The national coordinate system

A global position in a local system

By the early 1990s, New Zealand surveyors could determine positions from satellites with increasing precision, but the national coordinate framework had been built for a different era. The New Zealand Geodetic Datum 1949 had provided the basis for surveying and mapping for decades, while the New Zealand Map Grid carried national topographic mapping into metric production. Neither had been designed around a global geocentric reference system. GPS observations arrived naturally connected to international reference frames whose origin was at the centre of the Earth.

The change affected how coordinates were related to the Earth. New Zealand lies across the boundary between the Australian and Pacific plates, and positions measured in an international reference frame change as the crust moves. The motion is large enough to become visible in precision surveying over a relatively short period. A local datum that treated its control network as fixed could continue to support ordinary mapping, but increasingly accurate GPS observations would expose distortions accumulated through tectonic movement and the history of the older survey network.

NZGD49 had served national surveying and mapping for decades. It had been built from the survey technology, observations and national requirements available in the middle of the twentieth century and supported the cadastral and mapping systems described earlier in this book. The problem in the 1990s was that a datum designed around terrestrial control had to coexist with a positioning technology tied directly to global reference frames. New Zealand also needed a framework that could recognise that the ground itself was moving.

The replacement programme therefore began with measurement. Before a new national datum could be declared, surveyors and geodesists had to observe a network that tied New Zealand into the international system and provided enough repeated data to estimate how positions were changing.

The repeat GPS network

Five repeat first-order GPS campaigns were carried out between 1992 and 1998. The initial NZGD2000 realisation used 29 primary First Order 2000 stations, all of which were also existing NZGD49 control stations. Reoccupying the older marks allowed the new satellite-based observations to be connected with the control system already used by surveyors and mapping organisations.

LINZ contracted Dr Peter Morgan of the University of Canberra to process the first-order campaign data using GAMIT, GLOBK and GLOGR. Network observations were processed together and related to international reference stations. The work produced coordinates for the defining network in the International Terrestrial Reference Frame 1996, initially at an observation epoch around 1996.5, before relating them to the new datum’s reference epoch.

The survey programme extended beyond the 29 defining stations. Second-order network design and observation followed in 1995 and 1996, with later regional campaigns carrying the new framework through third and lower-order control. Existing observations were reused where their quality was adequate, while some marks had to be reobserved. A national datum only became operational for everyday survey work when the high-order framework could be connected to the denser control on which local surveys depended.

The programme involved more people than the authors named on the technical reports. Field crews occupied control marks across the country, geodetic staff planned networks and observation schedules, analysts processed repeated campaigns, and database staff carried revised coordinates into the national geodetic records. Equipment and processing methods had changed sharply since the triangulation era, but the job still required a national chain of observations that other survey work could trust.

Designing NZGD2000

LINZ introduced the New Zealand Geodetic Datum 2000 in 1998. It used the Geodetic Reference System 1980 ellipsoid and was aligned to ITRF96 at the reference date of 1 January 2000, or epoch 2000.0. Its origin was geocentric rather than defined through a local fitting of the ellipsoid to New Zealand. GPS observations could therefore be related much more directly to the national datum than they had been under NZGD49.

Graeme Blick and Don Grant were among the LINZ geodesists documenting the change as it developed. Merrin Pearse worked on the realisation of the new datum, while John Beavan at the Institute of Geological and Nuclear Sciences contributed to the deformation modelling. The published development papers also record Peter Morgan’s processing work on the first-order network. These named contributors sat within a wider programme involving LINZ, GNS and the survey community rather than forming a small group that created the system in isolation.

The datum also differed from NZGD49 in the way it treated movement. The design incorporated a deformation model used to relate observations made at different times to the reference epoch. LINZ described NZGD2000 as semi-dynamic. Official coordinates could remain expressed at epoch 2000.0 even though the marks and the ground on which they sat continued to move.

In practical terms, an observation made years after 2000 could be adjusted using the deformation model to calculate the coordinate that represented the point in the datum’s reference frame. The coordinate stored in a database could therefore appear stable while the system retained a model of ongoing horizontal motion behind it. New Zealand had not adopted a fully dynamic system in which every official mark coordinate continuously changed with time. It had built time into the process used to relate observations to the official coordinate.

