My final project was to examine crime incidence in the city of San Francisco. According to FBI reports this past year San Francisco is a huge city with a populace of over 805,235 people. It is also a minority- majority city where there are more ethnic populations than white as a whole.This presents an opportunity to examine where factors of poverty based on median income, high school drop out rate, and distance to police stations can influence the locations and incidence of crime. By examine several of these factors, I hoped to pinpoint areas of need and consideration for the city of San Francisco in terms of future crime prevention and areas of focus in resource allocation regarding public safety.
Map 1: Crime Incidents in San Francisco 2003
Map 1 displays the hotspots for crime incidence in San Francisco. In this large city, in 2003 there were nearly 15,000 crime incidents reported. This includes both violent and nonviolent crimes. As you can see from the map, they are concentrated mostly in the defined planning districts of the Downtown/Civic Center, South of Market, Chinatown, Bayview, and the Mission district. Data for the crime incidents in 2003 was found at https://data.sfgov.org/ and a point shapefile was derived from the latitude and longitude of this data to analyze it through a kernel density raster to find hotspots.
Map 2 is a kernel density analysis for hotspots of crime in 2012. By examining the hotspots of crime, although 2012 had only about 10,000 incidents of crime in comparison to 2003, it is clear that the hotspots for crime remain almost the same. Still in 2012, the areas of the Mission District, South of Market, Downtown/Civic Center, and Bayview seem to have the highest incidence of crime. In addition the Haight and Ashbury area and the Castro area have further defined areas of high crime in comparison to 2003. Overall in all types of crimes, Map 2 shows little difference in terms of areas of focus for crime prevention from 2003 to 2012. I was hoping to show a reduction in hotspot areas because according to report regarding FBI crime data for counties and cities, San Francisco’s crime rate was greatly reduced from 2004-2011. Map 2 was also derived from data from https://data.sfgov.org/ for 2012 crime and made into a point file which was used for the kernel density analysis.
Map 3 : Daytime Violent Crime Hotspots in 2012
To further examine violent crimes: murder, rape, assault and robbery, I took a look at the density of violent crime during the day and the evening for the 2012 data only. To first weed out only the violent crimes, I used an excel formula function to select those that can be defined as violent crimes. I exported that into a separate excel sheet. I also coded the excel crimes that could be considered daytime crimes and nighttime crimes and left them in the violent crime data.To show only the violent crime during the day, I used the select by attribute tool in ArcMap to select only the “day” labeled crimes for the kernel density analysis.
As you can see from map 2, the hotspots of violent crime during the day are fairly consistent with the incidence of all crimes in 2012. So it goes to show that even during the day, violent crimes occur within the same location parameters as all crime occurrence.
Map 4: Daytime Violent Crime Hotspots in Proximity to Police Stations
In Map 4, I took that same violent crime data for daytime incidences and layered a buffer of San Francisco Police Stations. The light yellow buffers surrounding the police stations show a proximity of a 1 mile difference. The data for the police stations was taken from the http://sf-police.org website which allowed me to create an excel sheet and add in location and address information and geocode the information to create a new point shapefile through ArcGIS.
Looking at the police station proximity, one can see that in the highest risk areas, of the Mission District and Downtown/Civic Areas, and Bayview areas, they are all within a 1 mile distance from police stations. That doesn’t seem to be a deterrent for incidence of violent crime during the day.
Map 5: Night-time Violent Crime Hotspots 2012
Map 5 is taking a look at violent crime incidents at night from the 2012 data. Since I already defined night-time and daytime incidents, I selected only the incidents that occurred at night with the select by tool in ArcMap to create this kernel density analysis of hotspots of violent crime incidence at night. I thought there might be a larger area of night time incidents because overall there were more night-time violent incidents, but it was consistent with Map 2 and the overall hotspots identified for 2012.
Map 6: Night-time Violent Crime Hotspots in Proximity to Police Stations
Map 6 shows the proximity of violent crime hotspots from night-time incidents to 1 mile buffered areas around police stations. So you can see that that even though to a certain extent, police do have areas that cover most of San Francisco within 1 mile buffered zones, that doesn’t seem to influence or affect the hotspot areas indicated by violent crimes that occurred at night.
Map 7: Population Density and Day-time Violent Crimes
Map 7 shows the population density for San Francisco by census tract which shows a graduated density with the deeper purple areas as the areas of most population density. One of the factors that could lead to more crime in any area could be population density and this map looks at the possible association. However, the violent crimes during the day are not inherently consistent with the most populated areas. For example, the Outer Sunset and Parkside areas are very populous but do not have as many incidences of crime as the aforementioned Civic/Center and Mission areas.
