Philosophy and Climate Science
by Eric Winsberg
Cambridge University Press, 2018
A journalist once wrote that anthropogenesis (human-caused climate warming) is as “obvious as 1+1=2”.
Eric Winsberg, in his book Philosophy and Climate Science, points out that this trivialises the issue and encourages laypeople (including climate sceptics writing for newspapers) to think they can evaluate the arguments themselves and find simple flaws in them.
Winsberg’s book looks at how the scientific arguments about climate change are constructed, and investigates four issues:
- climate data, and its independence from models
- climate models and climate simulations
- confidence and uncertainty
- verification of climate models
The relationship between data and models is nicely illustrated by the mismatch, reported in the year 2000, between surface temperature and the troposphere (the bottom 20km of the atmosphere). Climate models predicted the troposphere should be warming, but satellite data suggested it was cooling. Was the data wrong or the climate model?
Satellites orbit at a height of 35,000km – a long way from the troposphere – and they are used to measure the microwave radiation emitted from oxygen models in the atmosphere. From this data, scientists found a way to infer troposphere temperature, assuming, among other things, the correct amount of interference from the higher layers of the atmosphere. After half-a-dozen years of academic wrangling, all the scientists involved in the controversy agreed that the inference from the raw data (microwave measurements) to the temperature of the troposphere was flawed (partly because the satellites had drifted out of their initial orbits).
In fact, the data matched the climate model.
The point is that raw data must be modelled to be turned into climate data. This is true of microwaves measured by satellites, temperature measurements on land (how do we ensure the set of measurements truly reflects the average surface temperature of the earth?), ocean data, ice cores and all kinds of proxy data for climate variables.
Winsberg’s discussion of climate models and climate simulations starts with a very simple model of the earth’s climate. The model starts from the amount of energy reaching the earth from the sun, and subtracts from it the amount of “black-body” radiation given off from the earth (or any other hot body, as described by the “Stefan-Boltzmann” law). The temperature of the earth must be such that the energy radiating in from the sun must exactly balance that given off from the earth. From this, the temperature of the earth can be inferred to be -18°C ☹. The reason our temperature is actually about 15°C is because solar radiation is very short wave and goes straight through the atmosphere, while Stefan-Boltzmann radiation is long wave and tends to get absorbed by the atmosphere (and then re-emitted in different directions).
Winsberg terms such simple models as “mediating” models, which enable us to understand what factors need to be taken into account when developing models for specific purposes. In this case the mediating model reveals the importance of the greenhouse gas effect, due to ozone, water vapour, methane, pollutants and of course CO2, on the temperature of the earth. Based on an 1895 experimental result about CO2 absorption, we can predict the effect of doubling the concentration of CO2 in the atmosphere, assuming everything else stays the same. The result would be around a 1°C increase in temperature. However, as the world warms, this changes the impact of water vapour (even assuming relative humidity stays constant), which doubles the climate sensitivity.
The challenge is to understand all the changes that would occur in attempting to reach a climate equilibrium (a necessary upper limit of 6°C increase, based on historical records, is uncontroversial), and what happens “dynamically” on the way there. If we wish to take action to reduce climate change, the dynamic aspects are critical.
A climate simulation takes a climate model and runs it from an initial state over a simulated time period. Given models of atmospheric CO2, water vapour etc., a simulation divides the earth up into discrete cells (for example each about 100km by 100km) and uses model equations to infer the exchange of radiation, heat, moisture and air (mass and momentum) between neighbouring cells. As models have become more sophisticated they have been able to take more and more into account (starting with a model one cell deep of the atmosphere, then adding a model of ocean temperatures, ice-melt, trees and land-use, volcanoes, sunspots etc.).
With this basic understanding of models and simulation, Winsberg goes on to examine probability, and confidence. A single model will yield different outcomes based on the initial state and an element of unpredictability within the model itself, and thus a model produces different possible outcomes with different probabilities. However, the model itself may be right or wrong, or nearer or further from the “truth”: thus we also have a degree of confidence in the model itself. While probabilities can be quantified objectively, it is much harder to achieve an objectively supported level of confidence in a model.
To improve confidence, climate modellers often run a portfolio of multiple models. The extent to which they agree can give us confidence – though if all the models share the same underlying mediating models their agreement may result from the same flaw in the underlying model ☹.
Thus Winsberg completes the book with a discussion about verification and validation of climate models. One type of verification is to run a model against the past – with an initial state representing the year 1990, for example, and running it up to 2010 to see whether the predictions of the model accord with reality. Nevertheless, as any philosopher will tell you, just because something worked in the past doesn’t guarantee it will still work the same way in the future. Indeed as each year passes and the world warms, we are entering new states for which we have nothing to compare them with – or at least no past state on the earth for which we have enough data to make useful inferences.
Winsberg’s book is eye-opening in revealing the depth of the discussions which are carried on continually between climate scientists. It makes the important point that simple reasoning cannot enable non-experts (like me) to prove anthropogenic climate change, or that the earth’s temperature will increase by a certain amount given certain political decisions.
However, it also makes the point that committed, expert scientists have been measuring, modelling and testing theories about the climate, based on natural science, probability and statistics, and a huge amount of hard grunt work to get the right data in the right form, for many years. One thing has become overwhelmingly clear: we can and must act soon to avoid anthropogenic climate change shifting the earth into a new state of equilibrium which could no longer support today’s human society.
Winsberg’s Philosophy and Climate Science was published by Cambridge University Press in 2018