Environmental and Management Drivers of Magnolia silvioi Mortality in a Tropical Andean City

Abstract Background Evidence of urban tree mortality is mostly limited to temperate regions; little is known about mortality in tropical cities or threatened native trees used in greening programs. Magnolia silvioi (IUCN Endangered) has been widely planted in Medellín since 2009, outside its natural range. Methods We combined Medellín’s Urban Tree System (SAU) records with field assessments of 178 individuals to score survival, phytosanitary condition, site context, and management history. Random Forest classification models (500 trees) were trained using tree, site, spatial, and maintenance-related predictors. Model performance was evaluated with out-of-bag (OOB) error and class errors. Results Overall mortality was high (119/178; 66.8%). Of the 59 survivors, 54% showed decline symptoms. Mortality remained similarly high across natural, infrastructure, and built settings (approximately 65% to 71%). A reduced model with 6 predictors (latitude, longitude, pollution index, pruning, interventions, and planting date) improved performance (OOB = 20.8%), with latitude, pruning, longitude, and interventions as top predictors. Trees receiving more interventions and pruning exhibited lower predicted mortality. Conclusions Magnolia silvioi shows high mortality and widespread decline symptoms across urban areas of Medellín. Spatial variables were the strongest predictors of mortality, likely reflecting differences in geomorphology and jurisdictional management while consistent maintenance mitigates risk more than microsite variables. Management intensity, particularly pruning and repeated interventions, was consistently associated with reduced mortality risk. Findings support adaptive monitoring, risk-based maintenance, and cross-municipal coordination to sustain conservation-priority trees in tropical cities.