4.3. Local Colocation Patterns
A global measure of colocation patterns implies that the patterns are and remain stationary over space. However, most geographical processes are spatially heterogeneous [34
]. For example, demographic composition differs across neighborhoods, which leads to inhomogeneous distributions of criminal opportunities. Thus, the colocation patterns of crime and land-use features also might vary by location. The inability to detect inhomogeneity might lead to false conclusions about the ways that objects correlate with each other in a local context.
Therefore, we performed a local colocation pattern analysis. The LCLQ calculates the intensity with which a land-use feature invites or deters a given crime (LCLQcrime→land-use feature), and the statistical significance of this relationship (p-value). A LCLQ value significantly greater than 1 means that the occurrence of a given crime strongly depends on the presence of particular land-use features and a LCLQ significantly less than 1 means that the occurrence of a given crime is strongly deterred by particular land-use features.
Due to the spatial homogeneity, individual crimes of the same type might be attracted to or deterred by land-use features at different levels of significance. To obtain an overview, we summarized the LCLQs between the types of crime and types of land-use features, illustrated as horizontal bars in Figure 4
. The grey portion of each bar indicates the non-significant percentage of the LCLQs associated with the crimes and land-use features. The remaining colored portions represent the significant percentages (p
We classified the significant LCLQs into five groups based on the intensities (i.e., values of the LCLQ). Each group has been assigned a specific color for readability of the results. The dark blue portions of the bars in Figure 4
show the percentages of the LCLQs less than 0.5, which indicate a strong deterrence to crimes by the land-use features. The light blue portions of the bars (0.5 ≤ LCLQ ≤ 0.9) indicate weak deterrence, the green portion (0.9 < LCLQ < 1.1) indicates lack of influence, the yellow color (1.1 ≤ LCLQ ≤ 2.0) indicates weak attraction, and the red area (LCLQ > 2.0) indicates a strong attraction.
illustrates that the LCLQs of the three types of crime and the types of land-use features have similar structures. This similarity might be caused by the characteristics of these particular types of crime because all of them are property crimes that might be similarly influenced by land-use features.
However, particular types of crime are attracted to or deterred by different land-use features with different intensities. E-bike thefts were the most common of the three types of crimes. E-bikes offer inexpensive and convenient forms of transportation used by many residents, and they are particularly popular in families. People often ride e-bikes to work or for fun. E-bike thefts are likely to be found near stores, banks, restaurants, and governmental facilities. The overall local colocation patterns between e-bike theft and stores was hierarchal. Stores were the most common land-use feature of the 22 categories of land-use features. These stores are usually privately owned small shops that cater to relatively less affluent customers.
The crime dataset indicated that many victims had intended to temporarily park their e-bikes outside these stores. They had planned to buy something and shortly return to their e-bikes, so many of them left their e-bikes unlocked and forgot to monitor their bicycles while shopping. This carelessness provides opportunities for thieves. Banks are places where people routinely go to manage their money, ordinary restaurants are places where people go to dine, and governmental facilities are places where people conduct public or private business. These four types of land-use features are public places with large numbers of people that are difficult to regulate. This might partly explain why these land-use features attract e-bike thefts.
Only a small portion of industrial plants, service facilities, hostels, schools, commercial buildings, hotels, hospitals, markets, and universities strongly attracted e-bike thefts. The reason for that might be that few citizens traveled to these places by e-bike. In contrast, the remaining types of land-use features mainly deterred crime. These land-use features are either high-class places with security measures, including entertainment venues, office buildings, Internet cafés, cultural facilities, professional facilities, sports facilities, or places unlikely to park e-bikes, such as parking lots, transportation stations, and gas stations.
The local colocation patterns of burglaries with land-use features were similar to those of e-bike thefts. These two types are common property crimes with high exposure rates and low detection rates. Their similar spatial density distributions of crime are shown in Figure 2
a,b above, which, along with their similar characteristics, might result in similar local colocation patterns. However, despite the overall similarity, there are differences. Stores, banks, restaurants, service facilities, commercial buildings, high-class hotels, office buildings, and cultural facilities had higher rates of strong deterrence to burglaries than they had to e-bike thefts. Most of these land-use features are in the middle and southern areas, where there is a large commercial center, a busy pedestrian street, and several scenic spots. These areas are mostly commercial sites, and the proportion of residences is low. Burglaries are crimes that occur in residential areas, and they are less likely to occur in commercial areas.
