pith. machine review for the scientific record. sign in

arxiv: 2604.21233 · v1 · submitted 2026-04-23 · ⚛️ physics.ao-ph · cs.LG· physics.data-an· physics.geo-ph

Recognition: unknown

Assessing Emulator Design and Training for Modal Aerosol Microphysics Parameterizations in E3SMv2

Ann S. Almgren, Hui Wan, Kai Zhang, Kezhen Chong, Mohammad Taufiq Hassan Mozumder, Panos Stinis, Saad Qadeer, Shady E. Ahmed

Authors on Pith no claims yet

Pith reviewed 2026-05-08 13:22 UTC · model grok-4.3

classification ⚛️ physics.ao-ph cs.LGphysics.data-anphysics.geo-ph
keywords aerosol microphysicsneural network emulatorE3SMv2MAM4SciMLemulation accuracyscaling strategytraining convergence
0
0 comments X

The pith

A simple feedforward neural network with proper scaling and training convergence can accurately emulate key aerosol microphysics changes in E3SMv2.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests neural network emulators for aerosol microphysics processes inside the E3SMv2 climate model under cloud-free conditions. It systematically varies network size, input and output scaling, and training procedures on data from the MAM4 module, while tracking whether optimization reaches convergence. Results indicate that scaling choices and convergence status control how well the emulator reproduces shifts in aerosol particle concentrations driven by microphysics. When scaling works and training converges, even a basic architecture and moderate network size succeed at capturing the main features of those concentration changes. This approach matters because direct microphysics calculations are expensive; a working emulator could allow faster yet still realistic aerosol representations in long global simulations.

Core claim

Optimization convergence, scaling strategy, and network complexity strongly influence emulation accuracy. When effective scaling is applied and convergence is achieved, the relatively simple architecture, used together with a moderate network size, can reproduce key features of the microphysics-induced aerosol concentration changes with promising accuracy. These findings supply practical guidance for further emulator development and broader insights for emulating other aerosol processes and multi-scale atmospheric physics.

What carries the argument

A feedforward neural network emulator for the 4-mode Modal Aerosol Module (MAM4) microphysics under cloud-free conditions, trained with variable normalization and monitored for optimization convergence.

If this is right

  • Emulator development for this process can rely on relatively simple architectures once scaling and convergence are secured.
  • The same design principles are expected to apply to emulating other aerosol processes and multi-scale atmospheric physics.
  • Successful emulators of this kind can improve the numerical representation of aerosol processes inside global atmospheric models.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the emulator remains accurate when inserted into the full model that includes clouds and other processes, it could substantially lower the computational cost of aerosol treatments in climate simulations.
  • The emphasis on scaling and convergence monitoring may transfer to SciML emulators for other expensive sub-grid physics in earth system models.
  • Coupling the emulator back into E3SMv2 and measuring overall model performance would provide a direct test of practical utility.

Load-bearing premise

That achieving convergence and effective scaling on the training dataset will make the emulator accurate for every condition encountered in full E3SMv2 runs.

What would settle it

Apply the trained emulator to input states drawn from a broader range of aerosol loadings, relative humidities, or temperatures than those used in training and compare its predicted aerosol concentration changes against the full MAM4 microphysics solver.

Figures

Figures reproduced from arXiv: 2604.21233 by Ann S. Almgren, Hui Wan, Kai Zhang, Kezhen Chong, Mohammad Taufiq Hassan Mozumder, Panos Stinis, Saad Qadeer, Shady E. Ahmed.

Figure 1
Figure 1. Figure 1: Evolution of the MSE loss (left) and the suboptimality gap (right), namely, the difference between the loss at view at source ↗
Figure 2
Figure 2. Figure 2: Heatmaps of the performance metrics (at validation dataset) for different emulators with varying widths and view at source ↗
Figure 3
Figure 3. Figure 3: Heatmaps of the norm-based performance metrics (at validation dataset) for different emulators with varying widths and depths. 9 view at source ↗
Figure 4
Figure 4. Figure 4: Heatmaps of the R2 (at validation dataset) for each target variable with varying widths and depths. 10 view at source ↗
Figure 5
Figure 5. Figure 5: R2 values of ML emulator for each of the 20 target variables. Bar height corresponds to the average daily score across from January 1 to January 10, and error bars designate the minimum and maximum scores obtained in these 10 days. Another important perspective for assessing the capabilities and limitations of ML-based emulation is to examine how predictive accuracy varies across value ranges view at source ↗
Figure 6
Figure 6. Figure 6: History of training and validation losses for each of the view at source ↗
Figure 7
Figure 7. Figure 7: 2D histograms comparing the one-timestep mixing ratio changes in the testing dataset predicted by E3SMv2 view at source ↗
read the original abstract

