Elliptic moving window for focal statistics with rotation angle from a table?

[stefansteger]

Hello to all!

For a long time I have been searching for a solution of my problem...

The problem is as follows:

I need a elliptical search window for focal statistics (Max), where the major axis is oriented due to an angle from a table (in my case from aspect values so that the ellipse is oriented downslope).

I thought about creating a kernel file for irregular neighborhoods (which is ok for my purpose), but there is no possiblity to rotate this due to a table. I have very large data (>500km², 10m*10m), so a somehow manual creation of ellipses isn't possible, since i have to check that for every pixel... For Wedge Neighborhoods there exists the possiblity to have an angle input and to orient the window, but not for rectangular or irregular.

For your information: I am dealing with landslides (idealised form: ellipse) and I have absolute no idea about programming :(.
Would it be possible to create a phyton script (for toolboxes) by somehow adding the possibility to rotate the window due to an additional input? Is that a big task or only a matter of a few minutes?

I just saw that some interpolation techniques in arcgis have the possibility of a user defined elliptic search window (e.g.: http://resources.arcgis.com/de/help/main/10.1/index.html#//00300000000n000000)...

I would be very glad if someone can help me!

Thanks you a lot!

e.

[curtvprice · Accepted solution]

I need a elliptical search window for focal statistics (Max), where the major axis is oriented due to an angle from a table (in my case from aspect values so that the ellipse is oriented downslope).

We have tackled a similar problem with oriented well buffers for water-quality data analysis.

Our solution was to generate a collection of kernel files (say, 36 of them, one for each 10 degrees of azimuth, ker0.txt to ker35.txt), calculate azimuth codes (say - 0 to 35). Then you could set up a model that iterates on your azimuth code field values to run zonal statistics, using the Calculate Value tool to generate the kernel file name from the azimuth code on each iteration.

[stefansteger]

We have tackled a similar problem with oriented well buffers for water-quality data analysis.

Our solution was to generate a collection of kernel files (say, 36 of them, one for each 10 degrees of azimuth, ker0.txt to ker35.txt), calculate azimuth codes (say - 0 to 35). Then you could set up a model that iterates on your azimuth code field values to run zonal statistics, using the Calculate Value tool to generate the kernel file name from the azimuth code on each iteration.

Thank you very much for your help!

The kernel iteration sounds very promising...

So, you created 10 different raster, right? How did you get your final raster?
In my case I need one raster with the maximum values of the corresponding ellipse... if I combine them simply by their maximum value I will get also ellipses on my map which dont represent the actual aspect of the slope... is their a possiblity to apply focal statistics only on defined raster cells? (e.g. raster calculator: If aspect ==10 then apply focal statistics with kernel10.txt etc...)??

[curtvprice]

So, you created 10 different rasters, right? How did you get your final raster?

That's right (but my suggestion was 35, not 10); say the rasters are: s1,s2, ... s35, each generated with zonalstatistics with the appropriate kernel file for that aspect code (1 to 35).

Then with Raster Calculator and the Con tool, use the aspect code (here in raster "asp", integers from 1 to 35) to pick which aspect results to use:

Con("asp" == 1, "s1", Con("asp" == 2, "s2", Con("asp" == 3, "s3" ... ))))))))

Another (quite nifty) approach is to use the Pick tool, again assuming the asp codes are 1 to 35:

Pick("asp", "s1", "s2", "s3", ..., "s35")

[stefansteger]

Then with Raster Calculator and the Con tool, use the aspect code (here in raster "asp", integers from 1 to 35) to pick which aspect results to use:

Con("asp" == 1, "s1", Con("asp" == 2, "s2", Con("asp" == 3, "s3" ... ))))))))

Another (quite nifty) approach is to use the Pick tool, again assuming the asp codes are 1 to 35:

Pick("asp", "s1", "s2", "s3", ..., "s35")

Thank you very much! That approach sounds feasible for me...You helped me a lot!!

Greets,
stefan