Often I generate hotspot surfaces at the point level, I'll quickly run through my workflow:

  1. by taking the points, making a copy
  2. running the integrate tool at a 1-mile XY tolerance
  3. I am calculating disease prevalence, so in order to get a point-level prevalence rate, I select by attributes where some disease = 1 to get my numerators. Then I run collect events on the selection. After it's done, I clear my selection, run collect events to get the denominator.
  4. Once those are both output features, I spatially join the numerator to the denominator, with the summary statistic set to minimum.
  5. Once both counts are together, I can calculate prevalence by creating a new field in that spatially joined feature called Prevalence, where MinIcount/Icount * 1000 is the prevalence rate
  6. From there, I have an analysis field to use for the optimized hotspot analysis tool. I run hotspots, get a point-level hotspot results w/ z-scores.
  7. I then run the hotspot output through spline, using the GiZscore field as the analysis field.
  8. From there I classify the raster surface the same way as the hotspot tool outputs the points, and that's my hotspot surface.

Here's a zoomed in example of what the surface looks like with class breaks shown: Optimzed Hotspot Analysis

This is where my question comes into play.

Sometimes, a month or two later, my work will ask for me to only take the "hot" hotspots (GiZscore > 1.95), and delineate them back to a vector. We use this to show blobs that are disease "hot" hotspots, so we can easily overlay them with other thematic choropleth maps.

My approach so far: Use Contour (Spatial Analyst) tool to take those and turn them into lines. When you use the Contour tool set to 1.95, all of the output are polyline but they include lots of contour lines within the > 1.95 boundaries. I then use the Feature to Polygon tool, and I end up having to use Editor because it includes all the class breaks after 1.95, so I have a lot of little polygons inside the big blobs I need to manually delete. I want to manually delete them because I don't want them to be included in the dissolve step I'll do next to dissolve all the delineated polygons into one big blob.

Here's a pic illustrating the areas I'd want to vectorize: To be vectorized

Is there a better way to do this?

  • In general your question is: "how to convert raster to vector with preselected pixel values?" Right? – Mr. Che Jan 30 '17 at 9:43
  • yes, into dissolved polygons – rachel.passer Jan 30 '17 at 17:05

To convert raster to vector with preselected pixel values follow this steps (Spatial Analyst license required):

1) Use Raster Calculator with expression (replace raster_layer, 10500 and 10000 to your values):

Con(( "raster_layer"  <= 10500 ) & ( "raster_layer"  >= 10000),  1, SetNull( "raster_layer",  "raster_layer", "NOT VALUE IS NULL" ))

2) Use Raster to Polygon tool.

Example of first step results: enter image description here

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.