Select to view content in your preferred language

Issue: Suitability Modeler - Issue with default transformation applied to a raster - Limited reclassification options

2490
3
Jump to solution
05-14-2021 02:52 PM
Labels (1)
ScottRothberg__Ph_D_
New Contributor II

I am posting here with the goal of learning the best method to apply reclassification and edit class values and suitability values associated with rasters in Suitability Modeler. I have encountered a limitation and I need to determine if it is a workflow limitation or a limitation of the new and continuously improving Suitability Modeler interface.

---For context, I am a GIS professional with +10 years of experience and I also teach advanced GIS courses at the University of California. Historically I have taught suitability analysis using classic approaches, such as using a combination of Reclassification tools and Weighted Sum tools in Model Builder. I have started to teach Suitability Modeler and I and my students have encountered a reoccurring issue I am hopeful to resolve---

In Suitability Modeler, I understand that when a new input raster is added to the Criteria list a default transformation is applied based on the type of the input values. I further understand that this default transformation is derived from one of three methods you can use to transform the criteria (Unique Categories, Range of Classes, Continuous Functions). My issue manifests in this transformation pane.

Let's use an example of the American Census Survey and Renter Occupied Housing Units. This is the percent of total housing units that are renter-occupied, on a scale of 0.0-100 percent, and these values are attributed to feature classes representing each census block group in the US. In this example, we perform a feature to raster conversion which gives us a new raster. When added to the map the default symbology doesn't normally make sense in the real world (like the first class being 0-3.39%), so we would adjust the symbology to have manual intervals with 10 classes such as 0-10, 10-20, etc. *These are the classified values that we would like to see as our classes in the Transformation Pane in Suitability modeler*. So at this time the raster is ready to be added as a criterion. When we add it to the Criteria list two things occur: 1) In the Contents Pane under the Suitability Modeler Group, our raster is added but it looses the manual intervals we had set up previously (aka the first class goes back to being 0-3.39), and 2) In the Transformation Pane, our unique categories transformation or range of classes transformation also replaces the manual intervals with new start and end values (aka the first class goes back to being 0-3.39).

In the Transformation Pane the start and end values are not editable, so I cannot build out the manual intervals; Also the classes are not editable, so I cannot merge classes or add new ones that would make sense for the dataset being used.

What is the appropriate workflow supposed to be?

 

https://pro.arcgis.com/en/pro-app/latest/help/analysis/spatial-analyst/suitability-modeler/the-gener...

In the link above, in Step 2 > Paragraph 2 > last sentence, you can see a different, perhaps simpler example using the Bobcat study example. I have read through this a few times and I don't see how it is possible to accomplish the task.

 

https://www.esri.com/videos/watch?videoid=Hsq3Man1iyU&title=suitability-modeling

 

Likewise in this link at minute 3:10, you can see an example of applying a linear transformation to slope, but that is once again not the same thing as establishing classes for slopes as 1-3, 4-6, and >25.

0 Kudos
1 Solution

Accepted Solutions
KevinJohnston2
New Contributor

Hi Scott,

You have identified a limitation in (what you correctly refer to as) the “continuously improving Suitability Modeler”. Editing the ranges of the classes in the Range of Classes method has been in plan from the initial release of the Modeler but we have not yet implemented the capability.

You can work around the problem by:

  • Outside the Suitability Modeler, reclassify the raster of the census blocks applying the ranges you want using the Reclassify tool. Accept the sequentially increasing Values
  • Optionally, add a Text field to the resulting raster and enter the assigned ranges for the Values, if you wish to track them in the Modeler.
  • Enter the resulting raster as a criterion into the Modeler. Think of it as the census blocks are not the criterion but the desired ranges of the census blocks is the criterion.
  • When you Transform the raster, by default the Unique Categories will be applied.
  • You can select the added Text field as the Field so you can see the ranges you set.
  • Assign the appropriate suitability values.

Because of your request, we have moved the priority of this editing capability to the 2.9 development que. You are correct, you should be able to enter the census raster directly in as a criterion. When you Transform it, you should be able to select Range of Classes and type in the desired ranges.

The Modeler works on the actual values of the input criteria. We do not plan to support the symbology set to the layer before it was entered in as a criterion. This interactive layer display capability is useful before you use the Reclassify tool, but the Modeler has the interactive display capability imbedded in the Modeler environment. We want the user to consciously set the ranges based on meaningful values relevant to the subject of the model.

Thanks.

Kevin J

View solution in original post

3 Replies
KevinJohnston2
New Contributor

Hi Scott,

You have identified a limitation in (what you correctly refer to as) the “continuously improving Suitability Modeler”. Editing the ranges of the classes in the Range of Classes method has been in plan from the initial release of the Modeler but we have not yet implemented the capability.

You can work around the problem by:

  • Outside the Suitability Modeler, reclassify the raster of the census blocks applying the ranges you want using the Reclassify tool. Accept the sequentially increasing Values
  • Optionally, add a Text field to the resulting raster and enter the assigned ranges for the Values, if you wish to track them in the Modeler.
  • Enter the resulting raster as a criterion into the Modeler. Think of it as the census blocks are not the criterion but the desired ranges of the census blocks is the criterion.
  • When you Transform the raster, by default the Unique Categories will be applied.
  • You can select the added Text field as the Field so you can see the ranges you set.
  • Assign the appropriate suitability values.

Because of your request, we have moved the priority of this editing capability to the 2.9 development que. You are correct, you should be able to enter the census raster directly in as a criterion. When you Transform it, you should be able to select Range of Classes and type in the desired ranges.

The Modeler works on the actual values of the input criteria. We do not plan to support the symbology set to the layer before it was entered in as a criterion. This interactive layer display capability is useful before you use the Reclassify tool, but the Modeler has the interactive display capability imbedded in the Modeler environment. We want the user to consciously set the ranges based on meaningful values relevant to the subject of the model.

Thanks.

Kevin J

ScottRothberg__Ph_D_
New Contributor II

It is good to hear that the issue we identified is in queue. Thank you very much Kevin for providing a list of workarounds, particularly using the Reclassify tool to organize our criteria rasters. Ultimately for my students it reinforces one of my key lessons: Widgets like Suitability Modeler and Geostatistical Wizard are excellent resources, but a good GIS practitioner needs to understand the fundamental tools that are automatically being deployed by these resources.

KevinMJohnston
Esri Contributor

Hi Scott,

We got the following into 2.9. In the Transformation pane, the following options are available for the Range of Classes transformation method: 

  • Enter suitability values directly into the transformation table 
  • Change the number of classes 
  • Specify the classification method to be Natural Breaks, Quantile, Equal Interval, Defined Interval, Manual Interval, Geometric Interval, or Standard Deviation 

I hope this will help.

0 Kudos