Figure 2
From: Widespread persistent changes to temperature extremes occurred earlier than predicted
Underestimated emergence of persistent changes to TN90p (hot nights) and TX90p (hot days) in CMIP5 ‘Historical’ simulations is linked to a combination of biases in the simulated change (‘signal’) and the simulated variability (‘noise’). Panels show the fraction of CMIP5 ‘Historical’ simulations with signal (a,b), noise (c,d) and signal-to-noise ratio (e,f) of TN90p (left panel) and TX90p (right panel) that would result in a delay of emergence or produce emergence in the opposite direction to observed (see Method). Signal is approximated as the absolute total linear trend in temperature extremes over 1921–2005 and noise as the standard deviation of residuals after removing this linear trend. Stippling indicates where the linear trend in HadEX2 observations is consistent with cooling rather than warming (i.e., a negative trend for TN90p and TX90p). Gray color marks regions for which there is no emergence in HadEX2 observations by the year 2000. White regions have no data. See Fig. S10 for TN10p (cold nights) and TX10p (cold days). The map is produced using R version 3.0.3 software (https://www.r-project.org/).
