Uncertainties of climatic change impacts in Finnish watersheds: a Bayesian network analysis of expert knowledge

Climatic change impact studies are among the most complicated and uncertain environmental assessments scientists have ever faced. Not only are possible scenarios on key changes and impacts needed, but also estimates of their probabilities, which are of particular relevance to policy making. In modeling, the key uncertainties lie in inaccuracies and errors in parameters, and above all (but often not fully discussed) in assumed causalities (model structure). In the present study, the views of eight Finnish experts on interactions between climatic and aquatic systems were analyzed. They were asked to assess the states of 24 key variables in the climate-water system in southern Finland, using subjective probability distributions and the causalities between those variables. The methodological approach used was based on Bayesian belief networks. The highest uncertainties were seen in the changes of floods, water pH and oxygen concentrations, problems to constructions, and the recreational value of watersheds. Positive impacts can be expected on transportation and hydropower production. The relative importance of different causalities was analyzed separately for hydrological, limnological, and interest variables. The causalities between climate and hydrology did not appear very important from the interest standpoint, because uncertainty deriving from other sources masked their effects. The causality between temperature and precipitation was important throughout the model, while expected changes in temperature and precipitation were also important in the case of most variables.