“灾害–政策–生态”耦合响应下的农业抗灾韧性动态评估——以南京市2022旱情恢复为例

Dynamic Assessment of Agricultural Disaster Resilience under the Coupling Response of “Disaster-Policy-Ecology”—A Case Study of Drought Recovery in Nanjing City in 2022

摘要: 全球极端气候事件频发背景下，丘陵平原交错区农业生态系统稳定性面临严峻挑战，开展农业抗灾韧性精准评估具有重要理论与实践价值。针对传统评估依赖统计数据、分辨率低、动态性不足的局限，以2022年极端高温干旱事件前后的南京市为研究区，构建“高分辨率遥感 + 随机森林分类 + 多指数恢复力模型”的技术框架：基于Sentinel-2遥感影像(10 m分辨率)提取归一化植被指数(NDVI)等核心特征，通过随机森林算法实现耕地、林地等5类土地利用类型精准分类(总体精度89.7%，Kappa系数0.86)，结合归一化恢复力指数模型量化农业生态系统韧性变化。结果表明：① 2022年极端高温干旱导致耕地、林地NDVI分别骤降27.3%、15.4%，土地利用结构失衡；② 2023~2024年在政策调控与自然修复协同作用下，耕地、林地实现“超补偿”恢复，2024年NDVI较2021年分别提升65%、25%，恢复力指数达4.42、2.66 (高韧性等级)，土地利用结构回归均衡；③ 农业韧性呈显著空间分异，平原区恢复速率高于丘陵区，水体、裸地及建设用地为韧性短板。研究揭示了“灾害冲击–政策干预–生态恢复”的耦合响应机制，为同类丘陵平原交错区农业防灾减灾与韧性提升提供了标准化技术路径与实证参考。

Abstract: Against the backdrop of frequent global extreme climate events, the stability of agricultural ecosystems in hilly and plain transitional areas is facing severe challenges. Conducting precise assessments of agricultural disaster resilience has important theoretical and practical value. In response to the limitations of traditional assessments relying on statistical data, low resolution, and insufficient dynamism, a technical framework of “high-resolution remote sensing + random forest classification + multi index resilience model” was constructed using Nanjing city before and after the extreme heat and drought event in 2022 as the research area. Based on Sentinel-2 remote sensing images (10 m resolution), core features such as normalized vegetation index (NDVI) were extracted, and the random forest algorithm was used to achieve accurate classification of five types of land use, including cultivated land and forest land, with an overall accuracy of 89.7% and a Kappa coefficient of 0.86. Combined with the normalized resilience index model, the resilience changes of agricultural ecosystems were quantified. The results showed that: ① In 2022, extreme high temperature and drought caused a sharp drop of 27.3% and 15.4% in NDVI of cultivated land and forest land, respectively, resulting in an imbalance in land use structure; ② From 2023 to 2024, under the synergistic effect of policy regulation and natural restoration, cultivated land and forest land will achieve “overcompensation” restoration. In 2024, the NDVI will increase by 65% and 25% respectively compared to 2021, and the resilience index will reach 4.42 and 2.66 (high resilience level), and the land use structure will return to equilibrium; ③ The resilience of agriculture shows significant spatial differentiation, with a higher recovery rate in plain areas than in hilly areas. Water bodies, bare land, and construction land are the weak links in resilience. The study reveals the coupled response mechanism of “disaster impact policy intervention ecological restoration”, providing a standardized technical path and empirical reference for agricultural disaster prevention, mitigation, and resilience improvement in similar hilly and plain transitional areas.