the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 24 Jul 2026
An Extended Modal Aerosol Dynamics Model (ExMADv1.0) for Simulating Early Nanoparticle Growth
Abstract. New Particle Formation (NPF) is one of the important sources of aerosol particles in the atmosphere and plays a significant role in global climate change and air quality. Accurately simulating the early growth of nanoparticles from NPF remains critical yet challenging. Specifically, the trimodal framework in current modal schemes often fails to resolve the distinct evolution of nucleation-mode particles and their separation from the background Aitken population. To address this, we developed an Extended Modal Aerosol Dynamics Model (ExMADv1.0) that explicitly incorporates a nucleation mode. This model can be completely driven by observational data for process analysis. The ExMAD model was evaluated against observational data from two representative NPF events in Nanjing, China. A nucleation-resolved decomposition of observed particle size distributions confirms that newly formed particles constitute a distinct sub-25 nm mode during NPF events. Compared with the conventional modal configuration, which captures only ~45 % of the observed sub-25 nm particle number concentrations, ExMAD reproduces ~80 % of the observed magnitude during the NPF events. ExMAD also reproduces the observed geometric mean diameter (Dg) of the nucleation mode within ~2 nm, a key metric reflecting early-stage growth dynamics. The extended scheme also improves the simulation of particle growth into Cloud Condensation Nuclei (CCN)-relevant size ranges, reducing the overestimation of >50 nm particle concentrations by nearly half relative to the conventional configuration. Sensitivity simulations further show that the Kelvin effect suppresses early-stage condensational growth, delays the transfer of particles from the nucleation mode to larger modes, and limits the uptake of low-volatility organics by the smallest particles. Compared with sectional simulations, the extended modal scheme provides a more realistic representation of sub-25 nm variability while requiring only a ~2.4 % runtime increase relative to the default modal configuration, far lower than the cost of sectional schemes. These results demonstrate that the extended scheme offers an optimal balance between physical realism and computational efficiency, supporting its future application in three-dimensional aerosol-climate simulations.
- Preprint
(1389 KB) - Metadata XML
-
Supplement
(582 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3636', Anonymous Referee #1, 28 Aug 2026
-
RC2: 'Comment on egusphere-2026-3636', Anonymous Referee #2, 14 Sep 2026
Review of the manuscript titled “An Extended Modal Aerosol Dynamics Model (ExMADv1.0) for Simulating Early Nanoparticle Growth” by Lin et al.
Lin et al. address an important but challenging topic within atmospheric modelling, how to model the survival and growth of nanoparticles originating from atmospheric new particle formation. The new Extended Modal Aerosol Dynamics Model (ExMADv1.0) provide an interesting and potentially important addition to the already existing schemes of aerosol dynamics models that represent nanoparticle growth dynamics. I value that the authors compare the modal approach against more detailed sectional schemes, with up to 100 size bins. The new model may be an important contribution to the research field but in the present form the novelty of the ExMADv1.0 is not convincingly demonstrated. Before this manuscript can be considered for publication in GMD some major revision and additional evaluation of the model performance are needed.
Major comments:
- On line 53-54 you write “Current modal schemes in atmospheric chemistry and climate models, typically adopt a trimodal framework consisting of Aitken, accumulation, and coarse modes” while on line 82-84 you also mention that “ several global modeling frameworks have incorporated a nucleation mode into modal aerosol representations (McErlich et al., 2025; Tegen et al., 2019; Vignati et al., 2004). Maybe the most widely used modal scheme in global scale models is different versions of the M7 scheme by Vignati et al. (2004). Like ExMADv1.0, M7 include a separate mode for the nucleation mode, account for the Kelvin effect and in addition includes both water soluble and water insoluble particles in the Aitken, Accumulation and Coarse mode, hence in total 7 modes. Since ExMADv1.0 lack representation of the aerosol coarse mode and primary aerosols, which also influences the aerosol dynamics and gas-particle partitioning of semi-volatile organics and inorganic acids like HNO3 it is unclear to me what benefits the new ExMADv1.0 scheme provide compared to the existing apparently more comprehensive modal schemes which already includes a nucleation mode, coarse mode particles and primary aerosols. This need to be discussed in the manuscript and preferentially the new ExMADv1.0 scheme should be compared against existing state-of-the-art aerosol dynamics modal schemes like M7 both with respect to the numerical methods, results and the computational costs. Now the ExMADv1.0 performance is compared against a bi-modal model without an explicit nucleation mode and it is not surprising that ExMADv1.0 can outperform the more simplified bi-modal Aitken and Accumulation mode approach in representing the concentration of sub 25 nm particles (nucleation mode particles).
