A Model-Driven Approach to Uncertainty Reduction in Environmental Data

Abstract

Environmental simulation modeling is inherently linked to observation data on the status of the environment. Data are a limiting factor in the selection of a model suitable for sustainable decision making. Models transform observation data into information by extracting aggregate values from raw data, projecting values of selected environmental indicators and detecting trends to track changes in environmental conditions. The chapter describes the framework for a model-driven development of sampling programs on the quality of the aquatic environment. Monitoring designs are determined as solutions of the operation research models articulated based on the cost-effectiveness analysis. The uncertainty of the estimates derived from monitoring data is used as a measure of the effectiveness of a monitoring design. Since data collected for one set of objectives can be used for the purposes which were not even considered at the planning stage, simple random designs are preferable. The proposed approach takes into account existing sampling procedures, models used for data analysis and uncertainty associated with the collected data. The approach has been used to develop efficient simple random designs common for all water quality parameters which are detected from the same water sample. The designs were built using linear regression models which improved the observation programs by reducing the numbers of required samples.