, thereby, further improve the role of eutrophication models in water management.Summaries of the papers can be found in the Appendices.