Fundamentals of building large terrain datasets

Complexity: Beginner Data Requirement: Use your own data

Building a large terrain dataset

A terrain dataset is built from multiple data sources such as lidar mass point collections, 3D breaklines, and 3D-based survey observations. The data sources used to create terrain datasets are managed as a set of integrated feature classes in the geodatabase. Building a terrain dataset to efficiently visualize and store a large amount of source measurements can be a lengthy process. To create larger terrains, it is recommended to build the terrain dataset in phases by appending to referenced or embedded feature classes. The primary reason is to avoid extremely long processes where everything would be lost if a system failure were to occur.

The following overview steps provide a general workflow when working with mass points of terrain data:

Steps:

1. Create one multipoint feature class with as many points as you are willing to allocate to one process.
2. Build the terrain dataset , and optionally indicate you want the multipoint feature class to be embedded.
3. After the build is complete, delete the source multipoint feature class, if desired, to save disk space if the source points were embedded.
4. Create another feature class with additional points that are contiguous to the previous set.
5. If working in ArcSDE, register the feature dataset as versioned.
6. Use either the Append Terrain Points or the Append geoprocessing tools to append the additional points to the terrain's embedded feature class.
7. If in ArcSDE, unregister the feature dataset as versioned and compress edits to base.
8. Use the Build Terrain geoprocessing tool to update the terrain with the appended points.
9. Delete the source multipoint feature class, if desired, to save disk space.
10. Repeat steps 4–9 as needed to add more data.

Improving terrain dataset build performance

• The larger the z-tolerance or window size of the finest-resolution pyramid level, the better the performance will be.
• Fewer source feature classes will improve performance.
• A slow client/network can degrade performance since the client is involved in building the terrain pyramid.