Vegetation zoning in the Loess Plateau of northern Shaanxi based on Self-Organized Mapping neural network
摘要
内部生境相对一致的植被分区是开展生态恢复重要依据.黄土高原作为国家生态安全屏障和区域协调发展关键区,生态治理成效显著,但大规模植被工程推进过程中,由于缺乏科学的生境适配机制,出现植被过度扩张、恢复植被与立地条件错位等新问题.制定与环境条件相匹配的现势性强、生态适应性佳的植被分区方案,成为提升植被恢复成效的关键.本文以陕北黄土高原为研究区,利用2000-2023年气象数据(≥10℃积温、1月和7月均温、干燥度、降雨量),结合DEM、土壤、植被和地貌类型等多源数据,构建三级植被分区指标体系,采用能够捕捉复杂数据内在结构的无监督聚类模型——自组织映射神经网络(Self-Organizing Feature Map,SOFM),结合GIS进行分区制图,并利用轮廓系数(Silhouette Coefficient,SC)和戴维森堡丁指数(Davies-Bouldin Index,DBI)进行交叉验证,通过多指标动态平衡策略提高分区方案的稳健性.结果表明:①陕北黄土高原水热梯度分异显著,地貌、土壤、植被格局与水热梯度高度耦合,空间分异特征清晰;②基于加权评分确定最佳聚类效果,将研究区划分为2个植被区域、5个植被地带、11个植被区.研究提出的多尺度分区框架,揭示了不同分区植被类型与生境因子的空间适配规律,可为区域植被恢复精准施策、生态系统优化及可持续发展提供科学依据.
Vegetation zones with relatively consistent internal habitats serve as a crucial basis for ecological restoration.As a national ecological security barrier and key region for coordinat-ed regional development,the Loess Plateau has achieved remarkable ecological governance out-comes.However,in large-scale vegetation restoration projects,the absence of a scientific habi-tat adaptation mechanism has led to new challenges such as excessive vegetation expansion and mismatches between restored vegetation and site conditions.Developing a vegetation zoning plan that is environmentally aligned,targeted,and ecologically adaptable has become critical for improving vegetation restoration outcomes.This study focuses on the Loess Plateau of northern Shaanxi Province.Utilizing meteorological data from 2000 to 2023(including accu-mulated temperature ≥10℃,average temperatures in January and July,aridity index,and pre-cipitation),combined with multi-source data such as DEM,soil,vegetation,and landform types,a three-tier vegetation zoning indicator system was constructed.An unsupervised cluster-ing model—Self-Organizing Feature Map(SOFM)was applied to map the zones.Cross-valida-tion was performed using the Silhouette Coefficient(SC)and Davies-Bouldin Index(DBI),while a multi-indicator dynamic balancing strategy enhanced the robustness of the zoning scheme.The results showed that:(1)The hydrothermal gradient of the study area was signifi-cantly differentiated,and the landform,soil,and vegetation patterns were highly coupled with the hydrothermal gradient,with distinct spatial differentiation;(2)Based on weighted scoring,the optimal clustering effect was determined,and the study area was divided into two vegeta-tion regions,five vegetation zones,and 11 vegetation areas.The proposed multi-scale zoning framework reveals spatial adaptation patterns between vegetation types and habitat factors across different zones,providing scientific basis for precision vegetation restoration engineer-ing,ecosystem structure optimization,and regional sustainable development.