Interpolation & Gridding
This is where a point cloud becomes a map layer. These tools take the points and produce the continuous surfaces you actually work with: a terrain model, a surface model, a canopy height model, contour lines, a triangulated mesh.
Everything else in the LiDAR toolbox exists to make this step give a good answer.
The two models everyone needs
Section titled “The two models everyone needs”A terrain model, the DTM. Grid only the ground points, taking the lowest
return per cell. Lidar Block Minimum does it directly, and the interpolators give a
smoother result.
A surface model, the DSM. Grid the highest return per cell, whatever its
class, so roofs and treetops are included. Lidar Block Maximum or Lidar Digital Surface Model.
Subtract one from the other and you have the height of everything standing on the ground. If your cloud is already normalised, the surface model is the canopy height model.
Choosing an interpolator
Section titled “Choosing an interpolator”Five methods, differing in how they fill the space between points.
| Method | How it works | Use it for |
|---|---|---|
| TIN gridding | Triangulates the points and reads the value off the triangle plane. | The default for terrain. Honours the measured heights exactly and copes well with uneven coverage. |
| Nearest neighbour | Takes the closest point within a radius. | Very dense clouds, and quick previews. Fast, no smoothing, visible cell structure. |
| IDW | Weighted mean of nearby points, closer points weighing more. | Smooth surfaces where you want control through the distance exponent. |
| Sibson, natural neighbour | Voronoi based weighting of the natural neighbours. | Smooth, natural looking surfaces without the artefacts IDW leaves at point locations. |
| Radial basis functions | Fits smooth functions capturing local curvature. | The highest quality gridding, for micro topography. The slowest. |
For a first terrain model, use TIN gridding. Move to Sibson or RBF when the result needs to be smooth enough for curvature or hydrological analysis.
Choosing a cell size
Section titled “Choosing a cell size”The most consequential parameter, and it comes from the data, not from taste. Run Lidar Point Density first, in Analysis Metrics.
A cell should contain several points. Below roughly one point per cell, the interpolator is inventing rather than measuring, and the result looks detailed while being partly fictional. A cloud at 4 points per square metre supports 50 centimetre cells comfortably.
Start here
Section titled “Start here”Lidar Tin Gridding
Section titled “Lidar Tin Gridding”The workhorse. Delaunay triangulation of the points, then the value of each cell read from the plane of its triangle. Exact at the measured points, well behaved between them, tolerant of gaps.
Lidar Thin and Lidar Thin High Density
Section titled “Lidar Thin and Lidar Thin High Density”Reduce the number of points before gridding, which is often what makes a large tile workable in the browser. Lidar Thin keeps at most one point per cell, choosing the first, last, lowest, highest or nearest. Lidar Thin High Density is adaptive: it thins the over dense areas, typically the flight line overlaps, and leaves the sparse ones alone.
Lidar Contour
Section titled “Lidar Contour”Produces contour lines directly from the point cloud, through a triangulation, at the interval you set. Handy when the deliverable is a contour map rather than a raster.
All the tools
Section titled “All the tools”| Tool | What it does |
|---|---|
| Lidar Tin Gridding | Interpolates through a Delaunay triangulation, honouring the measured heights. |
| Lidar Idw Interpolation | Inverse distance weighted interpolation, with a controllable exponent. |
| Lidar Sibson Interpolation | Natural neighbour interpolation, smooth and artefact free. |
| Lidar Radial Basis Function Interpolation | Radial basis interpolation, the highest quality and the slowest. |
| Lidar Nearest Neighbour Gridding | Assigns the nearest point within a search radius. Fast, no smoothing. |
| Lidar Block Maximum | Highest return per cell. Surface model, canopy tops. |
| Lidar Block Minimum | Lowest return per cell. Terrain model, ground baseline. |
| Lidar Digital Surface Model | Surface model from the local highest points, through a triangulation. |
| Lidar Contour | Contour lines from the point cloud, at a chosen interval. |
| Lidar Construct Vector Tin | Delaunay triangulation returned as a vector mesh. |
| Lidar Hex Bin | Aggregates points into hexagons with per cell summaries. |
| Lidar Hillshade | Shades the cloud from per point normals and stores the result as colour. |
| Lidar Thin | Keeps at most one point per cell, by a strategy you choose. |
| Lidar Thin High Density | Thins over dense areas while preserving sparse ones. |
| Flightline Overlap | Counts the distinct flight lines contributing to each cell. |
| Lidar Tile Footprint | Returns the bounding box or convex hull of a cloud, as a polygon. |
Grid the right points. A terrain model built from an unfiltered cloud is a surface model with holes in it. Filter to the ground class first, in Filtering & Classification.
What lands on the map. The gridding tools produce rasters, which are added as layers. The contour and TIN tools produce vector layers. Everything else is a file.
Then use the Terrain toolbox. A DTM produced here is an ordinary elevation raster, so slope, hillshade, curvature and the whole Terrain toolbox apply to it.