Aerosol Effective Radiative Forcing in the Online Aerosol Coupled CAS-FGOALS-f3-L Climate Model

The effective radiative forcing (ERF) of anthropogenic aerosol can be more representative of the eventual climate response than other radiative forcing. We incorporate aerosol–cloud interaction into the Chinese Academy of Sciences Flexible Global Ocean–Atmosphere–Land System (CAS-FGOALS-f3-L) by coupling an existing aerosol module named the Spectral Radiation Transport Model for Aerosol Species (SPRINTARS) and quantified the ERF and its primary components (i.e., effective radiative forcing of aerosol-radiation interactions (ERFari) and aerosol-cloud interactions (ERFaci)) based on the protocol of current Coupled Model Intercomparison Project phase 6 (CMIP6). The spatial distribution of the shortwave ERFari and ERFaci in CAS-FGOALS-f3-L are comparable with that of most available CMIP6 models. The global mean 2014–1850 shortwave ERFari in CAS-FGOALS-f3-L (−0.27 W m−2) is close to the multi-model means in 4 available models (−0.29 W m−2), whereas the assessing shortwave ERFaci (−1.04 W m−2) and shortwave ERF (−1.36 W m−2) are slightly stronger than the multi-model means, illustrating that the CAS-FGOALS-f3-L can reproduce the aerosol radiation effect reasonably well. However, significant diversity exists in the ERF, especially in the dominated component ERFaci, implying that the uncertainty is still large.


Introduction
Aerosols impact the global climate by changing the Earth's radiation budget, which can not only scatter and absorb solar radiation directly [1], but also modify the cloud macro and micro physical properties by serving as the cloud condensation nuclei (CCN) and ice nuclei (IN) to indirectly perturb the Earth's radiation budget [2,3]. In addition, absorbing aerosols deposited on the ice and snow surface can also change the albedo of the Earth's surface [4]. Due to these comprehensive aerosol effects, there exist complex non-liner mechanisms, leading to the aerosol particles influencing radiative where ∆ is the difference between present day and preindustrial day, ERFari is the shortwave effective radiative forcing (ERF) of aerosol-radiation interactions (ari) or the shortwave direct radiative forcing, ERFaci is the shortwave effective radiative forcing (ERF) of aerosol-cloud interactions (aci) or the shortwave cloud radiative forcing under clean sky condition, and ∆SRE SW is the surface albedo radiative forcing. The three components are further calculated as follows: where F is the net shortwave radiation flux at TOA, F clean is the net shortwave flux at TOA under clean sky condition, F clean,clear is the net shortwave flux at TOA under clean clear sky condition. The diagnosis method of longwave effective radiative forcing (ERF LW ) is same as ERF SW . Following the simulation specification of AerChemMIP [11] and RFMIP [8], ERF is estimated using fixed SSTs and sea ice simulations, which are provided by the model's own preindustrial climatology data. Present day and preindustrial day refer to the years 2014 and 1850. Greenhouse gas concentrations are prescribed using 1850 climatology of CMIP6. The emissions of anthropogenic aerosols are taken from Hoesly et al. 2018 [40]. The only difference of two simulation experiments is the emission of anthropogenic aerosols. Each experiment is run for 30 years with the appropriate resolution of 200 km, and the annual means are used for analysis.

Results
We focus on assessing the effective radiative forcing of anthropogenic aerosol, which includes two main components, namely the ERFari and ERFaci. Before the discussion of each component, we first analysis the context related to the corresponding forcing component for a better understanding of how aerosols impact the climate system, such as the sulfate and carbonaceous aerosol optical depth, which related to the ERF ari , and the cloud micro and macro physics properties, which related to the ERF aci .

