Speed up Build Terrain tool

[tigerwoulds]

I am building a large terrain dataset (roughly 2000 square miles) in ArcGIS Pro and need some advice on how to speed things up. I've ran through this article on Fundamentals of Building Large Terrains.

The terrain will have 6 pyramid levels and span two counties so I've divided my multipoints into quadrants. My plan is to build the first terrain, the append the next set of multipoints and rebuild. Then continue this till all the multipoints have been incorporated into the terrain. I started the first process and the speed of the Build Terrain tool is going extremely slow. Looks like it will take multiple weeks to build one quadrant. I have 64GB ram available on my machine but Pro is only using about 3.

Anything I can do to speed up the process?

Given an unlimited budget (just an example here) , can someone give a basic breakdown of what ideal specs a computer should have for building a large terrain? My IT guys mentioned they can create a VM with beefed up specs or possibly purchase a machine strictly for terrain processing. I'd like to give them some 'best case' scenarios and work from there.

Thanks!

[MarcSwartz]

I feel your pain.  I have generated the Terrain Datasets for each county in North Carolina (100 Counties) from both 2 point per square meter Lidar and 8 point per square meter Lidar.  The Build Terrain tool does NOT honor the parallelProcessingFactor.  Basically it run on one core of a CPU.  Our solution was to run the Build Terrain tool on our server as it has the most reliability, especially from power interruptions.  Yes, it did take weeks, sorry.  

My recommendation is to run much smaller areas with a bit of overlap between each area.  And it is sad that this has been an issue form me since 2015 when I first started working with large lidar collections.

As far as a good CPU to run on, try to find one with the highest single threaded speed.  google "passmark single thread performance"

[tigerwoulds]

Thanks for your input! This was posted last month and since then I was able to get the terrain to build all at once. I did try building one quadrant, then using Append Tool to add more points > then rebuild. I kept getting an error 99999 when trying to run an append so eventually gave up on that.

Another thing I tried was the Replace Terrain Points and specified the 2nd quadrant as my processing extent (where no points existed from the first build). This actually worked and appended my 2nd set of points to the terrain. However when I built the terrain, it came out with some pretty serious issues. 

So I just figured building it all at once would have the be move here. It took about 4 week and ran. Unfortunately I had to make some revisions and re-build the whole terrain again. It's on week 1 now..wish me luck!

[MarcSwartz]

Good luck. 

Just curious, what products will you be producing from you Terrain Dataset?  We generated a statewide 10ft grids of elevations and have bare earth points by pyramid from the Terrains for our customer needs.  For CADD we are a Microstation shop and full resolution Lidar is normally to much for it to handle for the size of projects we do.  

[tigerwoulds]

1 -3ft DEMs, TINs, and Contours used for civil engineering projects - mainly Hydrology and Hydraulics. The goal is to create this larger composite terrain, then create derivative products for each project for various smaller study area. Ideally we wont have to process lidar for each project, we can just create a raster or TIN or whatever using the larger Terrain Dataset.

We've used LASDatasets before but the Terrain Dataset performance is better and if we need to add breaklines, or tweak elevation values the terrain dataset shines here. 

Our CAD guys use Microstation as well so if I send something to them, it's usually 3D contours. Been experimenting with LandXMLs too