Map 8: Population Density and Daytime Violent Crimes with Proximity to Police Stations
Map 8’s data looks at the population density and the proximity to police stations interposed on the population density. Since police station jurisdiction ideally would influence the occurrence of violent crime, we can see that despite that, the areas that are most populated don’t exactly correspond to areas of highest crime.
Map 10:Population Density and Night-time Violent crimes with proximity to Police Stations.
Maps 9 and 10 examine population density and areas of violent crime occurrence during the night-time hours. These maps were taken by taking census data for population in census tracts and creating a graduated polygon ArcGIS shapefile. By looking at population density and the occurrence of night-time violent crimes, it is apparent that population density cannot be one or the only key factor to influence incidence of violent crimes.
Map 12: Service Areas of Police Stations and Violent Crime Incidence in 2012
Maps 11 and 12 are looking at a network analysis by juxtaposing San Francisco roads to create an area of service for the police stations. Starting with the geocoded police stations, I was able to find a roads shapefile from https://data.sfgov.org/ to create a network analysis file. From there using the network analysis tools for service areas around the defined locations of the police stations, I defined 2 minute, 4 minute and 6 minute service areas that could be easily reached from the police stations. The caveat being that network analysis does not always take into consideration traffic. I could not find all current traffic conditions to change the miles per hour, so from the network analysis most of the whole city of San Francisco can be reached within 6 minutes from the defined centers of the police stations.
Thus looking at all the incidences of violent crime for 2012, at first glance it seems that police stations for the most part were reachable within the 2-4 minute ranges. So once again it is seen that even with very close proximity and access to police stations, the violent crimes occurred regardless.
Map 13: Raster Analysis of Median Income, Unemployment Rate, Distance to Police Stations
Map 13 examines the potential for crime occurrence based on the factors of Median Income, Proximity to Police Stations, Unemployment Rate, and High School Drop out Rate. This information was derived from census data and modeled through a composite raster of these 4 determinants. First the unemployment rate, high school drop out rate, and unemployment rate were changed to rasters through the feature to raster function. The proximity to police stations was derived by measuring distance of each centroid of census tract to the police stations. Then all 4 were reclassified through a model and weighted equally to create this composite raster analysis of areas of greatest need.
Since population density, or proximity to police stations wasn’t enough to determine areas of greater crime, adding all four of these determinants allowed for a fuller picture of where could be the greatest areas of risk for the city of San Francisco. However, they do not exactly correspond with the incidence of hotspots for crimes in 2012. Still, though, gradation shows that the areas that are red and orange show the greatest need and those do correspond with the areas of need in the Civic/Center Downtown and Mission areas, which have been seen to have the greatest incidences of crime.
Just to examine a possible 3D analysis I took a terrain shapefile from https://data.sfgov.org/ to show hills and higher elevations. However, the 3D analysis in Arcscene didn’t exactly work out to create a 3D file. However, you can see the blue dots which are violent crime incidences can be seen as occurring on the flatter terrain versus the more hilly terrains. This doesn’t prove anything per se, but it was interesting to see that outcome.
Findings and Further Examination:
Through this project I was able to look at a variety of conditions and variables that could influence or indicate why crime was occurring in specific areas. For an analysis of where crime could occur based on median income, proximity to police stations, unemployment rate, and high school drop out rate, specific areas of need were determined.CRIMEgisStellaKIM2 Because these determinants could also contribute to other needs based socioeconomic factors, it would be interesting for further study to examine these areas more in depth and see if there are other mappable factors that are present besides crime incidence. For example, perhaps lower test scores ,less access to schools and other factors could be explored. As for the major areas of crime incidence, for future analysis, the data presented only represents reported crime incidence. However, there is data available in the crime data that indicates if that crime was “solved” or if someone was caught at the scene of the crime. It would be interesting with further study to analyze what areas have not only the highest rates of crime, but also to see the areas where the highest areas of crime “solving”.
Arcmap is always a bit fickle with data, and even though data was thoroughly prepped with excel beforehand, transferring the XY latitude longitude data from a spreadsheet to ArcGIS presented problems because of the projections. Even with the correct XY latitude longitudes, and the correct and consistent projections, the data kept on not meshing on the same plane. So that took hours and hours in order to make the rasters work. All in all, although I may not use all of these skills, in the future I hope to be able to map community needs through these mapping skills.