Extended Data Fig. 5: Correlation between the future frequency of compound hot-dry events (f_HD) and changes in mean temperature and precipitation in idealised experiments.

(Note that an in-depth interpretation of the figure is provided in the Supplementary Material.) Pairs of temperature T and precipitation P are simulated from a bivariate Gaussian distribution with a given cor(T, P) which considers an expected future change in mean precipitation and temperature and variability around this change. For a given mean temperature change of +2 ^∘C and no change in mean precipitation, panel a,b show how future f_HD depends on the exact change in temperature and precipitation, respectively (given cor(T, P) = -0.5). For different values of cor(T, P) of -0.5 (c,d), 0 (e,f), and 0.5 (g,h), shading shows the correlation between the future f_HD and the change in temperature (left column) and precipitation (right column) at given levels of expected changes in mean temperature (shown on the x-axis) and mean precipitation (y-axis). For example, the correlation coefficient of the pairs in a is reported in panel c. Axes, green lines, and closed contours are the same as in Extended Data Figure 3. Stippling indicates where at least 90% of the f_HD values from the Gaussian distribution are equal to 0%.