An illustration of the trough-to-peak trick often used to state a false increase in the “trend” of global temperatures.
Innocent sounding periods like “the latter half of 20th c. ” or “last 50 years” are used to give the impression that it is an arbitrarily chosen round number, in reality it is a choice which determines the result.
The title of the graph was a “finding” of recent paper [*] but, as is shown here, a simple cosine which has no long term “trend” can be used to produce the same result.
There is a roughly repetitive pattern that is a significant part of NH temperatures, with about this period and timing.
This shows how totally false conclusions can be drawn after fitting a “trend” to such a choice of the time periods.
Fitting a linear “trend” model to data from such a system, that is anything but linear in its behaviour, is invalid. Yet this practice is ubiquitous in climate science, even in peer-reviewed literature such as this:
[*] Compo et al 2013: