Understanding uncertainties in Arctic aerosol representation in climate models

Chandrasekharan, Arundathi; Bergman, Tommi; Nordling, Kalle; Bouchahmoud, Meryem; Merikanto, Joonas; Makkonen, Risto

doi:10.5194/egusphere-2026-2122

Preprints

https://doi.org/10.5194/egusphere-2026-2122

Preprints

27 Apr 2026

| 27 Apr 2026

Understanding uncertainties in Arctic aerosol representation in climate models

Arundathi Chandrasekharan, Tommi Bergman, Kalle Nordling, Meryem Bouchahmoud, Joonas Merikanto, and Risto Makkonen

Abstract. Arctic amplification and its persistent underestimation in climate models underscore the importance of accurate representation of local Arctic feedback processes. Previous studies evaluating model data against measurements showed the importance of including local emissions, such as iodic acid and organic vapours, for an accurate representation of aerosols in the high Arctic. The MOSAiC expedition has produced a full year of data in the high Arctic, providing an opportunity to evaluate the performance of climate models in this region across strongly contrasting seasonal conditions. We evaluate four CMIP6 models and the chemistry-transport model TM5 using this data. CMIP6 models fail to capture the observed seasonal cycle and generally underestimate aerosol number concentration (CN), with the strongest underestimation in summer. To understand the cause of these model deficiencies, we conduct a sensitivity analysis using an ensemble of TM5 experiments by perturbing individual parameters and three reasons were identified. In summer, missing regional new particle formation (NPF) sources are the primary cause of the underestimation. Including methanesulphonic acid driven NPF improved the magnitude and seasonality of simulated CN. In winter and early spring, the model is missing aerosol sources such as blowing snow and lead emissions. During the Arctic haze period, the model underestimates the aerosol background concentration, possibly due to an underestimation of long-range transported aerosols. With cloud condensation nuclei (CCN), we observe a persistent underestimation even during periods of CN overestimation. These results identify gaps in Arctic aerosol representation in climate models that need to be addressed to improve climate projections.

Received: 14 Apr 2026 – Discussion started: 27 Apr 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 2171 KB)

Supplement (721 KB)

Download & links

Arundathi Chandrasekharan, Tommi Bergman, Kalle Nordling, Meryem Bouchahmoud, Joonas Merikanto, and Risto Makkonen

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-2122', Anonymous Referee #1, 19 May 2026
General comments
The authors evaluate the Arctic aerosols representation in four CMIP6 models (CESM2, MRI-ESM2, MIROC-ES2H and UKESM1) and a chemical transport model (TM5) using the one-year dataset generated during the MOSAiC expedition. They observe that CMIP6 models underestimate aerosol concentration and fail to reproduce the aerosols seasonal cycle, and then use several TM5 experiments tuning different individual parameters (MSA treatment, DMS emissions, nucleation rate, sea spray aerosols emissions), to investigate the reasons of the discrepancies between models and MOSAiC observations of aerosol number concentration (CN) and cloud condensation nuclei (CCN).
Methanesulphonic acid (MSA) and iodic acid are the main aerosol precursor in spring, summer and autumn during the MOSAiC campaign. Therefore, the authors discover that implementing an MSA dependent nucleation scheme in their TM5 model improves the seasonality of the CN values, and greatly improves the comparison with observations. The implementation of an iodic acid dependent nucleation scheme could also help to improve the aerosols representation in the TM5 model.
I think the authors make good use of the data obtained during the MOSAiC campaign, using the dataset to investigate the current deficiencies in the representation of Arctic aerosols in the studied CMIP6 models. Therefore, I recommend the manuscript to be published after some points are addressed:

Specific comments
L69: “By using a set of simulations with individually perturbed parameters,…” Please, mention here what individual parameter have been perturbed.

L83: What are differences between the CB05 carbon-bond mechanism chemistry scheme and the modified version mCB05? These should be mentioned in the text.

L104 and Eq.2: the oxidation scheme for DMS employed in the TM5 model seems too simple and outdated. There have been several updates in the last years (e.g. Kilgour et al., 2024 (https://doi.org/10.5194/acp-25-1931-2025).

L112-121, section 2.1.2: This section is too short, relying mainly on previous literature. It should at least include a brief description of the implemented emissions, instead of just directing the reader to the literature on which they are based.

L127: 9-month spin-up period. I think that is too short a period of time. Did the authors checked that that time is sufficient? Typically, 1-2 years spin-up is employed for troposphere stabilization.

L204-208: at least a brief description of the instruments and methods employed to create the MOSAiC dataset should be included. Simply listing the references is insufficient.

L219: “..while TM5 both underestimates or overestimates the observed monthly CN, depending on the experiment.” To understand what experiment underestimate o overestimate observations, it will be helpful if figure 2 differentiated the different TM5 experiments, instead of grouping them all with the same symbol (pink cross mark).

