Representing mixing state of black carbon (BC) is challenging for global climate models (GCMs). The Community Atmosphere Model version 6 (CAM6) with the four-mode version of the Modal Aerosol Module (MAM4) represents aerosols as fully internal mixtures with uniform composition within each aerosol mode, resulting in high degree of internal mixing of BC with non-BC species and large mass ratio of coating to BC (R_BC, the mass ratio of non-BC species to BC in BC-containing particles). To improve BC mixing state representation, we couple a machine learning (ML) model of BC mixing state index into CAM6 to develop a ML-based BC mixing state parameterization (MAM4-ML). In MAM4-ML, we use R_BC to partition accumulation mode particles into two new modes, BC-free particles and BC-containing particles. We adjust R_BC to make the modeled BC mixing state index (χ_mode) match the one predicted by the ML model (χ_ML). On a global average, the χ_mode decreases from 78% (MAM4-default) to 63% (MAM4-ML, 19% reduction) and agrees well with χ_ML (66%). The R_BC decreases by 52% for accumulation mode and better agrees with observations. The decrease in R_BC leads to a 9% decrease in BC-containing particles in accumulation mode and a 20% decrease in BC activation, globally. Moreover, the global mean BC mass absorption cross section (MAC) decreases by 16% and better agrees with observations, leading to a 16% decrease in global BC direct radiative effect (DRE).

China has implemented strict emission control measures, but it’s unclear how they affect BC aging and light absorption capacity. We then use the hybrid CAM6 with MAM4-ML to simulate BC aging under China’s recent clean air actions (2011-2018) and the future carbon neutrality scenario (2050/2100, SSP1-2.6). During 2011-2018, the R_BC widely increased (5.4% yr^-1) over the east of China. The increased secondary organic aerosol (SOA) coatings dominate (88%) the increased R_BC, while sulfate coatings decrease. The drivers of BC coating changes come from the different magnitudes of emission reductions in secondary aerosol precursors (i.e., VOCs and SO₂) and BC. During 2011-2018, the increased R_BC enhances BC mass absorption cross section (MAC, 0.7% yr^-1). In 2050/2100 for SSP1-2.6, emission control leads to further increased R_BC (95%/145%) and BC MAC (12%/17%). For both 2011-2018 and 2050/2100, the enhanced BC MAC partly offsets the declining direct radiative effect (DRE) of BC due to direct emission reduction. As a result, the full impact of direct emission reductions of BC on BC DRE is only 75% for 2011-2018 and 90%/94% for 2050/2100.

We highlight that coupling ML into GCMs is an advanced approach for improving key aerosol process modeling and for assessing the implications of emission control policies for BC climate effects.

黑碳,混合态,气候模式,机器学习,减排