Arcgis Online Inaccuracies

[ChrisBowles]

Hello,

I've jsut created a map for our Historic Environment Record.  The trouble is all points and polygons seem to out by about 80-90 metres from the features they relate to in the basemaps.  The points shapefile has a GB Ordnance Survey coordinate system, while the shapefiles are in WGS 84.  Both are having the same problem.  Are the basemaps off or am I?  Any help on how to fix this would be very helpful as I would like to go public with the map, but won't until accurate.

Thanks,
Chris

[ChrisBowles]

Hi,

It would be great to get a response to this so I can go live.  The basic problem is that both my polygon and point shapefiles are projecting about 80 metres east of where they should be no matter which open source basemap I choose.  As I said earlier, I'm using a combination of WGS84 and GB OS coordinate systems and both are having the same problem.  These project perfectly in my ArcMap projects which uses an Ordnance Survey mastermap and OS GB coordinate system.  Any help is greatly appreciated.

[MikeMinami]

These issues are very difficult to troubleshoot. Other users have discussed offsets, but on the order of 10 meters.

http://forums.arcgis.com/threads/42696-Projecting-data-correctly-in-Arcgis-basemaps.

Are you adding shapefiles to your webmap? Can you share your shapefiles here?

You mention that these display fine in ArcMap. However, the coordinate system you are using is OS GB coordinate system. What happens if you change the data frame coordinate system to web mercator Aux sphere? Do you see a similar problem? The problem seems like a coordinate system transformation issue.

Thanks,

Mike

[ChrisBowles]

Hi Mike,

Thanks for the response!  I tried to convert into WGS 84 in ArcMap (9.2) and the points and polygons I use projected as normal.  I've attached two shapefiles, one in OSGB and one in WGS 84. Both of these have the offset issue in Explorer.

Thanks!
Chris

[MikeMinami]

Unfortunately, I'm not a  coordinate system expert. However, I believe you have to apply a transformation to your data to get it to align correctly. This text is from Wikipedia, which appears to describe the shift you are seeing.

http://en.wikipedia.org/wiki/Ordnance_Survey_National_Grid#Datum_shift_between_OSGB_36_and_WGS_84

In Cornwall the WGS 84 longitude lines are about 70 metres east of their OSGB 36 equivalents, this value rising gradually to about 120 m east on the east coast of East Anglia.

A geodetic transformation between OSGB 36 and other terrestrial reference systems (like ITRF2000, ETRS89, or WGS 84) can become quite tedious if attempted manually. The most common transformation is called the Helmert datum transformation, which results in a typical 7 m error from true. The definitive transformation from ETRS89 that is published by the OSGB is called the National Grid Transformation OSTN02.[2] This models the detailed distortions in the 1936�??1962 retriangulation, and achieves backwards compatibility in grid coordinates to sub-metre accuracy.

Hope this helps,

Mike

[MelitaKennedy]

Supposedly, I am a coordinate systems expert, and it does look like there's a missing transformation.

Both posted shapefiles are still using British National Grid (based upon OSGB 1936).

If I add "Imagery" basemap which is using WGS 1984 Web Mercator, then add the shapefiles, I get the warning message about different geographic coordinate systems. I then set a transformation from OSGB 1936 to WGS 1984 using:

OSGB_1936_To_WGS_1984_Petroleum

I don't know if this transformation is available in the version you're using.

When I okay that dialog and compare the Shannabank Hill, Fort poly/point against the imagery, I can see that they're overlaying within a few meters. If I use OSGB_1936_To_WGS_1984_1 (a not-as-accurate transformation), the offset is more, closer to 7-8 meters as opposed to 4m.

When you converted the shapefiles to WGS84, did you use the Project Tool or the Define Projection tool or the data's property page in ArcCatalog? The latter two update the coordinate system information only, they don't change the coordinate values to the 'new' coordinate system.

Melita

[chrischappell]

I almost want to start a new thread since my issue is with attribute inaccuracies or errors. Positionally, my data is great bringing it into ArcGIS.com. But on this point layer, when I classify and change the symbols, the attributes seem to swap with other points in the feature. For all points. (see attached image for before / after classification and symbol change.)

This has become very frustrating as it's a pretty simple feature with only two classifications. But once I go there the data gets all weird. (specifically notice the address and Parcel_ID. As many of the attributes are the same for each point.)


[ATTACH=CONFIG]0[/ATTACH]

[MikeMinami]

ckchappell,

Can you share the webmap with the issue publicly and post a link to it here?
What web browser and version are you using?

Thanks,

Mike

[chrischappell]

http://apps.arcgis.com/hosted/OnePane/basicviewer/index.html?appid=d193de1feaef42c6b80bf008ae1c3ba7

Tried in both Chrome and Firefox with the same results. The 'example' above is something I just put together to illustrate the issue. (I had to redo the original removing any classifications.) In this 'example' I have two layers "before classification" and "After Classification" which are the exact same shapefiles. But when identifying you can see how the attributes are different. (By clicking on panel / page 1 then 2)

[MikeMinami]

Chris Chappell,

Can we get the source data for this layer? It appears that the 28th feature is missing geometry.

Thanks,

Mike