by Alberto Zaragoza Comendador
How sensitive is the Earth’s climate to greenhouse gases? Speaking about carbon dioxide in particular, how much would air temperatures increase if we doubled atmospheric concentrations of said gas?
This question lies at the heart of climate science. It is to climate what GDP is to economics – the central concept. So central that it’s very difficult to have a coherent discussion of climate issues if one does not know about sensitivity. But there is a crucial difference between these measures: the layman is somewhat familiar with GDP but not at all familiar with climate sensitivity.
Okay, many or most people cannot tell you exactly what GDP is. But many others will give you a crude, approximately correct definition. Furthermore, even those who cannot define it intuitively get the implications of both slow and fast GDP growth, and could tell you at least its order of magnitude (i.e. it happens at rates of 1% or 3% a year; not 0.1% or 30%).
As for climate sensitivity, I can report anecdotal experience from the Madrid area: not a single person that I’ve talked to has a clue what it is. I don’t mean that they fail to provide the technical definition – they don’t even know what it’s about. Ask about sensibilidad climática and people will think you’re talking about how humans react to temperatures, not how the atmosphere reacts to greenhouse gases.
It’s a pity because, at its core, climate sensitivity is an easy concept. And the way it’s calculated is easier than GDPs. You just need to grasp the interplay between:
- Temperatures: duh
- Forcing: it would be easier if we just called it “impact” or something to that effect, but still, not that hard. The more forcing, the more warming.
- Ocean heat uptake and the corresponding energy imbalance: perhaps the least intuitive part of the calculation. Nevertheless, it’s only necessary in order to estimate equilibrium climate sensitivity (ECS); for the transient climate response (TCR) all you need are temperatures and forcing.
People even talk about climate sensitivity without realizing it. For instance, one common argument among climate skeptics is that emissions of CO2 were small prior to 1950, and thus the warming that took place before that year could not have been due to man-made CO2 emissions. But what people making this argument mean, even without putting it that way, is that if CO2 drove the early 20th century warming then climate sensitivity must have been high. And yet, the evolution of temperatures since 1950 suggests a lower climate sensitivity. Skeptics making this argument are implying that it makes no sense for climate sensitivity to have been much lower since 1950 than before, and therefore something else must have been driving warming before 1950.
The problem with this kind of arguments (from all sides of the debate) is not that they use numbers, but that they’re not numerical enough. Or perhaps I should say rigorous enough. Now, I don’t have anything in particular against the early-20th-century-warming argument; I find it to be good example because it’s common. There are three issues plaguing this kind of argument:
- What matters for warming is not CO2 emissions or concentrations, but radiative forcing. This may seem obvious but even some sophisticated authors don’t actually to look at forcing – here’s a recent example.
- Proponents of the argument usually don’t even try to quantify the non-CO2 forcings. Okay, there are probably natural forcings (e.g. clouds) that we cannot quantify because there are no records until the last couple decades. But that doesn’t mean you shouldn’t account for the known forcings! Methane has a warming effect, aerosols have a cooling effect, etc. You have to take these into account if you want to know exactly why the climate did what it did.
- You then have to compare the evolution of these two inputs (forcing and temperatures) in order to arrive at an output, which is the amount of warming per unit of forcing. This ratio is essentially the TCR. Most proponents of the early-20th-century-warming argument cannot calculate a TCR because they don’t use radiative forcing as an input, and some don’t really check temperatures either (they just eyeball a temperature chart).
Nevertheless, it’s easy to understand why people would eyeball temperature charts, and easier to understand why they don’t look at radiative forcing. Even though the websites for temperature records are available to anyone with an internet connection, they require some sacrifice in terms of learning to navigate the data; if you don’t know that Wood for Trees exists, figuring out the exact temperature change between two points in time can be difficult.
As for radiative forcing, the “official” sources (e.g. the IPCC AR5 report) are updated every five or six years; if you want an estimate between those, you have to check the scientific literature. And unlike the Met Office or GISS, there isn’t any organization regularly pushing out press releases on what the latest forcing levels are. There is also the issue of aggregating the dozen or so different forcings into a single time series – one more obstacle for anyone who wants to calculate sensitivity.
Wouldn’t it be great if an app could check all this? You tell the app what years you’re interested in, and the app gives you:
- Temperatures or, more accurately, temperature anomalies
- Forcing levels. Not just for CO2, but for the aggregate of all forcings we more or less know about.
- Energy imbalance – derived from a closely related measure, ocean heat uptake
The app could also tell you what is the difference in these measures between two periods; that is to say, the app could inform the user that forcing between two points in time increased by A, temperatures increased by B, and energy imbalance increased by C. Going even further, the app could then spit out estimates of TCR and ECS.
