Map of auckland using suburb boundaries. Boundary data is provided by the NZFS under license.

The boundaries of New Zealand suburbs must be publicly released.

The boundaries of New Zealand suburbs and localities is held by the New Zealand Fire Service. For years, the NZFS has refused to provide this data under any terms except a restrictive license, and it has to stop.

In 2009, NZ data provider Koordinates, who offer a stunning array of geographic data to the New Zealand public for free, requested the boundaries for the suburbs and localities of New Zealand from the NZFS under OIA. The OIA was refused unless Koordinates were willing to agree to the license, a license which includes some key terms:

  1. You must update your data when boundary updates are released (about quarterly)
  2. You’re not allowed to provide the data to anyone else in any form that might possibly be reverse-engineered, unless they also agree to the same license.

The first point is the justification for the license. The concern, expressed by the NZFS, is that if the data were released without this requirement, online services would fail to keep pace with changes in boundaries, and thus people calling to report fires would accidentally give a neighbouring suburb.

The second point is ostensibly necessary to ensure that nobody simply rips the dataset off, however its actual effect is to drastically reduce the usefulness of the dataset — modern mapping solutions provide extremely high resolution imagery and any attempt to accurately render boundaries onto them is necessarily providing sufficient information to reverse engineer a functional equivalent.

New Zealand is losing out as a result. Many applications that should make use of suburb shapes — even simple ones that would never in their lives be used to determine the location for a callout (data visualisations, suburb selectors etc) — are not being made, or are being made with less useful units, because the barrier to obtaining this data is too high.

When Koordinates applied to the Ombudsman to resolve the issue in 2014, the Ombudsman refused the request. The justification given was that on balance, the NZ Fire Service had a point about the health and safety issue and that this outweighed the benefit of the public availability of the information.

Times have changed.

  1. NZ Locality information is now available, for any address in New Zealand, publicly via the NZ Street Address dataset available on LINZ.
  2. Many areas in New Zealand publicly gazette their boundaries, for example Wellington City.
  3. Other countries release this data, Southern Australia do a superb job.

With the availability of this information the concerns of the NZFS are no longer valid. Incorrectly specified suburbs or localities can come just as easily from the street address dataset as it could from boundaries — more so in fact since geocoding to an address is far more likely than checking boundaries. Similarly, an application using various currently gazetted boundaries could suffer from the same update problem the NZFS purports to solve via their license.

And this leads us to the main reasons why this is now urgent:

  1. The data is necessary for many useful applications and visualisations.
  2. Sufficient data is now publicly available to construct a competing dataset from open sources.

If #2 happens — and it will because #1 is highly motivating — the NZFS will have unwittingly created its worst nightmare. Rather than getting reports from a matching dataset that is merely months or years out of date on a boundary set that barely changes, they will instead be getting reports from a dataset that was constructed in a completely different way with far more chances for error.

In their response to the Ombudsman, the New Zealand Fire Service Commission stated:

The consequences of people relying on products that used different definitions of locality boundaries had been seen when the ambulance services used a different product for registering the location of an incident to that used by the Fire Service, and an ambulance dispatched to the same incident as the Fire Service had gone to a different location.

By not releasing the NZ Localities dataset, the New Zealand Fire Service is actively generating the conditions for this situation to re-occur, and on a much wider scale.

In the final decision the Ombudsman stated:

In summary, I am not persuaded that the public interest considerations in favour of disclosure, under s 9(1) of the OIA, outweigh the public interest in avoiding prejudice to the measures designed to protect public health and safety, and to prevent or mitigate material loss to members of the public. The effectiveness of those measures depends on those using it being required to use the most up-to-date (and unmodified) version of the dataset. There is no doubt that completely unlicensed distribution of the dataset would undermine this.

This is no longer a valid calculus. We are at the cusp of seeing two competing datasets exist, a scenario far worse for New Zealanders than a lack of updates.

In the same response, the NZFSC stated:

It is expected that prior to end of 2014, and subsequent to establishing NZFS as custodian, NZFS will release the NZ Localities dataset under a new Creative Commons licence (CC variation yet to be defined). The NZ Localities dataset will then be released through the Department of Internal Affairs website and will be available at no cost to the public (as per current access protocols) directly from NZFS under CC licensing.”

This never happened, but it needs to happen now, before it’s too late. Get the data out there, via demonstrably responsible parties like Koordinates and LINZ who can offer the data to application developers in forms that are easy to update or self-updating, and we minimise the negatives and avoid a disaster scenario.

I’m not the only one who is pointing this out, Hamish Campbell has been on this for years. See his comment to this article for more details.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.