Voxel layer visualisation problem

2294
13
05-03-2021 03:02 AM
FrederikkeHansen
New Contributor II

I have created a netcdf file and loaded it into ArcGIS Pro as a voxel layer. I can see that it is being read correctly because the symbology section shows an appropriate histogram for the data, and loading the file as a table also gives the correct data. However,it does not show up on my scene, no matter what I do. Originally I thought the file was too large, so I split it into sections (each about 33MB), and I have the same issue. What can I do to figure out what is causing this issue?

I have attached one of the smaller files if anyone would like to try to load it themselves.

0 Kudos
13 Replies
DanPatterson
MVP Esteemed Contributor

Add a voxel layer to a local scene—ArcGIS Pro | Documentation

Frequently asked questions—ArcGIS Pro | Documentation

Could you elaborate on what you did to whether any of the FAQ's are related


... sort of retired...
0 Kudos
FrederikkeHansen
New Contributor II

Yes, I've been trying to add a cross section of the subsurface with X, Y, and Z coordinates to a local scene. After discussion on my last community question, I was led towards voxel layers as the best way to achieve what I wanted. I followed this guide: https://www.esri.com/arcgis-blog/products/arcgis/data-management/creating-netcdf-files-for-analysis-... to make the netCDF file, albeit in MATLAB. The I loaded the file as a voxel layer, exactly as stated on the first link you mention. I have attached a screencap of where the layer should be appearing (and this is also the area that appears when using the zoom to layer function, so the coordinates are correct). I get no errors when loading the netCDF file, except it cannot read one of the attributes containing information about the coordinate system, but it should load it in correctly anyway, according to the FAQ you link. I don't see anything in the FAQ which is related to my issue. The coordinates of the voxel grid are [z,y,x] in my netCDF file, and I'm not getting any error messages as previously mentioned.

FrederikkeHansen_0-1620039136743.png

 

0 Kudos
Andrew--Johnson
Esri Regular Contributor

Hi,

I think the issue is the elevation surface is hiding the voxel layer. If you turn off the elevation surface/basemaps and zoom to layer then you see the voxel layer.

Andrew--Johnson_0-1620054886099.png

 

FrederikkeHansen
New Contributor II

Thanks Andrew! I tried that, but without success. I've tried some new things, which only add to my confusion.

1. I can load the data on my much less powerful laptop. I assume this is also why you were able to see the data when you loaded it, so there must be something weird about the way ArcGIS Pro is set up on my desktop...

2. Loading the data as a continuous data type on my laptop only gives me 4 pillarlike structures (like what you show) when there should be an entire section.

3. Loading the data as discrete data gives me the entire section, but only with -1, 0, or +1 values, where any value that isn't -1 or +1 is assigned 0.

I assume this is the intention for discrete data, but even so, why would discrete data give me the entire section, while continuous data only gives me these 4 separate pillars? Furthermore, why is none of this working on my desktop?

0 Kudos
Andrew--Johnson
Esri Regular Contributor

Hi,

I tried on two separate computers and it displayed on both. You may want to check the graphics driver and ensure it's up to date on the machine where it's not visualizing.

Looking at this dataset there are a lot of NaN values which is problematic when trying to visualize this as a voxel layer. I also tested on a 3rd party viewer and there it doesn't even draw so I think the best path forward is to fill in these NaN values with legit values. I'm not familiar with matlab so i cannot give you guidance there. I can point you to various tools to interpolate your data. Here's some learn lessons that you may want to take a quick look to see if this works for your case.

https://learn.arcgis.com/en/projects/visualize-and-explore-coral-reef-habitats-with-voxels/

https://learn.arcgis.com/en/projects/explore-and-animate-geological-data-with-voxels/

https://learn.arcgis.com/en/projects/visualize-ecological-marine-units-with-voxels/

https://learn.arcgis.com/en/projects/visualize-social-distancing-across-california/

 

let me know if you have any other questions.

thanks,

Andrewnan.png

0 Kudos
FrederikkeHansen
New Contributor II

Absolutely, I was aware of the NaN values. When I try to fill them with fill values, I don't see an option to exclude the fill values of the netCDF file from visualisation. I tried to clip them out using the symbology section, but this just resulted in ArcGIS Pro freezing completely. I will do a bit more testing and see if it's a GPU issue.

While I have your attention, is a voxel layer really the only way to accomplish what I want to do? All my coordinates are along a single line on the surface, and as such, I don't really see this as a 3D dataset, but rather a 2D slice of the subsurface. Is it really not possible to add a vertical-plane 2D raster? Adding it as a voxel layer requires me to fill an entire cube with inconsequential data (NaN or fill values), just to get a single cross section of the subsurface into my scene.

0 Kudos
Andrew--Johnson
Esri Regular Contributor

Hi,

Unfortunately displaying vertical rasters is not fully supported so I only see filling in the NaN values and then creating a section in the voxel layer at the location you want to visualize this. If there's anything else I can think of i'll post back here.

 

thanks,

Andrew

0 Kudos
FrederikkeHansen
New Contributor II

What a shame! Thanks for looking into it for me though 🙂

Kind regards,

Frederikke

0 Kudos
Andrew--Johnson
Esri Regular Contributor

Hi,

Yes I agree the voxel layer is designed to visualize a volume and this is just a cross-section so really the best route is to just visualize it as a vertical raster. I'm not sure though if there are any limitations with visualizing a vertical raster so let me reach out to the raster team to confirm.

thanks,

Andrew