The deformation model

The original deformation model represented broad crustal motion across the country. New Zealand spans two major plates, and the relative movement is not uniform from one end of the country to the other. In a global frame, positions in New Zealand move by several centimetres a year, with direction and rate varying across the plate boundary zone. The model allowed GPS observations collected at different dates to be related consistently to epoch 2000.0.

That approach depended on the quality of the velocity estimates used to build the model. The first model drew heavily on GPS observations from the 1990s, including a limited time span by later standards. Continuous GPS networks later supplied much longer and denser records of crustal movement. The model could therefore be improved as additional observations accumulated.

The model also had limits around sudden deformation. Plate motion can be approximated over time, but an earthquake can displace an area in seconds and create highly variable movement near faults. A datum that accounts for steady deformation still requires intervention when the ground jumps outside the assumptions of the existing model. That problem became operationally familiar during the following two decades.

NZGD2000 was designed to allow coordinates and the deformation model to be updated when new observations, earthquakes or local movement required it. The official coordinate system was therefore not a frozen table inherited from 1998. Its rules included mechanisms for maintaining the relationship between measured positions and the reference epoch.

Datum and projection

NZGD2000 provided the datum; a map projection was needed to produce rectangular eastings and northings. A datum defines the reference framework for position. A map projection converts positions from the curved reference surface into coordinates suitable for a flat map or GIS. The two are related but they are not the same object.

For national mapping, the New Zealand Map Grid had been tied to NZGD49. Its specialised polynomial formulation had served the country well and was used for the NZMS260 1:50,000 series. It was not compatible with the new datum simply by changing a label. LINZ adopted New Zealand Transverse Mercator 2000 for national mapping in 2001 after consultation.

NZTM2000 used the new datum and a conventional transverse Mercator projection. Coordinates were expressed as eastings and northings in metres and could be handled readily by mainstream GIS and mapping software. Cadastral surveying followed a related but different path through the new NZGD2000 meridional circuits, introduced in 2000 for local survey work.

The change therefore produced several linked names that later users could easily confuse. NZGD2000 was the datum. NZTM2000 was the national mapping projection. The meridional circuits were projected coordinate systems used for cadastral surveying. Software needed to know which combination applied before coordinates from different sources could be overlaid correctly.

Moving the data

A national datum change is partly a surveying exercise and partly a data-conversion programme. By the late 1990s government agencies, councils, utilities, consultants and research organisations already held large digital spatial databases. Those records had been captured in NZMG, local circuits or other coordinate systems tied to NZGD49. Moving into the new framework required more than changing metadata in a file header.

Coordinates had to be transformed using the correct operation for the source and destination systems. Software configurations, projection libraries and database definitions had to recognise the new datum and projections. Maps, standard templates, export routines and documentation also had to change. Organisations could hold old and new coordinate environments in parallel while applications, datasets and users moved at different speeds.

Rotorua District Council left a neat snapshot of that overlap on 5 April 2002. A council map of marae says it was produced on CIVIC INFO, the council's Geographic Information System, and its production path ends in an ArcMap `.mxd` file. The same map still declares New Zealand Geodetic Datum 1949 and New Zealand Map Grid, with Landsat 7 imagery from September 2000 underneath. New desktop software had arrived; the old national coordinate framework had not yet left. Migration happened layer by layer rather than at midnight on a changeover date.

Transformation changed the coordinate reference while preserving errors inherited from the source geometry. A boundary digitised from a small-scale map could be transformed accurately into NZTM2000 and remain no more accurate on the ground than it had been in NZMG. The same was true of early cadastral database geometry. Changing the coordinate framework corrected the reference system, not every positional weakness in the source record.

Automated GIS transformations hid more of the reference-frame work from users. A user could add two layers in different coordinate systems and watch one appear to move into place on the screen. The mathematics could be correct while the datasets still represented very different source scales, survey methods and accuracies. The new datum removed one class of incompatibility without removing the history of the data being transformed.

For national mapping, the change eventually appeared in the Topo50 series and other datasets produced in NZTM2000. Chapter 15 follows that cartographic history. For survey and GIS users, the earlier work consisted of updating coordinate definitions, transforming stored data and ensuring that new observations were not mixed with older coordinates under ambiguous labels.

