Re: [OSM-talk] What does WGS84 mean for openstreetmap these days?

[Simon Poole]

It's not quite the same thing as uncertainty in the datum itself.

Crust movements simply lead to things being somewhere else relative to a
global datum (aka they have moved), so a new measurement of the position
for the same object would return the correct current position and
theoretically if the original position was measured you could simply
apply a translation based on the crust movement to get the correct
current coordinates.

Am 20.12.2019 um 15:36 schrieb Jóhannes Birgir Jensson:
> Well the current issue in Iceland is a error of 50 cm between 1993 and 2016 
> due to crust movements. So it's less than 2 meters but more than one cm. What 
> is your accuracy limit if 2m is just unacceptable but a centimeter is?
>
>
> 19. desember 2019 kl. 17:58, skrifaði "Greg Troxel" :
>
>> Interesting about the datum history. But this is the cm strawman I
>> wasn't talking about, not the 2m issue.
> ___
> talk mailing list
> talk@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk



signature.asc
Description: OpenPGP digital signature
___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] What does WGS84 mean for openstreetmap these days?

[Greg Troxel]

"Jóhannes Birgir Jensson"  writes:

> Well the current issue in Iceland is a error of 50 cm between 1993 and
> 2016 due to crust movements. So it's less than 2 meters but more than
> one cm.

That's interesting and a useful data point for later discussion about
the points that my message said this discussino isn't about!

> What is your accuracy limit if 2m is just unacceptable but a
> centimeter is?

There are several issues that must be addressed over time for improved
accuracy.  I am not claiming that any particular accuracy is required
and I am very definitely NOT suggesting that OSM adopt accuracy
requirements for data.

I am not following where "1 cm" came from in this discussion and don't
think it is relevant.  (I do think that asking the question of how to
get to 1 cm eventually is interesting, but I see that as a separate
issue.)


My point is that our current definition of "WGS84" has ~2m of fuzz
*intrinsic to the definition* because (as discussed at length on the
proj list) saying "WGS84" means "these coordinates are in one of six
datums and I am not telling you which".  There is no need for OSM to
have this definitional uncertainty, and I think that is both the biggest
issue and the easiest to address.

Regardless of the definition, it is clear that there will be data of
varying accuracy in OSM.  I am not objecting to that reality, or asking
that OSM adopt accuracy standards.  There seems to be a shared norm that
more accurate locations of  nodes are preferred to less accurate
locations.

I think there's also a shared belief that wildly inaccurate nodes are
not helpful; if I added a POI that was 10 km off (in an area where there
are many things between the mapped location and the real location), then
that is probably a bad thing to have done.  But if I add a POI that is
100m off, someone might fix it, but I don't expect they would tell me
that I should not have added it.  Certainly this is true at 10m error
(the lack of outrage; it might or might not get improved).


So about the narrow issue of the definition of OSM's coordinate
system, do you prefer

  leaving it defined as "WGS84", so that coordinates have to be treated
  as having an intrinsic uncertainty of several meters

  changing it to "WGS84, and in particular the revision currently in use
  by GPS", allowing one to treat the datum as relatively precise and
  thus only question the accuracy of the coordinates themselves?

  something else?


Note that the second option allows one to transform e.g. precise
coordinates from a geodetic control point to add to OSM.  (Yes, I
realize one could transform to a particular realization anyway.)

In all seriousness, I cannot tell if you object to this first step, or
are making comments about some later step that might happen, or at
talking about the issues of OSM perhaps adopting accuracy requirements
(not on the table!0, or something else.

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] What does WGS84 mean for openstreetmap these days?

[Jóhannes Birgir Jensson]

Well the current issue in Iceland is a error of 50 cm between 1993 and 2016 due 
to crust movements. So it's less than 2 meters but more than one cm. What is 
your accuracy limit if 2m is just unacceptable but a centimeter is?


19. desember 2019 kl. 17:58, skrifaði "Greg Troxel" :

> Interesting about the datum history. But this is the cm strawman I
> wasn't talking about, not the 2m issue.

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] What does WGS84 mean for openstreetmap these days?

