Policing and public safety

By the mid-1990s New Zealand Police was rebuilding the way it recorded and retrieved operational information. The important geographic decision was made inside that wider records environment. A later practitioner account by Andy Gilmour and Jill Barclay records that geographic analysis was identified as a key reporting capability while the new system was being developed. Police wanted analysts to ask what had happened, where it had happened, when it had happened and what else had occurred nearby. That requirement placed location alongside offence type, time, people and other recorded attributes as a working part of crime analysis.

Earlier Police systems preceded MAPS. Police developed a client GIS and rolled it out to 19 analyst sites in early 1999, later expanding the deployment to 33 sites. It was a specialist environment. Each site needed a suitably configured workstation and printer, crime and mapping data had to be loaded locally, software versions could diverge and support was spread across geographically dispersed installations. Intelligence analysts were the principal users, and the practitioner history records limited familiarity with GIS among some of those expected to use it.

A photograph taken at the Police College on 2 November 1998 provides an earlier visual trace of this emerging practice. It shows a CRT displaying a geographic hot-spot map associated with unlawful taking. The surviving caption and image leave the software unidentified while confirming that screen-based crime-pattern mapping was already part of Police analytical work before the better documented MAPS milestones of 1999 and 2000.

The early client system made a simple operational change possible. Recorded incidents could be brought together according to place and time rather than read only as separate reports. An analyst could compare what had happened around a street, neighbourhood or patrol area, then add other attributes from the records. The map was no longer only the output of an enquiry. It became a way of constructing the enquiry.

That distinction was practical rather than philosophical. A burglary report contained an address, date, method of entry and details of what was taken. A list could sort those records by time or offence code, while location allowed the analyst to examine proximity, concentration and movement. Once events were geocoded, a cluster of burglaries could be compared with surrounding streets, with other recent events and with the timing or characteristics of those offences. Geocoding made geographic relationships easier to inspect while crime itself remained inherently uncertain.

The limits of the first deployment also became obvious. A client application installed at dozens of analyst sites was expensive to support and difficult to keep consistent. Local hardware mattered because the software demanded more capable machines than ordinary office applications, and maps also created a continuing need for printers. Data loading and version control added work that had little to do with crime analysis itself. Police had found a useful analytical method, but the first delivery model kept it tied closely to specialist locations and specialist staff.

By November 1999 the Map-based Analytical Policing System, MAPS, was operating in Wellington. Contemporary government material described the system as having proved useful there and said a nationwide expansion was proposed during 2000. The proposal set out a planned national rollout. The more detailed practitioner history later records a separate step in the sequence, with the general browser-based MAPS release dated to late 2000.

The surviving accounts use the name differently at these two dates. The MAPS name or operational lineage was in use in Wellington by November 1999. The 2008 practitioner account says the inaugural browser version was released in late 2000 after Police decided to replace the locally installed client application. Without a version-level technical record, it is safer to retain both facts than to invent a single clean launch date.

The Office of the Auditor-General’s 2001 report on dwelling burglary describes how Police used MAPS. The report was concerned with burglary investigation and prevention, not GIS history, which makes its operational examples especially useful. It describes MAPS as a more sophisticated system being tested in Lower Hutt during the audit and records its ability to analyse incidents across several dimensions. Location, time of day, similar methods of offending and repeated involvement of particular victims or victim types could be examined together.

One Auditor-General example begins with a burglary in Mt Victoria. An investigating officer could ask MAPS to display all burglaries within a defined radius of the address over the previous month, then separate those incidents according to whether a suspect had already been identified, what had been stolen and how entry had been gained. The point on the map was therefore only the starting location for a much wider query. That small example does more to explain operational GIS than a diagram of the software architecture: the map became useful when it helped an officer ask a better question about a real event.

The same records could be used differently for prevention. A Community Constable preparing to speak with a Neighbourhood Watch group could examine when recent burglaries had occurred and how homes had been entered. The geography defined the area of interest, while the incident attributes provided the detail needed for a practical discussion about local security. The system could also display offence density at a very local scale, including street-level hot spots, and identify addresses associated with repeated call-outs.

That workflow changed the role of the map. A paper wall map with pins could show where burglaries had occurred, but each extra question required someone to reorganise information manually. MAPS allowed the same body of records to be filtered repeatedly. A radius could be altered, the time window changed, offence characteristics selected and repeat locations highlighted without rebuilding the whole map. Crime geography became an interactive part of analysis rather than a static record of where events had been reported.

