LiDAR mapping

Christchurch, July 2003

For four days in July 2003, an aircraft crossed Christchurch and part of the adjoining Waimakariri River corridor carrying a laser scanner. Christchurch City Council commissioned the survey, flown by AAM GeoScan from 6 to 9 July. The scanner measured elevation through airborne laser returns, which were processed into a digital model of the ground and objects above it. The survey provides a detailed New Zealand operational example of airborne LiDAR in 2003.

The aircraft position was controlled using GPS, and ground checks assessed the survey. The laser returns were classified into ground and non-ground points. Staff reviewed and corrected algorithmic classifications against intensity information before generating the terrain products. Coordinates were transformed into the client’s mapping system, and local height control set the vertical reference. The delivered model was the end of an acquisition and processing chain, rather than a raw stream of laser pulses.

A point cloud can look almost photographic when displayed in three dimensions, but it is a set of measured returns with positions and heights. To produce a bare-earth model, software and people have to decide which returns represent the ground and which belong to vegetation, buildings or other objects. Interpolation, thinning, editing and gridding then turn selected points into a surface that ordinary GIS software can use. The apparently simple hillshade or digital elevation model on screen can conceal a considerable amount of judgement. LiDAR made terrain far more measurable, but it did not remove the need to understand how the terrain model was made.

Andrew Steffert’s 2012 account explains the development of the Horizons Flood Plain Mapping Project following the February 2004 floods. The first LiDAR survey covered about 1,200 square kilometres in 2005, with later work extending the coverage. DHI’s Terry van Kalken and Carl Johnson prepared the 2006–07 modelling and mapping report. Elevation information was combined with gauging records, hydrographs, photographs and observed flood extents to calibrate hydraulic models and map scenarios including 1-in-50, 1-in-100 and 1-in-200 events. The outputs supported evacuation planning, stopbank works and land-use decisions. The elevation data also found uses in drainage, soil and land-use capability assessment, coastal hazards, transport and erosion work. The people named in the later reports documented those stages; the original 2005 commissioning team remains to be identified.

From contours to dense measurements

GIS had used elevation long before airborne laser scanning arrived. Contours, spot heights, photogrammetric terrain models and interpolated surfaces could represent the shape of the land, and New Zealand mapping agencies had decades of experience producing them. What LiDAR changed was the density and regularity of the observations. Instead of relying primarily on lines and selected heights from which a surface was inferred, an airborne laser survey could collect a dense cloud of three-dimensional points across a large area. That made subtle slopes, banks, channels and surface breaks easier to model.

The improvement involved resolution, coverage, repeatability and the growing ability to reuse elevation data across several applications. A detailed terrain surface could support questions that had previously been awkward, expensive or dependent on substantial field survey. Water could be modelled across urban streets and properties. Small changes in ground height could be compared between repeated surveys. Engineers could derive sections and slopes over wide areas. Planners and environmental teams could inspect terrain independently of the contours printed on a topographic map. The same acquisition could be reused for several tasks, which gradually changed the economics of elevation data.

LiDAR also separated two ideas that were easily blurred in everyday GIS work. A digital surface model represents the upper surface detected by the sensor, potentially including buildings and vegetation. A bare-earth elevation model attempts to represent the terrain beneath those objects. A flood model usually needs something much closer to the ground surface than a model that leaves tree canopies and roofs standing in the way of the water.

Waimakariri and regional reuse

The Christchurch survey was followed by further Canterbury acquisitions. AAM captured LiDAR from 21 to 24 July 2005 for Environment Canterbury and Waimakariri District Council, and another survey from 6 to 11 February 2008 involved Environment Canterbury and Selwyn District Council. Within a few years LiDAR was being commissioned by several councils for adjoining parts of the same region. Elevation was becoming a reusable regional asset rather than an isolated experimental dataset.

