Figure 15. Boxplots of noise (unit: K) in TXx (a) and TNn (b)
calculated over the period 1950-2005 across 10 Australian regions, for
CanESM5-LE (cyan) and MIROC6-LE (green). 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.
4 Conclusions
In this study, we analyzed the projected changes for the temperature
extremes under future scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0 and
SSP5-8.5) from the Tier 1 experiment in ScenarioMIP, which is compared
with RCP2.6, RCP4.5 and RCP8.5 in CMIP5. We then use an SNR framework to
estimate the time when the signal of climate change for TXx and TNn
emerges from the internal variability in the two CMIP ensemble. In
addition, two LEs in CMIP6 are employed to estimate the effect of
internal variability on the projected changes and TOE/SNR.
The projected changes for the multi-model medians of the extremes under
the highest scenario show the strongest warming, and the warming for the
indices under SSP3-7.0 fills the gap between SSP2-4.5and SSP5-8.5, with
SSP1-2.6 showing the least warming, especially in the end of this
century. For some extreme indices (TXx, TXn, TNx, TNn, WSDI and CSDI),
although the spatial patterns of warming can be different, there usually
projects “warm-get-warmer” pattern over Australia. As for the spread
in the projections of temperature extremes, they broadly span narrower
envelopes for most indices under lower scenarios in the end of this
century. If we take a more sustainable pathway (SSP1-2.6), although it
may take two or three decades to take effects, the narrower spreads and
weaker projected changes pose relatively less challenge for adaptation
decisions compared to other scenarios. Compared to other regions, TA
usually shows highest warming. However, as the performance of the models
over TA usually shows lower scores (Deng et al., 2021), the projected
changes for the medians and the spread for the extremes may not be
robust (Pierce et al., 2009), which is also applied to other regions
such as SSA and SEA.
Compared to the counterpart future pathways in CMIP5, the spread in the
CMIP6 SSPs are commonly wider than RCPs; and for some extremes (e.g.,
TXx and TNn), the multi-model medians in SPPs are usually higher as
well. This is likely caused by different forcings and higher ECS in some
CMIP6 models (e.g., Fyfe et al., 2021; Palmer et al., 2021; Tebaldi et
al., 2021). For example, Fyfe et al. (2021) concluded that despite the
partly countervailing effect by the background stratospheric aerosols,
the higher amount of CO2 can lead to stronger warming in
SSPs. In this study, we also find that for some indices (e.g., TXx), it
is the models with higher ECS that usually show warmer evolution than
the multi-model medians in SSP5-8.5 (not shown). To further figure out
relative importance of each factor, more experiments based on CMIP6
models forced by CMIP5 RCP scenarios and/or CMIP5 models forced by CMIP6
SSP scenarios needed be conducted and added to the collection in
ScenarioMIP (Fyfe et al., 2021; Tebaldi et al., 2021).
We also demonstrate that the medians of SNR for both TXx and TNn in SSPs
are commonly higher than in RCPs; and the uncertainty for the SNR of TNn
is wider. It is noted that the spreads of SNR for both indices decrease
under lower scenarios, which confirms the benefits of lower emission
future pathways. Furthermore, the large uncertainty in time of emergence
(TOE) result from the inter-model spread of both signal and noise, which
is consistent with Hawkins and Sutton (2012). As previous studies
concluded that the statistical fit used in the SNR framework can
attribute internal variability to the signals (e.g., Hawkins & Sutton,
2012; Kumar & Ganguly, 2018; Lehner et al., 2020), we further
illustrate that internal variability can also influence the ranges of
noise. To better isolate forced response, dynamical adjustment or LEs
can be used (e.g., Lehner et al., 2020; Merrifield et al., 2020). In
contrast, using the mean across the range of noise in a LE may be a more
appropriate way to represent the expected noise for the model, which
needs further investigation.
This study suggests that for different extreme temperature indices, the
patterns for projected changes and TOE over Australia can be different,
which poses large challenge for stakeholders and policymakers. A further
effort is to improve the climate models in simulating the physical
processes and the internal variability. Unless they are better
understood and constrained, the uncertainty of projected changes and TOE
will likely continue over future model generations.
Conflict of Interest
The authors declare no financial or other conflicts of interests that
could have appeared to influence the work reported in this paper.
Acknowledgments
We acknowledge two anonymous reviewers for their constructive comments.
We thank Edward Hawkins for feedback and comments. This research/project
was undertaken with the assistance of resources and services from the
National Computational Infrastructure (NCI), which is supported by the
Australian Government. We thank the World Climate Research Programme’s
Working Group on Coupled Modelling, which is responsible for CMIP and
coordinated CMIP5 and CMIP6. We further acknowledge the climate modeling
groups for producing and making available their model output, the Earth
System Grid Federation (ESGF) for archiving the data and providing
access, and the multiple funding agencies who support CMIP and ESGF.
S.E.P-K. is supported by ARC grant number FT170100106 and CLEX grant
number CE170100023.