Select to view content in your preferred language

# Question about classification in weighted overlay

3190
4
09-27-2020 07:44 AM
New Contributor II

Hi, I am doing an weighted overlay analysis with five rasters. I am using a common scale from 1 to 10 (10 being most suitable and 1 being the least suitable). I have some questions about two of the rasters:

One raster is originally a vector dataset of buildings I want to stay away from as much as possible, but not exclude completely. Thus, the raster cells containing building are classified to 1 (less suitable), but my question is, what do I classify the surrounding raster cells where there are no buildings? Is it correct to classify them to 10 - most suitable?

I also have a raster where the situation is the opposite: this dataset contains an area I want the analysis to stay as close to as possible (originally a vector dataset now converted to raster). Therefore, the raster cells which contains this wanted area is classified to 10, but what do I classify the surrounding cells to? In this case the surrounding raster cells are not the least suitable, they are just not as suitable as the cells with the wanted area...In my head it would be wrong to classify them as 1 (least suitable).

1 Solution

Accepted Solutions
Regular Contributor

If the suitability is gradually increased as one moves away from buildings, the you could try running the new Distance Accumulation GP tool (or the legacy Euclidean Distance GP tool) with your building as feature source data. The other input fields are optional. The output will be a raster surface with the lowest value of 0 (your buildings) with values increasing as you move away from your buildings.

Next you can use the Rescale By Function GP tool to reclassify all your values between 10 and 1. See Rescale by Function (Spatial Analyst)—ArcGIS Pro | Documentation

There are various transformation functions that you can use and these relate to the kind of suitability and how quickly it increases or decreases.  Do see this link for a detailed explanation of options The transformation functions available for Rescale by Function—ArcGIS Pro | Documentation

Then you can use the same approach for your second raster but reverse the From scale and To scale values in the Rescale by Function tool.

Good luck.

4 Replies
Regular Contributor

If the suitability is gradually increased as one moves away from buildings, the you could try running the new Distance Accumulation GP tool (or the legacy Euclidean Distance GP tool) with your building as feature source data. The other input fields are optional. The output will be a raster surface with the lowest value of 0 (your buildings) with values increasing as you move away from your buildings.

Next you can use the Rescale By Function GP tool to reclassify all your values between 10 and 1. See Rescale by Function (Spatial Analyst)—ArcGIS Pro | Documentation

There are various transformation functions that you can use and these relate to the kind of suitability and how quickly it increases or decreases.  Do see this link for a detailed explanation of options The transformation functions available for Rescale by Function—ArcGIS Pro | Documentation

Then you can use the same approach for your second raster but reverse the From scale and To scale values in the Rescale by Function tool.

Good luck.

New Contributor II

For the buildings raster I have already applied a buffer of 100 meters, which reflects the distance I want to stay outside of. So all cells outside of these buffers are suitable per se.

Is it correct in this case to classify all surrounding (cells outside of this 100 m buffer) to 10 - most suitable?

For my second raster it would perhaps be more relevant to use the distance approach because I have not applied a buffer already. Also, this is not a raster with cells that should be avoided, but there are cells that should be prioritised more than surrounding cells.

Regular Contributor

So if the suitability is the same outside of your 100m buffer, then it should be appropriate to assign them all the same value of 10.

For your second raster it makes sense to use the distance approach from your polygon features provided there is gradient of decreasing suitability from your source polygons.

New Contributor II

Thank you, this really helped! I am going to use both methods and examine the results and the differences. Therefore I am marking your first answer as the "correct" answer, but both answers helped me.