Detailed Description:

High-resolution, digital, remotely sensed imagery not only offers the ability to rapidly capture and update urban landscape information but it permits the identification and mapping of urban features associated with criminal activities not frequently recorded on traditional GIS maps. The Omega Group, a San Diego, California, software development and consulting firm specializing in creating crime analysis software applications for law enforcement agencies requested NASA ARC assistance in exploring the crime analysis potential of remotely sensed imagery. An ARC project, through San Diego State University, was initiated in April of 1999 to evaluate the feasibility and utility of using digital imagery to identify geographic or "place" features commonly associated with criminal activity. The study also evaluated the ability to assign crime event probabilities according to locational characteristics derived from high-resolution, remotely sensed imagery in combination with census and ground-based information.

The City of Carlsbad, California, in cooperation with The Omega Group, offered to participate in the study and provided GIS map layers and computerized crime data. Only Part I crimes were considered in the study as they are considered by the FBI to be the most serious of criminal events and consist of murder or negligent manslaughter, rape, robbery, assault, burglary, larceny, vehicle theft and arson. Crime event records for the period from December 1995 through June 1998 were used in the study and consisted of over 10,300 crime events.

Three types of analyses were conducted as part of this ARC project. We first examined the ability to identify land cover features as surrogate identifiers of crime event attractors. We sought to determine if there was a surrogate relationship between urban features and the location of crime events. Next, we examined the use of remotely sensed images to create new variables that may help to identify geographic locations that have a lower or higher probability for crime. We also examined the use of imagery to measure nighttime illumination in relation to crime events.

The results of our first analysis revealed that property crime locations are not spatially random and that place characteristics almost certainly influence the decision to commit a crime. The study also suggests that certain land use activities are more attractive to criminal activity that others. The data suggest that types of land uses such as shopping centers are magnets for large numbers of people with a resulting increase in the likelihood that crime will occur. Locational factors such as proximity to freeway on/off ramps, location adjacent to major thoroughfares, and the presence of parking lots that serve multiple businesses contribute to the propensity of crime.

The study suggests that parcels occupied by middle and high schools are unique crime attractors. Residential parcels in close proximity to school parcels experience an increase in crime. There is evidence that multi-residency apartments in proximity to parking lots may also be crime attractors. Locational factors that make land parcels attractive to a crime can be observed and mapped from high-resolution remotely sensed imagery.

The second analysis examined the use of remotely sensed images to create new variables to help identify geographic locations with a lower or higher probability for crime. The result demonstrates the potential usefulness to crime analysts of combining census data (or other demographic data) with variables derived from remotely sensed images to assess where crime is occurring. The demographic data profiled the propensity for crime in an area, and the remote images further pin-pointed the opportunity for crime, drawing upon features such as non-vegetated areas, parking lots, and commercial areas that are generally associated with elevated risks of criminal activity.

The third analysis examined the use of imagery to measure nighttime illumination in relation to crime events. The research team photographed the nighttime illumination of the study area where crime rates are known to be highest. We then mapped the illumination pattern derived from the imagery and related the intensity of the observed light to the number and location of crime events for the same area. The percentage of crimes occurring within the lighted areas in the daytime (when lighting would presumably not be a factor deterring crime) were compared to the same locations at night (when presumably lighting would serve as a deterrent). The results suggest that nighttime lighting may be associated with a reduction in crime in those lighted places. The study further differentiated between non-residential (largely commercial) and residential areas (based on the land use classifications). In non-residential areas, 25 percent of all crimes occurring during the daylight hours occurred within the polygons bounded by the areas that are lighted at night. However, at night, only 13 percent of crimes were committed in those lighted areas. Clearly the lighting does not eliminate crime, but these numbers suggest a 50 percent reduction in crime in the lighted areas at night, compared to those same places during the day. In the residential areas, the data also suggest a deterrent effect from the lighting. The overall conclusion is that remotely sensed images add value to policing and to crime analysis. We have shown that remotely sensed images can be a useful addition to the mapping and interpretation of crime data; and that remotely sensed data can be combined in a GIS with other coverages and other sources of data to increase the understanding of criminal activity in a community. The uses range from the simple, but effective, use of a digital imagery as a background to a street map for police tactical work, to the more complex process of classifying remotely sensed images and using the resulting data as variables in models that predict the location of criminal activity.