Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Article
  • Published:

Machine learning reveals climate forcing from aerosols is dominated by increased cloud cover

A Publisher Correction to this article was published on 17 August 2022

This article has been updated

Abstract

Aerosol–cloud interactions have a potentially large impact on climate but are poorly quantified and thus contribute a substantial and long-standing uncertainty in climate projections. The impacts derived from climate models are poorly constrained by observations because retrieving robust large-scale signals of aerosol–cloud interactions is frequently hampered by the considerable noise associated with meteorological co-variability. The 2014 Holuhraun effusive eruption in Iceland resulted in a massive aerosol plume in an otherwise near-pristine environment and thus provided an ideal natural experiment to quantify cloud responses to aerosol perturbations. Here we disentangle significant signals from the noise of meteorological co-variability using a satellite-based machine-learning approach. Our analysis shows that aerosols from the eruption increased cloud cover by approximately 10%, and this appears to be the leading cause of climate forcing, rather than cloud brightening as previously thought. We find that volcanic aerosols do brighten clouds by reducing droplet size, but this has a notably smaller radiative impact than changes in cloud fraction. These results add substantial observational constraints on the cooling impact of aerosols. Such constraints are critical for improving climate models, which still inadequately represent the complex macro-physical and microphysical impacts of aerosol–cloud interactions.

This is a preview of subscription content, access via your institution

Access options

Rent or buy this article

Prices vary by article type

from$1.95

to$39.95

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Comparison between ML-MODIS predictions and MODIS observations.
Fig. 2: Changes in cloud properties caused by the volcanic perturbation estimated using machine-learning predictions and MODIS observations for October 2014.
Fig. 3: Responses of cloud properties to the volcanic aerosol perturbation in October 2014.

Similar content being viewed by others

Data availability

The MODIS cloud and aerosol products from Aqua (MYD08_L3) and Terra (MOD08_L3) used in this study are available from the Atmosphere Archive and Distribution System Distributed Active Archive Center of National Aeronautics and Space Administration (LAADS-DAAC, NASA), https://ladsweb.modaps.eosdis.nasa.gov. ERA5 datasets are available from the European Centre for Medium-range Weather Forecast (ECMWF) archive, https://cds.climate.copernicus.eu. Source data are provided with this paper.

Code availability

Code is available from the corresponding author on reasonable request.

Change history

References

  1. Lohmann, U. & Feichter, J. Global indirect aerosol effects: a review. Atmos. Chem. Phys. 5, 715–737 (2005).

    Article  Google Scholar 

  2. L’Ecuyer, T. S., Hang, Y., Matus, A. V. & Wang, Z. Reassessing the effect of cloud type on Earth’s energy balance in the age of active spaceborne observations. Part I: top of atmosphere and surface. J. Clim. 32, 6197–6217 (2019).

    Article  Google Scholar 

  3. Latham, J. et al. Global temperature stabilization via controlled albedo enhancement of low-level maritime clouds. Phil. Trans. R. Soc. 366, 3969–3987 (2008).

    Article  Google Scholar 

  4. Chen, Y.-C., Christensen, M. W., Stephens, G. L. & Seinfeld, J. H. Satellite-based estimate of global aerosol–cloud radiative forcing by marine warm clouds. Nat. Geosci. 7, 643–646 (2014).

    Article  Google Scholar 

  5. Twomey, S. Pollution and the planetary albedo. Atmos. Environ. 8, 1251–1256 (1974).

    Article  Google Scholar 

  6. Albrecht, B. A. Aerosols, cloud microphysics, and fractional cloudiness. Science 245, 1227–1230 (1989).

    Article  Google Scholar 

  7. Boucher, O. et al. in Climate Change 2013: The Physical Science Basis (eds Stocker, T. F. et al.) 571–657 (Cambridge Univ. Press, 2013).

  8. Toll, V., Christensen, M., Quaas, J. & Bellouin, N. Weak average liquid-cloud–water response to anthropogenic aerosols. Nature 572, 51–55 (2019).

    Article  Google Scholar 

  9. Bellouin, N. et al. Bounding global aerosol radiative forcing of climate change. Rev. Geophys. 58, e2019RG000660 (2020).

  10. IPCC Climate Change 2021: The Physical Science Basis (eds Masson-Delmotte, V. et al.) (Cambridge Univ. Press, 2021).

  11. Rogelj, J., Forster, P. M., Kriegler, E., Smith, C. J. & Séférian, R. Estimating and tracking the remaining carbon budget for stringent climate targets. Nature 571, 335–342 (2019).

    Article  Google Scholar 

  12. Andreae, M. O., Jones, C. D. & Cox, P. M. Strong present-day aerosol cooling implies a hot future. Nature 435, 1187–1190 (2005).

    Article  Google Scholar 

  13. Seinfeld, J. H. et al. Improving our fundamental understanding of the role of aerosol−cloud interactions in the climate system. Proc. Natl. Acad. Sci. USA 113, 5781–5790 (2016).

    Article  Google Scholar 

  14. Ghan, S. et al. Challenges in constraining anthropogenic aerosol effects on cloud radiative forcing using present-day spatiotemporal variability. Proc. Natl. Acad. Sci. USA 113, 5804–5811 (2016).

    Article  Google Scholar 

  15. Malavelle, F. F. et al. Strong constraints on aerosol–cloud interactions from volcanic eruptions. Nature 546, 485–491 (2017).

    Article  Google Scholar 

  16. Kaufman, Y. J., Koren, I., Remer, L. A., Rosenfeld, D. & Rudich, Y. The effect of smoke, dust, and pollution aerosol on shallow cloud development over the Atlantic Ocean. Proc. Natl. Acad. Sci. USA 102, 11207–11212 (2005).

    Article  Google Scholar 

  17. McCoy, D. T. & Hartmann, D. L. Observations of a substantial cloud–aerosol indirect effect during the 2014–2015 Bárðarbunga-Veiðivötn fissure eruption in Iceland. Geophys. Res. Lett. 42, 409–410,414 (2015).

    Article  Google Scholar 

  18. Toll, V., Christensen, M., Gassó, S. & Bellouin, N. Volcano and ship tracks indicate excessive aerosol-induced cloud water increases in a climate model. Geophys. Res. Lett. 44, 492–412,500 (2017).

    Article  Google Scholar 

  19. Diamond, M. S., Director, H. M., Eastman, R., Possner, A. & Wood, R. Substantial cloud brightening from shipping in subtropical low clouds. AGU Adv. 1, e2019AV000111 (2020).

    Article  Google Scholar 

  20. Gryspeerdt, E. et al. Constraining the aerosol influence on cloud liquid water path. Atmos. Chem. Phys. 19, 5331–5347 (2019).

    Article  Google Scholar 

  21. Possner, A., Eastman, R., Bender, F. & Glassmeier, F. Deconvolution of boundary layer depth and aerosol constraints on cloud water path in subtropical stratocumulus decks. Atmos. Chem. Phys. 20, 3609–3621 (2020).

    Article  Google Scholar 

  22. Ackerman, A. S., Kirkpatrick, M. P., Stevens, D. E. & Toon, O. B. The impact of humidity above stratiform clouds on indirect aerosol climate forcing. Nature 432, 1014–1017 (2004).

    Article  Google Scholar 

  23. Stevens, B. & Feingold, G. Untangling aerosol effects on clouds and precipitation in a buffered system. Nature 461, 607–613 (2009).

    Article  Google Scholar 

  24. Lebo, Z. J. & Feingold, G. On the relationship between responses in cloud water and precipitation to changes in aerosol. Atmos. Chem. Phys. 14, 11817–11831 (2014).

    Article  Google Scholar 

  25. Koren, I., Dagan, G. & Altaratz, O. From aerosol-limited to invigoration of warm convective clouds. Science 344, 1143–1146 (2014).

    Article  Google Scholar 

  26. Seifert, A., Heus, T., Pincus, R. & Stevens, B. Large-eddy simulation of the transient and near-equilibrium behavior of precipitating shallow convection. J. Adv. Modeling Earth Syst. 7, 1918–1937 (2015).

    Article  Google Scholar 

  27. Mauger, G. S. & Norris, J. R. Meteorological bias in satellite estimates of aerosol–cloud relationships. Geophys. Res. Lett. https://doi.org/10.1029/2007GL029952 (2007).

  28. Kaufman, Y. J. & Koren, I. Smoke and pollution aerosol effect on cloud cover. Science 313, 655–658 (2006).

    Article  Google Scholar 

  29. Gryspeerdt, E., Quaas, J. & Bellouin, N. Constraining the aerosol influence on cloud fraction. J. Geophys. Res. Atmos. 121, 3566–3583 (2016).

    Article  Google Scholar 

  30. Rosenfeld, D. et al. Aerosol-driven droplet concentrations dominate coverage and water of oceanic low-level clouds. Science 363, eaav0566 (2019).

    Article  Google Scholar 

  31. Christensen, M. W., Jones, W. K. & Stier, P. Aerosols enhance cloud lifetime and brightness along the stratus-to-cumulus transition. Proc. Natl. Acad. Sci. USA 117, 17591–17598 (2020).

    Article  Google Scholar 

  32. Breen, K. H., Barahona, D., Yuan, T., Bian, H. & James, S. C. Effect of volcanic emissions on clouds during the 2008 and 2018 Kilauea degassing events. Atmos. Chem. Phys. 21, 7749–7771 (2021).

    Article  Google Scholar 

  33. Glassmeier, F. et al. Aerosol–cloud–climate cooling overestimated by ship-track data. Science 371, 485–489 (2021).

    Article  Google Scholar 

  34. Christensen, M. W. et al. Opportunistic experiments to constrain aerosol effective radiative forcing. Atmos. Chem. Phys. 22, 641–674 (2022).

    Article  Google Scholar 

  35. Bender, F. A. M., Frey, L., McCoy, D. T., Grosvenor, D. P. & Mohrmann, J. K. Assessment of aerosol–cloud–radiation correlations in satellite observations, climate models and reanalysis. Clim. Dyn. 52, 4371–4392 (2019).

    Article  Google Scholar 

  36. Fuchs, J., Cermak, J. & Andersen, H. Building a cloud in the southeast Atlantic: understanding low-cloud controls based on satellite observations with machine learning. Atmos. Chem. Phys. 18, 16537–16552 (2018).

    Article  Google Scholar 

  37. Oreopoulos, L., Cho, N. & Lee, D. A global survey of apparent aerosol–cloud interaction signals. J. Geophys. Res. Atmos. 125, e2019JD031287 (2020).

    Article  Google Scholar 

  38. Fan, J., Wang, Y., Rosenfeld, D. & Liu, X. Review of aerosol–cloud interactions: mechanisms, significance, and challenges. J. Atmos. Sci. 73, 4221–4252 (2016).

    Article  Google Scholar 

  39. Gettelman, A., Schmidt, A. & Egill Kristjánsson, J. Icelandic volcanic emissions and climate. Nat. Geosci. 8, 243–243 (2015).

    Article  Google Scholar 

  40. Oreopoulos, L., Cho, N., Lee, D. & Kato, S. Radiative effects of global MODIS cloud regimes. J. Geophys. Res. Atmos. 121, 2299–2317 (2016).

    Article  Google Scholar 

  41. Mastrandrea, M.D., et al. Guidance Note for Lead Authors of the IPCC Fifth Assessment Report on Consistent Treatment of Uncertainties. Intergovernmental Panel on Climate Change (IPCC). Available at https://www.ipcc.ch/site/assets/uploads/2017/08/AR5_Uncertainty_Guidance_Note.pdf (2010).

  42. Grist, J. P. et al. Extreme air–sea interaction over the North Atlantic subpolar gyre during the winter of 2013–2014 and its sub-surface legacy. Clim. Dyn. 46, 4027–4045 (2016).

    Article  Google Scholar 

  43. Grosvenor, D. P. et al. Remote sensing of droplet number concentration in warm clouds: a review of the current state of knowledge and perspectives. Rev. Geophys. 56, 409–453 (2018).

    Article  Google Scholar 

  44. Platnick, A. S. et al. MODIS Cloud Optical Properties: User Guide for the Collection 6/6.1 Level-2 MOD06/MYD06 Product and Associated Level-3 Datasets. https://atmosphere-imager.gsfc.nasa.gov/sites/default/files/ModAtmo/MODISCloudOpticalPropertyUserGuideFinal_v1.1_1.pdf (NASA, 2018).

  45. Platnick, S. et al. The MODIS cloud optical and microphysical products: collection 6 updates and examples from Terra and Aqua. IEEE Trans. Geosci. Remote Sens. 55, 502–525 (2017).

    Article  Google Scholar 

  46. Hubanks, P., Platnick, A. S., King, M. & Ridgway, B. MODIS Atmosphere L3 Gridded Product Algorithm Theoretical Basis Document (ATBD) & Users Guide. https://modis-images.gsfc.nasa.gov/_docs/L3_ATBD_C6.pdf (NASA, 2016).

  47. Maddux, B. C., Ackerman, S. A. & Platnick, S. Viewing geometry dependencies in MODIS cloud products. J. Atmos. Ocean. Technol. 27, 1519–1528 (2010).

    Article  Google Scholar 

  48. Quaas, J., Boucher, O., Bellouin, N. & Kinne, S. Satellite-based estimate of the direct and indirect aerosol climate forcing. J. Geophys. Res. Atmos. https://doi.org/10.1029/2007JD008962 (2008).

  49. Quaas, J., Boucher, O. & Lohmann, U. Constraining the total aerosol indirect effect in the LMDZ and ECHAM4 GCMs using MODIS satellite data. Atmos. Chem. Phys. 6, 947–955 (2006).

    Article  Google Scholar 

  50. Hersbach, H. et al. The ERA5 global reanalysis. Q. J. R. Meteorol. Soc. 146, 1999–2049 (2020).

    Article  Google Scholar 

  51. Andersen, H., Cermak, J., Fuchs, J., Knutti, R. & Lohmann, U. Understanding the drivers of marine liquid-water cloud occurrence and properties with global observations using neural networks. Atmos. Chem. Phys. 17, 9535–9546 (2017).

    Article  Google Scholar 

  52. Breiman, L. Random forests. Mach. Learn. 45, 5–32 (2001).

    Article  Google Scholar 

  53. Biau, G. & Scornet, E. A random forest guided tour. Test 25, 197–227 (2016).

    Article  Google Scholar 

  54. Grange, S. K., Carslaw, D. C., Lewis, A. C., Boleti, E. & Hueglin, C. Random forest meteorological normalisation models for Swiss PM10 trend analysis. Atmos. Chem. Phys. 18, 6223–6239 (2018).

    Article  Google Scholar 

  55. Shi, Z. et al. Abrupt but smaller than expected changes in surface air quality attributable to COVID-19 lockdowns. Sci. Adv. 7, eabd6696 (2021).

    Article  Google Scholar 

  56. Yang, J. et al. From COVID-19 to future electrification: assessing traffic impacts on air quality by a machine-learning model. Proc. Natl. Acad. Sci. USA 118, e2102705118 (2021).

    Article  Google Scholar 

  57. Chicco, D., Warrens, M. J. & Jurman, G. The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation. PeerJ Comput. Sci. 7, e623 (2021).