From marks to active control

The first realisation of NZGD2000 depended heavily on periodic survey campaigns. By 2001 LINZ and the Institute of Geological and Nuclear Sciences had begun developing PositioNZ, a national active-control network of continuously tracking GPS stations. Instead of returning to the same high-order marks only during a campaign, a permanent receiver could collect observations every day.

The early system distributed observations as RINEX files. A 2003 LINZ technical report described fifteen permanent tracking stations in the North Island as the first phase, with hourly one-hour files and daily 24-hour files made available through the LINZ website. The files recorded observations at 30-second intervals and could be processed with a user’s own GPS data. The service was free, which made the national control network directly useful outside LINZ and GNS.

Physical marks and local survey control remained necessary for cadastral and engineering work. PositioNZ strengthened that wider network by supplying continuous high-quality observations that could be related to the marks and used to monitor the datum.

The partnership with GNS also joined geodetic and tectonic monitoring. Continuous stations measured position for surveying while also recording crustal deformation useful to earth-science work. The same observations could contribute to understanding plate motion, checking the deformation model and maintaining the national coordinate framework.

Real time

LINZ began trialling one-second real-time streaming from PositioNZ stations in 2006. The early trials were limited to a small user group and did not promise continuous service. By 2010, twenty stations were reported as streaming real-time data, while communications from remote sites remained a practical constraint. Some stations depended on long radio links or locations with limited broadband access.

The network was subsequently upgraded for wider real-time use. LINZ and GNS completed the upgrade in February 2011, according to the LINZ 2010/11 annual report. Communications were improved and the network was moving from GPS-only equipment toward multi-constellation GNSS capability. The real-time service allowed compatible equipment to receive live corrections from national reference stations rather than relying only on later post-processing.

The infrastructure changed what national geodetic control could deliver. An earlier surveyor connected to the national framework through a chain of physical marks and observations. A later user could also connect to continuously observed stations through archived RINEX or live data streams. The reference system remained national, but access to it increasingly occurred through network services as well as monuments in the ground.

The 2009 Fiordland earthquake exposed another use for that network. A magnitude 7.7 earthquake on 15 July produced measurable shifts, including movements large enough to exceed datum accuracy standards in parts of the country. LINZ geodetic staff and GNS scientists developed a localised deformation patch that could be added to the national model. Rachelle Winefield, Chris Crook and John Beavan later documented the work and the method used to incorporate earthquake deformation into LINZ processing.

Patches rather than replacement

The Fiordland work demonstrated how the semi-dynamic framework could be altered without abandoning the datum after every earthquake. The national model could retain broad plate-motion behaviour while a localised patch represented a deformation event affecting part of the country. Observations made after the earthquake could then be related more consistently to epoch 2000.0 coordinates.

The approach also avoided treating every movement in the same way. Some applications could continue using existing coordinates where ground movement was smaller than the accuracy of the data. High-accuracy surveying and engineering could require post-earthquake coordinates or direct connection to updated control. GIS datasets captured at metre-level accuracy did not necessarily need centimetre-level earthquake corrections.

New Zealand’s coordinate framework was therefore becoming more explicitly dependent on the intended use of the data. The datum provided the national reference, but the practical response to movement varied with survey order, source accuracy and application. That distinction would become much more visible after Kaikōura.

Kaikōura moves the control

The magnitude 7.8 Kaikōura earthquake struck shortly after midnight on 14 November 2016. Fault rupture and associated deformation affected a large part of the north-eastern South Island and lower North Island. LINZ rapidly published provisional horizontal and vertical movement maps for affected localities and guidance for surveyors working with pre- and post-earthquake coordinates. For most map users, an earthquake changes roads, coastlines and buildings. For surveyors it can also move the coordinate framework beneath the job, turning yesterday’s reliable control into something that has to be questioned.

The movement required an update to the existing deformation model. Near the faults, displacement varied sharply over short distances, and some high-order control marks could no longer support their previous coordinate orders. LINZ and GNS collected new observations and modelled the earthquake deformation while surveyors and infrastructure projects continued to need usable coordinates.

