BCC-ESM1 Model Datasets for the CMIP6 Aerosol Chemistry Model Intercomparison Project (AerChemMIP)

Jie ZHANG, Tongwen WU, Fang ZHANG, Kalli FURTADO, Xiaoge XIN, Xueli SHI, Jianglong LI,Min CHU, Li ZHANG, Qianxia LIU, Jinghui Yan, Min WEI, and Qiang MA

1Beijing Climate Center, China Meteorological Administration, Beijing 100081, China

2Met Office Hadley Centre, Exeter EX1 3PB, UK

3National Meteorological Information Center, China Meteorological Administration, Beijing 100081, China

ABSTRACT

Key words:BCC-ESM1,CMIP6,AerChemMIP,climate sensitivity,precipitation,surface air temperature,sulfate

1.Introduction

Climate impacts of atmospheric chemistry, especially the contribution of tropospheric ozone precursors to radiative forcing, have been estimated in all the Intergovernmental Panel on Climate Change (IPCC) reports. However,even the models used in the IPCC’s fourth report were mostly offline chemistry transport models, and none of the climate models included tropospheric ozone chemistry(Collins et al., 2017). The Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP) was launched in 2011 and aimed to document changes in atmospheric composition and the associated radiative forcings in CMIP5 models (Lamarque et al., 2013). However, the model setups for ACCMIP were generally different from CMIP5 (with lower resolution but more complex chemistry)and only four of the models could be used to estimate future climate changes in surface ozone (Schnell et al., 2016). To build on the experiences of previous programs, the CMIP6-endorsed Aerosol Chemistry Model Intercomparison Project(AerChemMIP) is a series experiments with the goal of quantifying the effects of near-term climate forcers (NTCFs: methane, tropospheric aerosols and ozone, and their precursors)and reactive well-mixed greenhouse gases (WMGHGs) on climate and air quality (Collins et al., 2017). AerChemMIP is designed to answer four scientific questions: (1) how have anthropogenic emissions contributed to global radiative forcing and affected regional climate over the historical period? (2) How might future policies (on climate, air quality, and land use) affect the abundances of NTCFs and their climate impacts? (3) How do uncertainties in historical NTCF emissions affect radiative forcing estimates? (4) How important are climate feedbacks to natural NTCF emissions,atmospheric composition, and radiative effects?

BCC-ESM1 is a new version of the Beijing Climate Center’s ESM that includes interactive atmospheric chemistry and aerosols. Its details are described in Wu et al. (2020a).BCC-ESM1, along with 14 other ESMs, have participated in AerChemMIP, including CESM2-WACCM, CNRMESM2-1, EC-Earth3-AerChem, EMAC-2-54-AerChem,GFDL-ESM4, GISS-E2.1, IPSL-CM5A2-INCA, IPSLCM6A-AER-LR, MIROC6, MIROC-ES2L, MPI-ESM-1-2-HAM, MRI-ESM2-0, NorESM2, and UKESM1-0-LL. All of these ESMs are run in AerChemMIP and the other programs under CMIP6 [DECK (Diagnostic, Evaluation and Characterization of Klima) and the other CMIP6-endorsed MIPs] with their maximum available complexity. Therefore, their results can be shared across MIPs and used for more comprehensive assessments. The AerChemMIP experiments are classified into three tiers by their priority. The Aer-ChemMIP is the first international MIP that BCC-ESM1 has been involved in, and thus far it has conducted all its Tier 1 experiments and seven experiments in lower priorities. BCC-ESM1 has also run the DECK and CMIP historical simulations, which serve as the “entry cards” for models participating in CMIP6 (Eyring et al., 2016).

AerChemMIP is the first chemistry program fully involved in the CMIP structure. Most of its experiments are newly designed and of special scientific interest. This paper introduces the experiment settings as well as the BCCESM1 outputs, to facilitate their use. Following this introduction, a brief overview of BCC-ESM1 is provided in section 2. The experiments run by BCC-ESM1 are also summarized in that section. Section 3 presents a technical validation of BCC-ESM1, including the model climate sensitivity and evaluations of surface air temperature (SAT), precipitation, and surface sulfate concentrations. Section 4 provides a summary, and section 5 shows some usage notes.

2.Model and experiments

2.1.Model configuration

BCC-ESM1 is based on the climate system model BCC-CSM2 (Wu et al., 2019a), but includes interactive representations of tropospheric chemistry and aerosols. As the successor of BCC-CSM1.1 (Wu et al., 2013), which participated in CMIP5, BCC-CSM2 included updates to the deep convection parameterization, considered the indirect effects of aerosols and gravity wave drag generated by deep convection, and introduced a new scheme for the cloud fraction.BCC-CSM2-MR, the medium resolution version of BCCCSM2, shows an overall improvement in many aspects,such as the tropospheric air temperature at the global and regional scale in East Asia, the stratospheric quasi-biennial oscillation, and the Madden–Julian Oscillation.

The atmospheric component of BCC-ESM1 is BCCAGCM3-Chem (Wu et al., 2020a), which includes an interactive tropospheric gas-phase and aerosol chemistry scheme.The model includes 66 gas-phase species and 13 prognostic aerosol species. The main atmospheric chemistry built in BCC-AGCM3-chem is based on MOZART2 (Model for OZone And Related chemical Tracers, version 2), a troposphere-only chemistry transport model developed by the National Center for Atmospheric Research (Horowitz et al.,2003). SO, dimethyl sulfide (DMS), and NHare the gas tracers newly added to BCC-ESM1. A bulk aerosol scheme is also newly introduced in BCC-ESM1, including the relative reactions, transport, emissions, and depositions. The aerosol species involved are sulfate (SO), hydrophobic and hydrophilic organic carbon (OC1 and OC2), hydrophobic and hydrophilic black carbon (BC1 and BC2), four categories of soil dust (DST01, DST02, DST03, DST04), and four categories of sea salt (SSLT01, SSLT02, SSLT03, SSLT04).Dust and sea salt aerosols are classified into four size bins by diameter. The mass-mixing ratios of each aerosol type are prognostic variables. Prognostic aerosol masses are used to estimate the liquid cloud droplet number concentration and the indirect effects of aerosols are considered.

To save computing resources and tune the model more efficiently, BCC-AGCM3-Chem has a relatively low horizontal resolution of T42 (about 280 km) and has 26 vertical levels with the top level at 2.914 hPa. The land component of BCC-ESM1 is the BCC Atmosphere and Vegetation Interaction Model, BCC-AVIM2.0, with terrestrial carbon cycle(Li et al., 2019). The oceanic component is the Modular Ocean Model, version 4, with 40 levels included (MOM4-L40; Griffies et al., 2005). The sea-ice component is the Sea Ice Simulator (SIS), with three vertical layers, one snow cover layer, and two ice layers (Winton, 2000). MOM4-L40 and SIS use a tripolar grid with a horizontal resolution of 1°longitude by 1/3° latitude between 10°S and 10°N, increasing to 1° at 30°S and 30°N.

2.2.Experiments

The experiments conducted by BCC-ESM1 are summarized in Table 1. The simulation data have been archived on the Earth System Grid Federation (ESGF) node (https://esgf-node.llnl.gov/projects/cmip6/). DECK includes four basic experiments: (a) a historical Atmospheric Model Intercomparison Project (amip) simulation; (b) a control simulation under pre-industrial conditions (piControl); (c) a simulation that branches from piControl in 1850 and is forced by an abruptly quadrupled global-mean concentration of CO(abrupt-4×CO2); and (d) a simulation that branches from piControl and is forced by a gradually increased COconcentration at a rate of 1% per year (1pctCO2). Amip is run by an atmospheric general circulation model (AGCM) with prescribed sea surface temperature (SST) and sea-ice concentration (SIC) and include three ensemble members. The other three experiments are all conducted by a coupled atmosphere–ocean general circulation model (AOGCM) and include only one ensemble member. The CMIP historical experiment (historical) is also conducted by the AOGCM. It branches from piControl and is forced by evolving external forcing. There are three ensemble members for historical with slightly varying initial conditions. Amip, piControl,and historical are baselines for many CMIP6-Endorsed MIPs. Abrupt-4×CO2 and 1pctCO2 are used to estimate model climate sensitivities and feedbacks.

