Resampling based on majority when there is a large difference in resolution

[SvenVerweij]

Hello!

For a project I am trying to resample a raster containing soil classes from 10 meter resolution to a 30 arc seconds resolution. I did this with the Resample tool and used as Resampled Technique: Majority, because I want the soil type that covers the largest area of the output cell. However, looking in the detail at the resampled raster, it seems it took the soil type at the centre of the cell and not the soil class covering the largest area of the cell. According to the documentation of the Resample tool, the Majority technique uses only the 4 by 4 cells closest to the middle of the cell, resulting practically the same as the Nearest Neighbour technique due to the large resolution difference. As an alternative, I used the Block Statistics tool, but here the same seems to happen. In the figure below this is illustrated. The 30 arc seconds cell gets soil type C, while it seems to make more sense it would get B (or A).

So to summarize my question: Does anybody know how to convert the resolution of discrete raster based on majority and not the value at the centre?

Thanks in advance!

Sven

[DanPatterson_Retired]

As soon as you mention arc second, then I think this should be the first step

    Project Raster—Help | ArcGIS for Desktop 

Or was the destination unit just for convenience? If it is indeed 30ish meters, I would still project it if you are going to geographic coordinates.

if you are actually downsizing and you have categorical data, then nearest neighbour is your only valid option

    Cell size and resampling in analysis—ArcGIS Help | ArcGIS for Desktop bottom paragraph

You have ruled out aggregation (Aggregate) I take it because you do have categorical data and changing the 30ish meter data to 10 m (via projection/resampling) is not an option?

[SvenVerweij]

Thanks for you answer, but the raster has to be changed from 10 meters to 30 arc second, because all the other layers and the analysis will be done at 30 arc second. Aggregate is not possible because it is categorical data, as you already mentioned. The raster is already projected to geographic coordinates, but 10 meter was the original resolution. It is a pity that only Nearest Neighbour is an option for categorical data and not the areas of the classes in the cell can be used.

[XanderBakker]

An option could be to have several steps of Resample (Majority) with a factor 2, but that is not something I recommend.

Depending the size, you could consider converting the raster to polygons and convert the polygons back to raster with the "MAXIMUM_COMBINED_AREA" cell assignment option and the proper cell size defined.

Polygon to Raster—Help | ArcGIS Desktop