Of course, these surveys have limited validity. A particular problem is that those who reply are self-chosen. In general, survey participants are probably more knowledgeable than many users of free desktop environments.
Still, the figures' general consistency suggests that they do have a rough validity, especially if you don't get down to decimal points. They suggest that either KDE remains relatively healthy or else, if it is in trouble, then the trend is too early to reliably detect.
Moreover, any decline may not be significant. KDE has reached lower levels of popularity in the past and recovered. Moreover, figures from KDE system administrators indicate that, at the same time that KDE's user figures were dropping in 2011, the monthly number of new KDE developers had reached pre-KDE 4 levels. If KDE really is less popular than before, apparently the project is in small danger of becoming less active.
So why have people been so quick to extrapolate from the change in Kubuntu's status to KDE's decline?
Perhaps people are believing what they want to believe. Yet I think the answer needs to go deeper than that.
To start with, Kubuntu's name recognition apparently deceives onlookers into thinking that it is more important to KDE that it actually is.
For another thing, KDE's popularity seems greater in Europe than North America. KDE e.V, the project's governing body, is based in Europe, and, increasingly, it seems, so are many of its larger projects. North Americans may underestimate KDE's popularity simply because, for them, GNOME has a larger local presence.
But the main reason is probably that, ever since the creation of the GNOME Foundation in 2000, GNOME has pursued ties with corporations much more aggressively than KDE has until very recently. The result of GNOME policy's is that, as Wallen points out, the major distributions install with GNOME as the default desktop environment.
The trouble is, the default desktop is a poor metric for success. All distributions ship with a choice of desktop via their repositories, and, increasingly, installation disks are available for both GNOME and KDE for distributions such as Linux Mint and openSUSE.
Even on a distribution like Fedora, which defaults to GNOME, the KDE spin is consistently the most popular variation, with over twice as many downloads as the next-popular Xfce spin. In many distributions, the main difference in status is that the default GNOME disk images are available a few days to a few weeks before the KDE images, and that might be due to a shortage of developers as much as the demand for each desktop environment.
But, whatever the reasons, all indications are that KDE, with or without Kubuntu, is much the same as it has ever been. If anything, GNOME might actually be in far worse shape, although a single year's decline can hardly be called a trend.
By all indications, KDE is within its normal range of popularity over the last five years. Any decline is slight, and, so far, has not affected KDE's ranking among desktop environments. Next year might be a different story, but, until then, any funeral arrangements or mourning for KDE seem premature.