The ability of a model to simulate the historical surface temperature record is generally regarded in the climate modelling community as an important, even essential, performance criterion for global climate models. UKESM1 has a particularly large cold bias in this quantity, peaking in the 1970s to 1980s with a negative bias approaching 0.5K (Sellar et al. 2019, see Figure 1). An overly strong aerosol forcing has been suggested as a likely leading candidate responsible for this bias consistent with increasing anthropogenic aerosol emissions during this period (Flynn and Mauritsen, 2020; Andrews et al, 2020). Anthropogenic emissions of SO2, a precursor to sulphate aerosol which acts to cool the climate, was the predominant source of anthropogenic aerosols during this time and UKESM1 has been shown to have a significant high bias in SO2 concentrations over anthropogenic source regions of Europe, north-east USA and China (Hardacre et al (2021); Jones et al.). Jones et al. demonstrated a significant improvement in simulated historical temperature record when an improved parameterization of SO2 dry deposition was implemented. If sink processes are too low in UKESM1 the atmospheric residence time of SO2 will be too long, more SO2 will be transported away from source regions and will be subsequently oxidised to aerosol or removed in more pristine remote regions which are more susceptible to aerosol forcing.
Following on from this we have developed a new configuration of UKESM, UKESM1.1, which includes the revised dry deposition of SO2 as its key science change. In brief, the updated dry deposition parameterization takes into account whether the underlying land, vegetated surface is wet or dry as a result of precipitation (Erisman et al., 1994a) and when the land surface is dry, the modelled surface resistance to SO2 deposition is now a positive function of near surface humidity (Erisman et al. 1994b). Given its high solubility the wetter the surface the higher the uptake of SO2. In addition, a bug was found in the surface resistance term for SO2 over water surfaces which further overly inhibited dry deposition. The new parameterization leads to a significant improvement (of the order of 50%) in the SO2 bias against ground-based observations in UKESM1 (Hardacre et al., 2021).
In addition to the SO2 changes we furthermore included various other smaller but desirable changes to the new configuration. A few bugs were discovered only after the freeze of UKESM1 and are corrected here. These include a correction to the updating of sulphuric acid vapour in the aerosol scheme when fluxes are passed from the UKCA chemistry; updates to the DMS chemical reactions and products in the full StratTrop chemistry scheme which make them consistent with the offline oxidant chemical mechanism (see Mulcahy et al. 2020); a fix for the vertical profile of cloud droplet number concentrations in the aerosol activation scheme and a correction to the coupled sea-ice heat fluxes in NEMO.
In addition, we revise some of the coupled tunings outlined in Sellar et al. (2019). The development and tuning of UKESM1 was done using a pre-industrial climate. Subsequent evaluation of these tuned parameters in present-day highlighted some retuning was desirable to improve agreement with observations. For example, the tuning of the burial of vegetation by snow and subsequent impact on shortwave clear-sky fluxes was found to be slightly positively biased in present-day simulations (Sellar et al. 2019). This parameter has been retuned to a lower value in the revised configuration. Simulated dust optical depth (DOD) was found to be biased low in UKESM1 (Mulcahy et al. 2020) and so a retuning of the dust was conducted during the UKESM1.1 development which increases DOD improving agreement with observations but roughly maintains global emission totals. The albedo of snow on sea ice was decreased in UKESM1 by 2% (Kuhlbrodt et al 2018) to compensate for deficient transport of warm Atlantic water into the Arctic in the lower resolution ocean model, ORCA1. As the climate is warmer with the revised dry deposition parameterization we revert to the original value as set in the ORCA025 model. Finally, tuning of the quasi-biennial oscillation (QBO), through modification of the frequency of convectively generated gravity waves, was erroneously omitted in UKESM1 leading to an excessively long QBO period. This has been rectified in UKESM1.1 and an improved (shorter) QBO period is seen.
Having run nearly 500 years of a stable piControl simulation with the final frozen configuration, we have also run 6 historical simulations, a 1%CO2 and 4xCO2 simulation and an AMIP simulation, completing the full set of DECK simulations. Each historical member was initialised from the piControl using restart files spaced 40 years apart. From the end of the first 3 historical members future scenario simulations were run for SSP1-2.6 and SSP3-7.0 scenarios.
Figure 1: Timeseries of the simulated historical mean surface air temperature anomaly relative to the 1850-1900 mean for UKESM1.1 and UKESM1 for (a) the full globe, (b) northern hemisphere extratropics, (c) tropics and (d) southern hemisphere extratropics. Observations plotted are from HadCRUT4 and Cowtan and Way datasets.
