Skip to content
Skip to main content

About this free course

Download this course

Share this free course

Could we control our climate?
Could we control our climate?

Start this free course now. Just create an account and sign in. Enrol and complete the course for a free statement of participation or digital badge if available.

2.1 Uncertainty in predictions

If we intervene in our climate, how confident are we in our ability to predict the resulting change?

Almost by definition, a model is a simplified version, an approximation that can never be perfect.

But their simplifications mean they can be used as tools to understand the world – i.e. they are ‘useful’ – as long as we are aware of their limitations.

We can reduce the uncertainty in climate model predictions in a number of ways.

One approach is to use many different climate models instead of just one. This means that we have multiple predictions and can take, for example, the mean and the 90% range of these, as you saw for the GMST predictions in Session 4, (repeated in Figure 4).

Figure 4 is a repeat of Figure 4 in Session 4. It is a line graph that plots past and projected annual mean GMST from the multi-model ensemble, relative to 1986–2005. Three lines are plotted: historical, RCP2.6 and RCP8.5. Historical data is plotted from 1950 to 2005, and RCP2.6 and RCP8.5 are plotted from 2005 to 2100. The graph shows Temperature change (in °C) on the y or vertical axis against year on the x or horizontal axis. The historical temperature change rises from a little below 0 °C in 1950 to just above 0 °C in 2005. RCP 2.6 shows the change rising gently to a peak around 2050 at close to 1 °C then it maintains steady value up to 2100. RCP 8.5 shows a steady rise to an change of about 4 °C in 2100. There are shaded regions around each line to show the 90% uncertainty range. This is typically around ± 0.2 °C or less for the historic data, around ± 0.5 °C for the RCP2.6 data by 2100, and around ±0.7 °C for the RCP8.5 data by 2100.
Figure 4 Repeat of Figure 4 in Session 4

Another way is to use multiple different versions of one model: changing the inputs slightly each time, to see the effect this has on the results. This is illustrated in the extract below from Tamsin Edwards’ (2015) work, published in the Guardian:

We used a computer model to simulate the Antarctic ice sheet from the recent past up to the year 2200: not just once, but 3000 times. Each version was slightly different to account for ‘known unknowns’ in the physical laws and simplifications describing how ice flows and slides, the map of the bedrock beneath the ice sheet, and when instability might be triggered in each region under [a] mid-high climate scenario…. This gave us a range of model predictions for sea level rise: three thousand possible futures fanning out from today.

Edwards (2015)

If you are interested, you can participate in the citizen science project ‘climateprediction.net [Tip: hold Ctrl and click a link to open it in a new tab. (Hide tip)] ’, which uses the public’s spare computing power to run many different versions of climate models.

Using many climate models, or many versions of one climate model, broadens the simulated climate distributions. We can compare simulations of the past with observations to test whether the models were successful.