Statistical methods for adaptive river basin management and monitoring

Decision-making at different phases of adaptive river basin management planning rely largely on the information that is gained through environmental monitoring. The aim of this thesis was to develop and test statistical assessment tools presumed to be particularly useful for evaluating existing monitoring designs, converting monitoring data into management information and quantifying uncertainties. River basin scale monitoring was performed using a wireless sensor network and a data quality control system and maintenance effort was assessed. National-scale, traditional monitoring data and linear mixed effect modelling were used to estimate the uncertainty in two status class metrics (total phosphorus, and chlorophyll-a) by quantifying temporal and spatial variance components. The relative sizes of the variance components were then used to determine how to efficiently allocate the monitoring resources. Nutrient and chlorophyll-a statuses were linked to external loading utilizing a large amount of monitoring data and a hierarchical Bayesian approach. This linkage was the basis for developing a practical assessment tool for lake management. To evaluate the network of relationships affecting phytoplankton development between water quality variables, structural equation modelling was used. Model residual and parameter uncertainty, and thus uncertainty in the assessment result, were estimated using probabilistic Bayesian modelling. In general, the results of this study suggest that the used statistical methods appear to be particularly useful for decision-making under an adaptive management framework, as they enabled predictions to be made based on existing monitoring data and have measures of uncertainty associated with the outcomes. The results suggest that the uncertainties often stem from the lack of input data or insufficiently allocated monitoring. Therefore, it should be ensured that information gaps in the nutrient loading values, as well as in other, especially biological variables, are sufficiently covered.