Extended Data Fig. 3: Uncertainty in the frequency of compound hot-dry events (f_HD) in idealised experiments.

(Note that an in-depth interpretation of the figure is provided in the Supplementary Material.) Given a present-day bivariate Gaussian distribution of temperature T and precipitation P with a correlation cor(T, P) of -0.5 (first row), 0 (second row), and 0.5 (third row), shading shows the uncertainty in the future f_HD associated with uncertainty in the change of mean temperature (left column) and mean precipitation (right column) at given levels of expected changes in mean temperature (shown on the x-axis) and mean precipitation (y-axis). Magenta isolines show the expected f_HD resulting from the expected changes in mean temperature and precipitation (they are the same on right and left columns for a given cor(T, P)). The second axes show changes in units of present-day standard deviations. The closed contour shows the kernel density containing 90% of the multimodel mean projected changes in mean temperature and precipitation in units of relative present-day standard deviations over land grid-points (actual changes in ^∘C and mm/day are shown in Extended Data Figure 2b and 7b, respectively). The green line indicates changes of equal magnitude in temperature and precipitation, in units of present-day standard deviations. (Note that the difference in magnitude of uncertainty from temperature (left column) and precipitation (right column) results from the fact that the uncertainty in the change of temperature is relatively large compared to the uncertainty in the change of precipitation).