The Auditor-General also records a movement analysis tool that could help work across Police districts by following recorded sightings of suspects or vehicles. For longer-term planning, MAPS could show changes in crime patterns over time and compare those patterns with Census demographic information. The report used changes in income and crime levels as an example of the kind of relationship that could be examined. These were analytical possibilities. Geography made comparison easier, while causal interpretation still depended on the quality of the records and the judgement of analysts.

Repeat locations were particularly well suited to spatial analysis. An address that had generated several call-outs in a given period could be identified as a repeated operational demand rather than appearing as separate incidents in separate reports. An area with many recorded offences could also be distinguished from a place where fewer offences involved greater harm. Those distinctions would become more formal in later Police analysis, but the underlying problem was already present in MAPS: counts, concentration, recurrence and severity describe different aspects of crime.

The 2001 audit also records the organisational requirements behind the software. Effective use depended on continuous input of current incident information, skilled intelligence staff and co-operation from patrol and other operational personnel. Reports had to be complete enough for analysis, and analysts needed feedback about whether the products they supplied were useful. The map could reveal patterns only to the extent that operational recording placed reliable information behind the coordinates.

Geocoding had become part of Police business process rather than a back-office technical detail. The later practitioner history records that early geocoding within the National Intelligence Application succeeded for only about half of records. Police changed data-entry requirements so crime locations were captured more structurally. Later geocoding was estimated at roughly 75 to 80 per cent. Improving the map therefore required changes at the point where an officer or staff member recorded the event.

About eighteen months after the early-1999 client deployment, Police decided to replace the locally installed application with a web-based mapping approach. The practitioner history places this decision within a wider thin-client strategy. A centrally managed browser system removed much of the need for dedicated GIS workstations, local software installations and separate versions at each analyst site. The general browser-based MAPS release is dated to late 2000.

The change widened access without making every user a GIS analyst. Intelligence staff remained the main analytical users, but guided queries and simpler mapping functions could be made available to operational staff through ordinary Police network computers. The practitioner account estimated a potential reach of about 5,500 networked computers supporting roughly 10,500 staff. GIS had moved from a small number of specialist installations towards an organisational information service.

MAPS also sat closer to operational systems. The practitioner account describes connections with the National Intelligence Application and Computer Aided Resource Dispatch, with information capable of reaching maps within about fifteen minutes of response or entry. That did not make the system a real-time common operating picture in the later disaster-management sense. It did make mapped operational information available much sooner than a periodic analytical report or printed crime map.

The technical architecture remained less important to users than the change in access. A browser let staff query geography without owning a specialist workstation or managing local spatial datasets. Central maintenance meant a change to a map service or analytical function could be made once rather than installed at dozens of sites. Police provides a clear operational reason for adopting it: geographic information was useful to more people than could reasonably be given a dedicated GIS installation.

The audit also caught MAPS while implementation was still uneven. At the time of field work Police had only about 30 computers capable of running the system, later increased to 51 machines with MAPS software. The Auditor-General did not treat hardware as the only constraint. It also pointed to the availability of skilled intelligence staff and the need for operational staff to provide full reports and respond to analytical products. A sophisticated map could not compensate for incomplete incident recording or a weak relationship between analysts and the officers who were expected to act on the analysis.

The report therefore places crime mapping inside an organisational workflow. Intelligence units produced analysis, operational staff supplied much of the source information, supervisors and managers decided how to use the results, and later reports fed new information back into the system. MAPS could make repeated victimisation or a street-level concentration visible, but somebody still had to decide whether that pattern justified patrol changes, a prevention visit or a longer investigation. This distinction is useful because later accounts of spatial policing can make the technology sound more autonomous than it was. By 2001, Police were using the tool in their regular work.

Specialists continued using desktop tools alongside browser access. The practitioner account describes Police retaining or later restoring higher-end analytical capability for intelligence specialists while web access served broader users. That mixed model is familiar in mature GIS organisations: a large group needs straightforward map and query access, while a much smaller group needs deeper analytical control. Browser delivery changed the distribution of geographic information without eliminating specialist GIS practice.

Fire and emergency organisations also used geographic information. Emergency organisations also depended on records that could be tied reliably to location, but the operational questions were different. A fire service needed to know where incidents had occurred, how demand varied and whether stations and appliances were positioned to meet response targets. Incident reporting, dispatch, addresses and resource planning all touched geography without being the same system.

The New Zealand Fire Service Commission’s 2000/01 annual report records implementation of FIRS 2000, the Fire Incident Reporting System. It provided online network input for emergency incident information and current reporting, with the stated aim of improving the speed and usability of incident data. That is an important information-system milestone, but the annual report does not describe FIRS 2000 itself as GIS. A later ability to export incident records with coordinates belongs to the later system history and leaves the earlier database’s role unchanged.

The same annual report does, however, document a clearly spatial planning workflow. The Fire Service had developed a resource-allocation model to provide a consistent basis for long-term resourcing decisions in metropolitan and major provincial centres. Its aim was to locate fire stations according to fire risk while supporting a target that 90 per cent of incidents should be reached within five minutes of firefighters leaving the station. Location was being used to test the shape of the response network itself.

The model used TransCAD, a geographic transportation and network-analysis package, together with the New Zealand deprivation index as an indicator of residential fire risk. It also incorporated commercial fire risk and projected community development and growth. The annual report describes the input factors as nationally consistent and independently verifiable so the model could be applied across centres on a common basis. The analysis supported operational decisions. It was a spatial allocation problem with a response-time objective.

A station location that looked sensible on a city map could perform poorly once the road network, population risk and growth were considered. Moving a station changed the area that could be reached within the target time. Adding future development changed expected demand. Combining risk measures with network travel therefore allowed the Fire Service to compare alternative station configurations against a nationally stated operational objective.

The model also shows why emergency GIS cannot be reduced to disaster maps. Much of the geographic work happens before an emergency. Stations, response areas, addresses, road networks and risk surfaces are maintained and analysed during ordinary operations so that the organisation is better placed when an incident occurs. The Civil Defence Emergency Management Act and National CDEM Strategy from 2002 later formalised a modern institutional setting for emergency management, but they did not create this use of spatial analysis and should not be treated as the cause of GIS adoption.

Councils also developed emergency applications for modelling and prediction. A 2009 Scion review records a Hastings District Council system called Firesim, developed by 1999 as an application of GenaMap. Staff could place an ignition point on a topographic map and display predicted elliptical fire perimeters derived from weather inputs and New Zealand Fire Danger Rating System values. The application also reported rate of spread, perimeter growth, head-fire intensity and likely suppression resources. It was intended for tactical wildfire planning but was also promoted for strategic planning, post-fire debriefs and training. The review points back to two 1999 council sources, including a paper by B. Robertson in the Resource Management Division's Office of Emergency Management. Hastings used GenaMap for fire prediction and council spatial data.

The Hutt Valley provides a smaller example of council GIS escaping its normal office. A 2001 Fire Service research report records that a Chilton Saint James school GIS team reused Wellington Regional Council's wildfire study and digital layers for a vegetation-fire project on the eastern Hutt hills. The regional council supplied land-cover, slope, solar-radiation and wildfire-risk data; Andrew Smith is named for the wildfire study and digital data. Hutt City Council's Land Information section supplied aerial photographs, while Greg Drummond of Hutt City Council is acknowledged as the GIS in School mentor. The students added GPS fieldwork and their own analysis. The public application served different users and tasks from the council’s internal emergency systems. It was council data, a regional risk model and practitioner time being reused as a local learning and fire-risk experiment.

Shared location references also mattered. From 2001 onward, emergency-services address work involving LINZ, Police and the Fire Service sought common core information for roads, addresses and place names. The national history of those datasets belongs with the address and property story, not here. For public-safety operations the narrower point is enough: an incident system and a dispatch process are only as useful as the organisation’s ability to agree on where the incident is.

A further operational dependency sat underneath both Police and Fire work: the ability to translate an ordinary description of place into a location that several systems could use consistently. Emergency-services address work in the early 2000s brought Police, the Fire Service and LINZ into a shared specification for roads, addresses and place names. The full history of that national reference information belongs with the address chapter, but its operational value is clear here. A dispatch centre, intelligence system and fire-resource model cannot reliably join information if the same road, locality or property is represented differently in each source.

