A spatial bayesian-network approach as a decision-making tool for ecological-risk prevention in land ecosystems

Prevention of ecological risks in land ecosystems is crucial for environmental protection and sustainable land use. With increasingly severe land degradation, new and effective methods must be developed for the management of ecological risks. In this study, a conceptual decision-making model in ecological risk prevention was developed using the Bayesian belief network with a geographic information system (GIS) for the regional-scale land ecosystem in the traditional mining city of Daye in Central China. Based on the results of a sensitivity analysis, the variable of eco-resilience reduction was identified as the most sensitive to habitat removal with the highest mutual information at 0.71. The two variables of soil pollution and water-quality deterioration were selected for a cross-validation analysis, and the changes in both the calibration and validation performance were very small. The scenarios we considered based on the interests of various stakeholders presented the spatial distribution of the following regulative effects of various management measures on a regional scale: (1) the variable of urbanisation showed that the probability of 11.5 % of all the grids decreased at a high state over an area of 177 km(2); (2) the variable of mining showed that the probability of 35.5 % of the all the grids at a high state decreased, over an area of 554 km(2); (3) the variable of habitat removal showed that the probability of 6.7 % of all the grids at a high state decreased, over an area of 87 km(2); and (4) the variable of health threats showed that the probability of 8.4 % of all the grids at a high state decreased, over an area of 135 km(2). The Bayesian-network-GIS based tools can support the decision-making process used for ecological-risk prevention in land ecosystems.