Figure 1a shows the global mean surface air temperature anomaly timeseries for 6 members of the UKESM1 and UKESM1.1 historical ensemble along with the ensemble means for both configurations. UKESM1.1 is warmer than UKESM1 throughout the historical period. The two models start to diverge from about 1920 coincident with a notable increase in anthropogenic SO2 emissions. A significant improvement (by more than 50%) is seen in the anomaly bias against both sets of observations. In particular, the cold “pot-hole” bias is improved with UKESM1.1 having much smaller cooling trends in the 1950-1980 period. After 1990 both models tend to warm at the same rate and UKESM1.1 is slightly too warm by the end of the historical period. This later period being dominated by the warming due to greenhouse gases hints at a similar climate sensitivity between the two models. Indeed, the ECS calculated from the 4xCO2 simulation is 5.27K compared with 5.36K in UKESM1. Figure 1b shows that most of this improvement is coming from the NH extratropical region with much smaller changes in the tropics and southern hemisphere (Figures 1c and d).
Further beneficial or neutral impacts are seen in other aspects of the simulated climate. Figure 2a shows the impact of UKESM1.1 on the global ocean heat content anomaly in the top 700m. Between 1971 and 1991 there is an improved agreement between UKESM1.1 and the observations in ocean heat uptake with UKESM1.1 having a larger heat uptake. From the late 1990s the rate of ocean warming between the two models is very similar. Examination of the regional ocean basins shows this stronger heat uptake predominantly takes place in the North Atlantic and North Pacific Ocean basins. Arctic sea ice is also improved, with reduced sea ice extent and volume improving agreement with observations (Figure 2b).
Figure 2: (a) Timeseries of the ocean heat content anomaly in top 700m normalised to 1971 and (b) Arctic sea ice extent for UKESM1 and UKESM1.1 and various observational datasets.
Figure 3 shows the near-surface temperature anomalies relative to an 1985-2014 mean for both the SSP3-7.0 and SSP1-2.6 scenarios for the 3-member ensemble in UKESM1 and UKESM1.1. While the warming rates are very similar in both scenarios post-2014, the UKESM1.1 model appears to warm slightly less in both cases. This could be due to the higher baseline aerosol optical depth in UKESM1.1 (driven largely by the higher DOD) driving a slightly stronger SW clear-sky cooling but more detailed investigations are ongoing.
Figure 3: Global near surface temperature anomalies from 1850 to 2100 relative to 1985-2014 mean from a 3 member ensemble of UKESM1 and UKESM1.1 for the SSP3-7.0 and SSP1-2.6 scenarios.
Andrews et al. Historical simulations with HadGEM3-GC3.1 for CMIP6. Journal of Advances in Modeling Earth Systems, 12, e2019MS001995. https://doi.org/10.1029/2019MS001995, 2020.
Flynn, C. M. and Mauritsen, T.: On the climate sensitivity and historical warming evolution in recent coupled model ensembles, Atmos. Chem. Phys., 20, 7829–7842, https://doi.org/10.5194/acp-20-7829-2020
Sellar, A. A., Jones, C. G., Mulcahy, J. P., Tang, Y., Yool, A., Wiltshire, A., et al. (2019). UKESM1: Description and evaluation of the U.K. Earth System Model. Journal of Advances in Modeling Earth Systems, 11, 4513– 4558. https://doi.org/10.1029/2019MS001739
Hardacre et al., Evaluation of SO2, SO42− and an updated SO2 dry deposition parameterization in UKESM1, Atmos. Chem. Phys. Discuss. [preprint], https://doi.org/10.5194/acp-2021-238, in review, 2021.
Mulcahy et al., Description and evaluation of aerosol in UKESM1 and HadGEM3-GC3.1 CMIP6 historical simulations, Geosci. Model Dev., 13, 6383–6423, https://doi.org/10.5194/gmd-13-6383-2020, 2020.
Erisman et al., Parametrization of surface resistance for the quantification of atmospheric deposition of acidifying pollutants and ozone, Atmospheric Environment, Volume 28, Issue 16, 2595-2607, https://doi.org/10.1016/1352-2310(94)90433-2, 1994a.
Erisman et al., The Elspeetsche Veld experiment on surface exchange of trace gases: Summary of results, Atmospheric Environment, Volume 28, Issue 3, Pages 487-496, https://doi.org/10.1016/1352-2310(94)90126-0, 1994b.