The same principle applied when emergency management became more formally organised under the Civil Defence Emergency Management Act 2002 and the first National CDEM Strategy. Those documents established responsibilities, planning expectations and national-local coordination, but they should not be used as a false starting date for emergency GIS. Spatial information was already present in public-safety work, and later CDEM organisations inherited council hazard layers, roads, facilities and other maintained datasets from systems developed for everyday business. The major disaster chapters later in the book will show what happened when those routine datasets had to be shared and updated under severe time pressure.

This distinction also keeps routine operational geography visible. Most days do not involve an earthquake or cyclone. Police still need to understand where repeated demand is occurring, Fire still needs stations and travel-time coverage, and emergency planners still need to know where roads, communities and facilities are. The spectacular map made during a disaster depends on years of quieter work that is rarely photographed: maintaining reference data, fixing addresses, training analysts, supporting servers and agreeing which version of a boundary or road network is authoritative.

By the 2010s Police spatial analysis was being used for a different kind of prioritisation. Sophie Curtis-Ham and Darren Walton’s 2017 Applied Geography paper compared two approaches at Census Area Unit level: the Priority Locations Index and the New Zealand Crime Harm Index. Both used geography to guide attention, but they measured different things. Treating them as alternative hot-spot maps would lose the point of the comparison.

The Priority Locations Index combined five sociodemographic variables with three crime variables, each indexed against national averages for Census Area Units. It was intended to identify communities experiencing a combination of social disadvantage and crime-related vulnerability. Police used the index as one input to resource allocation and the placement of Neighbourhood Policing Teams, alongside further local analysis. A place could therefore rank as a priority because of the interaction between social conditions and crime rather than raw offence counts alone.

The New Zealand Crime Harm Index approached the problem differently. Instead of giving each recorded offence equal weight, it assigned relative harm according to the seriousness of offence types using sentencing information. A location with fewer serious offences could therefore carry more measured harm than a location with a larger number of lower-harm offences. Crime count and crime harm were not interchangeable.

The study also distinguished these measures from density. A densely populated commercial or entertainment area could generate many incidents because large numbers of people passed through it. A socially disadvantaged residential community might have a different relationship between victimisation, harm and local vulnerability. GIS allowed the measures to be compared on the same geography, but it did not make them equivalent. The policy decision depended on which problem Police was trying to address.

The Priority Locations Index was calculated as an average of eight component measures: five demographic variables and three crime variables. Each component was converted to an index against the national average across Census Area Units, putting unlike measures onto a common scale before they were combined. The Crime Harm Index was built on a different foundation. Rather than averaging vulnerability indicators, it applied offence-specific weights derived from New Zealand sentencing information and then aggregated those weighted offences geographically. The authors calculated both total harm and population-controlled harm rates, which allowed a large population centre to be distinguished from a place experiencing unusually high harm relative to its population.

The authors then tested the relationship between the two indices rather than assuming they would identify the same places. The correlation was weak. Population size and urban-rural classification also explained part of the variation in Crime Harm Index scores, and mapped outliers showed Census Area Units where the two measures diverged substantially. Those differences are central to the historical story because they show Police geography moving beyond the question of where many offences occurred. The analytical framework could now ask whether a place had high offence volume, high weighted harm, high social vulnerability or some combination of the three.

This distinction marked a substantial change from the early MAPS scene. In 2001 an analyst could use incident points, time windows and Census information to examine local patterns and plan patrol or prevention activity. By 2017 spatial analysis could also combine nationally calculated measures to identify communities for sustained resource attention. The map had moved further upstream into prioritisation.

The change also made methodological choices more visible. An index built from sociodemographic and crime variables carries assumptions about which indicators represent vulnerability and how they should be combined. A harm index carries assumptions about how offence seriousness should be weighted. These indices supported decisions about Police resources; their results depended on the indicators and weighting selected. They are analytical tools designed to represent different operational concerns.

The 2017 paper tested how the methods overlapped and differed rather than declaring one universal answer. Some locations ranked highly under more than one measure, while others were elevated by one approach and not the other. That divergence is useful because it prevents a single word such as hot spot from flattening several quite different geographic problems. Density asks where incidents concentrate. Harm asks where the consequences of offending are greatest under the chosen weighting. Vulnerability asks where crime intersects with wider social conditions.

On 30 November 2016 Police launched policedata.nz, bringing another audience into the geographic information chain. The service provided interactive reports for partners, communities, researchers, media and other users who wanted to explore recorded crime data. Police described the reports as supporting planning, decision-making and policy development. The public service presented selected information through a separate interface.

