From points to pixels — creating digital elevation models from OpenTopography point clouds

Land Information New Zealand has a partnership with OpenTopography allowing 3D point cloud data to be downloaded, processed and visualized. This partnership complements the publishing of DEM and DSM products on the LINZ Data Service.

Not only can we download point cloud data from OpenTopography but we can also create Digital Elevation Models (DEM) and Digital Surface Models (DSM) directly from this point cloud data. Different methodologies can be used to create pixels from points, and in turn different visualizations of the point cloud data are available. Click on this link to find out how to download data from OpenTopography.

Digital Elevation Model of The Remarkables, Otago

DEM and DSM Nomenclature

Land Information New Zealand defines:

DEM (Digital Elevation Model) — a surface representing ground level elevation.

DSM (Digital Surface Model) — a surface representing the highest elevation (except for noise).

Open Topography defines a DEM to represent both a DEM and DSM, depending on what type of point classification is selected.

The two ‘DEM’ rasterization methods available on OpenTopography

Creating our DEMs and DSMs — Clarification

This hillshaded image of ground elevation (DEM) near Lake Ruataniwha highlights the Ohau A Hydro Station and the impact of the Ohau River on the shape of the land. Can you spot the fault traces in this image?

For example to generate a DEM in OpenTopography — points selected as input data need to be classified as ground.

To generate a DSM in OpenTopography is a little more complex — We need to grab points with the highest elevation values. For most LINZ datasets this would mean:

1. Selecting ground and unclassified points, as some areas would just be classified as ground and we do not want to lose detail in these areas. Ignoring water points as we just want to cover topography for these models.
2. Selecting the local binning method and making sure the maximum elevation value option has been checked.

From Points to Pixels

The two main methods of generating surfaces (rasters) from points in OpenTopography are:

• Delauney TIN (triangular irregular network) Method
• Local Binning Method

Before we start using these methods to creating raster surfaces from points, we need to make sure we have selected the appropriate points, in which the surface is generated from. For a DEM select ground return classification only, for a DSM select ground and unclassified points.

Choosing Ground — Return Classification means that all Elevation Models derived from the data represent ground points only (DEM). Conversely selecting Unclassified + Ground for this dataset, means that the Elevation Models derived represent the highest (DSM) providing that max (z) elevation value has been selected

Delauney Triangulation TIN Method

The TIN (triangular irregular network) method uses Delauney Triangulationto create a DEM surface. This surface formed by triangular planes of different shapes and sizes, defined by nodes and edges which is then rasterized. What makes Delauney Triangulation different from the other TIN methods, is that each triangle satisfies the Delaunay Triangle Criterion. This criterion states that no vertex lies within the interior of any of the circumcircles of the triangles in the network. This ensures the largest possible sized triangular area between points, therefore creating the ‘cleanest’ possible surface.

Delaunay Criterion 2D Example — Creating a planar surface from XYZ lidar points

This method is recommended for data with a low point density or a small scale area. For example an area with high vegetation such as a forest would most likely have low density as some of the light would not be able to penetrate and be reflected through the vegetation and canopy. These areas would benefit from a DEM created from the Delauney Triangulation TIN Method.

This point cloud tile showing ground points only. The data surveyed over dense vegetation has much less points versus areas that are classified as urban. This is showing an area north of Brooklyn, Tasman.

The Grid Resolution defines the size of each pixel in the output raster, and the Max. Triangle Size defines the maximum allowable length of a triangle. Any triangles with lengths larger than this will be omitted from the output raster, resulting in a DEM with null elevation values. Selecting an appropriate triangle size to avoid creating a raster with lots of null elevation values, takes a bit of trail and error. The suitable triangle size ultimately depends on the density of the selected points. Open Topography TIN rasters can be downloaded in Arc ASCII Grid, GeoTIFF and IMG formats.

Local Binning/Gridding Method — Our options

Circles with a specific radius are used to calculate elevation values.

OpenTopography’s local binning algorithm uses a circular search area, with a radius is defined by the user, to calculate the elevation model height at each pixel or grid cell. The resolution of each pixel or grid cell is defined by the user.

Each pixel node can be computed in five different ways in OpenTopography:-

The minimum elevation value (analogous to the ground level elevation or DEM) can be extracted in the search area. This can be useful for removing vegetation, and for displaying the morphology of flat, coastal areas. As well as exposing fault lines.

Irishman Creek Fault near Lake Tekapo, Canterbury unveiled by a DEM.

The maximum elevation value found in the search area. This can be useful for viewing vegetation and buildings in an area, or viewing the first return surface, creating a DSM.

Digital Surface Model (DSM) of Lincoln Township, Canterbury. On the top left hand corner you can see the old river channels flowing through the town, as well as the Lincoln University Campus (the largest building).

The mean elevation value found in the search area. DEM grids are assigned the mean elevation value of selected points within the user defined search radius. This method is effective if you want to smooth rough terrain.

The inverse distance weighted mean (idw) of the search area. For more information about the power of inverse distance weighting, check out this reference.

The number of points in the search area. (If the number of points in the search radius is 0, the grid node is assigned a null value. The null filling option uses a moving window of 3, 5, or 7 pixels if chosen by the user to fill null pixels in the grid). This is a great reference for exploring the level of detail available in an area.

Other methods are also available including standard deviation and a layer showing all these methods, separated in each pixel band, where each of these can be selected for visualization.

Tips using the Binning Method

If the pixel size is three to four times larger than the average point distance, you can safely use binning. If the cell size is smaller than that, you can try binning with void filling turned off.

If the resulting raster mainly contains voids and only a few single data cells, binning generally does not produce a meaningful elevation raster. You need to either increase the pixel size or switch to triangulation.

If the resulting raster shows enough content with some salt and pepper voids, and maybe a few larger voids, you can use binning with void filling turned on.

Please check out this OpenTopography link to find out more about using LiDAR Data and using Digital Elevation Models.

Bins versus Triangles — Pros and Cons

A literal representation of the differences between these methods.

The TIN method is a great method to use for a DEM. It generates a simplified, smooth surface, great for hydrological analyses. The TIN method can be useful for areas with sparse point cloud data.

Local Binning Methods are much more versatile and can create both a DEM and a DSM.

TINs create a continuous surface, where the local Binning Method by default only shows what data is available for each pixel. The latter is a more representative surface of the underlying data. Gaps or ‘Void Areas’ can be filled using the Null Filling option under the DEM (Local Bin Gridding Header) on the right hand side.

The morphology of the surfaces generated from these methods will be slightly different, looking at the example below the Local Binning Method (surface with gaps) has a slightly grainier appearance, because by default elevation value are deliminated by 1m cell radius, which happens to be the pixel resolution of this surface.

Variations between TIN (with no gaps) and Local Binning Method (with gaps)

Licensors and Attribution

The LiDAR data used in this story is licensed by Auckland Council, Environment Canterbury Regional Council, Nelson City Council, Otago Regional Council and Tasman District Council, for re-use under the Creative Commons Attribution 4.0 International license.

All of the New Zealand elevation data available on OpenTopography is available for free under the same license.

The licensor should be specified when attributing LiDAR Data. See Attributing elevation or aerial imagery data for more details.

Note: You can download and process up to 50 million points as a Open Topography guest. By creating a free OpenTopography account you can download and process up to 250 million points per job, and you also receive an archive of your jobs, and an share job results. Create your own OpenTopography account by scrolling to the top of the page and clicking on myOpenTopo. Below the Login fields you’ll find an option to create a new account. For users who need higher download and processing limits (up to 500 million points per job) you can request a Power User account by clicking the myOpenTopo Authorizations link on your myOpenTopo page.