Suitability Modeler - Modelling negative factors/influences

810
2
Jump to solution
02-08-2022 12:58 PM
RInfante
New Contributor II

Hello everyone,

So I'm evaluating the use of the Suitability Modeler for a site selection project we have going on and am wondering if there is a way to model a negative influence.

For example, our client has identified hazardous environmental setbacks that they want to have an active, negative influence in the process. 

My understanding is that the values in the "Suitability" column of the Transformation Pane are limited by whichever Suitability Scale was chosen in the Settings panel. None of these settings include negative ranges, which makes sense on a 'final output' level, but as it is right now the only way I see to model a negative influence at a parameter level is to pick a wide enough range (I'm using 1 to 10) that'll allow me breathing room in the upper numbers while any negative factors would be 'floored' to a value of 2 (I'm using 1 as the bottom 'NoData' values).

Is this correct? Is there no way to actively model detriment other than to set the values as 'not as good but still within a positive scale'?

I guess I understand the logic that that might be sufficient on a conceptual level, but in my conversations with the client they feel that there's a difference between including an active, negative influence at a parameter level and having to 'dilute' that factor due to a scale constraint.

I know that this could be handled if I programmed some raster math processes on my end, but yeah, we were hoping we could use the Modeler as an 'out of the box' solution.

Thanks!

 

 

 

 

 

 

1 Solution

Accepted Solutions
KevinMJohnston
Esri Contributor

The Suitability Modeler uses a weighted sum approach. The higher the value the better. The suitability scale is used to put the different criteria onto a common scale and the scale needs to be the same for all the criteria so they can be combined. The criterion cannot have different suitability scales.

One approach to incorporate negative influences is by reducing the suitability value by a specified amount. A second approach is having the negative influences contribute less to the final suitability score. The math is different between the two cases. Your logic follows the second approach. Negative influences could be assigned the lower suitability values. You could do it as you suggest selecting a large range but another possible approach might be is to use weights. For the later approach, you would need to have the negative influences as separate criteria from the positive. Positive criterion can be assigned higher weights (just a multiplier) and the negative factors assigned lower weights. Using weights might allow you to explore the impact of the negative effects. For example, in the most basic case, the positive criteria are assign a weight of 2 and the negative criteria a 1. But if the negative criteria are not that severe, you can reduce the weights of the positive criteria to say 1.5. Either assigning lower suitability values or using weights will take a fair amount of thought to achieve the desired effect.

As you have identified, you can do the subtractive approach outside the Suitability Modeler in Map Algebra or in ModelBuilder (or a hybrid of all three), but I assume you want the interactive aspects of the Modeler. But, currently in the Suitability Modeler, you can not reduce the suitability value by a specified amount (a negative value). You can only apply the second method - reduce the negative influences’ contribution to the suitability.

You mention NoData. NoData means there is something at a location that makes it so the location should not be considered (within a legal buffer of a wetland) when locating the subject. If any location in any of the input criteria is assigned NoData, that location will not be considered no matter what the suitability values are in the other criteria. This is different than assigning a location a “1”. A “1” means it is less suitable, not that the location should be removed from consideration. A location assigned a “1” in one criteria might be assigned high suitability values in the other criteria possibly making the location desirable. If that location is assigned NoData instead of a “1”, it will be removed from consideration no matter what the suitability is in the other criteria.

If there are locations that should be removed from consideration (e.g., within a wetland or is too steep), a more flexible way to remove them from consideration is using a mask. It is unclear if the environmental setbacks that you refer to are locations that should be removed from consideration or if they are a negative influence.

View solution in original post

2 Replies
KevinMJohnston
Esri Contributor

The Suitability Modeler uses a weighted sum approach. The higher the value the better. The suitability scale is used to put the different criteria onto a common scale and the scale needs to be the same for all the criteria so they can be combined. The criterion cannot have different suitability scales.

One approach to incorporate negative influences is by reducing the suitability value by a specified amount. A second approach is having the negative influences contribute less to the final suitability score. The math is different between the two cases. Your logic follows the second approach. Negative influences could be assigned the lower suitability values. You could do it as you suggest selecting a large range but another possible approach might be is to use weights. For the later approach, you would need to have the negative influences as separate criteria from the positive. Positive criterion can be assigned higher weights (just a multiplier) and the negative factors assigned lower weights. Using weights might allow you to explore the impact of the negative effects. For example, in the most basic case, the positive criteria are assign a weight of 2 and the negative criteria a 1. But if the negative criteria are not that severe, you can reduce the weights of the positive criteria to say 1.5. Either assigning lower suitability values or using weights will take a fair amount of thought to achieve the desired effect.

As you have identified, you can do the subtractive approach outside the Suitability Modeler in Map Algebra or in ModelBuilder (or a hybrid of all three), but I assume you want the interactive aspects of the Modeler. But, currently in the Suitability Modeler, you can not reduce the suitability value by a specified amount (a negative value). You can only apply the second method - reduce the negative influences’ contribution to the suitability.

You mention NoData. NoData means there is something at a location that makes it so the location should not be considered (within a legal buffer of a wetland) when locating the subject. If any location in any of the input criteria is assigned NoData, that location will not be considered no matter what the suitability values are in the other criteria. This is different than assigning a location a “1”. A “1” means it is less suitable, not that the location should be removed from consideration. A location assigned a “1” in one criteria might be assigned high suitability values in the other criteria possibly making the location desirable. If that location is assigned NoData instead of a “1”, it will be removed from consideration no matter what the suitability is in the other criteria.

If there are locations that should be removed from consideration (e.g., within a wetland or is too steep), a more flexible way to remove them from consideration is using a mask. It is unclear if the environmental setbacks that you refer to are locations that should be removed from consideration or if they are a negative influence.

RInfante
New Contributor II

Thank you for your reply Kevin! I apologize for seeing it so late...I don't frequent the forums and for some reason I assumed I'd get a notification that a reply had been posted.

The client was requesting for the setbacks to be counted as negative influences as opposed to omitted altogether. Now thanks to your answer I know that I'll have to find a non-direct way to model that (as you said: through lower weighting or suitability values).

So actually the reason I am using the NoData as '1' is because I am trying out a transit service analysis that uses different inputs that by themselves don't have equal coverage or extents (as opposed to say, slope or vegetation rasters that wouldn't normally have any 'gaps') .

So, for example: a specific service area for bus line X might not be present in a location, but a bunch of other ones (A,B,C) might. If I input the service area raster for X  and I don't modify the NoData to '1', then as you said I don't get to see how A,B,C factor in that location despite X's absence because X's NoData negates the rest of the inputs as well. If, on the other hand, I replace all NoData as '1' then at least I get a baseline floor upon which other factors can come into play.

I know that it's not the 'traditional' scenario that the Suitability Modeler might have been designed for (most examples are about locating this or that using 'naturally continuous' rasters), but yeah, at the end of the day I was assessing how the tool could be used for this kind of analysis.

Thanks again for your response and clarification!

 

 

 

0 Kudos