Well, such an app now exists.
Clisense: an app that does all the climate sensitivity math for you
If you google “climate sensitivity calculator” you’ll find several websites that use this term. However, they don’t actually do what I mentioned in the above section. This one, for example, simply shows how much temperatures will go up depending on climate sensitivity and CO2 concentrations (which are prescribed by the user); it’s actually a temperature calculator.
So, to the best of my knowledge, Clisense is the first app that allows the user to estimate climate sensitivity. The user selects two periods, which can actually be single years if you so wish (just select the same year as both the start and the end of a given period). The app will then estimate TCR and ECS while showing each step of the calculation. This allows users greater insight into just how we “know” that ECS is this low or that high.
Additionally, the app asks the user to prescribe:
- How much of the Earth’s energy imbalance is made up by ocean heat uptake. The IPCC’s AR5 report estimated 93% of planetary heat uptake was oceanic, and that is the default value used by Clisense, but this percentage is not totally certain.
- How “efficacious” the different forcings are. For the most part there is no reason to believe some forcings have greater efficacy than others, but the app allows users to play with the numbers and see how estimates vary. For instance, how would our estimates of sensitivity change if aerosols had greater efficacy than CO2?
The numbers will be meaningless to the average online Joe, so I also made this explanatory website. Having two websites is not the most elegant solution, but custom domains on Shiny Apps go for $300 a month so it will have to do.
Only as good as the data that goes in
To be clear, Clisense estimates climate sensitivity according to a variety of inputs. If our data on temperatures, forcing, and/or energy imbalance were significantly wrong, then the app’s estimates would also be wrong. Of course developing better estimates of all three inputs is one of the main goals in climate science, but the app cannot “guarantee” that the data is correct.
For temperature, I used HadCRUT as it’s the record I’m most familiar with. Going forward, at a minimum I would like to add the Cowtan & Way version, which shows greater warming.
For forcings, I took last year’s Lewis & Curry paper, which has data up to 2016. Since that paper also used HadCRUT for one of its set of results (for the other it used Cowtan & Way), it’s possible to replicate at least one of the paper’s TCR values. Clisense uses only arithmetic for now; I don’t calculate any confidence interval, I don’t use any Bayesian prior, etc. But the result Clisense gives is the same as in Lewis & Curry, which proves the paper’s estimates aren’t biased down by its use of Bayesian priors, as some online commentators had argued.
For ocean heat uptake, the source is Zanna et al 2019. This paper was really key for the app as it’s the only source I know of that offers yearly estimates going back to the 19th century; other ocean temperature records only go back to 1950 or so. Besides, Zanna et al offers a full-depth estimate of ocean heat uptake; using other datasets often involves adding or subtracting different sources to arrive at a complete estimate.
Zanna et al is the data source used in the app that I have greatest reservations about, because it shows a rate of ocean heat uptake of 0.3 or 0.4 w/m2 going all the way back to 1930. If you run the numbers with an initial period like 1930-1950 and a final period like 2007-2016, there is virtually no increase the rate of ocean heat uptake. This is hard to believe, as man-made forcing increased very strongly between 1930 and 2016. And the result is that is that for periods like those the ECS estimate is only marginally higher than that for TCR – or even lower, depending on the exact combination of periods. It’s not clear if the paper’s figures are too high for the first half of the 20th century or too low for the recent decades, but in either case the result would be to bias down ECS estimates.
It must also be said the actual ocean heat uptake data is not available; Clisense uses my digitization of the paper’s plots. In the future, I want to add other sources of data on ocean heat uptake, although that will probably mean the year range will be restricted.
Finally, a note of caution. The app, like the scientific literature, uses volcanoes and solar radiation as the only natural forcings. In other words, the app makes the assumption that other natural factors don’t matter. This is obviously absurd for short periods of time, due to oscillations like El Niño. That’s why users are asked to select not two years but two periods: so that natural variability evens out. However, just because one selects two periods, one cannot be sure that natural variability has been completely removed. If some natural factor (e.g. reduced cloud cover) has effectively acted as a positive forcing over decade or multi-decade timescales, then ECS estimates will be biased high, because we’re assuming all the warming was caused by man-made forcing when in fact part was caused by natural forcing. The reverse is true if natural variability has acted to cool the climate. I prefer not to go further down that hole, as the research is still very tentative.
The way forward
Let me emphasize that Clisense is a project developed in my spare time. I receive no funding – in fact it costs me money, both for the Shiny app and the WordPress site. If my personal situation were to change, I might find myself without sufficient time and energy to keep improving the app. My goal is to add a ton of features – I just cannot guarantee I will.
If you have comments, questions or feedback of any form, I encourage you to share them with me by writing to alberto.zaragoza.comendador at gmail.com.