topic Re: Focal statistics coordinate shift in ArcGIS Spatial Analyst Questions

Focal statistics coordinate shift

JohannaTorppa1 — Fri, 16 Jun 2017 10:13:15 GMT

In focal statistics, if the size of the neighbourhood is larger than a specific limit (for example 2 for circular radius), the output raster is shifted by the size of the neighbourhood... Seems like a bug, or is there a way to avoid this? ArcGIS Pro version 1.4.1.

Re: Focal statistics coordinate shift

JayantaPoddar — Fri, 16 Jun 2017 10:27:40 GMT

What happens if you use Snap Raster (Environment setting)—Geoprocessing | ArcGIS Desktop to the original Raster dataset?

Re: Focal statistics coordinate shift

DanPatterson_Retired — Fri, 16 Jun 2017 13:43:04 GMT

An image might help..

This goes beyond a snap raster issue perhaps, have you explicitly set all your environment options in the Environments tab of the tool? Cell size, extent, snap raster etc all need to be set explicitly (never assume anything). and the calculation is made for the core cell in the neighbourhood specified

Re: Focal statistics coordinate shift

JohannaTorppa1 — Tue, 11 Jul 2017 12:23:49 GMT

Thank you both Jayanta and Dan !

Below is a the result FS image (grayscale), input raster (layer below the FS image) and imput parameters + environmental variables.

How I see it is that the center point is shifted by radius-1 pixels to north-west. Output crd stystem is that of the input raster and Current Map, so no transformations are needed. The raster used in Snap and Mask coincides for sure with the input raster, and is in the same crd system as the input. The tool works fine for Radius=2, but for 3 and higher, the FS image is always shifted as radius-1.

Re: Focal statistics coordinate shift

DanPatterson_Retired — Tue, 11 Jul 2017 13:58:59 GMT

I would send it to support for confirmation Johanna

Re: Focal statistics coordinate shift

ChristopherDavid1 — Sun, 16 Jul 2017 19:06:07 GMT

Hi Johanna, I'm having the same exact problem running as a python script with the 10.5 arcpy library. The larger my neighborhood, the larger the shift (in my case also to the northwest). Looking forward to seeing the solution Esri support comes up with!

Re: Focal statistics coordinate shift

DanPatterson_Retired — Sun, 16 Jul 2017 22:32:49 GMT

You should report it to Christopher... the more confirmations, the bigger the problem, the faster the solution

Re: Focal statistics coordinate shift

ChristopherDavid1 — Tue, 18 Jul 2017 02:44:34 GMT

Thanks Dan. I will submit it. For what it's worth, my troubleshooting led me to the analysis mask. Removing the mask for focal stats allowed it to run without shifting the result, for some reason. I'm not sure why that worked, but hopefully Esri support will be able to figure it out. I added the mask back after running focal stats and it clipped the results of future tools to my study area. (Note: this worked for my workflow, since I was ignoring no data cells in focal stats)

Re: Focal statistics coordinate shift

MichaelShaw1 — Thu, 21 Nov 2019 19:16:17 GMT

This problem appears to exist within ArcGIS 10.6.1. Is there are work around for this? I have attempted to use the shift and snaps functions however does not resolve the problem.

Re: Focal statistics coordinate shift

JuanLaguna — Sat, 23 Nov 2019 00:49:11 GMT

Hello, Johanna.

Do you have access to more recent version of ArcGIS Pro? There is a similar-sounding issue to yours that was resolved for ArcGIS Pro 2.3 (and ArcMap 10.7). If you are able to try out your scenario with those versions of the software (or later), please let me know the outcome. Hopefully your issue will be resolved, but if not, we can work on getting your scenario delivered so we can investigate further here.

You can reach out to your Support contact to check on the original problem report, BUG-000106677.

Cheers,

Juan

Re: Focal statistics coordinate shift

Anne-MarieDubois1 — Tue, 05 May 2020 15:24:50 GMT

Hello,

It seems this issue has not been resolved in ArcMap 10.7 . I have applied a radius of 1 pixel with Focal stats on 3 different raster resolutions and I noticed that there is a shift on all my results. I am down to converting the rasters' extent (raw and filtered) to polygons and manually shifting the georeference. This is a SERIOUS waste of time.

