3.3.2: Sources of confidence and
uncertainty
Confidence
in climate model projections arises from many sources. First, climate models
are solidly based on physical laws and scientific understanding of physical
processes. Second, climate model results are evaluated in detail by comparing
model output to observations. The model evaluation chapter in the most recent
IPCC Assessment report provides many examples (Flato et al., 2013). Third, some
of the models used to make climate projections are also used to make seasonal
climate predictions whose skill is routinely evaluated (e.g., Kirtman et al.,
2013, Merryfield et al., 2013; Sigmond et al., 2013; Kharin et al., 2017).
There are,
however, uncertainties that have to be considered when using model projections.
These uncertainties stem from the fact that models cannot simulate all physical
processes exactly (and therefore must make approximations), and from internal
variability in both the simulated and the real climate system (see Chapter 2,
Box 2.5). The uncertainty due to approximations of physical processes can be
reduced, in principle, and models continue to improve in this regard (Flato et
al., 2013). However, it is impossible to reduce the uncertainty from internal
variability that is superimposed on the underlying forced climate change. In
addition, there is uncertainty about what future climate forcing (e.g., future
GHG emissions) will be, which is accounted for by making projections with a
range of forcing scenarios. These sources of uncertainty vary in importance
depending on the time and space scale under consideration — generally,
uncertainties diminish at larger spatial scales as internal variability
“averages out” to a certain degree when one considers larger regions (e.g.,
Haw-kins and Sutton, 2009). This also means that uncertainty is larger when one
looks at small regions or specific locations. In addition, at longer time
scales (say, by the end of the 21st century), uncertainty is dominated by
differences in the forcing scenarios and internal variability is, by
comparison, much smaller.
3.3.3: Global-scale climate
projections
As described
in Section 3.2, climate projections are a result of driving climate models with
different future forcing scenarios (RCPs, in the case of CMIP5). These
projections include the response of the climate system to external forcing
(e.g., changing GHG concentrations), internal variability, and uncertainties
associated with differences between models. These effects can be separated, to
some extent, by drawing upon projections from multiple models (e.g., Collins et
al., 2013). The multi-model average provides an estimate of the response of the
climate system to forcing, since internal variability and model differences are
“averaged out” to a large extent (see Box 3.2). The upper panel of Figure 3.6
shows the change over time in global mean surface air temperature, as simulated
by the CMIP5 models, spanning the period from 1950 to 2100. The heavy lines indicate
the multi-model average, and the shaded band represents the range of model
results around this average. Within this shaded band, each individual model
result would look like one of the individual coloured lines in Figure 3.5, but
for clarity this collection of individual lines is shown as a shaded band. The
high emission scenario (RCP8.5) results are shown by the red line and the
orange shaded band, whereas the low emission scenario (RCP2.6) results are
shown by the blue line and the blue shaded band.
There are
two key points illustrated in the upper panel of Figure 3.6. First, when
looking at projected climate change, the spread across models (the vertical
extent of the shaded bands) is smaller in the near term (to around 2040) than
it is toward the end of the 21st century, indicating that model uncertainty has
a larger effect further into the future. (Internal variability also contributes
to the width of the shaded bands, as described previously, but the size of this
contribution is not expected to change significantly in the future.) Second,
the differences among forcing scenarios are small in the near term, but become
large toward the end of the 21st century (as illustrated by the growing
separation between results for the low emission scenario [RCP2.6] and high
emission scenario [RCP8.5]). For simplicity, the medium emission scenarios
(RCP4.5 and RCP6.0) are not shown in the main part of the figure, but their
end-of-century results are shown on the right-hand side of the top panel for
comparison.
The spatial
patterns of projected temperature and precipitation change are shown in the
bottom panel. The large difference in mean change between the high and low
emission scenarios is clearly evident in the maps (darker colours indicate
larger change), but there is a marked similarity in pattern. For temperature,
changes are larger over land than over the adjacent ocean, and are larger at
high latitudes, particularly over the Arctic, an illustration of Arctic
amplification. As a result, projected warming in Canada is roughly double the
global mean. For precipitation, the pattern of change is more complex, with the
polar and equatorial regions projected to have increased annual precipitation,
whereas precipitation decreases are projected for much of the subtropics
(roughly 24º to 35º north and south latitude). For southern Canada, the
projected change in precipitation is rather small, but projected increases are
larger further north. (Changes in annual mean precipitation do not translate
directly into changes in seasonal snow cover or water availability.
On average,
the models project a future global mean temperature change (relative to the
1986–2005 reference period) of about 1ºC for the low emission scenario (RCP2.6)
and 3.7ºC for the high emission scenario (RCP 8.5) by the late 21st century,
with a 5%–95% range of about 1ºC above and below the multi-model average. This
change is over and above the 0.6ºC change that had already occurred from 1850
to the reference period. Therefore, the average projected change under the low
emission scenario is consistent with the global temperature target in the Paris
Agreement of limiting global warming to between 1.5ºC and 2.0ºC, although the
projected range from all models extends both below and above this target. The
low emission (RCP2.6) scenario requires emissions of CO2 to peak almost
immediately and reduce to near zero before the end of the century. Recent
studies (e.g., Millar et al., 2017) provide more detailed analysis of scenarios
that will limit warming to 1.5ºC, and these also involve very rapid and deep
emission reductions.
More details
regarding future projections, with a focus on Canada, are provided in other
chapters of this report. Confidence in climate change projections varies by
region and by climate variable. So, for example, confidence in temperature
change is higher than confidence in precipitation change. This is in large part
because temperature change is a direct consequence of radiative forcing,
whereas precipitation change is affected by a number of complex interactions,
including changes in the water-holding capacity of a warming atmosphere, in
global atmospheric circulation, in evaporation, and in other factors (e.g.,
Shepherd, 2014). Changes in snow and ice are a consequence of changes in both
temperature and precipitation and are discussed in more detail in Chapter 5.
Freshwater availability (see Chapter 6) and ocean changes are also affected by
changes in temperature and precipitation, as well as by other factors.
3.3.4: Compatible emissions
Earth system
models can be run in two different ways: one in which GHG concentrations are
set and another in which GHG emissions are set (both are available as part of
the RCP datasets). Concentration-driven simulations allow scientists to assess
the difference, from one model to another, in how the climate responds to
identical changes in GHG concentrations in the atmosphere. This helps separate
the response of the climate system to a change in forcing (e.g., change in GHG
concentrations) from the effect of carbon-cycle feedbacks involving the
terrestrial and oceanic biospheres. The response of these natural carbon sinks
to atmospheric CO2 levels and to climate change will influence anthropogenic
emissions compatible with a given CO2 pathway. Therefore, an interesting aspect
of these concentration-forced simulations is that global anthropogenic
emissions can be computed — emissions that are compatible with the prescribed
concentration pathway (e.g., Jones et al., 2013). The range in compatible CO2
emissions between different models provides a measure of the uncertainty
inherent in representing carbon-cycle feedbacks in models. Figure 3.7
illustrates results from compatible emission calculations and shows that, while
there is some variation, this group of models is consistent. For the RCP 2.6
scenario in which temperature is stabilized below about 2ºC, the models have
compatible emissions that start reducing immediately and reach near zero well
before the end of the century.
No comments:
Post a Comment