Table 1. Experiments conducted by BCC-ESM1.

The AerChemMIP experiments are run with the same complexity as in DECK and historical. The experiments are classified into three tiers by their priority. The Tier 1 experiments mainly focus on the impacts of NTCFs and reactive well-mixed gases on historical and future climate, as well as future NTCFs under different air quality policies. Tier 2 and 3 will examine the effects of NTCF emissions and climate feedback more specifically, to provide detailed uncertainty estimates.

BCC-ESM1 has run all the Tier 1 experiments and seven Tier 2 and 3 experiments:

● hist-piNTCF in Tier 1 parallels historical but with the anthropogenic emissions of aerosols, and ozone precursors fixed to their 1850 values. hist-piAer in Tier 3 is similar to hist-piNTCF but with the anthropogenic emissions of aerosol fixed to the 1850 values.By comparing with historical, these experiments can be used to estimate the climate impacts of anthropogenic emissions of NTCFs and aerosols. Three ensemble members are conducted for both histpiNTCF and hist-piAer with the same initial conditions as historical.

● piClim-control in Tier 1 is run under pre-industrial conditions for 31 years with prescribed SST and SIC from 30 years of piControl climatology. It is designed following the protocols of the Radiative Forcing Model Intercomparison Project (Pincus et al.,2016). Eight more piClim-X experiments are also conducted by BCC-ESM1 that branch from piClim-control: (a) piClim-NTCF and piClim-CH4 in Tier 1,which are the same as piClim-control but set NTCF emissions and the CHconcentration to present-day values, respectively; (b) piClim-aer, piClim-BC and piClim-O3 in Tier 2, which set the emissions of aerosol precursors, black carbon and ozone precursors to present-day values, respectively; (c) piClim-NOX,piClim-VOC and piClim-SO2 in Tier 3, which set the NOx, CO and VOCs, and SOemissions to presentday values, respectively. These experiments can help to quantify the effective radiative forcings (ERFs) at pre-industrial levels due to different drivers, as in Thornhill et al. (2020). The piClim-X experiments include only one ensemble per experiment in BCCESM1.

● histSST in Tier 1 also branches from piClim-control but with prescribed time-evolving SST and SIC from historical. All other forcings are the same as in historical. histSST-piNTCF and histSST-piCH4 in Tier 1 are the same as histSST and branch from the same point as histSST but with the NTCF precursors and methane fixed to their 1850 values, respectively.These experiments can help to diagnose the transient ERFs for NTCF and methane. The histSST-X experiments include only one ensemble per experiment.

● ssp370 in Tier 1 is a future simulation with the highest levels of short-lived climate pollutants without climate policy. It achieves forcing levels of 7.0 W min 2100. ssp370-lowNTCF in Tier 1 is a future simulation with strong levels of air quality control measures. Both experiments branch from the historical experiments in 2015 and are run for 50 years from 2015 to 2055 instead of 85 years from 2015 to 2100 as suggested by AerChemMIP, to reduce the computational expense. These experiments can help to estimate the impacts on future climate and air quality, as in Allen et al. (2020) and Turnock et al.(2020). Three ensemble members are conducted for both ssp370 and ssp370-lowNTCF.

● ssp370SST in Tier 1 is a transient future prescribed-SST simulation following the ssp370 scenario. It branches from the first member of amip in year 2015 with prescribed SST and SIC from the first member of the ssp370 experiment. ssp370SST-lowNTCF in Tier 1 branches at the same point as ssp370SST with prescribed SST and SIC from the first member of ssp370-lowNTCF following the ssp370-lowNTCF scenario. Different ERF changes between these future prescribed SST simulations and the relevant coupled model projections (ssp370 and ssp370-lowNTCF) help to quantify the efficacy of NTCFs to affect climate. Both experiments include only one ensemble.

3.Technical validation

3.1.Model climate sensitivities to idealized CO2 forcing

Climate sensitivity is an important metric for model evaluation. It is one of the major factors responsible for different model climate responses and uncertainty in future climate projections (Grose et al., 2018). Here, we estimate the global mean SAT response to idealized COforcing by using the 1pctCO2 and abrupt-4×CO2 experiments. Both experiments are initialized from the piControl experiment,which is used as the reference experiment.

The transient climate response (TCR) is calculated from the 1pctCO2 simulation in which COis gradually increased at a rate of 1% per year. The TCR is defined as the annual mean SAT change relative to the piControl equilibrium climate state, averaged over a 20-year period centered on the time at which the COconcentration was doubled(year 70 of the simulation). The 1pctCO2 forcing resembles the COforcing in the 21st century but is more idealized.As shown in Fig.1a, the evolution of SAT anomaly is linearly correlated with the COconcentration in the 1pctCO2 experiment. A high linear correlation is also evident in observations at both global and regional scales (Wu et al.,2019b). The TCR in BCC-ESM1 is about 1.78 K, which is 1.40 K for BCC-CSM2-MR (Shi et al., 2020). The TCRs in both BCC models lie within the “l(fā)ikely” range of 1.0–2.5 K suggested by IPCC AR5 (Bindoff et al., 2013). In the 1pctCO2 experiment and the real world, the distribution of heat flux at the atmosphere–ocean interface has not yet reached equilibrium.

The equilibrium climate sensitivity (ECS) is measured as the SAT changes in a double COexperiment when the climate system reaches its equilibrium climate state. The ECS in BCC-ESM1 is estimated by the abrupt-4xCO2 experiment. SAT changes in the abrupt-4xCO2 experiment are shown in Fig.1b. The model response to quadrupled COcan be separated into a fast-response stage in the first two decades and a slow-response stage afterward. Following the approach of Gregory et al. (2004), we regress the global annual mean net radiative-flux changes at the top of the atmosphere (TOA) on SAT changes and then extrapolate to zero net flux (Fig.2). The ECS is then defined as half of the xintercept concerning a COdoubling. This method yields an ECS estimate of 3.34 K for BCC-ESM1, which is 3.03 K for BCC-CSM2-MR (Wu et al., 2019a). ECSs in both BCCESM1 and BCC-CSM2-MR are within the “l(fā)ikely” range of 1.5–4.5 K suggested by IPCC AR5.

3.2.SAT in piControl and historical experiments

Fig.1. (a) Temporal evolution of globally averaged SAT changes relative to the reference experiment (piControl) for the 151 model years in the 1%CO2 experiment. (b) As in (a)but for SAT changes in the abrupt-4×CO2 experiment. Units: K.

Fig.2. Annual changes of the net TOA radiative flux and the global-mean SAT for the first 151 years of the abrupt CO2 quadrupling experiment (abrupt-4×CO2) relative to the preindustrial control run (piControl). The regression line is also calculated to estimate the ECS of BCC-ESM1.

Figure 3a shows the evolution of the global mean annual averaged SAT in the piControl run. In piControl,SAT reaches a quasi-equilibrium state with a slight increasing trend of about 0.016°C (100 yr), which is much smaller than the standard deviation of SAT (0.10°C yr). Time series of annual mean SAT anomalies from the three members of the CMIP6 historical simulations are shown in Fig.3b, as well as the observation from HadCRUT4(https://www.metoffice.gov.uk/hadobs/hadcrut4/). The anomalies are relative to the 1850–1950 mean. The SAT simulations show good agreement with HadCRUT in the first 100 years. However, there is a significant cooling bias in the second half of the 20th century, especially over the period from the 1960s to 1990s. We colloquially refer to this significant cooling bias as the “pot-hole” bias (because it represents an undesirable dip in the “road” of an accurately simulated SAT trend). It is a common problem in most of the ESMs participating in CMIP6. One possible reason for the pot-hole bias is that the prognostic aerosol schemes produce overly strong aerosol cooling (Flynn and Mauritsen, 2020; Meehl et al., 2020).

