[talk-au] Australia "changing coordinates"

Andrew Davidson theswavu at gmail.com
Tue Aug 2 03:57:24 UTC 2016

It's interesting because it highlights one of the foundation myths of 
OSM; which is that it uses the "WGS84" co-ordinate system. This is a 
convenient myth and if you're talking about only mapping to the nearest 
5m then it is in effect true. However, once you start talking about 
sub-metre accuracy it stops being true for a number of reasons:

1. Unless you are cleared by the US DoD you don't have access to this 
level of accuracy.

2. If the accuracy of your GPS is supplemented by external data then 
this is inevitably based on one of the realisations of the ITRS; which 
is not "WGS84".

3. WGS84 is a semi-dynamic datum. Every 1st of January the USAF uploads 
new co-ordinates for each of their ground control/monitoring stations 
(where they will be at 1st July). Which means that every location in 
WGS84 co-ordinates needs an epoch to indicate when the measurement was 
taken but the database schema for OpenStreetMap doesn't have an entry 
for this.

So what happens in OSM is that each country gets mapped to their local 
approximation of WGS84. In North America this is NAD83, Europe has the 
ETRS89 and the various national variations. The UK must be the most 
confusing as the OS seems to be maintaining two system OSGB36 and OS Net 
but a least they are far enough apart to be obvious.

In Australia it means that we've been happily mapping to GDA94 and 
ignoring the fact that we're racing north at something like 70-90 mm a 
year. People mapping with consumer grade GPS are only within 5m anyway 
so don't care. People with access to survey grade GPS understand the 
difference and convert back to GDA94.

What difference will it make to OSM? It depends. If you're mapping some 
where that has good quality aerial imagery and people have carefully 
traced things then it's going to be obvious (https://goo.gl/sWWfYm). On 
the other hand most of OSM in Australia can be described as roughly 
sketched and if all you have is Bing imagery and some crudely traced 
streets then I doubt you'll be able to tell (https://goo.gl/B7WVfz).

What to do about it? Unfortunately the changeover to GDA2020 is not 
going to happen overnight, the plan is for a three year transition 
starting in 2017. If it did happen overnight then we could just shift 
all of the nodes in Australia to their new positions can keep on 
mapping. What's actually going to happen is that the various different 
organizations are going to switch at different times and the situation 
for aerial imagery mosaics will be interesting if they aren't 
retrospectively re-projected (you'll have some places on GDA2020 and 
others still on GD94). So I'm guessing the approach of wait and see 
might be the one to take. Obviously if organizations start releasing 
data in GDA2020 and people start importing it into OSM we're going to 
have to make the call: move everything to GDA2020 or convert imports 
back to GDA94? It's only going to be a real issue once we get access to 
aerial imagery that's in GDA2020 or people get general access to GPS 
with decimetre accuracy.

Whatever we do it'll be a good test to see what should happen in OSM 
when NAD83 gets replaced by 2022 and everything in the USA moves 1~1.5m 

On 29/7/16 22:40, Andy Mabbett wrote:
> See you when you reach England ;-)
>    http://www.bbc.co.uk/news/technology-36912700
> But seriously: what impact might this have, on OSM?

More information about the Talk-au mailing list