Aerosol Optical Depth
Aerosols can scatter and absorb solar radiation directly. The aerosol optical depth (AOD) is an important aerosol optical property to measure how aerosols extinct the solar radiation in an atmospheric column. Figure 1 shows the comparison of spatial distribution of the modelled annual average AOD in sulfate and carbonaceous aerosol between the years 2014 and 1850. Obviously, the sulfate and carbonaceous AOD increased significantly in the year 2014 relative to the year 1850, which can be attributed to the large emissions of anthropogenic aerosols (e.g., BC, OC) and sulfate precursor (e.g., SO 2 ). The high values of sulfate AOD (>0.2) are mainly located in East Asia, South Asia, South Africa, and their downwind direction in the year 2014 (Figure 1a), corresponding to the areas with the strong increased sulfate AOD (Figure 1c). The high values (>0.2) of carbonaceous AOD can be mainly found in East Asia, South Asia, Central Africa and their downwind direction in the year 2014 (Figure 1d), and East Asia and South Asia contribute the most in the increased carbonaceous AOD (Figure 1f). For the natural aerosols (i.e., dust and sea salt), their distributions of AOD values change little (not show), whereas the dust global mean AOD decreases about 2% in the year 2014. This very weak decrease is also found in the Grandey et al. 2018 [46], which may be due in part to there is relatively little change in the surface temperature with the fixed SSTs, so there would not be expected to be much of a change in dust emission [47]. In sum, the total AOD (global mean, 0.111) in the year 2014 increases about 39% (0.031) relative to the year 1850 (global mean, 0.08). The sulfate AOD contributes to about 71% (0.022), and the carbonaceous AOD contributes to the remainder 29% (0.009) (Figure 1c,f,i). The increased species AOD concentrates most in the Northern Hemisphere (NH), especially in East Asia and South Asia. . The zonal means plot is located in the right of each subplot. The global mean value is shown the top right corner of each map, the second row (d, e, f) is same as the first row but for carbonaceous AOD (including pure black carbon and pure organic carbon, internal mixture black carbon and organic carbon), and the third row (g, h, i) is same as the first row but for the sum of sulfate and carbonaceous aerosols AOD. SU = sulfate, CA = carbonaceous aerosol, = the AOD difference between the years 2014 and 1850. Figure 2 shows the comparison of shortwave direct radiative forcing between CAS-FGOALS-f3-L and other 5 CMIP6 model results (All defined model abbreviations are summarized in Table 1 note ) . As can be seen, the global distributions of shortwave ERFari in CAS-FGOALS-f3-L (This study) are closely relative to distribution of the 2014-1850 increased AOD (Figure 1i), as expected. It is noted that the absorbing aerosols (i.e., BC) exert the "warming effect", leading to a positive shortwave ERFari, which can partly offset the "cooling effect" of scattering aerosols (i.e., sulfate and organic carbon aerosol). To understand ERFari differences, a comparison of single scatter albedo (SSA) assumed for anthropogenic aerosol might be interesting. However, due to lack of SSA in CMIP6 models, Figure S1 (only shows the changes of SSA in CAS-FGOALS-f3-L. A weak change of SSA in East Asia mainly caused by the offset of scattering aerosol and absorbing aerosol. Therefore, although the East Asia exists the strongest extinction (Figure 1i), the "cooling effect" seems not too strong in CAS-FGOALS-f3-L, even resulting in some models (i.e., GFDL-ESM4 and UKESM1-0-LL) including the "warming effect," namely positive direct radiative forcing. In general, the global distributions of shortwave ERFari in CAS-FGOALS-f3-L are consistent with that in other models, whereas the radiation cooling intensity of aerosols in East Asia (EA) and South Asia (SA) (CAS-FGOALS-f3-L, EA:  . The zonal means plot is located in the right of each subplot. The global mean value is shown the top right corner of each map, the second row (d-f) is same as the first row but for carbonaceous AOD (including pure black carbon and pure organic carbon, internal mixture black carbon and organic carbon), and the third row (g-i) is same as the first row but for the sum of sulfate and carbonaceous aerosols AOD. SU = sulfate, CA = carbonaceous aerosol, = the AOD difference between the years 2014 and 1850. Figure 2 shows the comparison of shortwave direct radiative forcing between CAS-FGOALS-f3-L and other 5 CMIP6 model results (All defined model abbreviations are summarized in Table 1 note). As can be seen, the global distributions of shortwave ERFari in CAS-FGOALS-f3-L (This study) are closely relative to distribution of the 2014-1850 increased AOD (Figure 1i), as expected. It is noted that the absorbing aerosols (i.e., BC) exert the "warming effect", leading to a positive shortwave ERFari, which can partly offset the "cooling effect" of scattering aerosols (i.e., sulfate and organic carbon aerosol). To understand ERFari differences, a comparison of single scatter albedo (SSA) assumed for anthropogenic aerosol might be interesting. However, due to lack of SSA in CMIP6 models, Figure S1 (only shows the changes of SSA in CAS-FGOALS-f3-L. A weak change of SSA in East Asia mainly caused by the offset of scattering aerosol and absorbing aerosol. Therefore, although the East Asia exists the strongest extinction (Figure 1i), the "cooling effect" seems not too strong in CAS-FGOALS-f3-L, even resulting in some models (i.e., GFDL-ESM4 and UKESM1-0-LL) including the "warming effect," namely positive direct radiative forcing. In general, the global distributions of shortwave ERFari in CAS-FGOALS-f3-L are consistent with that in other models, whereas the radiation cooling intensity of MRI-ESM2-0, and UKESM1-0-LL respectively. The CNRM_CM6-1 lacks apparent aerosol absorption (strong cooling near continental sources), whereas in GFDL-ESM4, the aerosol absorption is too strong, which is possibly due to more elevated absorbing aerosol above the clouds. In addition, since the outlier is the GFDL-ESM4 with an unlikely strong positive absorption (Figure 2d), the value of its shortwave ERFari are removed in statistical calculations (Table 1). Therefore, the multi-model mean value from the other four CMIP6 models is −0.29 W m −2 , with a small standard deviation (0.1) ( Table 1).   The results of ERFari and ERF in GFDL-ESM4 are removed in statistical calculations due to its unlikely strong positive shortwave ERFari; The results of ERFaci and ERF in MRI-ESM2-0 are removed due to its strong ERFari with added aerosol/ice-cloud interactions; Model NorESM2-LM and NorESM2-MM are also excluded in statistical calculations due to unreasonable results of ERFaci and ERFari. All available model results of CMIP6 are download from https://esgf-node.llnl.gov/search/cmip6/, and the calculation methods are same as Ghan et al. 2013 [45].

