[OSM-talk] The long tail - lowest common denominator

[Nick Black]

On 7/9/06, Lars Aronsson <lars at aronsson.se> wrote:
> Nick Black wrote:
>
> > I'm not sure that I am  following you Lars - do you mean that we could
> > use automatic feature detection to detect the edges of water features?
>
> Yes, that's what I was philosophizing about.  But I haven't tried
> it or even planned to try it.  I'm far too lazy for that.

One of the most significant features of the SRTM as a global dataset
is that it is self-consistent and internal errors fall within
specified and documneted thresholds.  This was acheived by a lot of
post-processing work.  Part of the work involved created a global
water bodies mask at the same scale (1 arc-second/3 arc-sec) as the
complete topographic dataset.

The mask is called the SWBD.  It is supplied in ESRI shapefile format
and is downloadable from the same ftp site as the SRTM data:

ftp://e0srp01u.ecs.nasa.gov/srtm/version2/

So thats cool - Free 3-arc-sec water body data.  Well, its only as
cool as the one degree tiles that it is supplied in.  If someone could
process this data to derive a complete water-bodies mask it would be
invaluable to OSM and to other projects.  A derived version of the UK
SWBD is already up on Chris's OSM freemap site, where is really shows
up the VMAP data (shows how coarse it is).

Nick

>
> The Landsat images are not perfect, of course, but they do show
> rivers and lakes in some greater detail than the VMap0 data.
> Except where white clouds hide every detail (such as over
> Trondheim), and except where some other dark features appear (such
> as railroads), the dark blue is water and the rest is land.  Just
> filter out the dark blue, and you have a rough estimate of where
> the water is.  Combine this rough estimate with the one given by
> VMap0, and you should be able to arrive at something better.
>
>
> --
>  Lars Aronsson (lars at aronsson.se)
>  Aronsson Datateknik - http://aronsson.se
>
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
>