NSW and Victoria just jumped 1.8 metres north

Depending on how they store imported data vs different Google generated data, you could be talking about a simple transformation parameter change to their projection, or you could be taking about some complex, varying corrections.

For instance, if Google was already using street view or proprietary aerial images for ground truthing, they may have already compensated for the shift in various data sets, in different amounts, depending on when it was collected, and how they handled location adjustments due to discrepancies.

But being Google maps, I wouldn't expect accuracy better than a few meters for any given dataset anyway, given past performance.

https://www.icsm.gov.au/australian-datums-and-how-move-betwe...

Data providers like Google may have already made the switch, or instead of using the static datums they might use a time-dependent datum where the coordinates change with time.

Thought we like to think of Google as a big data company, for many of its projects it's more an aggregator of other people's data. Yes, it does create original data. But a lot of its data is from outside sources. Very often data that you paid for with your tax dollars, and then get to pay for again with your privacy.

Most local datums are anchored to the local geology. So in the US, if you're using nad27 or nad83, the reference system is "moving" along with the continent (nad83 = North American Datum 1983), based on some weighted benchmark vectors and a lot of math. It's one reason why different datums with defined epochs exist. Some are static, but some are actually dynamic with time (eg ITRF).

But most people don't grok any of that and just stick with defaults, or guess-and-check during map setup and go with whatever looks "close enough".

In your case, chances are the map provider switched datums with the update, from nad27 to nad83, and your software wasn't told about the change. Or it was, but your road restriction data was drawn in the old projection, and not converted / transformed to the new one.

Or, you could have had data that initially was mislabeled, recorded with GPS (wgs84) and entered using one of the many datums that's similar, but not the same. Or maybe collected using one of the many state plane coordinate systems (which people tend to like because it's more or less Cartesian- east,north) which are themselves built in top of one of the datums. You can pick the right coordinate zone, but wrong underlying datum.

If any data provider or aggregator or if your company GIS guru ever selected the wrong projection (choices often are dozens of coded options that look nearly identical), or left it blank (undefined) because they weren't sure...

A instant "shift" would happen as soon as you "corrected" the error, (or upgraded software, or any other number of situations).

This is where a good GIS guru earns their keep, because conversions between projections and datums are just math, and data can often be corrected in seconds by running it through a manual "inverse" conversion, assuming one can recognize what happened.

Back to Australia: if people used GPS coordinates (wgs84) and didn't convert to a Australian centric datum (which is the default option for most collectors), the data would "stay still" while the continent moves. Which is one of many reasons why selecting an appropriate datum/ projection is so important.

For the US: While there is some intra-datum movement between nad27 and nad83, most of the differences are do to improvements in the quality and accuracy of surveys over half a century. So the differences between the two datums tend to grow as you get further from the beginning points of the surveys.

Which means some areas, like a lot of the East coast, have minimal differences, and areas out West have larger deltas, but it varies.

A couple of meters would be better explained by a bad datum transformation than by drift, which is likely an order of magnitude smaller at even the worst locations for the US.

However, what3words is only accurate to 3m, so this change may not have affected their system at all.

More that the other states haven't yet updated their data to account for the changes at all, yet. They have until June.

that map was produced a few years ago, when gda2020 was defined. since then, victoria and nsw have moved about another 14-21 cm, hence the 1.8 m figure. i think there's a little padding included too to account for future drift.

as to the question of if these two states alone jumped, yes, they would have, if you have a datasink that collects official maps and collates them. roads that align on the ground might not meet on your map, since victorian data is gda2020 and south australia is gda94.

how do you deal with that in real life? it's a real concern to real people, and it is not clickbaity to report it. the article did a good job of explaining how it matters to a normal person, but not a good job of explaining why it matters. that is the correct criticism.

So all of Australia will move by June, but so far only NSW and Victoria have.

I don't see why it's a good idea to update coordinates of different parts of a continent at different times. Likely a political decision, not a technical one.

The trickiest part is going to be that some previous GPS coordinates will have been "pinned" to the 1994 grid, and others will have been taken live over that time span.

See, surveyors have always known that you can't actually just write down the GPS coordinates of a boundary or intersection or bridge, because of this exact problem. So surveyors do the math (well, most likely their specialized surveying GPS equipment does the math) and write down the coordinates relative to the baseline for the continental plate in question.

If you have coordinates you know were LAT, LNG in the 1994 baseline, it's easy to apply the transform and get the LAT', LNG' in the new 2020 baseline.

But what happens if your coordinates weren't provided by a surveyor carefully surveying a location in the appropriate baseline? What if your coordinates were measured by a jerk walking to a place last year with a GPS and writing down whatever was displayed? Those coordinates don't need to be transformed, they were already basically correct for the 2020 baseline. If you add the correction to those numbers you're introducing a new error. And to make matters worse, people have been entering non-calibrated numbers into your data set for 25 years -- some numbers will have 5 years of drift, some 15 years. If you know when an uncalibrated measurement was made, you can kind-of interpolate the drift, but without that you're just SoL.

To make matters worse, a company like Google is using data from a wide variety of sources, some they gathered themselves, some they get from others. With your own data, you might at least know either whether it was calibrated to the previous baseline or at least when it was collected, but someone else's dataset might be messy in unknownable ways.

Say you previously got the location of a park entrance from a government parks department, in the 1994 coordinates. Now you go through your aggregated dataset and apply the transformation to the coordinates and get the new number in 2020 coordinates (which thankfully will match what a consumer GPS actually says, now). Next month, when you update your data from the government parks department, it gives you a different coordinate for the park entrance, which is neither the number you had previously for the entrance in 1994 coordinates nor the number you computed in 2020 coordinates.

Now, do you assume that new number is in the new coordinate system, or might it still have been measured in the old coordinate system, either having been measured prior to the changeover or just on old equipment that was not updated? Ideally your interchange format would include the time a GPS measurement was taken as well as any baseline it was calibrated to, but do you want to guess how often interchange formats are "ideal"?

So, long story short, it's going to be a long time before most of the old coordinates are flushed from the system, and you'll probably occasionally see a "wrong" coordinate that can be traced to this sort of thing every now and again for the rest of your life.