The 2005 Waimakariri data also demonstrate the long life that a good elevation survey can have. Later flood-hazard work continued to use the older LiDAR where newer coverage was absent, sometimes combining different survey years across an area. Where the landscape had changed, mixing survey years created a currency risk and showed why acquisition date and provenance have to travel with the data. A highly detailed surface can still be wrong for today's question if it records yesterday's riverbank, subdivision or stopbank. Currency became another dimension of elevation quality.

Regional reuse also raised the problem of consistency. Different contracts could use different specifications, point densities, classification rules, datums, output grids and file structures. Two detailed datasets are not automatically easy to merge just because both came from LiDAR. By the time national coordination became a realistic goal, the challenge was no longer simply obtaining laser-scanned elevation. It was obtaining data to sufficiently consistent specifications that neighbouring regions and later acquisitions could form part of a wider terrain framework.

Water finds the low places

North Shore City used LiDAR in an early operational workflow. A LiDAR survey flown in 2004 was used to create a digital terrain model with 2 metre grid nodes. From 2005, the council and its consultants used that surface with MIKE 21 hydraulic modelling to identify overland flow paths across the city. During the summer of 2005/06, field work was used to inspect and verify the modelled paths. The resulting information was then attached to property information and made available through the council's GIS environment.

The North Shore workflow linked laser survey, terrain modelling, hydraulic modelling, field verification and corporate GIS. Each stage turned the preceding output into something more directly related to a council decision. The elevation model helped predict where stormwater would travel when the piped network was exceeded. Field staff checked whether the model corresponded with the real city. GIS then made the resulting flow paths available alongside cadastre, roads, stormwater assets, aerial imagery and other council information.

Flood modelling required manual corrections to the terrain model. A later Auckland remapping project using 2016/17 LiDAR created a one-metre terrain model and still required thousands of manual adjustments to represent features that affected water movement. Kerbs, walls, culverts, underpasses and other structures can determine whether water crosses a surface in reality even when the raw elevation model indicates another path. A bare-earth surface is a model of the landscape rather than the landscape itself. Flood modelling depends on how that surface has been edited as much as on the density of the original point cloud.

LiDAR changed GIS practice most clearly at this point. It allowed councils to begin with a detailed measured terrain rather than construct every local height relationship from field survey or sparse elevation sources. That moved effort rather than eliminating it. Staff could spend more time interpreting flow routes, fixing model barriers and assessing consequences. The technology supplied a richer starting surface, while professional judgement remained necessary to turn it into a defensible model.

The point cloud becomes a dataset

Early LiDAR could be awkward for ordinary GIS systems because the source data were large. A point cloud may contain millions or billions of individual observations, far more than a conventional vector layer of roads or parcels. Storing, filtering, classifying and drawing those points required different software habits and often specialist processing. The 2007 SIRC work by M. Hamish Bowman on tools for aggregating and geoprocessing raw spatial data is a useful marker of this problem. LiDAR was already creating data volumes that required purpose-built approaches rather than treating every return like an ordinary GIS point feature.

For many users the practical answer was to work with derived rasters. A gridded bare-earth model could be shaded, contoured, queried and used in hydraulic or engineering models without loading the whole source point cloud. The 2003 Christchurch material includes two-metre gridded ground data among its products, while later national programmes standardised one-metre elevation and surface models. This is one reason LiDAR could spread into normal GIS practice even before every desktop handled point clouds elegantly. The specialist processing stage produced forms that existing analysis tools understood.

As hardware and software improved, the point cloud itself became more accessible. Analysts could examine individual returns, classify them, generate profiles, inspect vegetation or buildings and derive new surfaces rather than relying only on a contractor's original grid. LAS and later compressed LAZ became common exchange formats for point-cloud data. LiDAR produced new elevation models, and GIS gradually learned to treat the underlying three-dimensional measurements as reusable data in their own right.

The density of those measurements also increased. Later Canterbury surveys contained far more returns than the 2003 dataset over equivalent areas. That did not make the 2003 survey obsolete. It made comparison possible while also requiring analysts to understand that datasets from different years were not technically identical. A change map can show actual ground movement, differences in vegetation or structures, and differences caused by acquisition or processing. Repeated measurement is powerful precisely because it creates another layer of methodological responsibility.

