Arcpro crashing at step 3 of Build Network and Map Linkages

[amykeis...@gmail.com]

Hello,

I am using linkage mapper 3.0.0 and I was doing some testing for several weeks and things were working well. Then one time I canceled a run after noticing that I forgotten to put in a value for Bounding Circles Buffer distance. After I canceled and restarted that run, my Arcpro started crashing at step 3, giving this error:

The instruction at 0x0007FF8210912DE referenced memory at 0x000002341F11E000. The memory could not be read. 

I haven't successfully ran the tool on this machine since I first got the error. I always get this error now.

I tried uninstalling and re-installing arcpro, and I just got the error again.

Has anyone had this happen? Any ideas on how to fix? I feel like it isn't a problem with the tool itself, since it was running successfully before. 

Thanks,

Amy

[amykeis...@gmail.com]

And I am able to successfully run the tool with the same inputs and settings on a different machine, so this seems to be something going on with this particular laptop (so far! I sure hope it keeps working on my other machine :-)

[John Gallo]

Hi Amy,

Thanks for the update.  Glad all is running fine with Linkage Mapper.

I used to help people fix computers so here is a thought for what it is worth. Sounds like maybe a problem with your memory. When you restart your computer you get the option right away to use advance settings.  From there, I think you can find a tool that will scan your memory to see if it is broken, and it might have an option to try to fix it, or you can replace your memory stick.

Good luck with all,

John

John Gallo, Ph.D.
Senior Scientist
Conservation Biology Institute
136 SW Washington Avenue, Suite 202
Corvallis, OR 97333
Remote Office: Inglenook, CA

Phone: 707.962.9078
Twitter: @johnagallo

Updates: Action Research Newsletter

--
You received this message because you are subscribed to the Google Groups "Linkage Mapper" group.
To unsubscribe from this group and stop receiving emails from it, send an email to linkage-mappe...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/linkage-mapper/10dc9290-f9ed-43ca-b1a5-3d1891a2475an%40googlegroups.com.

[amykeis...@gmail.com]

Hey John,

Thanks so much for your help. 

I think I've narrowed it down to one issue. Every time I use a value in the "Bounding Circles Buffer Distance" box ArcPro crashes at the step where it is calculating bounding boxes for core areas with the error " The instruction at 0x0007FF8210912DE referenced memory at 0x000002341F11E000. The memory could not be read." 

Have you seen this before? Do you have ideas on why using the bounding circle buffer distance would cause this crash? I was hoping to use it to save processing time.

Thanks!

Amy

[John Gallo]

Hi Amy,

I'm back from vacation.  Sometimes this parameter throws problems when the mapping unit is ArcSeconds when people think it is in meters.  So you can check to make sure the mapping unit is what you think it is.  Second, and just for troubleshooting,  does it act the same if you make the bounding circle buffer smaller, such as 5-10 pixels?  If yes, please send the full log file as an attachment.

Thanks,

JOhn

[amykeis...@gmail.com]

Hi John! I hope you had a great vacation! I managed to get corridors ran over my whole area run on a test version of my resistance raster and corridors since I last posted! Woot!

I am doing a corridor analysis on a much larger area than I’ve ever done before. We are still working to improve our hubs and resistance surface, but in the meantime I am testing the corridor analysis on a draft versions.  All my data is in Albers equal area projection with meters as the unit. My resistance raster is 90 meter cells. My draft core layer has 1937 cores.

As I expected, I can’t run the whole area at the same time. I have tried doing this several times using a bounding circle to try to save space and processing time. When I do, it consistently either crashes without error or fails with an error message on step 3 when working on the first core area:

Target core areas for core area #1 = [ 6  7 13 14 15 31 32]------------------------------------------------------------------Geoprocessing error on **line 758** of s3_calcCwds.py in Linkage Mapper Version 3.0.0:---------------------------------ERROR 010236: Error in building VAT: MtVatBldNew: All cells in grid d:\lmfull\scratch\core1\addras  have NODATA value. VAT will not be built..ERROR 010067: Error in executing grid expression.Failed to execute (GetRasterProperties).

 

I have explored how to break up my area into subregions so I can run them. I’ve tested several sizes. I’m currently working with hexagon subregions that are 500k sq km in size. This is the largest size I’ve tested and breaks up my study area into 11 subregions.

As I mentioned above, trying to run the entire area at once always fails in step 3, however, it does correctly create a core area distance file. In my testing, I can use that core area distance file in subregion runs to save processing time. It also correctly creates a cost weighted allocation raster and a Euclidian allocation raster. I haven’t figured out how to re-use those in the subregion runs.

·         Is there a way to use cwd_alloc_ras created from the failed full run on subset run?

·         Is there a way to use euc_alloc_ras created from the failed full run on subset run?

To perform my subregion runs, I buffer each hexagon by 200 km, to try to minimize edge effects. I then clip the resistance raster to the buffered hexagon and I select hubs that intersect that hexagon to use as inputs in my subregion linkage mapper runs. I started by using the full resistance raster in the subregion runs. Sometimes I can use the full resistance raster and the hubs that intersect a hexagon and it works, but it seems like this was related to more crashing. Lately I’ve just been using the clipped resistance raster to try to narrow down what is causing crashes and I seem to be getting less crashing.

My goal was to run each of the buffered subregions using the following settings:

 

I’m using the bounding circles buffer distance to try to save hard drive space and processing time. Sometimes it crashes when I use the bounding circles buffer distance though. Sometimes it runs and sometimes it doesn’t. There are some hexagons that I have only been able to run if I leave the bounding circles buffer distance blank. I also played around with using a maximum Euclidian corridor distance of 200k meters, but sometimes those crashed. I figured I’d leave that blank for this testing because I can always go back in and remove longer corridors later. But then also sometimes it crashes even when I leave the bounding circles buffer distance blank too. I’ve had a hard time telling the difference between when ArcPro just crashes because ESRI gonna ESRI and when it is crashing because I’ve used a value in these fields that it doesn’t like.

I managed to get all my buffered hexagons to run now! And when I clipped the corridor outputs back down to the original hexagon and merged them all together, I’m not seeing any big edge effects. Woot Woot!

I’m curious what you think about these methods of splitting the area into subregions and buffering them. Have you done something like this before?

Do you have any tips or tricks for breaking up a large area to get it to run, limit edge effects, and minimize processing time?

Thank you for making this tool set! It is a life saver for me! 

~Amy

[amykeis...@gmail.com]

Here is a picture of the draft corridors :-)

[John Gallo]

HI Amy,

I'm just back from vacation.  I'll reply to you directly soon.

Thanks,

John

[John Gallo]

Hi Amy,

Congratulations!! This looks fantastic.

So glad you buffered the subregions sufficiently (200 km).  I don't see any edge effects either.

This is a great case study of how to do a mega analysis at high resolution over a huge area.

Regarding ideas for the future to limit the crashing, it would be to clip the resistance surface to the buffered subregion, as well as clipping the core surface.  Also, I would try to do this in a script, or semi-automated to have the hexagons smaller, and not use any tricks like bounding circles.

Congrats again,

John