| 耦合SHAP与地理探测器的青藏高原生态驱动归因 |
| 摘要点击 318 全文点击 8 投稿时间：2025-07-30 修订日期：2025-11-11 |
| 查看HTML全文
查看全文 查看/发表评论 下载PDF阅读器 |
| 中文关键词 青藏高原 生态环境状况指数 SHAP解释器 地理探测器 空间分异规律 非线性增强效应 |
| 英文关键词 Qinghai-Xizang Plateau ecological index SHAP explainer Geodetector spatial heterogeneity nonlinear enhancement effect |
| DOI 10.13227/j.hjkx.202507399 |
|
| 中文摘要 |
| 为系统揭示青藏高原县域生态环境状况的空间分异规律与复杂驱动机制，以《生态环境状况评价技术规范》（HJ 192—2015）为依据，对各县域单元的生态环境状况指数（EI）进行测算. 进而耦合SHAP解释器与地理探测器模型，从特征贡献与因子交互等多维度解析其核心驱动因子. 结果表明：① 青藏高原的生态环境状况呈现出由东南向西北递减的显著空间分异规律，EI优良等级区域集中于湿润的东南部林草过渡带，而较差与差等级区域则广泛分布于干旱高寒的西北腹地. ② EI的空间格局是多因子交互作用下非线性增强效应的直接体现. 其中，植被覆盖状况（NDVI和FVC）是居于主导地位的自然驱动力，其后依次为年均降水量（Pre）和土地利用类型（Land）. ③ 交互探测证实，植被因子（NDVI和FVC）是关键的放大器，任一因子与之交互均能显著增强对EI空间分异的整体解释力. SHAP模型进一步将此非线性关系进行量化归因，精准揭示了各驱动因子在不同情景下的贡献方向与强度. 研究结论指出，青藏高原生态环境状况是一个对多重驱动因子产生复杂非线性响应的综合体，其宏观格局由植被和气候系统主导，并受到土地利用格局的显著调制. |
| 英文摘要 |
| To uncover the spatial patterns and complex drivers of the county-level ecological environment status across the Qinghai-Tibet Plateau， we calculated the ecological index （EI） for each unit， following the “Technical Criterion for Ecosystem Status Evaluation”（HJ 192—2015）. A coupled SHapley Additive exPlanations （SHAP） explainer and Geodetector were then used to analyze the key driving factors through multi-dimensional analysis， including feature contribution and factor interaction. The results indicate that： ① The ecological environment status of the Qinghai-Tibet Plateau exhibited a significant spatial gradient， decreasing from southeast to northwest. Areas with “Excellent” and “Good” EI grades were concentrated in the humid southeastern forest-grassland transition zone， whereas those graded as “Poor” and “Bad” were widely distributed in the arid and alpine northwestern hinterland. ② The spatial pattern of the EI was a direct reflection of the nonlinear enhancement effect driven by multi-factor interactions. Among these， vegetation condition， characterized by normalized difference vegetation index （NDVI） and fractional vegetation coverage （FVC）， stood as the predominant natural driving force， its importance far surpassing that of secondary factors such as mean annual precipitation （Pre） and land use type （Land）. ③ Factor interaction detection confirmed that vegetation factors （NDVI and FVC） acted as a key amplifier， as their interaction with any other factor significantly enhanced the overall explanatory power for the spatial differentiation of the EI. Furthermore， the SHAP explainer quantitatively attributed this nonlinear relationship， precisely revealing the direction and magnitude of each driving factor's contribution under various contexts. This study profoundly concludes that the ecological environmental status of the Qinghai-Tibet Plateau is a composite system resulting from complex， nonlinear responses to multiple drivers， whose macro-scale pattern is dominated by the vegetation-climate system and significantly modulated by land use patterns. |
