Wednesday, August 17, 2016

Climate Change Interlude - On Science.

We are about to look at the various theories and opinions that proliferate in the debate on the origins of recent global warming. It is very easy to get lost in this. You read one article and think –yes, that looks convincing. Then you read an article saying the opposite and think – well yes they make some good points too. So what are we to think?

The debate is a scientific one, so first we should define what we mean by science. What is a scientific theory? how is a scientific theory evaluated?

To start with the fount of all modern knowledge, Wikipedia [1], “A scientific theory is a well-substantiated explanation of some aspect of the natural world that is acquired through the scientific method and repeatedly tested and confirmed, preferably using a written, pre-defined, protocol of observations and experiments.” Furthermore, “Scientific theories are usually testable and make falsifiable predictions”, and “Scientists use theories as a foundation to gain further scientific knowledge, as well as to accomplish goals such as inventing technology or curing disease.”

There are a couple of distinct strands here. Firstly, a theory must agree with lots of evidence acquired through experimentation, and it must make testable and falsifiable predictions. This is the classical definition most of us recognise from our school or university days. But there is a second element that we can use the theory to do or make something. In short, a theory should be useful. Note that it doesn’t have to be “correct” or “true”. It just has to explain enough of what we see consistently enough, and we have to be able to make predictions that are sufficiently reliable for us to be able to control the environment around us with confidence.

It is worth looking at another debate to help illuminate our thoughts on scientific theories. The validity of the Theory of Evolution is still debated in some parts of the world with Creationism being a preferred alternative. In the definition of a theory as given above the objection to Creationism is clear. It isn’t that Creationism is “wrong” – after all, if there is a higher all-powerful being that creates everything we see according to their higher plan, then who are we to refute that. The objection is that Creationism is not a scientific theory. It doesn’t allow or support enquiry on why we see the things we see; the things we discover are here because God put them here. It doesn’t allow us to speculate about the future; who are we to debate God’s plans for the world and its inhabitants? There are no experiments that can be performed, no predictions that can be tested. Ultimately the problem with Creationism is that it doesn't meet the criteria for a scientific theory. It is a religious theory, not a scientific one.

Computer models, fitting data, and estimation.

Science is very good at understanding systems of a single variable or a couple, but once you get a number of variables that are mutually dependent, things get very hard. Typically work progresses by fitting data to historic data and then using the best-fit parameters to extrapolate into the future. One of the things to watch out for is “over-fitting” which is that the more parameters you add the better the fit to historical data but often the fit to future data gets worse as your parameters are fitting to random elements of your data set. Parsimonious models with few parameters and simple relationships are generally preferred. If you try to extrapolate what happens into an area not covered by your data set (e.g. high CO2 levels) then your models may no longer work as new and un-predicted 

Another point about models is that fitting models to continuous distributions is problematic enough but fitting to distributions of binary events is horrendous; you need enormous amounts of data before you can make any solid statements. For instance, if your theory predicts changes in events of the order of a small number per century, you need several centuries of data before you can begin to get any certainty you are correct.

And so to climate change. There are two main problems that appear when we apply the expectations of scientific theories to climate.

Firstly, while the physics may be simple, the way the climate responds is very complicated; there are lots of variables, they differ for many parts of the world, and they are inter-dependent. Inevitably scientists end up fitting models to historic data. Scientists have become adept at developing theories that explain historical phenomena (see a previous post about the Hiatus), they are less good at predicting what is going to happen in the near-future both for the reasons given above and because we are going from a period of known climate into a future where some key variables have changed, and some other relationships may change in nature.

Secondly, we cannot easily test theories. We don’t have a spare earth we can treat differently. We have to take what nature has provided and try and draw scientific conclusions from that. This means that the proven ability of scientific climate theories to predict is much lower than for many areas of physics, for instance. Climate science is bit like doing particle physics but without a collider, instead restricted to observing particle interactions that take place naturally.

In summary, we are engaged in the business of science. We need to see theories explaining past measurements and predicting future ones. Experimentation is difficult – we need to take what nature has given us with no scope for independent assessment. Finally, theories that predict extreme events requires loads of data to have any scientific validity which may take us beyond the timescale required for action.

Next, a look at the Holocene; from the last ice-age to the present day.


No comments: