Thursday, 18 June 2015

If You F'ing Love Science, Try It

One of the most nauseating threads that keeps popping up on Facebook is "I F--king Love Science" on which typically eternal studenty types with Richard Dawkins delusions share snippets of pop-science and militant atheism, presumably trying to look intelligent. 

This often seems to completely miss the point of scientific endeavour, which if it's about anything at all must surely be about critical, constructive scepticism. You don't just believe something because "scientists" say so. Let's just take this story, shared on Facebook with a picture of planet earth in flames, and the headline that the first 5 months of 2015 are the "hottest on record."

The author does have the decency to link to the NASA data showing this apparent trend, but the rest of the article is sensationalist guff about melting roads, and only a vague mention of a Washington Post article which itself makes only a passing reference to carbon emissions as a possible cause of this, tagged on the end almost as an afterthought.

However the NASA data linked in the article is still not quite the whole story. Despite looking serious, the devil is in the detail. The top of the page has two notes - firstly that the base period is 1951 to 1980, and secondly that the records have been adjusted for "the elimination of outliers and homogeneity adjustment."

There's quite a good explanation here of why outliers are excluded, but it's not entirely clear that they should be in this case in this instance. While they probably wouldn't affect the trend, what they do show is that large variations are much more common than the adjusted data suggests. 

The homogeneity adjustment is a far more complex area, there's an explanation of it here. It takes a bit of brain-work but if you f'ing love science then it should be manageable. This is basically a very complex method of averaging 'urban' and 'rural' stations. It appears to average out recorded temperatures in a given area to account for localised heat from urban areas. This is an accepted statistical method, and it is clear why they have done it this time as there are far more urban areas generating a lot more heat today than there was when the data started in 1880. However it's not clear that this is an especially effective way of achieving this.

A bit of digging around brings you to the "long data" - that is to say the data as it was recorded without statistical manipulation. What it shows is that "outliers" are fairly common and this years deviation of 0.7°C is not especially extreme. January 1893 for instance was a full 1.44°C cooler than the this fairly arbitrary average, while February 1998 was nearly 1.2°C warmer. These could reflect inaccurate measurements, or they could simply reflect a temperature variation. Even 1.44°C, the biggest variation in the data, is hardly unheard of.

What this data set also lacks is context. 135 years is a long time by human standards, there have been huge advances in our ability to accurately measure temperature, and a far broader ranger of places we can do it. It wasn't until after world war 2 that there was any serious research presence in Antarctica. So while the data shows a very broad trend, trying to read anything into a deviation of 0.7°C seems a bit optimistic. In the context of something as multifaceted and complex as the climate of our 4.5 billion year old planet this is is a hopelessly small period of time over which to draw any meaningful conclusions about data.

None of that though makes a very shareable meme for the Facebook pages of those who f'ing love science. A sensationalist article with a clear narrative and a smug sense of intellectual righteousness does. It served to brilliantly highlight the relevance of this excellent piece by Johannes Bohannon, who concocted a cleverly designed study to show that eating chocolate could aid weight loss. Of course it is complete nonsense, but they did indeed set up a trial, use some statistical sleights of hand, a worthy sounding (though unrelated) PhD and a headline grabbing narrative. The article was eagerly accepted by scientific journals, and soon enough appeared as news on health blogs and in newspapers around the world. 

If you do love science, try it before swallowing wholesale any attention grabbing meme that pops up on your Facebook page. There does appear to be warming trend in the NASA data, but to take this, or an apparent scientific consensus as some sort of proof that man made global warming is an undeniable fact is not a love of science, which depends on scepticism and open enquiry. It is a love of bandwagons and sensationalist claptrap, driven by a desire to be seen as intelligent without putting in any of the effort required to look beyond a 3rd hand article designed to grab attention.