The authors mention and demonstrate (Fig. S4) the importance of iodic acid as a primary driver of nucleation in the Arctic from spring to autumn. Nevertheless, they do not include the formation mechanism in their TM5 model. It should not be difficult to implement the formation mechanism described in the Finkenzeller et al., 2022 (Nat. Chem. 15, 129–135 (2023). https://doi.org/10.1038/s41557-022-01067-z), using lower boundary conditions (LBC) for its precursors (e.g. iodine monoxide) from another model implementing an atmospheric halogen chemistry scheme. I think a sensitivity experiment implementing the formation of iodic acid would help to improve the comparison with observations during that time period, if they also implement an iodic acid dependent nucleation scheme. Although the latter is perhaps more difficult. That could help to give an answer for the question in L440: “Iodic Acid is a main driver of NPF in late summer and early autumn in the Arctic (Price et al. 2023) and could be the reason the model does not capture these events”

Technical corrections
Define CMIP and AMIP projects acronyms. It will be useful for non-modeller readers.

Figure 1: Font size in upper-right label is too small.

L127: describe what a “vertical resolution of 34 hybrid sigma pressure levels” is. It will be useful for non-modeller readers.

L230-231: “The main difference between the experiments is that the AMIP runs use observed SIC and SST for the plots, …”. In this statement the words “for the plots” can be removed.

L286: replace “1*10⁶ /cm³” by “1×10⁶ /cm³”

L356: Replace “Figure 5 shows the spatial distribution of CN…” by “Figure 5 shows the modelled spatial distribution of CN”.

Figure 6 and 7: font size of labels and colour bar numbers too small.

Figure 8: include (c) and (d) labels are missing in lower panels. The blank spaces at both the beginning and end of the four panels can be removed.

L411-414: that paragraphs is almost the caption of Figure 8, so it should either be removed or rephrased.

L440: replace “(Fig. S4)” by (Fig. S4d).
Citation: https://doi.org/10.5194/egusphere-2026-2122-RC1
RC2: 'Comment on egusphere-2026-2122', Anonymous Referee #2, 09 Jun 2026