    Article  Google Scholar 

  58. Cutler, A., Cutler, D. R. & Stevens, J. R. in Ensemble Machine Learning (eds Zhang, C. & Ma, Y.), 157–175, (Springer, 2012); https://doi.org/10.1007/978-1-4419-9326-7_5

  59. Bonte, S., Goethals, I. & Holen, R. V. Individual prediction of brain tumor histological grading using radiomics on structural MRI. In Proc. 2017 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC), pp 1–3, https://doi.org/10.1109/NSSMIC.2017.8532793 (2017).

  60. Bastos, L. S. & O’Hagan, A. Diagnostics for Gaussian process emulators. Technometrics 51, 425–438 (2009).

    Article  Google Scholar 

  61. Ackerman, A. S. et al. Effects of aerosols on cloud albedo: evaluation of Twomey’s parameterization of cloud susceptibility using measurements of ship tracks. J. Atmos. Sci. 57, 2684–2695 (2000).

    Article  Google Scholar 

  62. Jin, Z., Charlock, T. P., Smith, W. L. Jr & Rutledge, K. A parameterization of ocean surface albedo. Geophys. Res. Lett. https://doi.org/10.1029/2004GL021180 (2004).

Download references

Acknowledgements

We acknowledge the support of the UK Natural Environment Research Council (NERC) funded ADVANCE project (NE/T006897/1), which funded J.H., Y.C., D.P., A.S., D.G. and P.F. J.H., G.J. and F.M. were also partly funded by the EU’s Horizon 2020 research and innovation programme under the CONSTRAIN grant agreement 820829. J.H., P.F., G.J. and F.M. are supported by the Joint UK BEIS/Defra Met Office Hadley Centre Climate Programme (GA01101). Y.W. is grateful for support from P. Sarasin and the ETH Zurich Foundation (ETH Fellowship project: 2021-HS-332). D.G. is funded by the National Centre for Atmospheric Science (NCAS), one of the UK NERC’s research centres. J.d.L. acknowledges funding from the NERC funded V-PLUS grant NE/S00436X/1. N.C., L.O. and S.P. are funded by USA NASA programmes. The machine-learning training is performed using the Statistics and Machine Learning Toolbox in MATLAB (version R2019b, MathWorks Inc.). We thank K. Carslaw (University of Leeds) for co-developing and co-leading the Leeds aspect of ADVANCE project, and A. Jones (UK Met Office) for helpful discussions.

Author information

Authors and Affiliations

Authors

Contributions

Y.C. and J.H. conceived the study. Y.W. and Y.C. designed and developed the machine-learning approach used in this study with help from J.H. and J.F. J.H. led the ADVANCE project funded by UK NERC. Y.C., F.M., G.J. and J.H. performed the analysis of MODIS data with help from D.G., N.C., L.O. and S.P. N.C., L.O. and S.P. performed the cloud regime analysis. Y.C., J.H., Y.W., D.G., U.L., P.F., L.O., S.P., J.d.L., A.S., D.P. and J.F. contributed to the uncertainty discussion. Y.C. and J.H. performed the analyses and interpreted the results with inputs from all co-authors. Y.C. and J.H. led the manuscript writing with specific inputs and edits from D.G., L.O. and U.L.. All co-authors discussed the results and commented on the manuscript.

Corresponding author

Correspondence to Ying Chen.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Geoscience thanks Julia Fuchs, Velle Toll and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Tom Richardson, in collaboration with the Nature Geoscience team.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Relative frequency of occurrence (RFO) of cloud regimes.

The RFO values of the region studied here in September-October 2014 are given in red diamonds, data sourced from Malavelle et al.15. The RFO values during 2002–2014 globally are given in blue triangles, data sourced from Oreopoulos et al.40. CR6-CR11 are liquid-dominated cloud regimes, and the others are ice-dominated cloud regimes. The details of each cloud regime are given in the above references accordingly.

Source data

Extended Data Fig. 2 Correlation coefficient between machine-learning predictions and MODIS observations of cloud properties, including liquid cloud droplet number concentration (Nd), liquid droplet effective radius (reff), liquid water path (LWP) and liquid cloud fraction (CF).