The difference can be seen in the treatment of location and identity. Internal systems could contain detailed event records used for investigation, repeat-victim analysis and operational deployment. Public reports used anonymised information and disclosure controls. Police consulted Statistics New Zealand on official-statistics practice and the Office of the Privacy Commissioner on privacy issues. The service explicitly warned users not to attempt to re-identify people from released information.

Public geographic precision was therefore a policy choice as well as a technical one. A community needs enough location detail to understand patterns in its area, but publishing exact household-level victimisation can expose people. The current Police service excludes dwelling-based victimisations from some time-and-place public views, apart from burglary, while more complete information is handled through other controlled statistical products. The public map and the internal map answer different questions for different users.

The technology also changed after launch. Police records that policedata.nz moved from Flash to HTML on 31 October 2017 so the reports could be used on a wider range of devices. Further mobile-friendly reporting followed in April 2018. The changes widened access to public crime information. By then browser delivery was ordinary infrastructure. The operational decision was to make selected, privacy-managed Police information broadly usable outside the organisation.

The public reporting environment also changed the meaning of a crime map. An internal analyst can work with precise records because the organisation has lawful operational reasons to hold them and access can be controlled. A public report has to assume a very different audience, including people who know the local area well enough to infer identities from small numbers or precise addresses. Aggregation, suppression and the choice of which records to expose are therefore part of the geography. Less spatial precision can be a feature rather than a defect when the purpose is public statistics rather than investigation.

Police’s launch material made the intended change clear. Before policedata.nz, Police said it had not had an effective way to present recorded crime data online in an easily accessible format. The new service brought five interactive reports together and allowed users to combine information about victims, offenders, time, place and other characteristics without identifying individuals. Data that had previously been distributed through Statistics New Zealand were moved into a Police-hosted reporting environment, with Statistics New Zealand continuing to advise on the production of official statistics.

The later HTML and mobile changes extended that publication model rather than changing the underlying distinction. A partner organisation planning local services could use aggregated public reports without gaining access to intelligence files. A journalist could explore patterns without seeing the addresses and identities available to investigators. A resident could examine reported victimisation in an area without receiving a map of individual households. By this stage New Zealand Police was operating several geographic layers for different purposes, from internal operational detail to deliberately limited public statistics.

By the end of the 2010s New Zealand public-safety GIS had several distinct layers. Police analysts had moved from specialist client installations to centrally delivered browser tools, with location captured more deliberately in operational records. Spatial methods had expanded from local incident clusters to national approaches comparing density, harm and vulnerability. Selected crime information had also moved into public interactive reporting, where privacy and aggregation constrained what could be shown.

The Fire Service case followed a different path but reached the same basic dependency on maintained geography. An incident database improved the flow of emergency information, while a separate spatial model used risk, deprivation, roads and projected growth to test where stations should be placed. Emergency response depended on reliable address and road information long before anyone opened a crisis dashboard. GIS was embedded in preparation and resource allocation as well as response.

The major disasters that followed would place much heavier demands on these organisational capabilities. Canterbury, Kaikōura and Cyclone Gabrielle required shared operational pictures, rapidly changing road and cordon information, imagery, field updates and cross-agency data flows. Their roots include the quieter work described here: geocoded operational records, common reference data, browser access, maintained analytical capability and organisations accustomed to making decisions with geography.

Behind those systems was a growing community of people who learned to build, operate, teach and argue about them. By the time GIS had become part of crime analysis, emergency planning and public reporting, New Zealand no longer had only a small group of mapping specialists. It had practitioners with recognisable roles, user groups, conferences, professional networks and career paths. The technology was becoming ordinary enough that the field had to decide what counted as GIS work at all.

An official Police advisory for the November 2016 launch of policedata.nz names Chief Data Scientist Gavin Knight as the practitioner who would demonstrate the service in Wellington. The record identifies a specific contribution to public crime-data access, alongside the wider operational history of geographic information in policing.

Source WEB-POLICE-PEOPLE-2026 · New Zealand Police advisory, 29 November 2016

Duane Wilkins led the New Zealand Police geospatial team that implemented ArcGIS Enterprise in 2021, working with Sarah Hodgson and Aamir Bonamis.

Source author-police-2021 · Duane Wilkins, personal account