For a period, different coordinate references had to coexist deliberately. Survey guidance distinguished between existing pre-earthquake coordinates, provisional post-earthquake coordinates and positions derived from PositioNZ services. Users were expected to record which basis they had used. That was less tidy than a single instantly corrected database, but it reflected the reality that observations, models and official updates could not all be produced at the moment the faults moved.

The provisional movement maps made the coordinate problem visible. They showed horizontal and vertical displacement across the affected region rather than treating the earthquake only as damage to buildings or roads. Survey control had moved because the land carrying it had moved. The national system had to decide how much of that physical movement should appear directly as changed official coordinates and how much should remain represented through the deformation model.

January 2018

On 14 January 2018 LINZ released a new version of NZGD2000 and updated Landonline coordinates across the principal Kaikōura-affected area. The deformation model included patches for the 2016 earthquake and subsequent post-seismic deformation. Approximately 72,700 parcels in the Landonline survey and title database were updated as part of the change.

The horizontal changes were deliberately confined rather than applying every component of observed ground movement directly to stored coordinates across a much larger area. LINZ used a combination of reverse and forward patches within the deformation model. The approach limited disruption to existing GIS databases while still allowing new observations to be related to the revised datum correctly.

LINZ also supplied NTv2 grids so GIS custodians could update their own coordinates and keep external datasets aligned with the revised NZGD2000 environment. The change therefore propagated beyond the geodetic database and Landonline. Any organisation maintaining high-accuracy spatial information in the affected area had to decide whether its coordinates needed updating and how the new transformation should be applied.

The update also included heights where movement was large enough to require it. Horizontal and vertical effects did not share identical adjustment areas because the physical deformation and the way it was represented in the reference framework were different. Coordinate maintenance after Kaikōura was therefore not one uniform offset applied to every point.

Further refinement

Additional post-earthquake observations continued to arrive during 2018. LINZ incorporated new geodetic survey data into the National Geodetic Adjustment and refined the earthquake modelling again. A further update on 1 December 2018 changed geodetic coordinates and coordinate orders where the improved evidence justified it.

Across most of the national high-order network the later changes were small. LINZ reported that 98 per cent of Order 5 and better marks had horizontal changes below two centimetres. Larger adjustments remained concentrated in and around the earthquake-affected area. Some control marks near complex faulting were downgraded because the available model could not support their previous accuracy classification.

Previous coordinate values were retained in the Geodetic Database as historical values. That allowed users to trace which coordinate had applied before an update rather than erasing the earlier reference. By this point the national geodetic system was maintaining coordinate history as well as current values, observations and model versions.

Kaikōura required changes to coordinates and deformation modelling within NZGD2000. The response used mechanisms already anticipated by the semi-dynamic design: new observations, updated coordinates and modifications to the deformation model. The earthquake increased the scale and visibility of that maintenance work and forced the distinction between physical ground movement, official coordinates and GIS data alignment into normal operational practice.

A maintained national framework

By 2018 New Zealand’s coordinate framework looked very different from the system GPS had begun pressing against in the early 1990s. The national datum was geocentric and linked to an international reference frame. Mapping used NZTM2000 rather than NZMG. PositioNZ continuously observed the reference network and supplied archived and real-time data. The deformation model provided a formal way to relate observations at different times to epoch 2000.0.

The older physical infrastructure remained part of the system. Survey marks still connected local work to the national framework, and historical survey observations continued to contribute to geodetic adjustment. New infrastructure had been added around them: continuous stations, online observation archives, streaming services, model files, transformation grids and maintained coordinate databases.

The change also altered GIS practice. Coordinate reference information could no longer be treated as an incidental label attached to a dataset after it had been created. Datum, projection and, for high-accuracy work, epoch and deformation could affect whether two sets of coordinates represented the same physical place. The software made transformations easier, but the user still had to know what the source coordinates meant.

New Zealand’s national reference framework had become a maintained digital service as well as a survey network. Its coordinates could be revised after new observations or deformation, its model could be patched after earthquakes, and its active stations could supply data continuously. Aerial imagery was undergoing a similar change in use, moving from specialist survey and photogrammetric production into a routinely acquired, shared and refreshed layer used across GIS.