[Greg Troxel]

Yantisa Akhadi  writes:

> To add more challenges to this issue is imagery offset
> . The value
> can even be varied from tiles to tiles, that we often need to shift the
> object a couple of meters away. In a remote area, where there are no GPS
> traces as a reference, satellite imagery is often the only reference even
> when possibly it was a couple of meters off.

Certainly that is a challenge.  But, once the datum is defined clearly,
then it is work to figure out, but not a question of ambiguity as to
what is desired.

I am unclear on whether most imagery offset issues are due to imagery
being in a datum other than modern WGS84 (and not transformed), or just
due to errors.  In Massachusetts, US, it seems most of the imagery
sources are pretty well aligned.  But this cries out for actualy
checking better!


___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] What does WGS84 mean for openstreetmap these days?

[Yantisa Akhadi]

To add more challenges to this issue is imagery offset
. The value
can even be varied from tiles to tiles, that we often need to shift the
object a couple of meters away. In a remote area, where there are no GPS
traces as a reference, satellite imagery is often the only reference even
when possibly it was a couple of meters off.

Best,
Iyan

On Fri, Dec 20, 2019 at 12:59 AM Greg Troxel  wrote:

> "Jóhannes Birgir Jensson"  writes:
>
> > I don't think we can or will be providing accuracy up to cm when most
> > of the stuff we map from our chairs is off by a meter or two anyways -
> > the beauty is that it doesn't matter for 99,99% of users. If a
> > centimeter matters then we are probably dealing with legal matters and
> > there OSM makes it quite clear it is not suitable for such.
>
> My actual proposal, as opposed to the things I pointed out I wasn't
> proposing, is about removing the ~2m uncertainty that exists from our
> current definition.
>
> As for cm level, OSM does not have accuracy specifications and won't.
> Some people like to be accurate, and others like to add lots of detailed
> tags.  Between us we have great map.  I don't agree that anybody who is
> trying to be more accurate is necessarily concerned with something that
> is "legal".  I would expect many people would like to see better than 2m
> accuracy.
>
> Certainly cm-level is very difficult, and I see that as being pretty far
> out in the future.
>
> You didn't comment on the notion of defuzzing the reference to WGS84, so
> I'll assume you are ok with that.
>
> > Also regarding the accuracy, as another fast moving country Iceland is
> > actually splitting in the middle and so it edges west and east and
> > south as well, depending on where you are in the country. We've had 3
> > official national datums now, ISN93, ISN2004 and ISN2016 (helpfully
> > naming them after years). The fact is that pretty much everything is
> > still running in ISN93, ISN2004 saw very little uptake and ISN2016 has
> > started very slowly.
> >
> > So for Iceland we do know that we are never going to achieve a
> > centimeter accuracy, pretty much ever, and don't expect a free people
> > sourced geographical database to reach it.
>
> Interesting about the datum history.  But this is the cm strawman I
> wasn't talking about, not the 2m issue.
>
> I would not be surprised if in 20 years OSM had some approach to
> coordinates of crust-fixed points.
>
> ___
> talk mailing list
> talk@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk
>
___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] What does WGS84 mean for openstreetmap these days?

[Greg Troxel]

"Jóhannes Birgir Jensson"  writes:

> I don't think we can or will be providing accuracy up to cm when most
> of the stuff we map from our chairs is off by a meter or two anyways -
> the beauty is that it doesn't matter for 99,99% of users. If a
> centimeter matters then we are probably dealing with legal matters and
> there OSM makes it quite clear it is not suitable for such.

My actual proposal, as opposed to the things I pointed out I wasn't
proposing, is about removing the ~2m uncertainty that exists from our
current definition.

As for cm level, OSM does not have accuracy specifications and won't.
Some people like to be accurate, and others like to add lots of detailed
tags.  Between us we have great map.  I don't agree that anybody who is
trying to be more accurate is necessarily concerned with something that
is "legal".  I would expect many people would like to see better than 2m
accuracy.

Certainly cm-level is very difficult, and I see that as being pretty far
out in the future.

You didn't comment on the notion of defuzzing the reference to WGS84, so
I'll assume you are ok with that.

