1

I have two large rasters: One defining zones (private properties, with unique value for each property) and one of land cover change (deforestation data, with values for the year it has been deforested).

What I want is to calculate the amount of deforestation for each year on each property. I will later categorize the properties by property size, but I do not want to do it at this stage. And I would like to be able to plot the results spatially as well.

Also, I am interest on the size of the deforested patches, i.e., adjoining raster cells deforested on the same year. Is there a way to calculate this area without transforming into polygons?

I know that using polygons I could use intersect and it would give me exactly what I'm looking for. But these are VERY large rasters, it is simply not possible to use polygons for this. I looked into Zonal Statistics but as values for deforestation are categorical, that doesn't solve it. I'm currently trying to fix a bug while doing Tabulate Area (backgoung processing crashes), but still I wouldn't be able to plot the results.

Is there any different approach for this? I feel it's a simple query that I just haven't found the right tool(s) yet... I'm using ArGis 10.2.2

4
  • 1
    Why can't you use the Zonal Statistics tool? Does not matter if the raster is categorical simply multiply the number of cells by the area of a cell.
    – Hornbydd
    Dec 4 '17 at 14:37
  • Hi Hornbyd! But the zonal stats doesn't give me the number of cells of each year on each property- just the sum, max, min and other statistics that don't mean much on these data. Am I missing something?
    – Liz
    Dec 4 '17 at 14:56
  • A way around it is to reclassify your deforested area into a binary raster or 1 and 0, where 1 is deforested area. Then the sum is the number of cells you multiply by area. May be that can work for you?
    – Hornbydd
    Dec 4 '17 at 17:15
  • Maybe, but then I would have to create a new binary raster for each year, right?
    – Liz
    Dec 5 '17 at 18:07
2

Calculating the area of land cover data with each zone (private property) can be done easily using Tabulate Area tool.

Actually Tabulate Area is the typical tool for your question and since you mentioned that you are facing problem in using the tool in background, I will suggest to run the tool in foreground, sometimes the foreground geoprocessing can solve the problem of crashing issue.

To enable forground go to Geoprocessing Tab -> Geoprocessing Options -> Background Processing -> Uncheck Enable

enter image description here

Also try to make the cell size of both raster data same, you can use Resample tool to adjust the cell size.

2
  • Hi, thanks for the reply! I've tried disabling background activity, but the now the calculation just won't stop (it's been more than 24h). I believe I may have limits on my machine processing power and that's why it's not working. I'll try doing it on a subsample to see if it gives any results.
    – Liz
    Dec 5 '17 at 18:02
  • I already thought of a solution to plot this the way I want later, but what I haven't figure out yet is how to calculate the area of adjoing cells? As stated I do not want to transform to polygons, but I don't know any method that would do this is rasters... Tabulate area gives me the total area deforested say in 2001 in property X. But it's different if it was three different and distant areas of say 10m², or a big 30m² patch.
    – Liz
    Dec 5 '17 at 18:06
0

I figure out a way to accomplish what I wanted, and I'll post this answer for future references.

First I have to create a new binary (0 and 1) rasters for each deforestation year. So, instead of a single raster with values of the year that cell was deforested, I'll have on for each year of interest. Then Zonal Statistics will easily count the cells on each zone- wich gives me the area.

But as I'm also interested on the size of a deforestation patch (adjoing cells deforested on the same year), just Zonal Stats wouldn't do it. I mean, two properties may have deforested the same total area in a given year, but it make a difference if it was on a single big patch or several distant ones (hope I'm being clear). So I can use Region group (http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/region-group.htm) for each year, which will reclassify cells clusters and assign them with a unique value. The table produced can them be joined on the original property layer to produce the maps I'm looking for.

It's not very elegant, but worked on a small test area perfectly.

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.