Measuring change in Canterbury

Canterbury made that value unusually visible after the earthquakes of 2010 and 2011. The region already possessed pre-event LiDAR from 2003, 2005 and 2008. New surveys after the earthquakes could therefore be compared with earlier surfaces rather than interpreted in isolation. Researchers and agencies used repeated elevation models to measure subsidence and changes in flood vulnerability. A ground surface had become something that could be compared through time with enough precision to alter engineering and hazard analysis.

The wider earthquake response involved emergency systems, recovery mapping, data sharing and organisational consequences beyond the LiDAR story. Earlier surveys provided measurements of the landscape before the earthquakes. The 2003 survey was commissioned years before the earthquakes, yet it became a baseline against which later ground movement could be assessed. Several services could use the same terrain data.

Repeated surveys also highlighted the importance of a common coordinate and height framework. A difference of a few centimetres or decimetres is meaningless if two datasets are not on compatible datums or if vertical transformations are poorly controlled. New Zealand's geodetic framework therefore sits quietly underneath the LiDAR story. Dense elevation points do not escape the basic requirement to know where zero is and which coordinate reference system is being used. The more precise the comparison, the more obvious those foundations become.

This repeated-measurement capability extended beyond earthquakes. River channels shift, coastlines erode, subdivisions alter drainage, stopbanks are modified and forests change. Capturing the same area again can support change detection, but only when analysts control for acquisition date, classification, density, datum and processing. LiDAR did not make change self-evident. It made more kinds of change measurable.

Seeing the ground beneath vegetation

One of LiDAR's most useful properties is that an airborne pulse can produce more than one return as it passes through vegetation. Some energy may be reflected from the canopy and lower branches while another return reaches the ground. Where vegetation and survey conditions allow it, classification can therefore produce a terrain model that is far less dominated by tree crowns than a photograph or surface model. This capability is especially valuable in a country where forest, scrub and riparian vegetation can obscure landform. Vegetation remains part of the measurement environment, while classification gives analysts another way to infer the ground beneath it.

That capability changed the kinds of terrain features that could be inspected in GIS. Subtle channels, terraces, scarps and earthworks can become visible in a bare-earth hillshade even when the same forms are difficult to distinguish in conventional imagery. Forestry, geomorphology, archaeology and hazard work all gained from this, although each field has different standards for deciding what the revealed shape means. The laser does not identify a terrace as geological, archaeological or engineered. It records the geometry from which a specialist may form that interpretation.

A DSM can preserve the upper shape of forests and buildings while a DEM attempts to represent the ground below them. Comparing the two can support estimates of height or structure when the acquisition and classification are suitable for the task. LiDAR therefore created a family of possible elevation surfaces from the same acquisition. That flexibility helped make the point cloud valuable beyond the purpose for which the aircraft was originally flown.

Terrain becomes a shared input

Detailed elevation data were reused in many kinds of work. A drainage modeller could receive a terrain raster, an engineer a set of contours, a planner a hillshade, and a GIS analyst a slope layer. The acquisition technology might be invisible to the final user. This mirrors a broader pattern elsewhere in the book: mature geospatial infrastructure often becomes less visible precisely because its outputs become ordinary inputs to other work.

The derived products also changed organisational relationships. A council GIS team could hold the source survey and prepare simpler products for staff who did not need point-cloud software. Consultants could receive the same terrain basis as the client. Regional agencies could reuse a council acquisition for catchment work, while councils could later consume nationally published models. Elevation moved through organisations in several forms, from classified LAS files to grids, contours and shaded relief, each selected for a different audience.

This made provenance increasingly important. A user looking at a slope raster may be several processing steps removed from the aircraft and laser scanner that created the source observations. If the acquisition date, vertical datum, classification method or source survey identifier is lost along the way, the derived layer can become difficult to assess. National distribution later helped by making metadata and survey boundaries easier to retain with the products. The more ordinary elevation became, the more important it was to preserve the history hidden behind the raster.

