Improved back-propagation neural network in ecological vulnerability assessment of Zhalong wetland

Abstract

Combining GIS and RS technique in the study of ecological vulnerability assessment in Zhalong Wet Ecosystem, an improved Back-Propagation neural network was established to model and evaluate ecological vulnerability based on Press-State-Response model. The research educed that: the regions of moderate or severe vulnerability accounted for 18% while the regions of potential or minute vulnerability accounted for 63%; the potential fragility areas and the slight fragility areas with better health were all located in the internal marish while the moderate fragility areas and the serious fragility with poor health were mainly paddy fields; the vulnerability of the central area in the Reserve was not so obvious as the marginal area. The holistic ecological environment of Zhalong wetland was not optimistic. By modeling improved BP neural network and calculating the ecological vulnerability index EVI of each grid cell of Zhalong wetland, a reference of scientific significance was provided for the ecological environment protection and management of Zhalong wetland from the quantitative point of view.