基于SOFM神经网络的土地整治时空配置分区研究

研究目的:探索与完善土地整治时空配置分区的相关理论与方法,为新时期生态文明建设、"三生"空间重构等背景下土地整治的实施提供科学依据。研究方法:以江西省湘东区为例,从土地整治生态风险、迫切性和适应性三个维度,系统分析土地整治时空配置分区的影响因素,运用自组织特征映射(Self-organizing Feature Mapping,SOFM)神经网络方法实现土地整治时空配置分区。研究结果:(1)湘东区土地整治生态风险、迫切性和适宜性空间差异显著;(2)湘东区140个行政村被划分为8个项目区,并与传统分区方法进行对比分析,结合分区影响因素,将8个项目区归纳为近期优先、近期适度、中期紧缩和远期限制4个土地整治时空配置分区。研究结论:考虑生态风险的土地整治时空配置分区结果较传统未考虑生态风险的结果更具实地符合性;基于SOFM神经网络的土地整治时空配置分区方法将地理位置和空间属性有机结合,具有更强的科学性和应用性,将为新时期的土地整治时空配置分区研究提供新思路。

Study on Temporal-Spatial Allocation Zoning of Land Reclamation Based on SOFM Neural Network

The purpose of this study is to explore relevant theories and methods for the temporal-spatial allocation zoning of land reclamation in order to provide the theoretical and methodological basis for the scientific implementation of land reclamation under the background of ecological civilization construction and ecological-production-living spaces reconstruction in the new era. Taking Xiangdong District, Pingxiang City, Jiangxi Province as cases, this paper first systematically analyzed the influencing factors of temporal-spatial allocation zoning from the three dimensions of land reclamation ecological risks, the land reclamation urgency and the land reclamation suitability. And then, the temporal-spatial allocation zoning of land reclamation based on Self-organizing Feature Mapping (SOFM) neural network method were carried out. The results indicates that 1)there exists an significant spatial differences in ecological risk, urgency and suitability of land reclamation in the study area, and 2)the 140 villages in Xiangdong District could be divided into 8 zones, which were further categorized into near-term priority remediation area, near-term moderate remediation area, medium-term tightening remediation area and long-term limitation remediation area. Each area has different priorities for land reclamation. In conclusion, the temporal-spatial allocation zoning of land reclamation considering ecological risk is more reasonable. The zoning of land reclamation based on SOFM neural network is more scientific and practical because it can combine geographic and spatial attributes characteristics. There will be a new stream of thought for the study of temporal-spatial allocation zoning of land reclamation in the new period.