AI-based decision support for sustainable landscape regeneration in aging residential communities

Aging residential communities are increasingly confronted with significant landscape deficiencies, including inadequate tree canopy coverage, elevated land surface temperatures (LST), fragmented pedestrian networks, and insufficient outdoor rest facilities. These deficiencies disproportionately affect elderly residents who rely on walkable, thermally comfortable outdoor environments to maintain physical activity, social connections, and overall well-being. This study develops and evaluates a reproducible public-data framework for identifying and comparing landscape-regeneration needs in aging residential communities, utilizing directly measured accessibility, thermal, ecological, and service-access variables. We integrate multi-source geospatial data, including Sentinel-2 imagery, Landsat-8 thermal data, OpenStreetMap infrastructure, census demographics, and digital elevation models across 156 residential communities in Seoul, South Korea. An XGBoost-based prioritization model achieved an area under the receiver operating characteristic curve (AUC) of 0.891 and an F1-score of 0.847, outperforming logistic regression (AUC = 0.712), random forest (AUC = 0.856), and weighted index baselines (AUC = 0.783). SHapley Additive exPlanations analysis revealed that summer LST, tree canopy coverage below 15%, and walking distance to healthcare facilities exceeding 800 m were the strongest predictors of regeneration priority. Intervention scenario modeling demonstrated that combined tree planting and accessible path reconstruction were associated with the largest improvements in outdoor usability scores (38.2% increase) under standardized intervention intensities. This framework provides municipalities with transparent, data-driven tools for prioritizing landscape interventions in communities serving aging populations.