Toward the goal of using Scientific Machine Learning (SciML) emulators to improve the numerical representation of aerosol processes in global atmospheric models, we explore the emulation of aerosol microphysics processes under cloud-free conditions in the 4-mode Modal Aerosol Module (MAM4) within the Energy Exascale Earth System Model version 2 (E3SMv2). To develop an in-depth understanding of the challenges and opportunities in applying SciML to aerosol processes, we begin with a simple feedforward neural network architecture that has been used in earlier studies, but we systematically examine key emulator design choices, including architecture complexity and variable normalization, while closely monitoring training convergence behavior. Our results show that optimization convergence, scaling strategy, and network complexity strongly influence emulation accuracy. When effective scaling is applied and convergence is achieved, the relatively simple architecture, used together with a moderate network size, can reproduce key features of the microphysics-induced aerosol concentration changes with promising accuracy. These findings provide practical clues for the next stages of emulator development; they also provide general insights that are likely applicable to the emulation of other aerosol processes, as well as other atmospheric physics involving multi-scale variability.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper explores emulation of aerosol microphysics processes (under cloud-free conditions) in the MAM4 module of E3SMv2 using a simple feedforward neural network. It systematically examines the influence of architecture complexity, variable normalization/scaling, and training convergence on accuracy, concluding that effective scaling plus convergence allows the moderate-sized network to reproduce key features of microphysics-induced aerosol concentration changes with promising accuracy. The work aims to provide practical guidance for SciML emulator development in atmospheric models.

Significance. If the empirical findings hold under proper validation, the paper supplies useful practical insights on design choices (scaling, convergence monitoring) for emulating multi-scale aerosol processes. This could inform more efficient parameterizations in global models like E3SM, with potential applicability to other atmospheric physics. The systematic empirical assessment of training behavior is a clear strength.

major comments (2)
  1. [Results] Results section: the central claim that the emulator reproduces 'key features ... with promising accuracy' when scaling and convergence are achieved is not supported by any quantitative error metrics (e.g., RMSE, mean relative error, or bias on aerosol number/mass concentrations) or direct comparisons against the native MAM4 solver on held-out data. This leaves the performance level unsubstantiated.
  2. [Validation and generalization discussion] Validation and generalization discussion: the claim that convergence plus effective scaling on the chosen dataset suffices for accurate representation 'across the full range of conditions encountered in E3SMv2 simulations' is untested. No out-of-sample validation on independent E3SMv2 trajectories, no analysis of training-data coverage (e.g., nucleation rates, coagulation kernels, modal parameters under varying T/RH), and no discussion of extrapolation risk are provided.
minor comments (2)
  1. [Abstract] Abstract: 'key features of the microphysics-induced aerosol concentration changes' is not defined; specify which species or moments (e.g., number concentration in Aitken/accumulation modes) are being emulated.
  2. [Methods] Methods: provide explicit details on the 'moderate network size' (layer count, neurons per layer) and the exact normalization/scaling formulas applied to inputs/outputs.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful and constructive review. The comments correctly identify areas where our claims require stronger quantitative support and more explicit discussion of limitations. We address each major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Results] Results section: the central claim that the emulator reproduces 'key features ... with promising accuracy' when scaling and convergence are achieved is not supported by any quantitative error metrics (e.g., RMSE, mean relative error, or bias on aerosol number/mass concentrations) or direct comparisons against the native MAM4 solver on held-out data. This leaves the performance level unsubstantiated.

    Authors: We agree that the Results section would benefit from explicit quantitative metrics. While the manuscript includes visual comparisons demonstrating that key features of aerosol concentration changes are reproduced, these do not substitute for numerical error measures. In the revised version we will add RMSE, mean relative error, and bias statistics computed on held-out data, together with direct side-by-side comparisons against the native MAM4 solver outputs. revision: yes

  2. Referee: [Validation and generalization discussion] Validation and generalization discussion: the claim that convergence plus effective scaling on the chosen dataset suffices for accurate representation 'across the full range of conditions encountered in E3SMv2 simulations' is untested. No out-of-sample validation on independent E3SMv2 trajectories, no analysis of training-data coverage (e.g., nucleation rates, coagulation kernels, modal parameters under varying T/RH), and no discussion of extrapolation risk are provided.