> Also regarding the accuracy, as another fast moving country Iceland is
> actually splitting in the middle and so it edges west and east and
> south as well, depending on where you are in the country. We've had 3
> official national datums now, ISN93, ISN2004 and ISN2016 (helpfully
> naming them after years). The fact is that pretty much everything is
> still running in ISN93, ISN2004 saw very little uptake and ISN2016 has
> started very slowly.
>
> So for Iceland we do know that we are never going to achieve a
> centimeter accuracy, pretty much ever, and don't expect a free people
> sourced geographical database to reach it.

Interesting about the datum history.  But this is the cm strawman I
wasn't talking about, not the 2m issue.

I would not be surprised if in 20 years OSM had some approach to
coordinates of crust-fixed points.

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] What does WGS84 mean for openstreetmap these days?

[Greg Troxel]

Simon Poole  writes:

> Thus is a slightly tricky subject and it is not going away.
>
> For another aspect of it see
> https://www.openstreetmap.org/user/StephaneP/diary/390290

Thanks -- I had not seen that.

I would say that to be pedantic, there is a minor error in the post, in
that OSM coordinates are by definition WGS84.  Agreed that when people
add points with coordinates that are in other datums, then the points in
the database have errors.

I am in the process of figuring out how to deal with this, as accurate
locations in my state basically come from using the state's reference
network, which gets you NAD83(2011) epoch 2010.0.

> Essentially in some cases we are using imagery that isn't actually using
> WGS84 as if it was (fsvo of WGS84 as you correctly point out) and we
> currently don't actually have a way to correct this . And yes while
> continental shift is for most countries smaller than all the other ones
> when adding geometry, for Australia this not necessarily true.

That's true for how people with editors generate coordinates.  It seems
quite possible to adjust imagery to WGS(G1762) in editors, and arguably
that should happen.  I wonder though how often imagery is sufficiently
accurate in some national datum that this matters.  I suspect it's more
and more often.


In my message, was really trying to deal with the issue that by saying
"WGS84" instead of "WGS84(current realization)", OSM has a built-in
uncertainty of about 2m before we even start talking about where data
came from.  That seems easy to take off the table with zero workflow
changes.  At least then there will be a clear definition of what's
intended.

Actually changing the notions is much more difficult and I was trying to
separate the easy step from the harder ones.


___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] What does WGS84 mean for openstreetmap these days?

[Jóhannes Birgir Jensson]

Hello Greg

I don't think we can or will be providing accuracy up to cm when most of the 
stuff we map from our chairs is off by a meter or two anyways - the beauty is 
that it doesn't matter for 99,99% of users. If a centimeter matters then we are 
probably dealing with legal matters and there OSM makes it quite clear it is 
not suitable for such.

Also regarding the accuracy, as another fast moving country Iceland is actually 
splitting in the middle and so it edges west and east and south as well, 
depending on where you are in the country. We've had 3 official national datums 
now, ISN93, ISN2004 and ISN2016 (helpfully naming them after years). The fact 
is that pretty much everything is still running in ISN93, ISN2004 saw very 
little uptake and ISN2016 has started very slowly.

So for Iceland we do know that we are never going to achieve a centimeter 
accuracy, pretty much ever, and don't expect a free people sourced geographical 
database to reach it.

--Jóhannes / Stalfur


19. desember 2019 kl. 14:37, skrifaði "Greg Troxel" :

