New climate change datasets: boring but useful results. (updated)

The Berkeley Earth Surface Temperature project has just released its first data sets and analyses.  The aim of this project is to summarise all the temperature records around the world in a comprehensive and transparent way, both to get an estimate of changes in global temperature and to make it easy to see the impact of  data quality filters that have been applied by previous climate modelling projects.  So far, they have analysed all the land measurements; ocean measurements are coming next. You can download the data and code yourself, and check their findings or explore further (if you actually want 1.6 billion temperature measurements — don’t try this on a smartphone).

The main finding of the project won’t surprise most people: it’s getting hotter. More importantly, the estimates based on all available data agree almost perfectly with the previous estimates that were based on a small subset of ‘best’ weather stations.  Incorporating lower-quality stations doesn’t change the estimates. Even using just the low quality stations gives pretty much the same estimates. Other things that don’t affect the results include the urban heat island effect: cities are hotter than the countryside, but most of the world isn’t in a city. They’ve also made a neat movie of the climate since 1800: you can see the normal oscillations over time, and the heating trend that eventually swamps them.

Combining temperature records with varying quality, measurement frequency, and duration, is a major statistical task even without considering the volume of data involved. The statistical expertise on the Berkeley Earth team includes an Auckland Stats graduate (and Berkeley Stats PhD), Charlotte Wickham.

Updated to add: this is now in the Kiwi media: NZ Herald, Stuff, 3 News.