| 基于InVEST模型的生态系统服务驱动因子与权衡协同关系：以陕西渭北旱塬为例 |
| 摘要点击 232 全文点击 12 投稿时间：2025-04-02 修订日期：2025-07-21 |
| 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 InVEST模型 地理探测器 时空地理加权回归（GTWR）模型 XGBoost 驱动因素 权衡协同机制 生态系统服务 |
| 英文关键词 InVEST model Geodetector geographically and temporally weighted regression （GTWR） models XGBoost driving factors trade-off and synergy mechanism ecosystem services （ESs） |
| DOI 10.13227/j.hjkx.202504035 |
|
| 中文摘要 |
| 阐明生态系统服务之间的权衡与协同机制及其时空演化规律，是干旱区生态安全格局构建及多功能土地利用优化的关键科学问题. 以陕西渭北旱塬区为研究区，基于2002~2022年陕西渭北旱塬5期遥感及社会经济统计数据，集成InVEST、地理探测器、时空地理加权回归（GTWR）模型、XGBoost-SHAP解释框架及空间偏相关分析方法，系统揭示产水量（WY）、生境质量（HQ）和土壤保持（SC）这3类生态系统服务的动态演变及其驱动机制. 结果表明：①研究期内，产水量、生境质量和土壤保持生态系统服务整体均呈显著增长趋势，增量分别为521.87 mm、0.78和275.83 t·km-2，空间分异显示，延安南部和铜川丘陵地带为生态系统功能提升的核心区；②WY主要受土地利用（LULC）、坡度（SLOPE）和年降水量（PRE）影响；HQ受年均温（TMP）和归一化植被指数（NDVI）影响明显但交互作用趋于下降；SC对SLOPE因子响应稳定，呈“地形主导”结构；除此之外，驱动因子交互作用增强服务响应，能显著提升因子对服务解释力，表现出非线性增强与时序转变特征；③在GTWR模型拟合中，各因子对生态系统服务影响具有空间异质性与动态性；④XGBoost-SHAP模型重要性排序中，不同生态服务的因子重要性顺序存在明显区别，但其重要性排序在研究期未呈现显著的时序分异特征；⑤生态系统服务间交互作用呈现二元分化特征，WY与HQ以权衡关系为主，WY与SC以协同关系为主，HQ与SC则呈现协同与权衡空间均衡格局. 研究通过多模型耦合揭示了干旱区生态系统服务互馈机制的非线性特征和空间异质规律，可为区域生态安全屏障与国土空间精准治理提供量化决策依据. |
| 英文摘要 |
| Clarifying the trade-offs and synergies among ecosystem services and their spatiotemporal dynamics is a critical scientific issue for constructing ecological security patterns and optimizing multifunctional land use in arid regions. This study， focusing on the Weibei dry plateau region of Shaanxi Province， integrates the InVEST model， geographic detectors， geographically and temporally weighted regression （GTWR） models， XGBoost-SHAP explanation frameworks， and spatial partial correlation analyses， based on five phases of remote sensing and socioeconomic statistical data from 2002 to 2022. It systematically elucidates the dynamic evolution and driving mechanisms of three ecosystem services： water yield （WY）， habitat quality （HQ）， and soil conservation （SC）. The primary findings include： ① Temporal evolution characteristics revealed that WY， HQ， and SC significantly increased during the study period， with increments of 521.87 mm， 0.78， and 275.83 t·km-2， respectively. Spatial differentiation identifies southern Yanan and the hilly areas of Tongchuan as core zones of ecosystem service enhancement. ② Driver mechanism analysis indicated spatial variation in explanatory power among different service drivers. WY was mainly influenced by land use （LULC）， slope （SLOPE）， and annual average precipitation（PRE）. HQ was significantly affected by TMP and normalized difference vegetation index （NDVI）， though this influence was declining. SC consistently responded to slope， displaying a terrain-dominated structure. Furthermore， interactions between driving factors enhanced service responses， notably increasing explanatory power through nonlinear enhancement and temporal shifts. ③ Spatial heterogeneity verification via GTWR modeling demonstrated that the influence of each factor on ecosystem services exhibited spatial heterogeneity and dynamics. ④ Machine learning interpretation using the XGBoost-SHAP model revealed distinct importance rankings of factors across ecosystem services， though no significant temporal differentiation in importance ranking was observed during the study period. ⑤ Trade-offs and synergies showed binary differentiation among ecosystem services. WY and HQ predominantly displayed trade-offs， WY and SC exhibited synergistic relationships， while HQ and SC maintained a balanced spatial pattern of both synergy and trade-off interactions. By coupling multiple models， this study highlights the nonlinear and spatially heterogeneous interactions among ecosystem services in arid regions， providing a quantitative decision-making basis for regional ecological security barriers and precise territorial governance. |