> (This is a long and complicated subject and I am intentionally asking
> only part of the question.)
> 
> It's been said from the beginning that coordinates in the openstreetmap
> datbase are in "WGS84". That more or less meant "what a GPS receiver
> showed", back in the days when GPS was the GNSS system of choice and
> accuracies were low compared to talking about versions of WGS84.
> 
> In discussion on the proj list, it seems the consensus view is that
> WGS84 is now a term that refers to any one of the 6 realizations of
> WGS84 over time. This makes sense when you have data that is merely
> labeled WGS84, without a more specific label such as WGS84(G1762). This
> means that WGS84 is considered low accuracy (because the original was),
> and thus any transforms involving it are assigned high error values.
> 
> This page has a good overview of the various WGS84 realizations and
> their relationship to ITRF realizations:
> 
> https://www.e-education.psu.edu/geog862/node/1804
> 
> As normal people (or at least normal nerds) get access to more accurate
> positions, this question begins to matter, as in North America positions
> in original WGS84 and modern WGS84 differ by more than a meter.
> 
> I should note that now that WGS84 has converged to ITRF, and new ITRF
> realizations seem to be at most cm-level changes from previous ones, I
> do not expect future WGS84 revisions to be signficantly different from
> either the current one.
> 
> So, I wonder if we want to change the definition for OSM coordinates
> from "WGS84" to "the realization of WGS84 currently in use by GPS".
> That doesn't change older coordindates (and I am not suggesting any
> automated changes!!!). But it does give a notion of what coordinates
> should be, both in using them and in producing new ones for editing. I
> expect that this will have zero practical effect for most people, but
> will allow higher accuracy for those who are into extreme accuracy.
> 
> postscript:
> 
> I am intentionally leaving out of this discussion two more issues (which
> could result in further changes, with much more complexity). I list
> them so that those with some background in geodesy can begin to ponder,
> and to explain that my stopping at the proposal above was intentional.
> 
> 1) WGS84 is a US datum. BEIDOU, GALILEO, GLONASS use different datums.
> SBAS systems also use different datums -- WAAS seems to give
> coordinates in "ITRF2000 (current epoch)". It seems most are
> equivalent to some modern ITRF, with possibly differing epochs.
> 
> (I will assume for point 2 that there OSM redefines coordinates to be a
> particular ITRF at a particular epoch, probably matching the current
> WGS84.)
> 
> 2) ITRF is global, but objects we map are generally crust-fixed on some
> plate. The US has a (mostly, if you're not in CA) crust-fixed datum,
> NAD83, and other countries do too. This is particularly acute in
> Australia which is a notably fast-moving country :-) The modern trend is
> for stations to have velocities and not just coordinates. Over 20
> years, this starts to matter. Several countries are introducing new
> national datums that are intended to address some of these issues. I
> don't think it makes sense for OSM to deal with this issue for a few
> years.
> 
> ___
> talk mailing list
> talk@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

Re: [OSM-talk] What does WGS84 mean for openstreetmap these days?

[Simon Poole]

Thus is a slightly tricky subject and it is not going away.

For another aspect of it see
https://www.openstreetmap.org/user/StephaneP/diary/390290

Essentially in some cases we are using imagery that isn't actually using
WGS84 as if it was (fsvo of WGS84 as you correctly point out) and we
currently don't actually have a way to correct this . And yes while
continental shift is for most countries smaller than all the other ones
when adding geometry, for Australia this not necessarily true.

Simon

Am 19.12.2019 um 15:33 schrieb Greg Troxel:
> (This is a long and complicated subject and I am intentionally asking
> only part of the question.)
>
> It's been said from the beginning that coordinates in the openstreetmap
> datbase are in "WGS84".  That more or less meant "what a GPS receiver
> showed", back in the days when GPS was the GNSS system of choice and
> accuracies were low compared to talking about versions of WGS84.
>
> In discussion on the proj list, it seems the consensus view is that
> WGS84 is now a term that refers to any one of the 6 realizations of
> WGS84 over time.  This makes sense when you have data that is merely
> labeled WGS84, without a more specific label such as WGS84(G1762).  This
> means that WGS84 is considered low accuracy (because the original was),
> and thus any transforms involving it are assigned high error values.
>
> This page has a good overview of the various WGS84 realizations and
> their relationship to ITRF realizations:
>
>   https://www.e-education.psu.edu/geog862/node/1804
>
>
> As normal people (or at least normal nerds) get access to more accurate
> positions, this question begins to matter, as in North America positions
> in original WGS84 and modern WGS84 differ by more than a meter.
>
> I should note that now that WGS84 has converged to ITRF, and new ITRF
> realizations seem to be at most cm-level changes from previous ones, I
> do not expect future WGS84 revisions to be signficantly different from
> either the current one.
>
> So, I wonder if we want to change the definition for OSM coordinates
> from "WGS84" to "the realization of WGS84 currently in use by GPS".
> That doesn't change older coordindates (and I am not suggesting any
> automated changes!!!).  But it does give a notion of what coordinates
> should be, both in using them and in producing new ones for editing.  I
> expect that this will have zero practical effect for most people, but
> will allow higher accuracy for those who are into extreme accuracy.
>
>
> postscript:
>
> I am intentionally leaving out of this discussion two more issues (which
> could result in further changes, with much more complexity).  I list
> them so that those with some background in geodesy can begin to ponder,
> and to explain that my stopping at the proposal above was intentional.
>
> 1) WGS84 is a US datum.  BEIDOU, GALILEO, GLONASS use different datums.
>SBAS systems also use different datums -- WAAS seems to give
>coordinates in "ITRF2000 (current epoch)".  It seems most are
>equivalent to some modern ITRF, with possibly differing epochs.
>
> (I will assume for point 2 that there OSM redefines coordinates to be a
> particular ITRF at a particular epoch, probably matching the current
> WGS84.)
>
> 2) ITRF is global, but objects we map are generally crust-fixed on some
> plate.  The US has a (mostly, if you're not in CA) crust-fixed datum,
> NAD83, and other countries do too.  This is particularly acute in
> Australia which is a notably fast-moving country :-) The modern trend is
> for stations to have velocities and not just coordinates.  Over 20
> years, this starts to matter.  Several countries are introducing new
> national datums that are intended to address some of these issues.  I
> don't think it makes sense for OSM to deal with this issue for a few
> years.
>
> ___
> talk mailing list
> talk@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk



signature.asc
Description: OpenPGP digital signature
___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk

[OSM-talk] What does WGS84 mean for openstreetmap these days?

[Greg Troxel]

(This is a long and complicated subject and I am intentionally asking
only part of the question.)

It's been said from the beginning that coordinates in the openstreetmap
datbase are in "WGS84".  That more or less meant "what a GPS receiver
showed", back in the days when GPS was the GNSS system of choice and
accuracies were low compared to talking about versions of WGS84.

In discussion on the proj list, it seems the consensus view is that
WGS84 is now a term that refers to any one of the 6 realizations of
WGS84 over time.  This makes sense when you have data that is merely
labeled WGS84, without a more specific label such as WGS84(G1762).  This
means that WGS84 is considered low accuracy (because the original was),
and thus any transforms involving it are assigned high error values.

This page has a good overview of the various WGS84 realizations and
their relationship to ITRF realizations:

  https://www.e-education.psu.edu/geog862/node/1804


As normal people (or at least normal nerds) get access to more accurate
positions, this question begins to matter, as in North America positions
in original WGS84 and modern WGS84 differ by more than a meter.

I should note that now that WGS84 has converged to ITRF, and new ITRF
realizations seem to be at most cm-level changes from previous ones, I
do not expect future WGS84 revisions to be signficantly different from
either the current one.

So, I wonder if we want to change the definition for OSM coordinates
from "WGS84" to "the realization of WGS84 currently in use by GPS".
That doesn't change older coordindates (and I am not suggesting any
automated changes!!!).  But it does give a notion of what coordinates
should be, both in using them and in producing new ones for editing.  I
expect that this will have zero practical effect for most people, but
will allow higher accuracy for those who are into extreme accuracy.


postscript:

I am intentionally leaving out of this discussion two more issues (which
could result in further changes, with much more complexity).  I list
them so that those with some background in geodesy can begin to ponder,
and to explain that my stopping at the proposal above was intentional.

1) WGS84 is a US datum.  BEIDOU, GALILEO, GLONASS use different datums.
   SBAS systems also use different datums -- WAAS seems to give
   coordinates in "ITRF2000 (current epoch)".  It seems most are
   equivalent to some modern ITRF, with possibly differing epochs.

(I will assume for point 2 that there OSM redefines coordinates to be a
particular ITRF at a particular epoch, probably matching the current
WGS84.)

2) ITRF is global, but objects we map are generally crust-fixed on some
plate.  The US has a (mostly, if you're not in CA) crust-fixed datum,
NAD83, and other countries do too.  This is particularly acute in
Australia which is a notably fast-moving country :-) The modern trend is
for stations to have velocities and not just coordinates.  Over 20
years, this starts to matter.  Several countries are introducing new
national datums that are intended to address some of these issues.  I
don't think it makes sense for OSM to deal with this issue for a few
years.

___
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk