Optimizing urban agriculture for long-term ecosystem services: Integrating scenario-based suitability analysis and the landscape approach

Abstract

Uncontrolled urban sprawl intensifies socio-ecological pressures, demanding planning strategies that measurably enhance urban ecosystem services. Urban agriculture is a promising lever, yet its long-term ecosystem services contributions remain insufficiently quantified. This study addresses two critical questions: (1) How can suitability analysis guide the spatial integration of urban agriculture to optimize long-term ecological benefits? and (2) How can a landscape approach-based urban agriculture planning strategy be designed to align with ecosystem services enhancement goals? We develop a transparent, reproducible pipeline linking machine-learning suitability modeling (XGBoost with SHAP), deep-learning land use simulation, and monetary ecosystem services valuation. Using Rotterdam as a case study, we simulate three development scenarios for 2030 and 2050: Business-as-Usual (BAU), Suitability-Based Autonomous Transformation, and Suitability-Based Landscape Approach Transformation. Suitability-guided scenarios outperform BAU, with the landscape-approach scenario delivering the most stable multi-decadal outcomes for regulating and cultural services. However, provisioning services can plateau or even decline when ecological protection constraints limit intensive production, revealing the limits of land allocation alone. We conclude by offering thresholds and rules that translate suitability and scenario outputs into a transferable urban agriculture planning model, enabling planners to embed urban agriculture within a landscape approach as part of broader sustainable urban transformation.