General Comments
This manuscript by Chandrasekharan et al., titled "Understanding uncertainties in Arctic aerosol representation in climate models", evaluates the representation of Arctic aerosols in four CMIP6 models and the TM5 model using the full annual cycle of observations from the MOSAiC expedition. The topic is timely and of clear importance to the Arctic climate modelling community. The ensemble of sensitivity experiments of parameter purturbation is quite comprehensive, and the study sorted out the relative contributions of several key processes, including MSA-driven nucleation, DMS emissions, sea spray aerosol, and nucleation rate uncertainties, to Arctic aerosol number and CCN concentrations across seasons. These results provide valuable and actionable insight into the structural gaps in current climate models. However, several findings would benefit from more thorough physical interpretation before the manuscript is ready for publication (see detailed comments below). Overall, the writing can use some improvement since the manuscript is a bit hard to follow. I would recommend minor revision.
Specific Comments
Line 91 Please provide a brief description of the ELVOC treatment, including how their gas-phase concentrations are calculated and how they partition to the aerosol phase.
Line 108 The CCN diagnostic based on Petters and Kreidenweis (2007) is introduced without sufficient methodological detail. Please give a brief description. Also, please clarify (1) how the scheme treats coarse-mode particles (e.g., sea spray aerosol) as potential CCN, and (2) whether and how competition effects between particles of different sizes during activation are accounted for. This is relevant in the Arctic context where both freshly nucleated particles and large primary particles coexist.
Table 2 (BULKMSA) Consider replacing "bulk tracer" with "passive tracer" to more precisely convey that MSA in this experiment undergoes no further chemical processing or size-resolved partitioning.
Line 206 The authors state that aerosol precursor gases (H₂SO₄, HIO₃, and MSA) were measured during MOSAiC and used in the analysis, yet no model–observation comparison for these species is presented in either the main text or the supplement. It is unclear to reader how well the model reproduces precursor concentrations and therefore difficult to assess how model parameters should be adjusted relative to the BASECASE to improve CN and CCN simulations. Please include these comparisons in the main text (or did I overlook?).
Lines 220–221 It is worth noting that models such as CESM2 do not include MSA as an aerosol component or nucleation precursor. This omission is likely a primary reason for CESM2's severe underestimation of Arctic CN and CCN.
Line 262 Most CMIP6 models employ highly simplified chemistry schemes that are known to poorly represent long-range transport and ageing of anthropogenic aerosols even at mid-latitudes. Therefore, I wonder how Arctic haze is modeled. The authors should compare simulations of anthropogenic aerosols (e.g., sulfate) against MOSAiC, and discuss how these limitations in the CMIP6 models' chemical schemes may contribute to the wintertime and springtime CCN underestimation.
Line 269 The discussion of factors contributing to overestimated summer CN in TM5 should acknowledge the potential role of missing Arctic dust. Mineral dust can act as a condensation sink for H₂SO₄, MSA, and SOA; its absence in the model would reduce the condensation sink, leading to higher gas-phase precursor concentrations and consequently elevated nucleation rates and CN.
Lines 272–274 It looks like that the SS_HIGH run produces CN values closer to the MOSAiC observations in summer than the BASECASE (Fig. 3d), which may be good.
Line 281 (BULKMSA figure reference) There appears to be an error in the figure reference: BULKMSA is shown in Fig. 3a, not Fig. 3b as cited. Please verify and correct. Additionally, it would be helpful to clarify whether the BULKMSA line overlaps substantially with MSA_SVOClike in Fig. 3a, as the two lines are difficult to distinguish visually.
Line 281 (MSA_SVOClike figure reference) Similarly, MSA_SVOClike is plotted in Fig. 3a, not Fig. 3c as stated in the text. Please correct the figure reference throughout.
Line 283 The authors could make the interpretation of Fig. 3c more explicit. The difference between BASECASE and SO4NUCL represents the isolated contribution of MSA to nucleation, since SO4NUCL uses sulphuric acid as the only nucleation precursor while all other settings remain identical. As shown in Fig. 3c, this difference is substantial in summer, indicating that MSA contributes more to Arctic NPF than sulphuric acid during this season, consistent with the observational findings of Boyer et al. (2024). This point can be stated more explicitly.
Line 290 The DMS oxidation reaction producing SO₂ and MSA is given as Eq. 2 in the methods section, but is referenced here as Eq. 3. Please verify and correct.
Line 295 The authors state that modelled MSA concentrations are lower than observations by approximately one order of magnitude, which is a critical result bearing directly on the interpretation of the nucleation experiments. However, there is no figure of this model–observation MSA comparison in the main text or supplementary material. Please present this comparison explicitly.
Line 307 The comparison between DMS_LOW, NUCLRATE_LOW, and MOSAiC observations in Fig. S2 is difficult to assess because the lines for these experiments are too thin and closely spaced to distinguish clearly, particularly during May–June 2020 where the authors report the largest deviations. Please increase line thickness and use more contrasting colours or line styles in this figure to improve readability.
Line 349 The discussion of CCN underestimation focuses primarily on particle size and hygroscopicity biases, but several other potentially important contributors are not considered. These include: (1) the absence of anthropogenically influenced aerosols associated with Arctic haze, which are likely important CCN in winter and spring; (2) missing iodic acid-driven nucleation, which, while primarily a summer process, can produce particles that grow to CCN sizes; (3) Arctic mineral dust, which, once chemically aged through condensational growth of H₂SO₄ or organics, can become hygroscopic and CCN-active; and (4) marine-derived organic compounds such as methanethiol, which may contribute to secondary aerosol formation and CCN activity over the open Arctic Ocean. Please discuss these missing processes in the context of the CCN underestimation.
Line 350 Please briefly state what particle size and hygroscopicity (κ) thresholds correspond to the transition between sub-CCN and CCN sizes in the Petters and Kreidenweis (2007) framework as implemented in TM5 at the 0.2% supersaturation used for comparison with MOSAiC observations.
Line 371 The spatial maps in Fig. 6 show that NUCLRATE_HIGH leads to a decrease in CCN in some Arctic regions (Fig. 6l). Please provide a physical explanation for this result.
Line 421 The authors attribute the model's failure to capture the second blowing snow event to the low wind speed observed during that period, but do not speculate on what physical mechanism produced the CN spike in the observations. Was this event potentially associated with Arctic haze transport rather than local blowing snow? Or could it reflect a local emission mechanism not captured by wind-speed-dependent parameterisations, such as sublimation of blowing snow particles at low wind speeds? Please provide a more explicit discussion of the possible causes.

Citation: https://doi.org/10.5194/egusphere-2026-2122-RC2

Arundathi Chandrasekharan, Tommi Bergman, Kalle Nordling, Meryem Bouchahmoud, Joonas Merikanto, and Risto Makkonen

Supplement

https://doi.org/10.5194/egusphere-2026-2122-supplement

Arundathi Chandrasekharan, Tommi Bergman, Kalle Nordling, Meryem Bouchahmoud, Joonas Merikanto, and Risto Makkonen

Viewed

Total article views: 381 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
290	67	24	381	30	17	17

HTML: 290
PDF: 67
XML: 24
Total: 381
Supplement: 30
BibTeX: 17
EndNote: 17

Views and downloads (calculated since 27 Apr 2026)

Month	HTML	PDF	XML	Total
Apr 2026	101	35	5	141
May 2026	167	28	12	207
Jun 2026	22	4	7	33

Cumulative views and downloads (calculated since 27 Apr 2026)

Month	HTML	PDF	XML	Total
Apr 2026	101	35	5	141
May 2026	167	28	12	207
Jun 2026	22	4	7	33

Viewed (geographical distribution)

Total article views: 376 (including HTML, PDF, and XML) Thereof 376 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 14 Jun 2026

Short summary

This study examines how well CMIP6 models and the chemistry transport model TM5 simulate Arctic aerosols, using data collected during the MOSAiC expedition. Additionally, the study includes a sensitivity analysis to identify the sources of any inaccuracies in the model results. The findings offer suggestions for improving model performance in the Arctic and highlight areas that need further research.


Total:	0
HTML:	0
PDF:	0
XML:	0