The spatial structure of SAT in the historical experiment is examined over the period from 1986 to 2005 (Fig.4),as well as its biases compared with ERA5 reanalysis (Copernicus Climate Change Service, 2017). The spatial structure of SAT is well captured with a spatial correlation of 0.99. Simulated SAT is about 0.35 K higher than that in ERA5, attributable to the warmer tropics and subtropics. Generally, the biases in BCC-ESM1 show great similarity with those in BCC-CSM2-MR (Wu et al., 2019a), including the slight systematic warm biases over oceans, as well as the relatively large warming/cooling biases over the polar regions. The biases over the oceans and polar regions may partly come from the ocean and sea-ice modules, which are the same in BCC-ESM1 and BCC-CSM2-MR. Both models also show systematic cool biases over most land regions. The cooling bias may be partly attributable to the land surface modeling component, since it is also evident in AMIP simulations, or to other processes like errors in cloud properties. The cooling biases in BCC-ESM1 may also be related to the overestimated sulfate concentration in the atmosphere shown in subsection 3.4.

Fig.3. (a) Time series of the global mean SAT in the piControl simulation. (b) Time series of the global mean SAT anomalies in the historical simulations (three members, black lines) and the observation (red line) from HadCRUT4. The anomalies are relative to the 1850–1950 mean.

3.3.Precipitation and associated near-surface horizontal wind

Annual mean precipitation in the historical experiment and its comparison with GPCP observation (Adler et al.,2003) are examined and shown in Fig.5. The biases are generally within 1 mm dand the global mean precipitation in BCC-ESM1 is very close to the observation: 2.27 mm dversus 2.24 mm d. BCC-ESM1 still suffers from the double ITCZ problem: precipitation south of the equator in the eastern Pacific is overestimated and the SPCZ is too zonally elongated. Zhang et al. (2015) found that the double ITCZ related biases include sufficient precipitation in the southeastern Pacific, warmer SST, weaker easterly flow,and stronger meridional wind divergences away from the equator. However, no progress can be identified from the multimodel ensemble (MME) mean in CMIP3 to that in CMIP5.The double ITCZ measures are also examined for BCC-ESM1, as in Zhang et al. (2015). The SST bias in BCC-ESM1 closely resembles the SAT bias (not shown).As indicated by Fig.4c, the warmer SST south of the equator amplifies the precipitation over the Southeast Pacific and leads to the eastward-extended SPCZ. Over the eastern tropical Pacific, the near-surface zonal wind weakens and the meridional wind converges south of the equator(Fig.6). Both zonal and meridional wind biases contribute to the SPCZ biases in BCC-ESM1. Generally, biases of the double ITCZ measures in BCC-ESM1 are quite similar to those in the CMIP3 and CMIP5 MMEs. With great improvements in atmospheric physical parameterizations, the double ITCZ bias is much weaker in the newly developed high-resolution climate system model, BCC-CSM3-HR (Wu, 2020b).

Fig.4. Annual mean SAT in (a) BCC-ESM1 and (b) ERA5 reanalysis from 1986 to 2005. (c) The SAT biases. Units: °C.

Other model deficiencies in precipitation reproductions are evident, including the overestimations over eastern Indonesia, western Indian Ocean, southern Africa, and East Asia. Meanwhile, underestimations are pronounced north of the equator in the Pacific and Atlantic. Such biases seem to be common in CMIP3 and CMIP5 models (Oueslati and Bellon, 2015). The sufficient precipitation over eastern Indonesia and the western Indian Ocean accompanies stronger trade winds in the east. The anomalous easterlies converge toward the biased precipitation, west of the Maritime Continent and the eastern African coast. The underestimations north of the equator in the Pacific and Atlantic are asymmetric with the overestimations in the south, which may be attributable to the asymmetric warming amplitudes by the equator as well as the local meridional wind divergences(Zhang et al., 2015). The Southern Africa and East Asia are greatly affected by the global monsoon system, and the precipitation biases can be related to many physical and dynamic processes, but investigating these is beyond the scope of the present study.

Fig.5. Annual mean precipitation in (a) BCC-ESM1 and (b)GPCP from 1986 to 2005. (c) The precipitation biases. Units:mm d?1.

3.4.Surface sulfate and its variation

Sulfate is a major species of aerosol. Sulfate aerosols can cool the atmosphere by reflecting solar irradiation directly or by making clouds more reflective via aerosol indirect effects. The anthropogenic emissions of SOfrom fuel combustion and industrial activities constitute the main precursor of atmospheric sulfate. A large amount of sulfate in the atmosphere is believed to have significant cooling effects at regional and global scales, although large uncertainties remain in the magnitude of these effects. BCC-ESM1 can capture the major features of aerosols reasonably well in comparison with observed aerosol concentrations and their optical properties (Wu et al., 2020a). In this paper, we use ground-based monitoring networks to extend the evaluation of sulfate. Simulation from the historical experiment is examined here.

The sulfate observations are from three monitoring networks: the European Monitoring and Evaluation Program(EMEP; T?rseth et al., 2012); the United States Interagency Monitoring of Protected Visual Environment (IMPROVE;Malm et al., 2004) extensive ground-based networks; and the Acid Deposition Monitoring Network in East Asia(EANET, 2019). For each observation site, model data are selected only when the monthly observation is available. To maximize the amount of data available for comparison but also ensure the climatological representativeness of the observation, all 12 monthly means are required to calculate an annual mean, as in Mulcahy et al. (2020).

Fig.6. Annual mean near-surface zonal wind (UAS) in (a) BCC-ESM1 and (b) ERA5 from 1986 to 2005. (c) The UAS biases. (d–f) As in (a–c) but for the near-surface meridional wind (VAS). Units: m s?1.

The annual mean surface sulfate concentrations from 274 monitoring sites are calculated and shown in Fig.7a. Surface sulfate concentrations at EMAP, EANET and IMPROVE sites are about 3.33 μg m, 3.20 μg mand 1.35 μg m, respectively. Generally, sulfate is higher in southern Europe than in northern Europe and higher in the east of the U.S. than in the west. BCC-ESM1 tends to overestimate the observations by about 0.95 μg mbut shows a high spatial correlation of 0.77, significant at the 1% level using a Student’s t-test (Fig.7b). The surface sulfate concentrations are comparable with the observation at the EMEP sites (3.18 μg m) but overestimated by approximately 2 μg mat the EANET and IMPROVE sites. BCC-ESM1 can capture the observed sulfate centers well over the eastern U.S., southern Europe, and Japan.

Regional statistical evaluations are shown in Fig.8. Considering the relatively low model resolution and the high heterogeneous distribution of sulfate, simulations falling with the range of 50% to 200% of the observations (gray dashed lines in Fig.8) are considered to be reasonable. Over the EMEP sites (Europe), the model shows a significant correlation of 0.66, with a slight underestimation over the southern EMEP sites, where the measured sulfate concentration is relatively large. Over EANET sites (East Asia), the comparison shows a high degree of scattering, with a low correlation with the observation of 0.39. Over the western U.S., the sulfate concentration is about twice that of the observation.As one of the major emission centers, the eastern U.S. is more polluted than the west. Among the four regions, the simulations perform best over the eastern U.S., with the highest correlation coefficient with the observation (0.87) and the smallest root-mean-square error (RMSE; 1.04 μg m). Generally, the performance of BCC-ESM1 in reproducing the surface sulfate concentration is comparable to instrumental measurements, and its performance is comparable with that of UKESM1 (Mulcahy et al., 2020). The major discrepancy is that BCC-ESM1 tends to overestimate sulfate levels over the western U.S. and East Asia, which may partly contribute to the regional cooling biases as shown in Fig.4c.

