,
8 tweets,
3 min read
Read on Twitter
This is a big deal actually.
Getting a full assessment of all the sources of bias in global temperature products in a quantitative manner is hard. (Though it’s relatively easy to show multiple methods give similar answers). 
There are issues associated with coverage changes over time, homogenization choices in land and ocean data, interpolation and averaging issues, etc.
Estimating the total uncertainty is hard. But ultimately rewarding:
Estimating the total uncertainty is hard. But ultimately rewarding:
Other groups (HadCRUT4 and @BerkeleyEarth) have done similar analyses and while they have approached this similarly, the answers are slightly different.
These updated uncertainties allow us to be more quantitative about the annual uncertainties and the ranking of individual years. No surprise that 2016 is by far the likeliest ‘warmest year’ in the record.
Some of you might remember the results from 2014. Notably the #FailonSunday accused me of ‘admitting to a margin of error’ in the analysis (obviously it was a slow news day!). At the time, we’d stated that the likelihood of 2014 being the top year was ~38% (ahead of 2010).
But with the new error analysis, it turns out we were conservative! We now rank 2014 at 79% likely to have been the warmest at the time (though it was soon eclipsed by 2015 and the 2016).
But the biggest result is the one you knew already. Warming of the global climate system is unequivocal and far larger than the uncertainties in the global anomalies.
