[OSM-talk] Coastlines (was: Question about automatic data upload)

[Collinson Mike]

At 06:56 AM 13/09/2006, OJW wrote:
>(4) There was a big debate about shifting data a few metres in various
>directions - I thought some people were still of the opinion that we
>shouldn't upload an area until there are local OSM'ers doing clifftop+beach
>walks to verify the data (and I never did find out if it was acceptable to
>upload data without these corrections, nor get any theories on what
>correction might be valid over large areas)

I'm still doing some work on this.  Here's some preliminary data:

1) I can't see any bias towards land or sea in the PGS data, so if 
the coast is facing the east, west, north or south the shift appears 
to be the same in local areas.

2) I picked four sites up to 60 km apart around Sydney and digitised 
a few obvious landmarks from government 1:25,000 topographic maps and 
overlaid the data. Result: No significant systematic change in shift 
between sites (the PGS data always seemed be shifted 54 - 60 metres 
to the north as per GPS results).

3) I then picked another site another 240 km to the north and the 
result was entirely different.  The data overlies perfectly.  I 
bought the map so I did a whole stretch of coastline and 
embayments.  The area lies on the same publicly available Landsat image.

So the question I now want to answer is: does the change in shift 
abruptly alter somewhere along the coast or is it gradual.  The 
entire New South Wales topographic map collection is coming out on 
DVD in 2 weeks for 99 Aussie dollars so I'm going to wait for that, 
it will make life way easier, [eat your heart out UK Ordnance Survey 
users! :-) ]

Meanwhile, I've gone ahead and uploaded sections of uncorrected 
Australian coastline but been careful to leave in the same 
created_by=almien_coastlines tag so that it should be easy to write a 
script to pull them back out and apply an offset.

Mike
Oz