Earth temperature


The average surface temperature of our planet over the ~540 My of the Phanerozoic (since the first proliferation of complex life forms) is a fascinating topic, with some immediate relevance to what we face in the decades and centuries ahead. One of the more startling achievements of the last 30 years of climate science is the production of a large set of actual measurements of temperature history (from physical proxies), replacing much of the earlier geological induction (i.e. informed guesses).  A failing is that few – with the notable exception of James Hansen – have attempted a coherent integration of these diverse findings.

A decade ago I made this ugly thing at Wikipedia, by cobbling together a stack of Rob Rohde palaeotemperature graphs, pixel by pixel:

My purpose was at least partly to try to prompt Rob, then an active WP editor, to have a go at the problem himself … given that he obviously had much more graphical skill and climate understanding than me. Dr Rohde has been busy with other things (a PhD, then a pivotal role at the Berkeley Earth project), so he never got there.  As a result, my ugly and not very satisfactory effort has become uncomfortably widely referenced and used.

This week, Gavin Schmidt at Real Climate (and NASA Goddard) challenged readers there to help produce “some coherent, properly referenced, open-source, scalable graphics of global temperature history” to “out-compete the myriad inaccurate and misleading pictures that continually do the rounds on social media”. He might have been thinking of mine…

It’s clearly time to attempt an update. My first pass is above. There’s been plenty more published on the subject of palaeotemperatures in the interim (see Gavin’s post), which I’ve made a start at trying to include.


1.  The first and most glaring issue is the data miss-fit at the K-T boundary (~65 My). It was there in the original too, but neatly hidden by my rescaling and Rob Rohde’s multiple temperature scales.  It appears to arise because the Royer et al data represents shallow tropical sea temperature, which varies much less than the global average surface temperature (as reconstructed by Hansen / Zachos).  For now, I have factored Royer et al by 2.0 to obtain approximate agreement.

2.  The NGRIP data used is δ18O (oxygen isotope measurements) from the north Greenland ice core. Converting δ18O to temperature is a complex business. I’ve used a very old method from Johnsen et al (1989). The plotted valued is halved to approximately adjust for polar amplification.


Data sources are detailed at the Wikipedia page:

The source spreadsheet is available here:

1 comment to Earth temperature