- The selected scenarios consist of two different NPF events at the SORPES station in Nanjing. If I understand the method correctly, ExMADv1.0 is setup as a stationary zero-dimensional box model, without addition of primary aerosol particle emissions and is run with sulfuric acid (SA) and oxygenated organic molecule (OOM) concentration timeseries from measurements at SORPES. I assume that the measurements were performed within the surface layer, near the ground or? With this 0D-setup you inherently need to assume that:
- the measured air mass is the same thought the entire NPF events (no horizontal transport or vertical mixing),
- that primary particle emissions upwind and at the station have negligible influence on the particle number size distributions,
- that the observed OOM and SA concentrations are representative for the entire atmospheric boundary layer that effectively determines the aerosol particle growth rates over the simulated 10 hour periods.
- That the assigned volatility distribution of the measured OOM is realistic.
Concerning the first 3 points I doubt that this is the case and for point nr 4 I miss information about how the volatility distribution (VBS) of the measured OOM is derived and what uncertainties that exists in the translation of the measured OOM to volatility distributions. The OOM were measured with a CI-Api-TOF which has uncertainties concerning the absolute OOM concentrations and secondly these instruments do not provide information about the exact isomer structure, including the functional groups that governs the volatility. Some assumptions must have been made when translating the OOM mass spectrums to VBS distributions.
- Since ExMADv1.0 is setup as a 0D-box model without representation of primary aerosol sources I think that the selected (only two) atmospheric NPF cases and the evaluation of the different model versions performance against these observations provide little information and justification of the novelty and performance of ExMADv1.0 compared with the bimodal scheme and more detailed sectional schemes. The apparently worse performance of the Sec100 scheme compared to ExMADv1.0 concerning the observed evolution of the N4-25 nm particles can as an example be because the models are run with the presumably high OOM concentrations measured close to the surface, which likely are not fully representing the average OOM concentration within the entire boundary layer. This would result in two rapid growth of the nucleation mode particles into the Aitken mode size range in the models. Considering the limitation of the 0-dimensional ExMADv1.0 setup without primary aerosol particle sources and the complexity of the aerosol sources and transport patterns governing the aerosol size distributions at SORPES I suggest that you instead evaluate the modelled nanoparticle growth and PSD evolution against some idealized smog chamber (atmospheric simulation chamber) experiments. A possible alternative would be to run the model as a Lagrangian model towards the SORPES station and include emissions and deposition of gases and particles and boundary layer dynamics along the simulated air mass trajectories.
- I find no information about how the model represent the initial particle size distribution chemical composition in each mode for the simulated NPF events or how/if the model represents the hygroscopic growth of the aerosol particles which also influences the coagulation and condensation sink. The chemical composition of the aerosol particles will govern the partitioning of the OOM according to Raoult’s law. I miss information how the aerosol composition influences the gas-particle partitioning of the different OOM, which aerosol species apart from the OOM and SA that the model considers and how the aerosol hygroscopic growth is modelled.
Minor comments:
- Line 48. Deposition also influence the aerosol population.
- Line 61-62. “freshly nucleated particles” are also secondary particles. Reformulate and write ... aged secondary particles, freshly nucleated particles exhibit ...
- Line 121. Why is the coarse mode excluded from the current simulations? Is the ExMADv1.0 a modification of an existing trimodal model with Aitken, Accumulation and Coarse mode, replacing the coarse mode with a nucleation mode instead? Otherwise, I see no point in excluding the coarse mode since this reduces the applicability of the ExMADv1.0 for any global model simulations focusing on aerosol radiative forcing or air pollution, including PM2.5 and PM10. Please clarify.
- I lack details about the exact model assumptions and setup for the two simulated nucleation events.
- No information is provided about the sectional method used in the Sec20, Sec50 and Sec100 approach.
- Line 470-475 “Both scenarios overestimate 𝑁>100 in the two events, with smaller bias in ExtMod (shown in Fig. S3 and Table S6). These results illustrate that incorporating the early growth dynamics is critical to mitigate the overestimation of particle survival to CCN-relevant sizes (Blichner et al., 2021; Pierce and Adams, 2009). However, the residual positive bias in ExtMod likely reflects the absence of explicit CCN activation and cloud-processing sinks in the present box model, which allows particles that have grown into the CCN-relevant size range to remain in the particle population rather than being activated and removed.” Was the selected NPF events influences by aerosol- cloud processing? Was it not clear sky conditions? In order to activate and remove aerosol particles by clouds you also need precipitation. Were the simulated NPF events influenced by precipitation?