The Monte Carlo results of ML-MODIS validation against MODIS observations without volcanic aerosol-perturbation are given in black. The variations of comparisons with volcanic aerosol-perturbation in October 2014 are given in pink. The boxplot shows 10th, 25th, median (Med.), 75th and 90th percentiles with the mean value indicated by a dot.

Source data

Extended Data Fig. 3 Anomalies in MODIS-Aqua cloud properties for October 2014.

The spatial distributions and zonal means of anomalies in Nd, reff and CF are shown in the panels a-c. Anomalies correspond to the deviation from the 2002–2020 climatology (excluding the 2014 eruption year). The positive anomalies are shown in red and negative ones in blue. The standard deviation is shown by the grey shading.

Source data

Extended Data Fig. 4 Change (a) and anomaly (b) in liquid water path (LWP).

a) Similar to Fig. 2, changes are detected using machine-learning; b) similar to Extended Data Fig. 3, anomaly corresponds to the deviation from 2002–2020 climatology.

Source data

Extended Data Fig. 5 Similar to Extended Data Fig. 3.

but show anomaly in sea-surface temperature (a), anomaly in ice-cloud fraction in October 2014 (b), and climatological anomaly of low-level cloud cover in October 2014 using ERA5 reanalysis (c).

Source data

Extended Data Fig. 6 The top-10 most important features for machine-learning to predict unperturbed liquid cloud fraction in October.

The feature importance is normalized with the maximum as 100%. The value of these features in 2014 are entirely within the variation range of machine-learning training dataset, see Extended Data Fig. 10.

Source data

Extended Data Fig. 7

Similar to Fig. 1, but show results in September 2014.

Source data

Extended Data Fig. 8 Similar to Fig. 3.

Panel a shows results in September 2014. Panel b shows results in October 2014 but excluding the regions where the cold anomalous sea surface temperatures were outside the variation range at the same location.

Source data

Extended Data Fig. 9 Cloud responses to Holuhraun volcanic aerosol in September 2014.

Left panels a-c (similar to Fig. 2but for September 2014) show cloud responses to volcanic aerosol using machine-learning (ML) approach. Right panels d-f (similar to Extended Data Fig. 3but for September 2014) show anomalies in cloud properties.

Source data

Extended Data Fig. 10 Probability distribution of the top-10 most important features.

as shown in Extended Data Fig. 6. Red bars indicate the counts (scaled by 0.1 to fit the display range) of the training data in each bin, which covers the entire variability range of black and blue bars; black bars indicate the data counts from the entire studied region in October 2014; and blue bars indicate the counts from the sea surface temperature anomaly region only. Note that the counts per longitude are different, because we only consider data over the oceans.

Source data

Supplementary information

Supplementary Information

Supplementary Discussion sections 1 and 2 and Table 1.

Source data

Source Data Fig. 1

Statistical source data.

Source Data Fig. 2

Statistical source data.

Source Data Fig. 3

Statistical source data.

Source Data Extended Data Fig. 1

Statistical source data.

Source Data Extended Data Fig. 2

Statistical source data.

Source Data Extended Data Fig. 3

Statistical source data.

Source Data Extended Data Fig. 4

Statistical source data.

Source Data Extended Data Fig. 5

Statistical source data.

Source Data Extended Data Fig. 6

Statistical source data.

Source Data Extended Data Fig. 7

Statistical source data.

Source Data Extended Data Fig. 8

Statistical source data.

Source Data Extended Data Fig. 9

Statistical source data.

Source Data Extended Data Fig. 10

Statistical source data.

Rights and permissions

Springer Nature or its licensor holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Chen, Y., Haywood, J., Wang, Y. et al. Machine learning reveals climate forcing from aerosols is dominated by increased cloud cover. Nat. Geosci. 15, 609–614 (2022). https://doi.org/10.1038/s41561-022-00991-6

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41561-022-00991-6

This article is cited by

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing