once and for all: what is the most accurate way to calculate the area of intersecting polygons (i.e. how much of one polygon is within another)?

[AurelieShapiro]

This is probably the most important operation anyone has to do with GIS: calculate the area of something within something else. I can't think of a single day when I don't have to do this. So why is there no reliable method?

I am calculating the total area of forest within provinces and districts. I don't use tabulate area because you have to convert the outputs from m2 to something else and it's a pain. I saw the Summarize Within tool in ArcGIS Pro and thought gee that does exactly what I need. 

All my data are in cylindrical equal area projection.

The results I get from using Summarize Within however are greatly different from if I clip my polygon to a province area and add geometry attributes with geodesic area. We are talking large differences here - 20% or more. I have lots of provinces and districts and clipping to each one will take an eternity. 

How are people supposed to reliably calculate area if the GIS tools provided give completely different results? Is there reliable information as to which tool to use and which is more accurate? I am wondering now how much error there are in all the forest area reporting out there when using different tools seems to create huge differences.  

[MervynLotter]

Are you referring to the Tabulate Intersection GP tool  when you mentioned "Tabulate area"?

I use the Tabulate Intersection—Help | ArcGIS Desktop  for these kinds of analyses and it lets you specify the output units. Unfortunately this tool does require a ArcGIS Desktop Advanced licence.

[AurelieShapiro]

I was talking about tabulate area in ArcGIS, but Tabulate Intersection does the same thing - and provides different results/calculations than other tools. The question is, which is more accurate??

[DanPatterson_Retired]

Provide a sample of some geometry you are working with.

The geometric calculation of area is often dependent on the geometry and how it is projected... even with geodesic methods.

[AurelieShapiro]

yes of course, but when running two tools on the same dataset in the map project and the same projection, the results should in theory be the same? Or closer? here are some data: provinces and forest. area - Google Drive 

The differences get much larger over bigger areas...

[MervynLotter]

I was under the impression that the Tabulate Area tool was for use with raster data but I now see it can do both. 

I took a closer a look and when working with vector data, both the Tabulate Area and Summarise Within tools rasterize the data and a cell size needs to be set for Tabulate Areas. 

I checked these three tools against my own data and the difference was minor, around 0.007%. Tabulate Areas was less similar than the other two but I suspect if I had to decrease cell size, it would match up.

I then checked these three tools on the dataset you provided on Mangroves in central Mozambique and again the results were consistent. 

For example:

Tabulate Intersection: 85 376 ha of mangroves

Tabulate Areas (50 ha cell size given fragmented extent of mangroves): 85 333 ha (I converted to ha)

Summarise Within: 85 376 ha

Therefore results from Tabulate Intersection and Summarise Within are identical and Tabulate Areas only 0.05% off. 

Perhaps try again in a new view? 

[AurelieShapiro]

Thank you very much - my question was which method is more accurate? Because with larger areas the differences are much larger...this was only a subset of my dataset. Over the entire globe the differences between the methods ends up much larger...

[MervynLotter]

Hmm, I would go for Tabulate Intersection as data does not need to be rasterised. It is also the only tool where you can set an output tolerance. 

I suspect the discrepancies you have seen may be related to the calculation of area on a planar surface versus a geodesic surface. However I suspect the tools would all be using a planar surface. Something to look into using the Calculate Geometry tool to compare with various outputs. 

I also noticed that I meant to write that the output cell size for the Tabulate Areas tool was set to 50 m not 50 ha. 

[AurelieShapiro]

Thanks again!

So this alludes to my original comment - why can't we simply select planar or geodesic area (which according to what I have read is more accurate) as an option for these tools? And when we work with rasters it's a whole other mess...

[AurelieShapiro]

So I am using tabulate intersection to calculate the forest cover for different areas.

My forest cover is a polygon for which I have calculated geodesic area.

Tabulate area seems to sum up the geodesic area (I selected SUM for my AREA_GEO column), but the tool also produces an "Area" column, which in the geo tool I specified I wanted in hectares. These two values are different - about 2-10% different: the Area column different from the sum of the geodesic area. And these values are also different from the results I get if I just CLIP the layers and tally the area. So which is correct??