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).