Scholarlink

<p>Land use and cover change has garnered widespread attention due to its profound impacts on habitat quality and ecosystem stability. In this study, a comprehensive land use-driven habitat quality assessment framework was proposed to systematically evaluate these dynamic impacts. The primary innovation of this research lies in the development of a coupled spatiotemporal simulation framework that integrates Artificial Neural Network-Cellular Automata (ANN-CA), Shapley Additive Explanations (SHAP), the Patch-generating Land Use Simulation (PLUS) model, and the InVEST model to provide highly accurate scenario projections combined with globally explainable driving mechanisms. The Yellow River Delta, a region experiencing intensified socio-economic activities driven by abundant land, energy, water, and ocean resources, was selected as the study area. Three distinct development scenarios were systematically designed to capture the uncertainties of future land use dynamics up to the year 2030, thereby revealing the fundamental trade-offs between economic growth and ecological security. The results indicated that under the natural development scenario, cultivated land would be maintained at approximately 63.90% of the total area, while critical ecological lands, specifically grassland and woodland, would experience a substantial decline by 2030. Furthermore, low-quality habitat areas would expand to encompass over 30% of the total regional area under both natural and urban development scenarios. Conversely, the ecological protection scenario demonstrated that proactive regional environmental management could significantly improve the connectivity of high-quality habitats and enhance overall ecosystem resilience by actively restoring grasslands and water bodies while strictly restricting the outward expansion of construction land. Consequently, the proposed framework establishes a novel, transparent, and robust methodological paradigm for rigorously quantifying the impacts of land use change on biodiversity, yielding critical predictive insights and actionable guidelines for regional ecological conservation.</p> Graphical Abstract <p></p> <p>This visual summary serves as a pivotal entry point into the research, offering a concise overview of the explainable spatial explicit approach for assessing the impact of land use and cover change on habitat quality in the Yellow River Delta. Regarding the data, Part 1 encompasses preprocessing procedures that unify multisource spatiotemporal datasets from the 2000 to 2020 period, including land cover records, socioeconomic metrics, soil vegetation indices, and physical geographic variables. For the analyses and model components, Part 2 illustrates the prediction process of LULC utilizing the ANN-CA model. This specific model demonstrates robust nonlinear capabilities to generate high precision land use quantity demands for the year 2030. To resolve the inherent opacity of complex computational techniques, SHAP analysis is implemented to quantify the contribution of various driving factors, thereby establishing clear global interpretability. Part 3 details the multiscenario simulation process. Crucially, the multiscenario analysis is executed within the PLUS model, which allocates spatial patterns strictly utilizing the quantity thresholds of the natural development scenario previously predicted by the ANN-CA model. The study systematically designs three future pathways: natural development, urban development, and ecological protection. Focusing on the results, Part 4 highlights the integration with the InVEST model to evaluate dynamic habitat responses. The resulting spatial maps present detailed habitat quality patterns across the distinct future scenarios, identifying specific vulnerable ecological zones. The overall conclusion indicates that coupling the precise quantity prediction advantage of the ANN-CA model with the spatial allocation of the PLUS model provides an academically rigorous framework. This methodology effectively quantifies biodiversity dynamics, delivering essential predictive insights for sustainable land management and ecological conservation in the Yellow River Delta.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Explainable Machine Learning-Driven Spatial Explicit Approach for Projecting the Impact of Land Use/Cover Change on Habitat Quality: A Case Study of the Yellow River Delta, China

  • 期刊论文

  • 发表日期：

    2026年6月15日

查看全文
  • Shaohong Guan,
  • Weizhi Gao,
  • Bingyi Wang,
  • Yuncheng Jiang,
  • Zhaoli Wang,
  • Yulei Xie

摘要

Land use and cover change has garnered widespread attention due to its profound impacts on habitat quality and ecosystem stability. In this study, a comprehensive land use-driven habitat quality assessment framework was proposed to systematically evaluate these dynamic impacts. The primary innovation of this research lies in the development of a coupled spatiotemporal simulation framework that integrates Artificial Neural Network-Cellular Automata (ANN-CA), Shapley Additive Explanations (SHAP), the Patch-generating Land Use Simulation (PLUS) model, and the InVEST model to provide highly accurate scenario projections combined with globally explainable driving mechanisms. The Yellow River Delta, a region experiencing intensified socio-economic activities driven by abundant land, energy, water, and ocean resources, was selected as the study area. Three distinct development scenarios were systematically designed to capture the uncertainties of future land use dynamics up to the year 2030, thereby revealing the fundamental trade-offs between economic growth and ecological security. The results indicated that under the natural development scenario, cultivated land would be maintained at approximately 63.90% of the total area, while critical ecological lands, specifically grassland and woodland, would experience a substantial decline by 2030. Furthermore, low-quality habitat areas would expand to encompass over 30% of the total regional area under both natural and urban development scenarios. Conversely, the ecological protection scenario demonstrated that proactive regional environmental management could significantly improve the connectivity of high-quality habitats and enhance overall ecosystem resilience by actively restoring grasslands and water bodies while strictly restricting the outward expansion of construction land. Consequently, the proposed framework establishes a novel, transparent, and robust methodological paradigm for rigorously quantifying the impacts of land use change on biodiversity, yielding critical predictive insights and actionable guidelines for regional ecological conservation.

Graphical Abstract

This visual summary serves as a pivotal entry point into the research, offering a concise overview of the explainable spatial explicit approach for assessing the impact of land use and cover change on habitat quality in the Yellow River Delta. Regarding the data, Part 1 encompasses preprocessing procedures that unify multisource spatiotemporal datasets from the 2000 to 2020 period, including land cover records, socioeconomic metrics, soil vegetation indices, and physical geographic variables. For the analyses and model components, Part 2 illustrates the prediction process of LULC utilizing the ANN-CA model. This specific model demonstrates robust nonlinear capabilities to generate high precision land use quantity demands for the year 2030. To resolve the inherent opacity of complex computational techniques, SHAP analysis is implemented to quantify the contribution of various driving factors, thereby establishing clear global interpretability. Part 3 details the multiscenario simulation process. Crucially, the multiscenario analysis is executed within the PLUS model, which allocates spatial patterns strictly utilizing the quantity thresholds of the natural development scenario previously predicted by the ANN-CA model. The study systematically designs three future pathways: natural development, urban development, and ecological protection. Focusing on the results, Part 4 highlights the integration with the InVEST model to evaluate dynamic habitat responses. The resulting spatial maps present detailed habitat quality patterns across the distinct future scenarios, identifying specific vulnerable ecological zones. The overall conclusion indicates that coupling the precise quantity prediction advantage of the ANN-CA model with the spatial allocation of the PLUS model provides an academically rigorous framework. This methodology effectively quantifies biodiversity dynamics, delivering essential predictive insights for sustainable land management and ecological conservation in the Yellow River Delta.