Citation: https://doi.org/10.5194/egusphere-2026-3636-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 245 | 87 | 48 | 380 | 48 | 47 | 44 |
- HTML: 245
- PDF: 87
- XML: 48
- Total: 380
- Supplement: 48
- BibTeX: 47
- EndNote: 44
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Lin et al. present a modal aerosol dynamics model framework ExMAD that incorporates a nucleation mode to represent ~sub-25 nm particles originating from new-particle formation (NPF). In conventional modal models, the nucleation mode is merged with larger particles in the Aitken mode. This distorts the modeled size distribution and affects the particle number predictions. The new model is evaluated against two observed NPF events in Nanjing, China, and also compared to a sectional aerosol dynamics model for the same events.
Adding a nucleation mode is expected to improve the representation of NPF by modal models with a minor computational cost, and the work is thus relevant to modeling of aerosol concentrations and their effects. The model shows improved performance compared to a modal framework without a nucleation-resolving mode especially in the sub-25 nm size range. The manuscript is very clearly written, and the results are well presented and suitable for the journal. I find, however, the model evaluation quite limited as detailed below.
Main comments:
(1) I find the evaluation of the model inadequate: the performance is only compared to two 1-day NPF events, primarily to measured PNSD and additionally to the same events modeled by a single sectional model. This doesn’t give a proper view of model performance. Numerical GDE models are normally evaluated against more accurate benchmark model(s) and/or analytical solution when available, at various ambient conditions.
Specifically, the two evaluation cases are very similar, showing strong NPF typical in polluted environments in e.g. Chinese megacities. The model needs to be evaluated in different types of environments, especially if it is to be applied in regional and global large-scale models, including less polluted and pristine environments (e.g. marine, rural, less polluted background). Extensive case studies may be out of the scope of the current work, but some comparison to e.g a sectional benchmark model would be needed, including:
i. representative cleaner environments, and/or
ii. NPF with different nucleation and growth rates, not only intense NPF and rapid growth.
(2) Regarding the comparison of ExMAD and a sectional model (Section 3.4): I wonder if ExMAD is “better” for right reasons? It doesn’t sound correct that a parameterized modal representation is “more accurate”. Evaluating the models by comparing both of them to a given field measurement doesn’t necessarily give information of their physical goodness, as the observation is likely affected by factors that are not included in the models (e.g. emissions and transport, as also noted in the manuscript). It’s even discussed in the Intro that sectional models are suitable for NPF and UFP modeling. Please comment on this.
(3) I’d also suggest some more discussion on the advantages and disadvantages of the modal and sectional GDE model approaches for NPF modeling. For example, modal models have typically issues with NPF as the averaging “pulls back” the size distribution at high concentrations of nucleated particles. This effect should remain also when the nucleation mode is added, even if its magnitude is decreased. I suggest to summarize in the Introduction the types of biases that can be expected for modal vs. sectional models. Also note that there are different types of sectional models with different advantages (e.g. fixed grid, hybrid bin).
(4) There are some references to previous works that have incorporated a nucleation mode in a modal model (L83-84). How do these models and results compare to the current work?
(5) The effect of the Kelvin factor for evaporation is assessed by comparing ExMAD results with and without the factor. This doesn’t give info on how well the Kelvin effect is captured by the weighted average Kelvin factor in the nucleation mode, applied in the model. The size-dependence of the factor is very strong at the smallest sizes. How does the Kelvin factor affect model results in a sectional framework, in which its size-dependence may be more accurately described?
Other comments:
(1) Discussion on Fig. 5 on p. 15: The low bias of ExtMAD for N(25-100) is assumed to be due to missing primary emissions and transport. Can it also be contributed by the modeled growth rates? In Fig. 4 it looks like the modeled growth of nucleated particles is slower than what is observed.
(2) The sentence about Fig. 6c-d on L442-L443 is not fully justifiable, it’s not obvious from the figure that ExMAD would capture the temporal variability (even considering a possible temporal mismatch). The modeled and observed mean diameters in the Aitken mode look very different.
(3) Regarding modeling of larger CCN-relevant sizes N(>100) (Fig. S3), it’s stated that the model overestimates N(>100). It should also be noted that the observed N(>100) seems to be approximately unaffected by the NPF events, while the model predicts a notable (up to over ~3-fold) increase in N(>100).
Technical:
(1) In Figs. 2 and S1, panels (c) (3-modal fit) look much worse than panels (b) (2-modal fit); should the panels be the other way around?
(2) Figs. 3 and S2: What are the units of the axes?
(3) Fig. 5: The explanations for the line styles are opposite in the figure panels and in the caption.
(4) L569: There is some error message in the text.