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).