Has the HF spectrum become noisier over the decades due to man-made interference? Is this a reason why HF band activity appears to be declining?
HF propagation is not only limited by manmade interference but also by natural phenomena like sunstorms.
Anyway, I'd account that to social and technological reasons:
- HF activity suffers very much from the old men's club's image phenomenon (a term that I just invented!): How attractive is it to join a communication group that consists of old men?
- Old men get older and die. That reduces the pool of people able to use HF.
- Ham radio generally is losing a bit of appeal because, well, it's lost all specialness to be able to talk someone at the other end of the globe, because that's what we do all day using the internet. Thus, flaky speech or Morse communication itself has not much benefit.
- good HF antennas tend to be large. Compare to the overall trend of migration to metropolitan areas.
- Rent is pretty much going up everywhere. Having a garden with a shed increases in price!
- HF comms do bear attractiveness to engineer types due to the fact that there's much to be technically improved (compare: WSPR, FT8), but that's usually not what people have in mind when they say "HF band activity".
So, yeah, doing long-range bad-quality speech communication with old men simply isn't that attractive these days.
I find you're in the wrong with your observation of declining activity, however. Last time I talked to larger groups of hams (which was in Friedrichshafen at HAMRADIO'18), they were all totally astonished by the count of contacts made in the last months – all due to FT8. FT8 offers a new exciting technological aspect to ham radio, and hence is used far and wide, with hundreds of thousands of contacts¹.
So, if you ask me, this is all a self-made problem of how ham education is done: Instead of focusing on modern technology, people are thought CW, and proper voice comms conduct, and outdated tech. That does a tremendous job at keeping the hobby's reality compatible to techniques and technology of the 1940s to 50s, but guess what, people's interests moved on with the technology that around that time left ham radio in the dust.
¹ Technologically, FT8 is not even that exciting, and it's quite obvious that it's cleverer than the classical digital modes and by far far far cleverer than trying to do analog modes² over low-SNR links. It's early 1980's state-of-the art communications, with computer frontends, at best. The only cool thing is the channel code, an LDPC, but older channel codes could have achieved similar performance, albeit at the cost of higher computational complexity (not that it matters much, at these symbol rates)
² the discussion whether analog or digital modes offer better communication quality for speech has been pretty much scientifically answered in the mid-80s. Digital wins. There's no good reason people are still doing AM or FM communications for voice other than cost in the case of least-end handsets (albeit even the cheapest Baofeng handhelds for 20€ do DSP for audio, so even that doesn't count anymore) and backwards compatibility with the 1940s.
It is a measurable fact that there is more electronic noise across the radio spectrum now than ever before.
However, I doubt that it has directly had much effect on the amount of amateur radio activity on the HF bands. There are many other factors (see Marcus' answer for some of those) that contribute to the decline of activity on any given band or group of bands.
Finally, to address the question in the title: We have been at an extreme solar minimum for several years now, the likes of which has very likely never been seen since the discovery of radio. This is (in my opinion) "what has happened to HF" recently. There have been extended periods of no sunspot activity on the sun at all, and this all contributes to making HF conditions miserable.
Before looking for reasons why HF activity "appears to be declining", we should look for evidence that HF activity is declining — so let's look at some available sources of data. While contests may not be fully representative of the broader HF community, they do involve a lot of people, and they provide a ready source of reliable data that goes further back in time than things like lotw/eqsl/pskreporter. So let's see.
CQ WW DX CW:
- 1987 - 2,602 entrants, #1 had 12,387 QSOs, #100 had 2,281 QSOs
- 1997 - 3,063 entrants, #1 had 14,770 QSOs, #100 had 3,561 QSOs
- 2007 - 4,705 entrants, #1 had 14,642 QSOs, #100 had 4,285 QSOs
- 2017 - 7,940 entrants, #1 had 12,892 QSOs, #100 had 4,850 QSOs
CQ WW DX SSB:
- 1987 - 2,615 entrants, #2 had 15,602 QSOs*, #100 had 2,882 QSOs
- 1997 - 3,354 entrants, #1 had 14,927 QSOs, #100 had 3,514 QSOs
- 2007 - 4,873 entrants, #1 had 12,637 QSOs, #100 had 4,280 QSOs
- 2017 - 8,124 entrants, #1 had 13,872 QSOs, #100 had 4,344 QSOs
*: The #1 entrant in 1987 is listed as having 32,767 QSOs, which is a supicious number, being 2^15-1, and probably a data error.
ARRL International DX Contest (CW):
- 2007 - 2,569 entrants, winner had 5,326 QSOs
- 2017 - 3,837 entrants, winner had 6,788 QSOs
What I see in this survey of HF contest results is a mix — the top performers are flat to maybe slightly down, but the number of competitors submitting logs has shown a sharp increase, resulting in many more operators/teams getting within striking distance of the winners (as reflected in the consistent increase in the 100th-place QSO totals for CQWW). What I don't see is any evidence that band performance is so bad as to severely hamper the chances of making a contact, or that the pool of HF operators (even code-qualified ones!) available to enter contests is shrinking.
Man-made noise has some detriment to HF communications. ITU-R P.372-13 illustrates:
Note how the man-made noise at a quiet receiving site (C) is above the minimum atmospheric noise (B) from 70 kHz to 5 MHz. The full document has more figures that characterize noise specifically by time of day, season, and geography. But the gist is that at the quietest sites in this band (which includes the amateur 80 and 60 meter allocations), about half the time the noise floor is dominated by man-made noise.
At higher frequencies the man-made noise drops below natural sources (galactic noise, D) at quiet locations, but the man-made noise in cities (E) is on the order of 20 dB higher in cities, even worse indoors or in densely populated areas. That's well above natural sources. This graph ends at 10 MHz, but the trend continues with city man-made noise being about 20 dB higher than natural sources for the rest of the HF band.
It's hard to say if this is "a reason why HF band activity appears to be declining". The question has been asked before, though the poster of that question asserts the problem is not man-made noise. The trouble is in the word "appears". A lot of things contribute to appearances, quite likely many of them social as Marcus Müller explains.