From Health Impact Assessment to Decision Support: An Optimization Framework for Prioritizing Urban Greening Interventions Under Spatial Constraints

Abstract

Introduction

Urban greening has become a key strategy for promoting population health and supporting sustainable urban planning. Although Health Impact Assessments (HIAs) can quantify the potential health benefits of alternative greening scenarios, they rarely provide practical guidance on how limited green infrastructure should be spatially allocated to maximize public health gains under real-world planning constraints.

Objectives

This study aimed to develop a spatial optimization framework to prioritize urban greening interventions at the census tracts level in the municipality of Florence (Italy). The proposed algorithm allocates additional tree cover while accounting for land availability constraints and maximizing the reduction in attributable deaths (AD) associated with greenness exposure.

Methods

Deaths in 2025 among residents aged ≥35 years were redistributed from the municipal to the census tracts level using a multinomial allocation weighted by the local age and sex population structure. Exposure in 2023 was measured as mean Tree Cover Density (TCD). Since the exposure-response function, derived from the literature, was estimated for the Normalized Difference Vegetation Index (NDVI), a generalized additive model (GAM) was used to calibrate TCD into NDVI. Uncertainty in the exposure-response function and stochastic mortality allocation was propagated through Monte Carlo simulations and incorporated into the optimization framework. An iterative greedy optimization algorithm was then developed to allocate additional tree cover by maximizing the marginal reduction in AD per square meter of new tree cover. At each iteration, tree cover was incrementally assigned to the census tract with the highest expected marginal health benefit, subject to realistic spatial constraints, including the available plantable area and a maximum target of 10% TCD, consistent with the 2024 Nature Restoration Law.

Results

Among the 2,168 census tracts in Florence, 395 had already achieved the 10% TCD target in 2023, leaving 1,773 sections eligible for additional greening interventions. The proposed optimization framework allocated new tree cover to 1,740 census tracts while using approximately 8.5% of the total potentially plantable area available across the municipality. Overall, the optimized allocation increased TCD and NDVI across eligible census tracts and was associated with an estimated reduction of approximately 48 deaths among residents aged ≥35 years. The prioritization strategy consistently selected census sections with the highest expected marginal health benefit while satisfying all spatial constraints.

Conclusions

The proposed framework extends conventional HIA by integrating statistical modelling with spatial optimization to support evidence-based urban greening policies. Although still largely theoretical, this approach represents a step toward a practical decision-support tool for identifying where limited greening resources can generate the greatest public health benefit under realistic implementation constraints.