Wow - so much misinformation on a single page! (A few have it right, though).
So - to dispel various myths above:
* GPS operates in 3-space, not 2-space. [Actually, for purists who may disagree, GPS operates in 4-space, but I think I'll leave that out of here]. Some cheaper units without barometers are poorer at resolving the vertical component, but most are not. Some may only use X and Y for calculations, but those should probably be tossed on a heap right now. With any decent GPS, going downhill/uphill will not affect the trueness of the reported speed to the extent suggested (i.e. that it just ignores the Z component).
* Even many small handheld units will report velocity quite accurately, assuming they have a fairly clear view of the sky. With most units, if you have a clear view of much of the sky, it will detect a movement as small as a few centimeters, certainly a meter. Sure - there are some vagueries in the system that are unpredictable, and these can minorly affect precision. However, *in practice*, most GPS units with a clearish view of the sky are plenty precise enough to determine the difference between e.g. 26 and 30 (or even 27) mph averaged over just 1 or 2 seconds.
* As someone did mention (using different words), there is a difference between "precise" and "accurate". Even if the GPS accuracy is off by say 15 metres at the time this does not mean its precision is off by the same amount - it is quite likely to be very much less than that, and *precision* is what is important for speed determination. Accuracy is irrelevant to precise (and accurate) speed determination.
* Clouds hardly affect GPS at all. Trees can. Buildings can. Clouds - no.
* Hand-held speed guns - including laser and radar - do NOT measure instantaneous speed in a markedly different way to GPS, as someone suggested. Instantaneous speed is all but a mathematical on-paper concept. All of them - including those using the Doppler effect - rely in concept on *sampling*; i.e. taking more than one measurement over a period of time. With Doppler, you still have to measure the frequency or wavelength of the returning wave, and you cannot do this instantaneously, only by looking at the wave for at least one wave cycle. This may seem like splitting hairs, but rather I am trying to point out that laser and radar similarly rely on high-precision timing systems that are also subject to inaccuracy; i.e. they are not perfect just because they are laser, etc.
* Someone suggests the unit dropping from visibility of 4 to 3 satellites causes a large loss of precision. This is correct; however - most units will be typically seeing around 10 satellites most of the time (once "warmed up"), constantly dropping some and re-using others as location changes. The loss of precision dropping from 10 to 9 or even 8 satellites is negligible compared to going from 4 to 3.
* The intentional random inaccuracy for non-military applications (SA = Selective Availability) was removed as one of Clinton's last-minute decrees before leaving office. It does not apply anymore. The US have said it will never apply again; whether that turns out to be true is moot.
* GPS signals can be used to locate the receiver precisely-enough for these purposes at worst every 1 second, not the 6/30/5/other guesses above. GPS satellites transmit various signals; the one used primarily once other data has been established using other signals (the "warmup" period) transmits every millisecond.