The local colocation patterns of robbery with land-use features clearly differed from those of e-bike theft and burglary. Overall, the percentages of significant LCLQs for robbery with individual types of land-use features were much smaller than for the other two types of crime. Moreover, different land-use features had different influences on robbery than on the other two types of crime. Stores, banks, and restaurants more strongly attracted robbery, which is a reasonable result because stores and restaurants are places with frequent cash transactions and cashiers and customers are potential victims. Bank employees and customers are easy targets for robbery because they tend to have money on hand. Governmental facilities attract robbery because they are responsible for some financial transactions involving cash, such as payment of water, electricity, heat, cable, and telephone bills.
Of note, a much larger percentage of transit stations invited robberies than e-bike thefts or burglaries. According to our data, almost 20% of all robberies in the study area during the study period were near bus stations, and 31 of them were near Hankou Railway Station. Bus stations are places where people gather, and the bustling crowd and lack of self-protection when getting on or off a bus provide opportunities for robbers. Hankou Railway Station is one of China’s most important railway hubs, located adjacent to Second Ring Road, surrounded by a large long-distance bus station and several local bus stations. The crowds of people in this area make it an ideal place for crime. However, the remaining types of land-use features mainly deterred robbery.
Another important benefit of using LCLQs is that the results are mappable, which provides the opportunity for visual interpretation and differences in the LCLQs can be observed over space. In other words, the spatial heterogeneity of the attraction/detraction intensities of the land-use features on crimes can be intuitively interpreted. Figure 5
presents maps of the local colocation patterns. To conserve space, four of the 22 types of land-use features are shown. Figure 5
shows that the colocation patterns of the types of crime and the types of land-use features were not homogeneous over space, which the global CLQ did not recognize.
A strong attraction of stores to e-bike thefts was mainly found in three areas, identified by the ovals numbered 1, 2, and 3 in Figure 5
a. A large urban village with several wholesale markets and two main streets is located in area 1. Area 2 is the site of a large general hospital and a provincial bus terminal. The southern part of Jianghan District (Area 3) includes Hanzheng Street, which is a historically-prosperous center of trade. As shown in Figure 5
d, banks attracted e-bike thefts mostly in the mid-southern and southern areas, which are the main commercial areas of the district.
Regarding restaurants’ attraction of e-bike thefts, there were three main areas, shown as the areas in ovals 4, 5, and 6 of Figure 5
g. Wuhan No. 1 High School, the Party School of C. P. C. Wuhan Municipal Committee, and several communities are located in Area 4. Areas 5 and 6 are commercial areas, and there is a flourishing commercial pedestrian street in Area 6. Transit stations mainly presented a strong deterrence to e-bike theft (Figure 5
and Figure 5
b,e,h,k illustrate that the influences of land-use features on burglary were similar to those on e-bike theft. However, Figure 5
b shows that stores deterred burglary in the area inside oval 7. This difference was because several railways surround this area of crowded, shabby houses. The relatively poor transportation and economic conditions of this area are unlikely to encourage some types of land-use features, such as stores, banks, restaurants, and stations. On the other hand, the lack of public security measures provides opportunities for burglars. Overall, the land-use features strongly deterred burglaries in this area.
In contrast to e-bike theft and burglary, stores, banks, and restaurants mainly invite robbery. These influences tended to be distributed in the mid-southern and southern areas of the district, as shown in Figure 5
c,f,i. However, Figure 5
l, regarding transit stations, is clearly different because transit stations are far more attractive to robbery than to burglary or e-bike theft. Ovals numbered 8, 9, and 10 show three concentrated areas of attraction. Hankou Railway Station, a long-distance bus station, and several local bus stations are in area 8. Area 9 is between two parks, which are traversed from east to west by a main road. There is also a large department store in Area 9. Area 10 is the channel to a river-spanning bridge, and many bus transfer stations are located there. Area 10 is also near a dock, and it seems reasonable that these transportation hubs are places that should be protected from crime.