Return Time Plots with Two Sets of Data

If two return time plots are shown on the same graph it is very easy to compare how the likelihood of a certain extreme event is different in each plot.

In our climate attribution studies, where we are trying to find out if climate change has changed the risk of extreme weather events occurring, we plot the results from two model ensembles, each made up of tens of thousands of model simulations.

One ensemble represents the world as it actually is, including the effects of man-made climate change. The other ensemble represents the climate “that might have been”, without man-made climate change.

Since each plot shows the chance of an extreme weather event occurring in a different “world”, the difference between the two curves tells us how the probability of the extreme weather event has changed due to the human influence on the climate.

This allows us to answer the crucial question:  “has the likelihood of a particular weather event occurring changed due to climate change?”

When you look at the return time plot for any of our climate attribution studies, such as the weather@home 2014 UK Flooding experiment, the results are presented as a return time plot with two curves – blue and green, plus some light green dots:

All the dots of the same dark colour (blue or green) belong to an ensemble of climate models with identical climate conditions. In this example, dark blue is the winter as observed; dark green is the “world that might have been” without man-made climate change.

The only difference between these two ensembles is the human influence on the climate system. The “world that might have been” models are carefully designed to remove the human drivers of climate change, such as greenhouse gas emissions.

Since we have no observations of the hypothetical “world that might have been”,  the uncertainty is higher than for the world that was actually observed. To account for this uncertainty we do not just make one ensemble with the human signal removed from the climate drivers, but several ensembles with different possible ways of removing that signal. These are represented by the light green dots. The light green dots, therefore, represent our understanding of the uncertainty in the “world that might have been” model.

How can we tell if climate change has influenced the chance of extreme weather occurring?

If, when you compare the results from the two types of simulations in this return time plot, you see that the blue curve (observed winter) is lying above the green curve (“world that might have been”, without climate change) then you can say that the in the observed winter, a threshold for extreme rainfall is exceeded more frequently than in the “world that might have been, without climate change”.

That is, climate change slightly increased the probability of extreme rainfall occurring that winter.

If the blue curve is below the green, then the probability of the rainfall threshold being exceeded has been made less likely due to climate change.

In our weather@home 2014 UK Flooding experiment, we found that the blue “observed winter” curve did indeed lie above the green “world that might have been” curve. More specifically, we found that a 1-in-100-year winter rainfall event (ie. 1% risk of extreme rainfall in the winter of any given year) is now estimated to be a 1-in-80 year event (i.e. 1.25% risk of extreme rainfall in any given winter) so the risk of a very wet winter has increased by around 25%. This change is statistically significant thanks to the number of computer simulations we were able to run– over 33,000 computer models run in the experiment.