Fig.7. (a) Annual mean surface sulfate concentration in 2000–09 at 274 monitoring sites in the three ground-based networks: EMEP in Europe, IMPROVE in the U.S., and EANET in East Asia. (b) Corresponding simulation in BCCESM1. All 12 monthly means are required to calculate annual mean results. Units: μg m?3.

Fig.8. Comparison of simulated annual mean SO42? from BCC-ESM1 against ground-based measurements from (a)EMEP (Europe), (b) EANET (East Asia), (c) West IMPROVE(western U.S.), and (d) East IMPROVE (eastern U.S.)networks. Observations and model data cover the years 1978 to 2009 for EMEP, 2000 to 2014 for EANET, and 1990 to 2014 for IMPROVE. NMB stands for normalized mean bias,RMSE for root-mean-square error, and COR for correlation coefficient. Linear regression statistics are also included.Units: μg m?3.

Figure 9 shows the evolution of the simulated and observed annual mean sulfate concentration, averaged across all the available ground-based observations, for all available years in the four typical regions. The uncertainty is calculated as ±1 standard deviation of the observed annual means across all the measurement sites. Simulations are suggested to be comparable with observations if the evolutions of the mean series of all sites are similar and the bands of uncertainty ranges are well overlapped. Generally, despite the systematic overestimation in East Asia and the western U.S., BCC-ESM1 can reproduce the changes in sulfate concentration at regional scales reasonably well.

Over Europe (EMEP), the sulfate concentration decreases by about 4 μg mfrom 1978 to 2009. BCC-ESM1 captures the reduction and the changes are well within the observed uncertainty range, albeit the negative trend is slightly smaller. This underestimation of the annual mean surface sulfate concentration over Europe is also evident in UKESM1 simulations (Mulcahy et al., 2020) and is suggested to be related to a low bias in wintertime in UKESM1(Turnock et al., 2015).

Over EANET sites, the surface sulfate concentration increases slightly with large interannual variability and uncertainties (about 3–4 μg m). It increases in the early 2000s,decreases by about 1.5 μg mfrom 2006 to 2012, and then bounces back afterward. The model generally tracks the observation quite well, despite a systematic overestimation by about 2 μg m. The simulation also shows large variations across all the measurement locations. Note that the EANET sites here are mostly in Japan and the southern countries.China is also a main sulfate emissions center with large spatial heterogeneity. Local evaluation over China will be interesting when high coverage monitoring data are available.

Fig.9. Time series of annual mean observed (red lines) and simulated (blue lines) sulfate concentrations averaged across all measurement locations in each network for a particular year. Shaded areas show ±1 standard deviation of the observed and modeled annual mean values across all the measurement locations. Units: μg m?3.

Over the western U.S., the sulfate concentration is relatively small with little to no trend. The simulated sulfate concentration is about twice that of the observation with a small decreasing trend. This region is generally remote from source regions and differences may be attributable to the biases in transportation from the source region, oxidants, or removal rates (Mulcahy et al., 2020).

Over the eastern U.S., one of the major anthropogenic source regions, a significant negative trend is observed due to reductions in emissions. Although the model generally overestimates concentrations by about 0.8 μg m, the decreasing rate and the uncertainty range in BCC-ESM1 show reasonable agreement with the observations.

4.Summary and relevant studies

This paper summarizes the experiments and datasets of BCC-ESM1, specifically for the CMIP6-endorsed AerChem-MIP. The basic performance of BCC-ESM1 is evaluated,revealing that: the climate sensitivities in BCC-ESM1 are well within the likely ranges suggested by IPCC AR5; SAT and precipitation can be reasonably represented (the biases of which are discussed); and the present-day surface sulfate concentrations also compare well with observations, including their spatial structure and regional trends.

