Ask A Cartographer

Analyzing Bike/Ped Crash Data

November 02 2010 | 0 comments
Categories: Symbology

I would like to analyze the location of about 320 bicycle and pedestrian crashes in a city over a 6 year period. About 2/3 of the points have XY values. The other third, I manually placed using a description of the location. I would like to display the locations in a more meaningful way than just points. Would I be able to do a dot density map or some type of heat map? Attached is a map I found in another plan. Is it possible to replicate this in GIS? The points all have dates associated with them. I have Arc10 and went to the training and wrote down "generate schematics." I do not have much experience with the spatial analysis tools and am not even sure what that means. Any and all suggestions are most welcome.

Mapping Center Answer:

The method shown in your attachment uses lines of varying width to represent the number of accidents on each segment of the road network. You can do this if you have a numeric attribute for each road segment that identifies the number of accidents for each. Then you can use the Quantities - Graduated Symbol renderer on the Symbology tab of the Layer Properties.

Graduated Symbols

Below is an example in which I used the FlowClass attribute to show the amount of flow in each river segment.

Graduated Lines

Since you have point data, you would have to assign each road segment the accident data that is contained in the points. There are a number of ways you can do this – if this is the solution you want to try, let us know and we can tell you how to go about this.

On the other hand, you mention a heat map approach, which I take to mean a density surface. That is a raster solution that would result in a surface that tells you how many points fall in a neighborhood around each sell in the raster. Below is an example in which I mapped the density of cities along the rivers.

Point Density

You also mention a dot density map – that is not the solution since that is a way to symbolize polygons based on a numeric value (see our blog entries on this on Mapping Center). You do not have polygon data so you do not need to consider this mapping method.

You also mention schematics – that is probably overkill for what you are trying to achieve here. There are easier and more straight forward ways to achieve a good looking and communicative map for your data. For example, you say that you do not want to simply map the points, but such a map, in which you simply map the location of each point, can be very illuminating. To display overlapping points so that they can be seen, use a casing (a thin outline in a different color) on the symbol.

Having said all that, I encourage you to look at some of the tools in the spatial statistics toolbox as these may provide an even better solution. There are several short videos to get you started at bit.ly/spatialstats. You could, for example, use the hot spot analysis tool to create a map showing statistically significant hot/cold spots. Given that you have 6 years of data, you could also use the new temporal mapping tools to animate the distributions over time. If you use the Hot Spot Analysis tool, you will want to aggregate your point data before running the analysis. Strategies for doing this are discussed in Learn more about how Hot Spot Analysis works (see the section titled Hot Spot Analysis Considerations).

If you would like to post a comment, please login.

Contact Us | Legal | Privacy |