From local contracts to national coordination

For much of the 2000s and early 2010s, New Zealand LiDAR remained a patchwork. Councils and other organisations commissioned surveys when they had money and a sufficiently strong flood, engineering, coastal or planning need. Main urban centres accumulated useful coverage, while many provincial and rural areas did not. Data could be held under different arrangements and produced to different specifications. The national map of available elevation was therefore shaped by local priorities and budgets.

The October 2018 Provincial Growth Fund decision altered that pattern. Government committed substantial funding for regional three-dimensional mapping, with LINZ managing a coordinated programme alongside regional partners. The public announcement in March 2020 described up to $19 million over five years and said LiDAR then covered about 10 per cent of New Zealand, with the programme expected to lift national coverage towards 80 per cent. The earlier inventory showed the main centres had useful coverage while provincial New Zealand remained sparse.

The programme brought together the Crown and partner organisations. Regional councils, local organisations and central government contributed to acquisitions, and the sequence of contracts reflected regional readiness and priorities. Marlborough, Tasman and Hawke's Bay were among the first contract areas announced, with further work across Bay of Plenty, the West Coast, Waikato, Canterbury, Southland and other regions. Agencies coordinated their surveys and data releases. National funding could fill gaps that were difficult for individual regions to justify alone.

Later coverage reporting needs to be read by date. In August 2024 LINZ said LiDAR for about 70 per cent of the country was available, with another 10 per cent due over the following months as the programme completed. Current programme reporting describes the partnership effort as bringing national coverage to more than 80 per cent. These figures describe the outcome of several years of acquisition and publication. Coverage was substantially lower when the funding decision was made in 2018.

A specification for a country

National coverage required national consistency. LINZ's National Aerial LiDAR Base Specification set common expectations for coordinate systems, vertical datum, classification, accuracy, project reporting and deliverables. Programme-era contracts produced classified point clouds alongside one-metre bare-earth elevation models and one-metre surface models. The programme allowed different local requirements while ensuring that regional datasets could be checked and combined without every user first reverse-engineering a local specification.

The specifications also evolved. Earlier versions set minimum pulse-density requirements that later versions increased, while accuracy controls and reporting requirements were refined. National elevation infrastructure was maintained through common technical specifications. A national model is credible only if users can understand how its component surveys meet common requirements. Standards became part of the data.

Quality control remained necessary even with a specification. LINZ describes centrally checking and editing programme data before publication. Classification errors, water surfaces, vegetation, bridges and complex terrain still need scrutiny. Hydroflattening and breaklines may be required so rivers, lakes and other surfaces behave sensibly in a derived elevation model. The move to national infrastructure therefore increased the importance of systematic quality assurance rather than making it disappear.

National elevation products combined derived grids from individual surveys, while the source LiDAR point clouds remained separate datasets. A one-metre national DEM is a derived product assembled from source surveys. It is extraordinarily useful for many GIS users who may never download a LAS file. Yet the source point cloud preserves much more information than the grid. Keeping both allows general users to work with a simple national terrain surface while specialists return to the measurements when a different classification or analysis is needed.

Open elevation

The other national change was access. Programme outputs were released through LINZ data services rather than retained only by the organisations that commissioned the flying. Current national products include the bare-earth DEM, surface model, hillshades and access to source point clouds. Individual source datasets may remain owned or licensed by the organisations that funded them, but they are distributed under open terms where specified. The result is a shared elevation resource whose reuse is no longer limited to the original flood model, engineering project or council.

Open access changed the practical audience. A university researcher could use a regional point cloud without negotiating a bespoke data transfer. A consultant could begin with an existing terrain model rather than commissioning a new survey simply to establish basic topography. Councils could combine neighbouring coverage. Community and environmental projects could use detailed terrain that would previously have been financially out of reach. None of this made LiDAR acquisition cheap, but coordinated public investment allowed the same expensive capture to support many later users.