BCC-ESM1 is one of the few ESMs participating in CMIP6 with interactive atmospheric chemistry, and its Aer-ChemMIP outputs have been widely used in recent atmospheric chemistry studies (i.e., Allen et al., 2020; Griffiths et al., 2020; Thornhill et al., 2020; Turnock et al., 2020).For example, in the special issue for AerChemMIP, BCCESM1 data are used to estimate the climate and air quality impact of mitigation policies. Turnock et al. (2020) used the outputs from BCC-ESM1 and nine other ESMs to make a first assessment of the impact on historical and future changes to air pollutants (Oand PM). CMIP6 models show a consistent overestimation of surface Oconcentration but an underestimation of surface PMwith a large diversity in the historical period. For scenario ssp370,which encompasses the highest levels of short-lived climate pollutants without climate policy, both surface Oand PMincrease across most regions. In BCC-ESM1, surface Ochanges are about 3 ppbv higher than the multimodel mean value for the historical period and the surface Oprediction is large in scenario ssp370. The large Oresponse in BCC-ESM1 may be related to the absence of NOtitration,which will reduce surface Oconcentrations. Allen et al.(2020) quantified the 2015–2055 climate and air quality effects of non-methane NTCFs with “weak” and “strong”levels of air quality controls. Their findings suggest air quality improvement with non-methane NTCF reductions, but additional surface warming and wetting may occur, particularly in Asia and the Arctic. Generally, results in BCCESM1 are consistent with the seven other ESMs participating in CMIP6, both in terms of global mean climate anomalies and air pollution anomalies due to NTCF mitigation.

5.Usage notes

The variables of Priority 1 in AerChemMIP have been generated, including meteorological variables at various time scales (monthly, daily, six-hourly, and hourly) and chemical variable outputs at the monthly scale. All the variables are listed at http://clipc-services.ceda.ac.uk/dreq/tabs03/expt_AerChemMIP_AerChemMIP_1_1.html. For brevity,we summarize the chemical outputs in Table 2, including the volume mixing ratios of gas-phase chemical species in the atmosphere, the mass mixing ratios of aerosols, atmospheric optical thickness, etc.

The horizontal resolution of all the variables is T42(approximately 280 km). Most of the 3D atmospheric outputs are on model atmosphere levels (26 levels in a hybrid sigma-pressure vertical coordinate system). To facilitate multimodel comparisons, 11 atmospheric variables are interpolated to 19 fixed pressure levels for monthly outputs, and 6 of them are interpolated to 8 fixed pressure levels for daily outputs (Table 3).

Acknowledgements.

This work was jointly supported by the National Key Research and Development Program of China (Grant No. 2016YFA0602103), the National Key Research and Development Program of China CERC-WET Project (Grant No.2018YFE0196000) and the National Natural Science Foundation of China (Grant No. 41805063).

Table 2. Chemical outputs of BCC-ESM1 in CMIP6. Variables marked with an asterisk (*) are only provided for the first member of historical experiments.

Table 2. (Continued.)

Table 3. Monthly outputs of 11 variables that have been interpolated to 19 fixed pressure levels at 1000, 925, 850, 700, 600, 500, 400,300, 250, 200, 150, 100, 70, 50, 30, 20, 10, 5, and 1 hPa. Daily outputs of 6 variables have been interpolated to 8 fixed pressure levels at 1000, 850, 700, 500, 250, 100, 50, and 10 hPa.

Data availability statement

The data for the BCC-ESM1 simulations (Zhang et al.,2019) are available on the ESGF nodes (https://esgfnode.llnl.gov/projects/cmip6/). The dataset is in Network Common Data Form (NetCDF), version 4.

Disclosure statement

No potential conflicts of interest are reported by the authors.

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source,provide a link to the Creative Commons license, and indicate if changes were made.