[Openstreetmap-dev] DOP and errors in meter

Preben Mikael Bohn preben at chin.dk
Tue Dec 27 09:35:51 GMT 2005

> I've never really 100% understood how DOP
> correlates to errors in meter, but....

In _theory_ the DOP is basically a factor you multiply to your error on
_each_ range measurement (to the satellite) in order to get your error. As
the DOP value is dependent on what error we wish to compute (3D, 2D,
horisontal, vertical, time, etc) you need to specify which error it is
computed for (usually specified by HDOP, GDOP, TDOP, etc).

The problem however is that the error on each range measurement is unknown
and not even Gaussian distributed or white; thus the DOP valu may be used
as a very rough indicator of satellite geometry problems (high value =
poor satellite geometry) but can not really for anything else.

The error estimates given by GPS receivers can be the DOP multiplied by an
estimated error on each range or it can be something completely different.
Values from two different receivers at the same point can not really be
compared without knowing the difference. In any case the error estimates
are just estimates and should thus be used in the same careful way as the
DOP values.

> 1. convert all dop values to meters
> 2. store both dop values from GPS and err. estimate in meters.
> 3. just use dop nothing else

2 is obviously the best choice. 1 is only possible if you know the
DOP-type and you can estimate the error on each range measurement. If I
were to choose between 1 and 3 I would select 3. :)

>   The error estimation is something we should make greater use of, a
> track point with error of 600m is not useful when the editor window is
> showing a image that is 100x100 meters. And likewise a Node created in
> an editor which has an image resolution of 100meters per pixel should
> be marked as such.

Agreed :)

Best regards Preben

PS. Would it be possible to setup the mailing list so that when you click
"Reply" the mail-client replies to the list, not the sender of the mail?

More information about the dev mailing list