Selecting defined area by suitability

Question

Hi,

I have calculated a raster layer on a scale of 0 to 1 defining suitability for a certain activity where 0 is least suitable and 1 is most suitable.

Is it possible to make a selection on the layer of the "best" 30 square miles for the activity? This isn't all going to take the value 1 and will go down to I imagine about 0.85 but I need to extract the sites that are most effective.

Many Thanks

MervynLotter · Answer

Esri have just very recently released a geoprocessing tool, Locate Regions, that will assist in selecting the best areas from the output of a suitability or similar analysis. It is currently only available in Pro 1.3 but I believe it would also be available in ArcMap 10.5, due for release at the end of this year.  This tool is part of the Spatial Analyst toolbox.

In the tool you can ask it to select the best site, or any number of best sites, of any given size, shape, distance from other best sites, etc.

It looks like a fantastic tool, do see Locate Regions—Help | ArcGIS for Desktop  and How the Locate Regions tool works—Help | ArcGIS for Desktop .

JayantaPoddar · Answer

Let's assume the total area is 120 sq. miles. To achieve the best 30 sq. miles (approx.), I would Reclassify the raster using Quantile classification with 4 classes. The class having the highest range of values is the most suitable ~30 sq. miles (top 25%).

The area won't be exact since many pixels would have same value (which could be the threshold value of a particular class).

DanPatterson_Retired · Answer

well, you just need to get some criteria... You have come up with a suitability and chances that people would agree to the same threshold value would be slim.  So   if you examine your histogram for the raster in the symbology you should see approximately where the midpoint is.  You could also do statistics on the raster and get the means and std deviations... perhaps the most suitable areas... in your opinion or others... have a threshold of mean + 1std (ie > 0.66).

So this small, process-stopping detail, is where you enter the realm of modelling.  Do it manually so you know what you are doing.  If you suspect that you are going to have to do this a number of times to get it right... you could automate the whole process in modelbuilder or a python script.

Have fun

PS a quick look at the initial map with an appropriate symbology and clicking around with the info tool would help... as would the contour tool... at least you have a starting point to see whether there is any 'clustering' of areas within the higher range... and it doesn't look lie a speckaled mess.