1

So I have a database of lats/longs representing incidents occurring on or near roadways.

My goal is to identify the street block areas with the highest number of incidents in the database. I don't have any other map layers I'm using for this.

What I'd like to do is, for each point, see how many other points are a given defined distance away (a quarter of a mile, for instance). The result would then show where the clusters are.

Any recommendations on the best way to do this? I have copies of both QGIS (with the spatial query plugin) and ArcMap, although I'm much more familiar with QGIS.

2

In ArcGIS, if you want to know the number of point at a specified distance to a given point, you can use the "Point distance" tool.

Another approach would consist in using the integrate tool in order to group points in a neighborhood to one location (WARNING: this will modify your data, use this on a copy of your feature class), then create a buffer around the remaining points and use spatial join to get the count.

However, in your case you then want to make clusters. So, in ArcGIS, you could use the point density tool, then reclassify and finally "raster to polygon. In QGIS you can do the same steps with the Heatmap plugin.

As a final remark, it is often better to avoid working in Lat/long when it comes to measuring distances. Try to project your points in a local XY coordinate system.

  • Thanks for your help. Unfortunately, my ArcGIS license doesn't include access to those features. I tried using the QGIS heatmap plugin but (as far as I can tell) it only generates an image. I need something similar to the Point Distance tool that creates output that can be used to determine the number of points within a given distance. Are there any other options for QGIS? Also, I should note that these are lat/longs for the entire country. Is there a best national coordinate system to use when measuring distances? – Mike Jun 2 '14 at 14:31
  • there are some free tools for arcgis in the spatial ecology toolset if you don't have advanced licence spatialecology.com/htools/pntdistmatrix.php. For the heatmap solution, you can use either the raster calculator or r.recode to make a categorical raster, then you can convert this raster to polygon (poligonize) – radouxju Jun 2 '14 at 18:03
  • for distance, the best is equidistant projection, but those only preserve distance along a set of principal lines. Usually I would trust the people who designed projections for their countries, because they are often optimal compromise. – radouxju Jun 2 '14 at 18:06
1

You're wanting to perform a density or possibly cluster analysis. In ArcGIS there are the Density Tools of Spatial Analyst, or the Analyzing Patterns Toolset and Mapping Clusters of Spatial Statistics, if you have access to either. I'm more familiar with ArcGIS, but I know in QGIS there is a Heatmap Plugin.

Most related question: How to build effective heat-maps?

Hopefully this, or searching on some of the terms contained therein, will point you in the right direction. You might also want to scan the Related link list on the right. Also, I'm assuming you know you'll need to convert your x/y (lat/lon) list into an actual point feature class if that isn't how it exists in the database and you haven't already done so.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.