The economics therefore resemble other parts of New Zealand’s geospatial history. Aircraft, sensors, flight planning, processing, classification, quality control, storage and professional expertise all remained real costs. The larger shift concerned who paid at the point of reuse. National coordination and open publication moved LiDAR from a series of project purchases towards a common spatial foundation.

There is also a maintenance problem. Terrain data can remain useful for years, but urban development, earthworks, river movement, forestry and natural hazards gradually age it. A national elevation model therefore needs acquisition dates and source boundaries as much as it needs seamless display. The smooth hillshade tempts users to forget that different parts of the country may have been captured in different seasons and years. Infrastructure is not the same thing as simultaneity.

LiDAR limitations

The dramatic appearance of LiDAR can encourage more confidence than the data deserve. A hillshade can reveal old channels, terraces, earthworks and subtle landforms that are difficult to see in aerial photography, especially when vegetation returns have been removed from a bare-earth model. Automated extraction still misclassified some shapes. A terrain feature still needs geological, archaeological, engineering or environmental context before it becomes evidence of what caused it.

Grid spacing and positional accuracy describe different properties. Grid spacing describes the raster cells in a derived product. Accuracy describes how close the measured or modelled position is expected to be to reality under specified conditions. Point density describes something else again. Mixing those concepts can make a detailed dataset sound more precise than it is. The national specification helps by separating them, but the user still has to read the metadata.

Water is another limitation. Conventional airborne topographic LiDAR does not generally map the bed beneath water in the same way that it maps dry land. Hydrographic or bathymetric LiDAR uses different wavelengths and specialised acquisition to penetrate suitable water conditions. New Zealand has used bathymetric LiDAR, including after Kaikōura. The ordinary national elevation model should not be imagined as a seamless laser measurement of land and underwater terrain.

Laser returns measured surfaces, while legal boundaries depended on survey and title records. Just as an orthophoto cannot decide where a cadastral boundary legally lies, a LiDAR point cloud measures surfaces rather than ownership. It can show a bank, fence line, building edge or road formation with impressive clarity. Those visible features may be useful evidence for many tasks, but they do not replace the survey record. High detail is not the same as legal authority.

A national terrain foundation

By the mid-2020s, LiDAR had travelled a long way from the local contracts of the early 2000s. The 2003 Christchurch survey was a carefully commissioned city project with specialist processing and local deliverables. Similar acquisitions spread through councils and regional programmes because flood modelling, engineering and environmental work could justify the cost. Repeated Canterbury surveys then showed the value of keeping measured terrain through time. The 2018 funding decision and subsequent regional partnerships turned that accumulated practice into a coordinated national programme.

Detailed terrain data became available for reuse across disciplines, projects and organisations. A planner, flood modeller, engineer, researcher or GIS analyst can begin with a nationally distributed terrain foundation whose source surveys, coordinate systems and specifications are documented. The point cloud remains available for questions that need the measurements themselves, while derived one-metre models make the data usable by a much wider audience. That combination of detailed source data and accessible standard products is what made LiDAR ordinary.

Earlier forestry, engineering, coastal or research LiDAR acquisitions may yet be identified in company and institutional records. The Christchurch survey provides a dated council case against which those earlier acquisitions can be compared.

LiDAR had made large-area terrain measurable as dense three-dimensional data. A very different acquisition model was also emerging. Small remotely piloted aircraft carrying ordinary digital cameras allowed individual teams to create their own high-resolution orthomosaics, point clouds and surface models over much smaller areas, shifting some detailed mapping from national and regional laser surveys to locally controlled drone capture.

LiDAR becomes practitioner training

By February 2021, LiDAR was also the subject of a GrowGISNZ GeoBites training session devoted to finding, accessing and working with New Zealand elevation data. The session sits after the national acquisition and open-data story: its emphasis was no longer whether LiDAR existed, but how ordinary practitioners could discover the available datasets, choose appropriate elevation products and use them in GIS. That shift from specialist acquisition to routine practitioner use is a useful marker in the technology’s New Zealand history.