While urban trees and park vegetation are widely recognized for their benefits to the environment and human well-being, the association between tree canopy cover and crime risk remains contested. Trees can help prevent crime by providing psychological benefits and enhancing natural surveillance; however, dense, unmanaged vegetation may also increase crime risk by obstructing visibility and creating secluded spaces. This dual nature complicates urban forestry and park management, yet many studies have assumed a linear relationship between tree canopies and crime risk, potentially leading to oversimplified or divergent conclusions. To address this limitation, we analyzed 2081 urban parks in Seoul, South Korea, first employed a generalized additive model to capture non-linear patterns and identify structural thresholds, followed by a semi-parametric specification using piecewise linear terms for tree cover to precisely estimate the magnitude of the relationship. Our results revealed distinct canopy cover thresholds linked to crime prevention: Crime risk was initially associated with 1.9% increase per 1 %-point increase in tree cover, but this association reversed beyond 48% cover with a 0.8% decline. Such negative association became stronger above 89% with a 2.9% decline. Additionally, we explored the interaction between trees and other park features, finding that low tree cover combined with moderate grass coverage was associated with a higher crime risk. These findings emphasize the importance of balancing tree cover with other environmental factors to optimize safety and sustainability in urban parks. By providing quantitative evidence of the role of tree cover in crime prevention at different levels, our study offers valuable insights for park design and urban greening.