The Tukituki River flood study used 120 million LiDAR points to build the terrain model that underpins the hydraulic analysis. That level of resolution, fine enough to resolve individual berms, road shoulders, and low-lying paddocks, fundamentally changes the quality of a flood hazard assessment compared to the coarser data available even five years ago. But LiDAR processing isn't plug-and-play: filtering, datum correction, and validation against survey benchmarks all determine whether the model is fit for consent purposes.
What LiDAR data actually is
LiDAR (Light Detection and Ranging) is an airborne survey technology that fires laser pulses at the ground from an aircraft and records the time each pulse takes to return. The result is a dense cloud of three-dimensional points, each with an easting, northing, and elevation. Modern LiDAR surveys in New Zealand typically achieve point densities of 2 to 20 points per square metre, with vertical accuracy of plus or minus 50 to 100 mm on bare ground.
For flood hazard assessment, the critical output is a bare-earth Digital Elevation Model (DEM), which represents the ground surface with vegetation, buildings, and other above-ground features removed. This DEM is the terrain surface that water flows across in the hydraulic model.
Why resolution matters for flood modelling
A flood model is only as good as its terrain data. On a flat floodplain, a 200 mm difference in ground elevation can determine whether a building platform is above or below the assessed flood level. Coarser terrain data, such as the 20 m contour data that was standard practice before LiDAR became widely available, simply cannot resolve features at that scale.
On the Tukituki study, the LiDAR-derived DEM resolved features that would have been invisible in older data:
- Stopbank profiles: The crest elevation and cross-sectional shape of existing stopbanks were captured at sub-metre resolution. A stopbank that is 100 mm lower than assumed in a coarser model could overtop at a lower return period, fundamentally changing the flood hazard assessment downstream.
- Road crowns and culvert approaches: Roads act as barriers to overland flow. The LiDAR data captured the road crown elevations and the low points where culverts pass beneath, allowing the model to correctly route floodwaters through the transport network.
- Drainage channels and farm races: Small channels that concentrate and direct flow were resolved, improving the accuracy of the overland flow paths between the river and the subject site.
The Glendale Road Auckland flood study similarly relied on LiDAR-derived terrain data to resolve overland flow paths through an urban catchment where the difference between a compliant and non-compliant floor level was less than 300 mm.
Processing LiDAR: from raw points to a useable DEM
Raw LiDAR data cannot be used directly in a hydraulic model. It requires several processing steps, each of which affects the quality of the final terrain surface.
Ground classification and filtering
The raw point cloud includes returns from vegetation canopies, rooftops, power lines, fences, and other above-ground features. Ground classification algorithms identify which points represent the actual ground surface and discard the rest. This is largely automated but requires manual checking in areas of dense vegetation or complex terrain. Mis-classified points (a tree return classified as ground, or a ground point classified as vegetation) create artefacts in the DEM that can redirect flow in the hydraulic model.
Datum and coordinate system verification
LiDAR data must be referenced to a consistent vertical datum. In New Zealand, this is typically NZVD2016 (New Zealand Vertical Datum 2016). Older LiDAR datasets may reference local datums or MSL (mean sea level) benchmarks. A datum offset of even 50 mm propagates directly into the flood level assessment. On the Tukituki study, the LiDAR data was verified against three survey benchmarks on the site and in the surrounding catchment to confirm datum consistency.
Hydro-enforcement
LiDAR does not penetrate water. River channels and ponds appear as flat surfaces in the raw data, at whatever level the water happened to be on the day of the survey. For hydraulic modelling, the channel bathymetry needs to be burned into the DEM using survey cross-sections or estimated channel geometry. Bridges and culverts also need to be cut through road embankments so that the model correctly routes flow through these structures.
DEM resolution selection
The raw point cloud is interpolated onto a regular grid (the DEM). Grid size is a trade-off between detail and computational cost. A 1 m grid preserves most of the terrain detail but can result in a very large model domain. A 2 m or 5 m grid reduces the computational burden but smooths out smaller features. For the Tukituki study, a 2 m grid was adopted for the main floodplain area, with 1 m resolution near the subject site where greater precision was needed.
Validation: is the DEM fit for purpose?
A DEM is only fit for consent purposes if its accuracy can be demonstrated. This typically involves:
- Benchmark comparison: Checking the DEM elevation at known survey points. Differences greater than plus or minus 100 mm on flat, bare ground indicate a problem with the data or processing.
- Visual inspection: Reviewing the DEM in 3D for artefacts such as vegetation remnants, building footprints left in the ground surface, or interpolation errors near water bodies.
- Cross-section comparison: Overlaying DEM-derived cross-sections with surveyed cross-sections at key locations (stopbank crests, channel inverts, building platforms) to confirm consistency.
Where to source LiDAR data in New Zealand
LINZ (Land Information New Zealand) hosts open-access LiDAR data through the LINZ Data Service. Coverage is expanding rapidly, with post-Cyclone Gabrielle datasets now available for much of the Hawke's Bay and East Coast regions. Regional councils also hold LiDAR data that may not yet be published on the LINZ portal. For site-specific studies, commissioning a new LiDAR survey is an option but typically costs $15,000 to $50,000 depending on the survey area.
LiDAR-derived terrain models have transformed flood hazard assessment in New Zealand. The resolution is fine enough to resolve stopbank crests, road crowns, and drainage channels that older data missed entirely. But the raw data requires careful processing, datum verification, and validation before it is fit for a consent-level assessment.