[OSM-dev] gpx elevation data for given bounding box?

[Andy Allan]

On Sun, Sep 14, 2008 at 8:46 AM, bvh <bvh-osm at irule.be> wrote:
> On Sat, Sep 13, 2008 at 08:41:28PM +0100, Tom Hughes wrote:
>> > The reason seems to be that some people *think* they are not accurate.
>> I think the experiment we conducted at the 4th birthday party showed
>> pretty conclusively that it's not accurate. We were only about 12 foot
>> above sea level (as measured by looking over the wall at the Thames) but
>> the various GPS units present reported values from about 0 to about 200
>> foot or so. They also varied radically over time despite the fact that
>> we were sat still.
>
> I think that is quite accurate actually. How many devices participated?
> What was the error distribution like?

>From memory it was about 8-10 devices, probably 2 or 3 from each
category: Garmin H series, Garmin non-H, others. Identical chipsets
were varying when compared with one another over the full range of
error. It was pretty bad really, but a great reaffirmation of why we
use other sources (e.g. SRTM) for height data and not GPS.

Would be interested to see how many devices you'd need over what time
span to get certainty close to 1m vertical, but that'll need a
statistion to help. I'm not even sure if 20 devices over 2 hours would
even vary usefully close - I suspect the standard deviation would
still be very large.

Cheers,
Andy