There, but rarely seen.

I undertook some more WSPR work at 28MHz yesterday.  Prompted by the not-unexpected finding that Es-mediated signals are extremely variable over short time periods, I thought I'd look at data that is rarely examined: frequency drift.

Why?

Well, it's obvious that the cause of the vast signal strength changes over short timescales is due to the irregular nature of the Es 'clouds', and that there is also organsed wave structure within them.  Also, the Es clouds move with a high velocity - perhaps from a few tens up to several hundred metres per second, where a Doppler shift would therefore be strongly expected.

To visualise the complexity involved, take a quick look at this fine video of noctilucent clouds, which are closely related to Es, at least in terms of structure and motion:

For this simple experiment, I chose TF3GHZ, as that station has an underlying zero frequency drift in TX.

I plot the results of the signal received (blue), together with the signal drift (orange), to which 10 has been added in order to bring the line away from the horizontal scale (i.e. where you see the orange line = +10, it was actually 0dB drift).  Blue boxes indicate periods when no signal received here:

TF3GHZ 1W 28MHz WSPR received by MW1CFN 2020 May 31-June 01.

With Iceland slightly to my NW, we would, simplistically, expect the signal to be coming from Es with a resultant motion (mindful I am looking 'broadside' to the motion, not in line with it), that yields a negative Doppler shift (i.e. a slight drop in frequency as the reflecting surface recedes).  And, indeed, that is what is seen: up to minus 4Hz shift.

Sticking 4Hz into the Doppler equations yields an effective 'broadside' motion of, very roughly, 40m per second, with the proper motion significantly higher than that.

Obviously, one would need to look at a large number of stations to see if this result is just fortuitous, or is repeated consistently.  Even then, motion in Es is, as the video shows, non-trivial, and so I expect it will be a very messy picture.

One obvious question to ask is whether the expected Doppler shifts are seen for receivers with transmitters to their west and east.  It can be done, but it needs more time than I have.  It also needs a pretty sophisticated statistical analysis to drag the real effect from what is a very messy picture, given we need to identify, for each TX-RX set-up, a zero underlying drift - itself a tedious exercise - as just a starting point!