Figure 12. Boxplots of projected changes in TXx and TNn over 2071–2100 (bold color) and 2031-2060 (light color) relative to the base period 1961–1990 across 10 Australian regions, for CanESM5-LE (cyan) and MIROC6-LE (green). (a) TXx under SSP5-8.5; (b) TNn under SSP5-8.5; (c, d) same as (a, b) but for SSP1-2.6. The boxes indicate the interquartile spreads (ranges between the 25th and 75th percentiles), the black lines within the boxes are the multi-member medians, the whiskers extend to the edges of 1.5 × interquartile ranges and “outliers” outside of the whiskers are denoted by diamonds.
The temporal evolution of signal and the boxplots of noise for TXx and TNn over Australian regions under SSP5-8.5 are shown in Figs. 13-15, and the resulting SNR in Figs. S29 and S30. The relative magnitudes of the ranges in signal and noise over the regions between the two LEs also resemble that for the spread of the TXx and TNn trends shown in Deng et al. (2021). This suggests that internal variability has impacts not only on the uncertainty of signal, but also on the ranges of noise, making the resulting spread of SNR (Figs. S29 and S30) wider or narrower than that for the corresponding signal (Figs. 13 and 14), which introduce further uncertainty in the ranges of TOE. Although the effects of internal variability on TXx and TNn are similar under SSP1-2.6, the temporal evolution of the SNR and the signal for TXx and TNn stabilizes and there are narrower spreads for SNR compared to SSP5-8.5, which is due to the lower magnitude in signal under the lower scenario (Figs. S31-34).