基于贝叶斯网络的城市生态红线划定方法

生态红线划定是保护生态安全,协调城市建设、基本农田保护和生态保护之间矛盾的重要方法。目前,有关生态红线的划定方法多基于生态适宜性评价,忽视对土地利用变化的探究,缺少与城市建设发展相互协调,导致生态用地经常被占用,生态红线保护效果不好。在综合分析生态用地历史变化过程和生态适宜性条件的基础上,提出了基于贝叶斯网络的城市生态红线划定方法,并以鄂州市为研究区验证了模型划定的效果。划定结果表明,该方法符合鄂州市城市发展的趋势和生态用地空间分布特性,既有利于稳定且生态服务价值高的区域划入红线,又保证了生态红线空间的落地实施。与传统生态评价方法相比,贝叶斯网络模型划定方法的实效性更强,可以为城市生态红线划定方法研究提供参考和借鉴。

A method of delimiting urban ecological red line protection area based on bayesian network

Delimiting ecological red lines is the key to answering spatial-allocation questions for land resources. At present, most studies on the delimitation of ecological red lines have been based on the evaluation of suitability conditions of ecological land, without considering the process of historical change. Therefore, it cannot be guaranteed that the zoning results will adapt to the trends in ecological land change in the area, leading to frequent designation and adjustment of ecological red lines. In this study, to guarantee the stability of ecological red lines, we examined a delimitation method for ecological red lines based on a Bayesian network model, which included both factors concerning ecological suitability conditions and dynamic change of the land. Next, the proposed model was applied to E-Zhou, a city near the mid-reaches of the Yangtze River. In this proposed model, ecological suitability factors, which represented the quality of the ecological land, and dynamic change factors, which illustrated the history of the ecological land, were obtained. Using the ecological land potential value as the target variable, we defined the structure of the network using expert knowledge. The Bayesian network was trained by the maximum likelihood method with 20000 random sample points. The results showed that only 65.5% of ecological land remained stable from 2004 to 2013 in E-Zhou City. For the two dynamic factors, the influence of farmland occupancy was the greatest, and accounted for 43.6%, whereas the influence of urban encroachment accounted for 10.2%. The results of the sensitivity analysis also indicated that farmland occupancy had the greatest influence on the potential value of the ecological land, with the highest variance reduction of 29.5%, followed by eco-environmental sensitivity, and the importance of ecosystem services, which exhibited variance reductions of 8.55% and 1.84%, respectively. The variance reductions for the distance to a water body and the ecological protection of a forest were greater than that of the distance to a road or a railway, change in traffic facilities, and urban construction, which all had little effect on the ecological land potential. A causal link between influence factors and the target variable was obtained during backward propagation in diagnostic analysis. Under the condition that the value of ecological land potential was "yes", the probability of the "extremely important" factors of ecosystem services and water conservation increased by 6% and 2.7%, respectively. This indicated that the contributions of ecosystem service values of a forest ecosystem and aquatic ecosystem increased. Furthermore, the probabilities of "highly sensitive" and "extremely sensitive" factors of ecological sensitivity decreased by 13.4% and 1.8%, respectively. This verified the improvement of the natural condition of the land by ecological land protection. Then, using the values for the ecological suitability factors in 2013, we obtain the relative potential value of each ecological land parcel as the ecological red lines through forward reasoning of the trained network, and delimiting ecological land with the potential value of the target variable. The results showed that this delimiting method worked in accordance with the trends in urban development in the area and spatial distribution of ecological land in E-Zhou City. This method can improve the stability of ecosystem services, as well as ensure the quantity and quality of ecological red lines. This new and effective model could serve as a support for other cities.