25 · LiDAR & drone imagery

Tuesday, Nov 24, 2026 · taught remotely

Materials for this session are not published yet. They appear here before class.

Objectives

By the end of this session you can:

  • Explain what LiDAR measures and how a cloud of laser returns becomes an elevation raster.
  • Build a terrain model, a surface model and a canopy-height raster from public LiDAR data with lidR and terra.
  • Compute slope from an elevation raster and summarise it over polygons.
  • Read a drone orthomosaic into R and line it up with other layers.
  • Say what plane and drone data resolve that satellite imagery cannot, and what they give up in area covered and visits repeated.

What we cover

  • Published work built on elevation and aerial imagery. Henderson, Regan & Venables (REStud, 2021) traced every Nairobi building from aerial photographs for 2003 and 2015, measured 2015 building heights with LiDAR, and tracked demolition, rebuilding and the conversion of slums. Saiz (QJE, 2010) measured the land too steep or wet to build on around US metro areas from elevation and water data, and showed that it limits housing supply. Duflo & Pande (QJE, 2007) computed slope across each Indian district and along its rivers from a global elevation grid, and used river gradient to estimate how large dams affect farm output and poverty.
  • How LiDAR works: a laser scanner on a plane times each reflected pulse and turns it into a distance. The result is a point cloud, millions of points with a location and a height, classified as ground, vegetation or buildings. Each echo is a return; the first often comes from treetops or roofs and the last from the ground, and pulses that slip between leaves give the terrain under a forest canopy. Points come as LAS files, or LAZ when compressed.
  • From points to rasters: a digital terrain model (DTM, often called a DEM), the bare ground’s height, from ground returns; a digital surface model, the height of whatever stands on top, from first returns; and canopy height as the difference, cell by cell. Both models must share a vertical datum, the zero level of all heights. Compare one LiDAR height with a building of known height.
  • Slope, how steeply the ground rises, computed from height differences between neighbouring cells and averaged over polygons.
  • Where to get it: public LiDAR collections cover much of the United States, along with the elevation rasters derived from them.
  • Drone imagery: photos with a ground sample distance (the ground width one pixel covers) of a few centimetres, stitched into an orthomosaic, one image with true map distances everywhere. It shows single trees, crop rows and roof damage that a satellite pixel averages together. Georeferencing pins it to the map at ground control points, spots with surveyed positions. Overlay a basemap to verify a road lines up.
  • Application: what each home can see, such as a lake or a new wind turbine, computed from the surface model and set beside sale prices; and buildings that changed between two LiDAR flights of the same area.

This page is written for self-study: every step runs from the code shown, on data downloaded within the page.