Ye, C., Chen, Y., & Li, J. (2018). Investigating the influences of tree coverage and road density on property crime. Isprs International Journal of Geo-Information, 7(3), 101. doi:10.3390/ijgi7030101

The research is conducted to discover the correlations between tree coverage and property crime, and road density and property crime in the city of Vancouver. The researchers argued that crime analysis, prevention, and mapping have benefited from the quickly developed GIS techniques. Previous studies have analyzed some factors that affect crime rates such as population density, poverty level, unemployment rate. However, the impacts of vegetation density and road density on the crime rate have long been under debate, especially in Canada. Hence, the research hoped to fill in the blank and further supported decision making in urban property crime prevention.

For the vegetation coverage data, the research obtained a 2*2m high spatial resolution 2013 orthophoto. Researchers classified the vegetation coverage area and conducted an accuracy assessment. The producer’s and users’ accuracies reached 96.9% and 99.9%, respectively. For road density, they get data from the road network data from the Statistic Canada, and the density was calculated as the ratio of the sum of the road lengths to the land area. Data also included other factors that have already been proved related to crime in the previous studies, including population density, income, a lone parent, unemployment rate, light density, and graffiti. All the data were used as the dependent variable of the crimes in the later regression models.

The research used the cross-sectional analysis by using GeoDash and ArcGIS software. First, it calculated the global Moran’s I in GeoDash to examine the spatial autocorrelation of crime incidents. Then, it conducted an Ordinary Least-Square (OLS) regression model in GeoDash to examine the correlation between crimes and dependent variables. The OLS model ran for total property crime (theft, mischief, and break-in), theft crime and break-in crime respectively. Then, since the OLS ignores the spatial autocorrelation of crime data, a spatial lag model was applied to examine the spatial dependence. Finally, Geographically Weighted Regression (GWR) tool in ArcGIS was used to examine the local correlation at the DA’s level.

The OLS results showed significant negative correlations for both vegetation coverage and road density with each type of crime rate. However, the R2 index for both three types of crime is lower than 0.2, meaning a low fitness of model. The spatial lag result increased the R-squared value, confirming a better performance. The GWR result showed an overall better performance compared to the OLS model. Hence it had various local coefficients. The GWR reflected a negative relationship between poverty crime rate and the vegetation coverage rate closer to the downtown area while showed a weak positive correlation in Stanley Park. Moreover, for road density, there is always a negative correlation.

Overall, the research accomplished the goal to discover the spatial pattern and variation between tree coverage and property crime, and road density and property crime in the city of Vancouver. They also admitted the limitations of research. For example, the study did not differentiate urban trees along streets and the trees in parks, Planning trees in a different location may have a different influence on the crime rate. The research gave suggestions on designing more complex road system and planning more trees in the downtown area in the aim of crime reduction. Overall, I give the research 8/10.