• Reviews 21 studies applying optimization frameworks to green roof systems. • Classifies simulation-based and surrogate-assisted multi-objective methods. • Identifies key trade-offs among energy, hydrology, cost, and thermal comfort. • Shows substrate, vegetation, and irrigation parameters dominate performance. • Highlights the lack of integrated multi-physics and data-driven frameworks. • Proposes a next-generation optimization model for multifunctional GR design. Green roofs (GRs) constitute a key nature-based solution for mitigating the adverse effects of urbanization, including heat accumulation, stormwater runoff, and biodiversity loss. Over the past two decades, research has progressed from empirical evaluations of GR performance toward computational frameworks that optimize their multifunctional behavior. This review systematically synthesizes and critically examines Multi-Objective Optimization methodologies applied to GR systems, with emphasis on Simulation-Based Optimization and Surrogate-Assisted Optimization approaches that enable efficient exploration of complex, high-dimensional, and multi-criteria design spaces. An extensive review of the available research identified 21 studies that explicitly applied optimization methodologies to GR systems, are analyzed according to their optimization objectives, methodological configurations, and spatial scale of application. Although recent advances have improved the analytical rigor of GR research, most optimization-oriented studies remain fragmented, prioritizing a limited subset of objectives—most frequently building energy performance and stormwater management. Thermal comfort and microclimate regulation are often treated as secondary objectives or embedded within energy- or spatial-cooling metrics, and cross-domain feedback between thermal, hydrological, and ecological processes remains insufficiently explored. A systematic synthesis of reviewed optimization frameworks reveals that integration between physical simulations and data-driven surrogates is often modular rather than fully coupled. Simulation engines and surrogate models are predominantly linked through sequential or hierarchical workflows rather than dynamic co-simulation, and few frameworks incorporate uncertainty quantification, life-cycle metrics, or adaptive vegetation dynamics. Nature, however, functions synthetically—through continuous interactions among thermal, hydrological, and ecological processes. GRs embody this principle, transforming inert surfaces into living systems that improve energy efficiency, water regulation, and carbon sequestration while enhancing urban well-being. Future optimization frameworks must therefore emulate this natural integrative logic through coupled multi-physics and data-driven modeling, positioning GRs as adaptive mediators within the urban energy–water–ecology continuum.
Mihalakakou et al. (Wed,) studied this question.