tl;dr
A neural network given ~2000 years of temperature records concludes global warming not caused by man
Image created by moi using the Deep Dream Generator starting with this annual global mean temperature image from wiki commons
Let me get this out of the way first - I am not a climate change denier nor a conspiracy theorist. The overwhelming scientific consensus remains that global warming is a real phenomenon, caused by human activities. Even a team of scientists funded by noted climate change skeptic Charles Koch (aka Satan incarnate along with his brother David) reached this conclusion in 2012 in a comprehensive review of the scientific studies that had been conducted to date on global warming.
Given that there is really no downside to reducing the carbon footprint of industrial societies and weaning them off fossil fuels (except for possibly reducing corporate profits slightly - gasp, the horror!), I remain a proponent for stronger environmental controls and an international regulatory framework that has some real teeth.
With all that said, I still wanted to talk about a recent study published by John Abbot and Jennifer Marohasy in the December 2017 issue of GeoResJ which calls into question the established scientific consensus around global warming and suggests the increases in temperature we are seeing are natural. Unfortunately I could not locate a .pdf of the paper that is free to download, but if you want to pony up $30 you can download the full paper here. There are a few articles circulating now that summarize the paper - this is a decent place to start.
The basic idea here is that Abbot and Marohasy fed a neural network temperature readings from AD 50 to 2000, and the neural network chewed over this data and used it to predict temperatures through the 20th century. The study authors claim that the predictions from the neural network closely match observed actual temperatures without taking into account man-made CO2 emissions, and in fact are a better fit with an observed slow down in temperature increase which the standard General Circulation Model cannot explain.
I find this interesting on several levels. First I am fascinated with the uses people are putting neural networks to in general. From beating human Go masters to turning your favorite selfie into a surrealistic nightmare that Dali would be proud of, neural networks are hard at work around the globe and it is no exaggeration to say that their potential is pretty much limitless. We are just scratching the surface of what will be accomplished within our lifetimes by neural networks / AI.
Secondly, I think this is a good illustration of something that a lot of people lose track of - the "theory" part of a scientific theory. Even an extremely well researched and supported scientific theory that has the support of an overwhelming majority of the scientific community is still just that - a theory. It is likely that nearly every currently accepted scientific theory will be modified or completely overturned in the future. It is always good to question the assumptions that underpin scientific theories, and poke and prod at things which are accepted as truth. Just because observed temperature increases nearly identically match the predictions of the impact of human activity on the climate does not mean that there is a causal link. Correlation ≠ causation (necessarily). There is certainly a possibility that the earth was going to warm up this amount anyway and the assumptions made of the impact of increased CO2 in the atmosphere are faulty.
In the spirit of poking and prodding at things, I should also link to a blog post which takes this new study with a large grain of salt and more or less dismisses it entirely. It is interesting to note that the author of this post does not address the study's claims that the predictions of the neural network are actually a better fit with observed climate change than the currently accepted model.
I also want to link to a blog post by Jennifer Marohasy, one of the authors of the paper. A couple of interesting quotes from Marohasy:
In our new paper in GeoResJ, we not only use the latest techniques in big data to show that there would very likely have been significant warming to at least 1980 in the absence of industrialisation, we also calculate an Equilibrium Climate Sensitivity (ECS) of 0.6°C. This is the temperature increase expected from a doubling of carbon dioxide concentrations in the atmosphere. This is an order of magnitude less than estimates from General Circulation Models.
The science is far from settled. In reality, some of the data is ‘problematic’, the underlying physical mechanisms are complex and poorly understood, the literature voluminous, and new alternative techniques (such as our method using ANNs) can give very different answers to those derived from General Circulation Models and remodelled proxy-temperature series.
Marohasy notes that their neural network has used similar techniques to forecast rainfall in Australia and has achieved a greater degree of accuracy than the Australian Bureau of Meteorology's predictions based on the general circulation model.
I am curious what you feel about this - is this "junk science"? Is this a legitimate technique for analyzing climate change which should prompt at the very least a reexamination of the assumptions underpinning the general circulation model?
Cheers - Carl
~ The above post is 100% my own words with the exception of quotes attributed to another author. Copy pasta is a dish best served... never ~
Steemit Logo created by @rmp