Aerosol-Radiation Interactions
Atmosphere 2020, 11, 1115 7 of 15 lacks apparent aerosol absorption (strong cooling near continental sources), whereas in GFDL-ESM4, the aerosol absorption is too strong, which is possibly due to more elevated absorbing aerosol above the clouds. In addition, since the outlier is the GFDL-ESM4 with an unlikely strong positive absorption (Figure 2d), the value of its shortwave ERFari are removed in statistical calculations (Table  1). Therefore, the multi-model mean value from the other four CMIP6 models is −0.29 W m −2 , with a small standard deviation (0.1) ( Table 1).   Figure 3 shows the comparison of the main cloud micro and macro physical properties for the year 2014 and year 1850. The aerosol particles, especially the hydrophilic particles (e.g., sulfate, OC, sea salt), can act as the cloud condensation nuclei (CCN) to impact the cloud properties. Therefore, we first show the surface CCN number concentration at a fixed supersaturation of 0.1% to look at how much CCN changes in the year 2014 relative to the year 1850, although we notice that it is not necessarily a very good indicator to reveal the relevant CCN number concentration in the real atmosphere [48]. The locations of the large increase in CCN number concentration (Figure 3c) are similar with the increase of the sulfate and carbonaceous AOD (Figure 1i). The global mean surface CCN number concentration is 69.3 cm −3 , which is slightly lower than that of the result of 97.8 cm −3 from Liu et al. 2012 [49] using the modal aerosol module (MAM7). The value of percentage increase is 101% (Figure 3c), which is close to the result of 115% from Grandey et al. 2018 [46], although their value was taken from the level of 860 hPa.  (Figure 1i), mainly due to the increase of sulfate and carbonaceous aerosol drive the increases of CDNC. The global mean cloud droplet effective radius at the cloud top above 273 K reduces from 9.9 µm to 9.2 µm (Figure 3g-i), which can be mainly attributed to the significant increases of CDNC when the cloud liquid water path (LWP) is assumed constant (i.e., Twomey effect), although the small increases exist in LWP (Figure 3l). Most of the decreased regions are located in the midlatitude of the Northern Hemisphere, especially in the regions emitting the anthropogenic aerosol heavily and their downwind direction. The weak increase of LWP (Figure 3l) is possibly caused by the decrease of precipitation rate (refers to the warm cloud precipitation, Figure S2, Supplementary Materials), which is related to the aerosol second indirect effect. Generally, an increase of the CDNC number concentration drives a decrease of precipitation rate, which are represented in the CAS-FGOALS-f3-L model using the Berry-type autoconversion scheme [43,44], although we note that the nonlinear CCN-precipitation mechanism may possibly more complicated in reality. In addition, cloud fraction (CF) has been shown Atmosphere 2020, 11, 1115 8 of 15 to have a potentially large impact on the ERFaci, a small increase in CF (0.2%, Figure 3o), especially in the low CF (0.3%, Figure S3i) are found in our model.

Aerosol-Cloud Interactions
The cloud, which accounts for about 67% of the Earth surface, can scatter lots of the solar radiation directly, exerting a strong "cooling effect" in the climate system. Assuming that anthropogenic aerosol only contributes by the fine-mode sizes (those smaller than 1 μm) and that only indirect effect through low altitude water clouds matter. Therefore, only shortwave radiative effects are discussed here; for climate implications, longwave effects are ignored. Figure 4 shows the

Aerosol-Cloud Interactions
The cloud, which accounts for about 67% of the Earth surface, can scatter lots of the solar radiation directly, exerting a strong "cooling effect" in the climate system. Assuming that anthropogenic aerosol only contributes by the fine-mode sizes (those smaller than 1 µm) and that only indirect effect through low altitude water clouds matter. Therefore, only shortwave radiative effects are discussed here; for climate implications, longwave effects are ignored. Figure 4 shows the diagnosis results of the shortwave ERFaci between CAS-FGOALS-f3-L and the other five CMIP6 models results. Generally, the increase in anthropogenic aerosols causes an increase in the cloud droplets number concentration and the cloud lifetime, leading to a higher cloud albedo and consequent negative radiation forcing. The distributions of shortwave ERFaci in CAS-FGOALS-f3-L, which are consistent with that of the other five available CMIP6 models, are closely associated with changes in column-integrated cloud droplet number concentration (CDNC), cloud drop effective radius, liquid water path (LWP), and cloud fraction (CF). The strong negative radiation forcing areas mainly locate in the East Asia diagnosis results of the shortwave ERFaci between CAS-FGOALS-f3-L and the other five CMIP6 models results. Generally, the increase in anthropogenic aerosols causes an increase in the cloud droplets number concentration and the cloud lifetime, leading to a higher cloud albedo and consequent negative radiation forcing. The distributions of shortwave ERFaci in CAS-FGOALS-f3-L, which are consistent with that of the other five available CMIP6 models, are closely associated with changes in column-integrated cloud droplet number concentration (CDNC), cloud drop effective radius, liquid water path (LWP), and cloud fraction (CF For the ice-cloud shortwave ERFaci, a Twomey-like effect may increase the albedo of ice cloud in the previous studies [50], whereas a strong variation in the response of ice cloud amount to aerosol still exist in the models, leaving the overall magnitude of the forcing uncertain [51]. In this study, the MRI-ESM2-0 model shows the strongest negative shortwave ERFaci (−2.45 W m −2 ), which may be partly attributed to it allows the interaction between aerosols and ice clouds [52]. Therefore, the ERFaci in MRI-ESM2-0 is excluded in statistical calculations (Table 1). In sum, the assessing shortwave ERFaci in CAS-FGOALS-f3-L (−1.04 W m −2 ) is slightly stronger than the multi-model average value from these CMIP6 models (−0.78 W m −2 ), with a small standard deviation (0.18) ( Table  1). The large diversities still exist in the local area, such as East Asia, South Asia, and Europe ( Figure  4), implying that the interaction between aerosols and clouds still needs to further research to reduce this uncertainty.   For the ice-cloud shortwave ERFaci, a Twomey-like effect may increase the albedo of ice cloud in the previous studies [50], whereas a strong variation in the response of ice cloud amount to aerosol still exist in the models, leaving the overall magnitude of the forcing uncertain [51]. In this study, the MRI-ESM2-0 model shows the strongest negative shortwave ERFaci (−2.45 W m −2 ), which may be partly attributed to it allows the interaction between aerosols and ice clouds [52]. Therefore, the ERFaci in MRI-ESM2-0 is excluded in statistical calculations (Table 1). In sum, the assessing shortwave ERFaci in CAS-FGOALS-f3-L (−1.04 W m −2 ) is slightly stronger than the multi-model average value from these CMIP6 models (−0.78 W m −2 ), with a small standard deviation (0.18) ( Table 1). The large diversities still exist in the local area, such as East Asia, South Asia, and Europe (Figure 4), implying that the interaction between aerosols and clouds still needs to further research to reduce this uncertainty. Excluding the outlier among these models (i.e., BCC-ESM1), the modeled global mean ERF SW value (−1.36 W m −2 ) is slightly stronger than the multi-model average value from CMIP6 models (−1.07 W m −2 ) ( Table 1), illustrating that the new coupled model can reproduce the aerosol radiation effect reasonably. The BCC-ESM1 model shows the strongest negative ERF SW (−3.9 W m −2 ), possibly due in part to it uses a simple empirical function to represent the aerosol active process [53], which is more than five times larger relative to the minimum (GFDL-ESM4, −0.59 W m −2 ) value of ERF SW among CMIP6 models, implying that the diversity in CMIP6 models still large. In addition, it is interesting that CAS-FGOALS-f3-L and MIROC6 applying the same SPRINTARS aerosol module, have a similar pattern but the ERF SW in CAS-FGOALS-f3-L (−1.36 W m −2 ) is slightly weaker than that of MIROC6 (−1.61 W m −2 ) (Figure 5a,m), which is possibly caused by the difference of carrier model. In order to maintain completeness of Formula (5), we also list the three components of the shortwave and longwave ERF ( Table 1). The longwave ERF (+0.2 W m −2 ) in AR5 is based on expert judgement, whereas Heyn et al. 2017 [10] found that the ERF in the terrestrial spectrum may be smaller in CMIP5 models. In this study, the assessing longwave ERF in CAS-FGOALS-f3-L is +0.09 W m −2 , which is slightly smaller than the CMIP6 multimodel means (+0.12 W m −2 ) and the AR5 results. In addition, it can be seen that the ERF is mainly dominated by the ERFaci, the ERFari second, and the SRF is the smallest. The ERFaci in CAS-FGOALS-f3-L is −1.1 W m −2 , which is slightly stronger relative to the previous observation-based studies (−1.0 to −0.2 W m −2 ) [51]. It is not clear that whether the observation-based estimates are more reliable with their own sources of uncertainty [51]. For the overall uncertainty in ERF, Bellouin et al. [54] recently provided a complete estimate in ERF (−2.0 to −0.35 W m −2 ) with a 90% likelihood based on models, theory, and observations. The ERF in CAS-FGOALS-f3-L (−1.27 W m −2 ) is in their estimate ranges, however, the model results may be better agree due to offsetting errors or because of same critical possibly false. The uncertainty of the ERF is still large, especially for the ERFaci.

Effective Radiative Forcing
Atmosphere 2020, 11, x FOR PEER REVIEW 12 of 17 effect reasonably. The BCC-ESM1 model shows the strongest negative ERFSW (−3.9 W m −2 ), possibly due in part to it uses a simple empirical function to represent the aerosol active process [53], which is more than five times larger relative to the minimum (GFDL-ESM4, −0.59 W m −2 ) value of ERFSW among CMIP6 models, implying that the diversity in CMIP6 models still large. In addition, it is interesting that CAS-FGOALS-f3-L and MIROC6 applying the same SPRINTARS aerosol module, have a similar pattern but the ERFSW in CAS-FGOALS-f3-L (−1.36 W m −2 ) is slightly weaker than that of MIROC6 (−1.61 W m −2 ) (Figure 5a,m), which is possibly caused by the difference of carrier model. In order to maintain completeness of Formula (5), we also list the three components of the shortwave and longwave ERF ( Table 1). The longwave ERF (+0.2 W m −2 ) in AR5 is based on expert judgement, whereas Heyn et al. 2017 [10] found that the ERF in the terrestrial spectrum may be smaller in CMIP5 models. In this study, the assessing longwave ERF in CAS-FGOALS-f3-L is +0.09 W m −2 , which is slightly smaller than the CMIP6 multimodel means (+0.12 W m −2 ) and the AR5 results. In addition, it can be seen that the ERF is mainly dominated by the ERFaci, the ERFari second, and the SRF is the smallest. The ERFaci in CAS-FGOALS-f3-L is −1.1 W m −2 , which is slightly stronger relative to the previous observation-based studies (−1.0 to −0.2 W m −2 ) [51]. It is not clear that whether the observation-based estimates are more reliable with their own sources of uncertainty [51]. For the overall uncertainty in ERF, Bellouin et al. [54] recently provided a complete estimate in ERF (−2.0 to −0.35 W m −2 ) with a 90% likelihood based on models, theory, and observations. The ERF in CAS-FGOALS-f3-L (−1.27 W m −2 ) is in their estimate ranges, however, the model results may be better agree due to offsetting errors or because of same critical possibly false. The uncertainty of the ERF is still large, especially for the ERFaci.

Discussion
The CAS-FGOALS-f3-L can only reproduce direct aerosols radiation effect offline and missing the interactions between aerosols and cloud in the previous version. In this paper, we incorporate the

Discussion
The CAS-FGOALS-f3-L can only reproduce direct aerosols radiation effect offline and missing the interactions between aerosols and cloud in the previous version. In this paper, we incorporate the aerosol-cloud interaction into the Chinese Academy of Sciences Flexible Global Ocean-Atmosphere-Land System (CAS-FGOALS-f3-L) by coupling an existing aerosol module named Spectral Radiation Transport Model for Aerosol Species (SPRINTARS), which can more realistically represent the climate impacts of aerosols. A pronounced feature is that the atmospheric component named version 2 of the Finite-volume Atmospheric Model (FAMIL2) can perform a higher resolution aerosol simulation with the updated dynamic core to reproduce the microphysical processes at smaller scales and study the aerosol transportation in the complex terrain areas, such as the Tibet Plateau. The aerosol-radiation interaction and the aerosol-cloud interaction have been preliminary realized in our model, giving a chance to study the aerosol climate effect. Although the global mean and distribution of aerosol direct radiative forcing, cloud radiative forcing, and effective radiative forcing are comparable to other models in CMIP6, the challenges still exist.
On the one hand, it is definitely clear that the two moments aerosol and cloud parameterized scheme (i.e., mass and number concentration of aerosol particles and cloud drops are both predicted), can be more realistically to character the aerosol-cloud interaction [55]. Currently, however, the aerosol module SPRINTARS and CAS-FGOALS-f3-L atmospheric component FAMIL2 are still using the single moment scheme to predict the mass concentration and diagnosis the number concentration with pre-prescribed size distribution, this may hamper us to further dig out the important key factors related to aerosol-cloud interaction and bring a large uncertainty to assess the aerosol effective radiative forcing. Therefore, a two moments aerosol and cloud parameterized scheme are urgently required to be incorporated into the CAS-FGOALS-f3-L in the further.
On the other hand, as we know that the dust and black carbon particles can act as the ice nucleus to modify the properties of high cloud, however, the aerosol-cloud interaction in CAS-FGOALS-f3-L only considers the large scale warm stratus cloud, the aerosol-ice cloud and aerosol-mixed phase cloud interaction are both missing in the model. It is known that the second organic aerosol (SOA) contribute to a large portion in fine-mode aerosols, and the nitrate aerosols may important in local areas. These missing hygroscopic growth species can also participate the formation of cloud, perturbing the Earth's energy balance. In addition, the ERF also highly depends on the specific aerosol active scheme and cloud-rain autoconversion scheme, which related to the aerosol indirect effect. Therefore, exploring and refining these processes in future will allow us more precisely quantize how climate system "feel" the aerosol perturbation.

Conclusions
In this paper, we realize the aerosol-cloud interaction by incorporating the ARG (Abdul-Razzak and Ghan, [42]) aerosol activation scheme and the berry-type autoconversion scheme [43,44] into the CAS-FGOALS-f3-L which has participated the CMIP6, and further assess the effective radiation forcing (ERF) of anthropogenic aerosols and its decomposed components produced by the new coupled model CAS-FGOALS-f3-L based on the protocol of current CMIP6. The estimated ERF values and their spatial distributions are compared with the CMIP6 multi-model results. The main findings are summarized as follows: The shortwave ERFaci is closely associated with changes in column-integrated cloud droplet number concentration (CDNC), cloud drop effective radius, liquid water path (LWP), and cloud fraction (CF). The MRI-ESM2-0 model shows the strongest negative shortwave ERFaci (−2.45 W m −2 ), which may be partly attributed to being coupled to the aerosols-ice cloud interaction [52]. (4) The global mean 2014-1850 shortwave effective radiative forcing (ERF SW ) produced by CAS-FGOALS-f3-L (−1.36 W m −2 ) shows a stronger negative forcing relative to the multi-model mean value from 11 available CMIP6 models (−1.07 W m −2 ), and the pattern are also comparable, illustrating that the new coupled model can reasonably reproduce the aerosol radiation effect. The ERF is mainly dominated by the ERFaci, and the uncertainty of the ERF is still large, especially for the ERFaci.
These results provide an important reference to study the aerosol climate effect with the new coupling system and necessitate model improvements in some key processes, such as a two-moments aerosol and cloud microphysical scheme, adding SOA and nitrate aerosol, and explicitly character aerosol-ice cloud interaction, to better quantify the climate impacts of aerosols.