During past decades, much progress has been made to elucidate the relationship between the quantity of urban green space and crimes, but less is known about how their spatial distribution is associated with crimes. Therefore, this study aims to understand how the composition and configuration of urban green space are related to crimes in metropolitan cities in the United States. Our study performed landscape analysis using 756 census tracts for violent crimes and non-violent crimes in Chicago from 2017 to 2021 while controlling socioeconomic status. Spatial lag models showed that tree canopies were negatively associated with crime rates, while grass areas were positively associated with crime rates. In terms of the spatial distribution, there were more crime incidents in irregularly shaped green spaces compared to reguraraly shaped green spaces. On the contrary, disconnected tree canopies were more likely to have lower crime rates. Aggregated (i.e., a single large) also tended to have less crime rates. The results of our study provide some evidence of the role of the spatial distribution of urban green space on crimes and shed light on the ways to reduce crimes through effective urban green space planning