Figure 2: Maps
showing the evolution of the spatial distribution of anthropogenic SO2
emissions in the pre-industrial era represented by 1850 (a); the peak
emissions era of 1970 (b); current emissions (c); and future emissions
under SSP 245 (d).
For the idealised CMIP simulations (abrupt-4xCO2 and1pctCO2 ) no emissions files are used and so the cumulative
anthropogenic CO2 emissions are calculated from the difference in the
diagnosed CO2 atmospheric mass concentrations in these and thepiControl experiment. Emissions of all other species are also
provided but set to zero (as they represent no change since the
pre-industrial).
2.1 Target ESM
As the model with the most relevant experiments completed, we use the
output from simulations performed by the NorESM2 model in its low
atmosphere-medium ocean resolution (LM) configuration (Seland et al.,
2020). This model consists of a fully coupled earth system with online
atmosphere, land, ocean, ice and biogeochemistry components. It shares
many components with the Community Earth System Model Version 2
(Danabasoglu et al., 2020) but has a replaced aerosol and atmospheric
chemistry scheme (including their interactions with clouds) and a
different ocean model. It has a relatively low equilibrium climate
sensitivity (ECS; equilibrium global mean temperature after a doubling
of CO2) of 2.5 K, particularly compared to the 5.3K of CESM2 (Gettelman
et al., 2019), which has been attributed to ocean heat uptake and
convective mixing in the Southern Ocean (Gjermundsen et al., 2021).
Combined with a s for 1850 to 2014), this likely accounts for the
somewhat anomalous cooling between 1950-1980 in the historical
simulations.
The output of these simulations are aggregated to annual mean values but
kept at their native spatial resolution (approximately 2°). The
temperature (T) and precipitation (P) are exactly equivalent to the
archived surface air temperature (tas) and total precipitation (pr)
output variables respectively. The DTR is calculated as the annual mean
difference in the daily maximum and minimum surface air temperatures
(tasmax – tasmin). The PR90 is calculated as the 90thpercentile of the daily precipitation in each year. The annual mean
baseline values (from piControl ) for each variable are then
subtracted from each experiment so that they represent a difference from
pre-industrial. Temperature changes under anthropogenic climate change
are routinely reported in this way, and it also makes the downstream
emulation task somewhat easier as it removes an offset. The values are
not scaled to have unit variance, but users of the dataset might choose
to do this with certain emulators. Samples of these output fields from
the target ssp245 dataset are shown in Figure 3. The relative increase
in warming in the northern polar regions (known as Arctic amplification)
is clearly seen in Fig. 3a, as well as the north Atlantic warming hole
(Woollings et al., 2012; Drijfhout et al., 2012; Manabe and Stouffer,
1993), the emergence of which is also affected by aerosol radiative
forcing (Dagan et al., 2020). Figure 3b shows the strong land/sea
contrast in DTR, since most of the change is confined to land, and
largely caused by changes in aerosol (particularly sulfate) forcing.
Most of the precipitation response shown in Figure 3c-d is due to the
shift in the inter-tropical convergence zone (ITCZ) which results from a
shift in the cross-equatorial energy balance under increased warming
(Schneider et al., 2014), but some features, particularly in South-East
Asia might be due to local aerosol effects (particularly due to BC;
e.g., Bollasina et al. 2014, Wilcox et al. 2020, Mansfield et al. 2020).