Has anyone planned on fixing this?

Kind regards,

Re: Focal statistics coordinate shift

curtvprice — Sun, 31 May 2020 23:39:16 GMT

Anne-Marie, I found both these bugs on support.esri.com:

ArcGIS Pro (fixed 2.3) BUG-000106677: Setting a Raster Analysis Mask in the Environment Se..

ArcMap 10.7 (not in product plan 6/2019): BUG-000101814: A raster shifts its geographic position when applyin..

So it appears the excellent Juan Laguna may have been incorrect, this was not fixed in 10.7. We have been warned that bug fixes to the old ArcMap platform are lower priority, all major development is now happening on Pro.

Do you have a vector mask? Worth trying a raster mask (better performance anyway) or no mask and see if you still get the shift. Or you could do it in ArcGIS Pro. I really like the 2.5 version, it has come very far since the early days.

Re: Focal statistics coordinate shift

JuanLaguna — Mon, 01 Jun 2020 09:51:54 GMT

Hi Anne-Marie.

I'm sorry you're having trouble with this. Is it possible you could send me a sample of your data and an exact scenario? I'd like to try out your scenario here. I am not able to reproduce a result that has an unexpected shift when I tested with ArcMap 10.7. If I could get could a snippet of your data with all the settings you are using, I will investigate further.

Thanks,

Juan

Re: Focal statistics coordinate shift

Anne-MarieDubois1 — Tue, 02 Jun 2020 20:50:40 GMT

Hi Juan,

I can't seem to understand how to share data on this page.

But here is what I was doing.

(extent of my original data)

(tool parameters)

(shaded part is my result)

What I find odd, is that the displacing does not seem to be equal. Here is a view from the opposite side of my raster where it is exactly the same. Somehow, I'm starting to think it has more to do with raster data extension then an actual shift.

Re: Focal statistics coordinate shift

JuanLaguna — Tue, 02 Jun 2020 22:00:14 GMT

Hi Anne-Marie.

Thanks for getting back to me.

Would it be possible for you to create a small zip file that contains a subset of your data that demonstrates the issue, and email it to me? I am at jlaguna@esri.com

You can also use Esri's file transfer site: https://mft.esri.com/EFTClient/Account/Login.htm

Download and install FileZilla (if you don't already have it): https://mft.esri.com/eftclient/Account/MFTFilezilla.zip

Instructions: https://mft.esri.com/eftclient/Account/Instructions.png

Thanks,

Juan

Re: Focal statistics coordinate shift

JuanLaguna — Wed, 03 Jun 2020 04:03:57 GMT

Anne-Marie Dubois‌, Curtis Price‌

Just to respond to the part of the issue about when the fix was available.

As was noted above, for ArcGIS Pro the Status notification indicates that a fix was implemented, and the Version Fixed was Pro 2.3.

https://support.esri.com/en/bugs/nimbus/QlVHLTAwMDEwNjY3Nw==

Those code changes were also implemented for ArcMap version 10.7. Unfortunately, this situation seems to have triggered a glitch or synchronization problem in our tracking system, where the Support Site is reporting that the Status is “Not in Current Product Plan”, and the Version Fixed field comes up as blank (even though the fix was actually made). We will look into rectifying the underlying cause for this, so thank you for bringing it to our attention.

Best regards,

Juan

Re: Focal statistics coordinate shift

curtvprice — Fri, 12 Jun 2020 17:21:15 GMT

Looking at these screen shots, it appears you may be doing some projection on the fly. I do not recommend doing this, as it is hard to control resampling on the fly - it's best to run the Resample tool on inputs (with a Snap Raster used to line things up) before you run Focal Statistics.

Re: Focal statistics coordinate shift

JuanLaguna — Fri, 12 Jun 2020 18:01:55 GMT

Hi Curtis.
Thanks for the the suggestion. Anne-Marie and I are working on the exact scenario being used, and we'll update here once we get to the bottom of it.

Cheers,

Juan