Behaviour of GNSS signals near and under tree canopy is a complex subject, and subject to many effects, some of which might be called multipath. Multipath is, as the name suggests, an effect due to a common signal arriving at a receiver via multiple paths. These paths may be caused by different effects, but, at ground level, are due to features on the ground near the receiver, such as ground surface, tree trunks, tree canopy and other reflecting features such as walls or rock faces.
In addition to reflected signals, the signal can be significantly attenuated by the water content of leaves in the tree canopy, and scattered by branches.
The result is a very complex variation in the received signals, which can weaken the signal (simple attenuation), cause fading (rise and fall of the signal due to additive and destructive interference of in-phase and anti-phase reflections), and smearing of the signal due to canopy scatter (causing small changes in signal path length). The effect of shadowing and fading caused by these effects increases the noise in the solution, resulting in a higher error measurement, which the receiver ought to report).
The influence of large-scale reflectors can be such that they are seen as the dominant signal (especially if the direct line-of-sight path is blocked). These non-line-of-sight reflections can distort the navigation solution to a significant degree.
There's a useful discussion of multipath and non-line-of-sight signals at
Inside GNSS. For a more theoretical look at the effect of trees, have a look at
this paper.
So, yes, Hugh; it's possible that the effect you saw was due to fading, shadowing and reflections caused by the forest environment. However, if you had just switched your receiver on, it could also have been due to a poor fix from a small number of satellites, as the receiver started to acquire them from a cold or warm start. Given the fact that the error resolved itself quite soon, I'm minded to suggest that it was the latter effect, and not multipath after all...
Coincidentally, I have recently been doing some trials of a GNSS receiver under tree cover, and, yes, it's very interesting to actually see the effects at work. One of the hardest things for these trials, though, is to create a ground truth with which to compare the GNSS results. GoogleEarth imagery is rather handy...
> I was aware of the possibility of satnav error caused by signal multipath but there is nothing like experiencing it to ram the message home!
And I'm one who keeps trying to ram home the message that GNSS may be wrong... Maybe I need to paraphrase the
Quaker Advice No 17: 'Think it possible that your GPS may be mistaken...'
