Data acquired:
- City Boundary SHP- City of Vancouver Open Data Catalogue
- Block Outlines SHP-City of Vancouver Open Data Catalogue
- Crime CSV File-City of Vancouver Open Data Catalogue
- Park(polygon features) SHP -City of Vancouver Open Data Catalogue
- Street Lighting Pole (node) SHP-City of Vancouver Open Data Catalogue
- Rapid Transit Lines SHP-City of Vancouver Open Data Catalogue
- Rapid Transit Stations SHP-City of Vancouver Open Data Catalogue
- Vancouver outline- UBC Geography File Database
- Dissemination Areas Boundary Files -Statcan
Seasons and Times
Months determined by meteorological seasons. Times determined by earliest sunset and latest sunrise per each season.
Step 1) Sorting and Filtering of CSV files.
The first step to this project was the sorting and filtering of the CSV file which included all of the 2016 Vancouver crime data. As the file included all of the crime data since 2003 the first issue was ensuring that only one years worth of data was included. I chose 2016 as it was still recent and completely finished. Though 2017 could have been used, it would not have included crime reports for November and December of 2017. In order to get the crime data for 2016 i filtered by year. The result was then a total of 34,612 crimes. From there i then sorted by crimes and created multiple work books for each season and time such that in total I had 8 workbooks.
Step 2) Excel work
As this project involved a large amount of data and occurred over an extended period of time, i decided it would be useful to visualize the results in graph form.Therefore, I created multiple graphs in excel to show how in each season crime changed between day and night time. As well, I wanted to show how crime was occurring at each of my measurable variables (Street lights, City Parks, SkyTrain Translink stations). This data to be as accurate as possible was shown both in the form of total sum and as a proportional measurement such that they could then be compared side by side. Lastly i then created a final graph to show how total crime varies between day and night hours, though as will be discussed later this was a point of uncertainty.All graphs from this step can be found in the menu under Graphs.
Step 3) Create a ‘base’ in ArcMap
Then next step was then to create a geodatabase and import all of the shapefiles noted within the Data Acquired. Following I then added the 8 aforementioned workbooks from Excel which will acted as all of the data for crime.I then created 8 data frames with one for each season/time. These 8 data frames were then the working areas for my preliminary maps.
Step 4) Buffering each variable
In order to encompass an area which could be studied for each variable I decided that it would be necessary to slightly buffer each point and polygon. My buffer for parks was 20m, which I decided upon as it was not a large buffer that I felt would impact results but was reasonable enough that it would be able to surround any possible crimes on the outskirts. This buffer as well was small enough that it did not change any of the visuals of the parks. Transit stations were buffered differently than parks as I wanted to take into account crime which may be occurring in the direct vicinity such that a block’s radius could be considered. For Transit stations I chose a buffer distance of 80m as when I measured Vancouver block lengths on Google Maps this distance appeared to allow for near blocks to be considered without creating a buffer which was too large that it would include further away blocks. One issue which did arise early in this project was how to properly buffer street lights. I took many different approaches to the buffering of street lights with each one incurring drastically different results. The logic behind the buffering of street lights was that lights illuminate the surrounding area and therefore should in turn deter crime where they shine. Unfortunately I was able to find no data on how far street lights shine,with the exception of a source citing hundreds of feet. The two buffer distances I originally tried were 1m and 10m, both of which created results which seemed incorrect as they either isolated the lights in an impractical manner or created large light tracts which as well seemed wrong. More over, I had to consider direct illumination which would be the focal spot of where the light shines compared to the dimming effect as the light reaches further away. I did eventually come to the almost arbitrary decision of 3m which though I think a strong case can be made for the consideration of where the light may shine brightest, it undoubtedly is a point of contention.
Step 5) Crime attributes and spatial joins
With the 8 excel crime tables I was able to import them into ArcMap as XY data as they had coordinates attached. Following that I then exported all of the data for the tables into shapefiles. This allowed for all of the crime data to become far more flexible in ArcMap. From there i was able to open up the properties and assign all of the values for the category of TYPE. This meant that the data could be shown as points on my map, and though it was not a point of interest I was as well able to see which points where which type of crimes. At this point I then had buffered points and polygons for my parks,stations and lights and a shapefile with all of the points for crimes. The last step before joining the layers was opening each crime shapefiles attribute table. In there I created a new column of data which i called Count. More over using the field calculator I then gave each point a value of 1. This allowed for each point to have an associated value which would later be used to calculate density. Following this step I then spatially joined the individual buffered layers for parks, lights and transit stations to the crime layer where I joined based on spatial location and summarized based on sum. This allowed for the total number of crimes occurring based on the count column to be calculated within the proximity of each buffer. The result of this can then be seen in the preliminary maps for the street lights, parks and SkyTrain stations. As well I exported the attribute tables from each of the resulting spatial joins into Excel. From there I graphed how many crimes were occurring in total for each category of buffer, and how these crimes were proportionally committed in order to compare results. The links to these graphs can be found below.
Step 6) Final Maps
Following the creation of all the preliminary maps and graphs I then created two final data frames named day and night. From each of the previous 8 data frames I copied the spatially joined layers of parks, lights and stations. As well I copied the basic layers needed for my maps into each data frame so that every data frame had the Vancouver base, dissemination areas, block outlines, and rail lines. At this point I wanted to create two maps which would show how crime changes temporally between day and night for all of the seasons combined. In order to do this i needed to do a series of spatial joins in order to ensure that only the locations where crimes occurred in every season would be presented. Therefore I did a series of spatial joins similar to the ones from earlier steps were i joined based upon spatial location and summarized based on sum. At this point I had two final maps which showed the locations where crimes committed between all four seasons at night and at day were. Due to the fact that in my preliminary maps it could already be largely inferred where these crimes were occurring I decided to change how the data Fiwas being portrayed. I decided upon two separate methods of a percentage analysis and a hot spot analysis. For both methods I decided to use 5 natural breaks. The percentage maps show using a sequential color scheme and graduated symbols (circles) what percentage of crime is occurring for the locations where crimes occur in every season. In turn, the hot spot analysis takes into account all of the locations where crimes have occurred and then presents the data through confidence intervals where hot spots of crime are likely, and where non significant spots are. These maps can be accessed by the link below.