    Authors: The manuscript presents a controlled study under cloud-free conditions and does not claim that the emulator has been validated for the full range of E3SMv2 conditions. Nevertheless, the discussion section can be improved by explicitly addressing data coverage, extrapolation risks, and the absence of independent trajectory validation. We will add a limitations paragraph that covers these points, clarifies the intended scope of the current work, and notes that broader out-of-sample testing is planned for subsequent studies. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical assessment of NN emulator training and accuracy

full rationale

The paper performs an empirical study of neural network design choices (architecture, scaling, convergence) for emulating MAM4 aerosol microphysics under cloud-free conditions. It reports observed training behavior and accuracy on key features without any derivation chain, uniqueness theorem, or fitted-parameter prediction that reduces to the inputs by construction. No self-definitional steps, no renaming of known results as new derivations, and no load-bearing self-citations that substitute for independent evidence. The central claim rests on experimental outcomes rather than tautological equivalence to the training data or prior author work.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on the empirical observation that scaling and convergence control accuracy; no new physical entities are introduced. The main unstated premises are that the training data distribution is representative and that convergence on the loss implies physical fidelity.

free parameters (2)
  • network size (moderate)
    Chosen by the authors as part of the architecture sweep; its value is not numerically specified in the abstract.
  • scaling/normalization strategy
    Treated as a tunable design choice whose effectiveness is assessed post-hoc.
axioms (1)
  • domain assumption A feedforward neural network can approximate the mapping from aerosol state variables to microphysics-induced concentration changes.
    Implicit in the choice of emulator architecture; standard universal-approximation assumption for this class of networks.

pith-pipeline@v0.9.0 · 5548 in / 1396 out tokens · 27292 ms · 2026-05-08T13:22:30.486034+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

44 extracted references · 15 canonical work pages

  1. [1]

    Pushing the frontiers in climate modelling and analysis with machine learning

    Veronika Eyring, William D Collins, Pierre Gentine, Elizabeth A Barnes, Marcelo Barreiro, Tom Beucler, Marc Bocquet, Christopher S Bretherton, Hannah M Christensen, Katherine Dagon, et al. Pushing the frontiers in climate modelling and analysis with machine learning. Nature Climate Change, 14(9):916–928, 2024

    2024

  2. [2]

    ACE: A fast, skillful learned global atmospheric model for climate prediction

    Oliver Watt-Meyer, Gideon Dresdner, Jeremy McGibbon, Spencer K Clark, Brian Henn, James Duncan, Noah D Brenowitz, Karthik Kashinath, Michael S Pritchard, Boris Bonev, et al. ACE: A fast, skillful learned global atmospheric model for climate prediction. arXiv preprint arXiv:2310.02074, 2023

    work page arXiv 2023

  3. [3]

    James PC Duncan, Elynn Wu, Jean-Christophe Golaz, Peter M Caldwell, Oliver Watt-Meyer, Spencer K Clark, Jeremy McGibbon, Gideon Dresdner, Karthik Kashinath, Boris Bonev, et al. Application of the AI2 climate emulator to E3SMv2’s global atmosphere model, with a focus on precipitation fidelity.Journal of Geophysical Research: Machine Learning and Computatio...

    2024

  4. [4]

    Neural general circulation models for weather and climate

    Dmitrii Kochkov, Janni Yuval, Ian Langmore, Peter Norgaard, Jamie Smith, Griffin Mooers, Milan Klöwer, James Lottes, Stephan Rasp, Peter Düben, et al. Neural general circulation models for weather and climate. Nature, 632 (8027):1060–1066, 2024. 13 ML Emulation of Aerosol Microphysics in E3SMv2A PREPRINT

    2024

  5. [5]

    Samudrace: Fast and accurate coupled climate modeling with 3d ocean and atmosphere emulators.arXiv preprint arXiv:2509.12490, 2025

    James PC Duncan, Elynn Wu, Surya Dheeshjith, Adam Subel, Troy Arcomano, Spencer K Clark, Brian Henn, Anna Kwa, Jeremy McGibbon, W Andre Perkins, et al. Samudrace: Fast and accurate coupled climate modeling with 3d ocean and atmosphere emulators. arXiv preprint arXiv:2509.12490, 2025

    work page arXiv 2025

  6. [6]

    Deep learning to represent subgrid processes in climate models

    Stephan Rasp, Michael S Pritchard, and Pierre Gentine. Deep learning to represent subgrid processes in climate models. Proceedings of the national academy of sciences, 115(39):9684–9689, 2018

    2018

  7. [7]

    Could machine learning break the convection parameterization deadlock? Geophysical Research Letters, 45(11):5742–5751, 2018

    Pierre Gentine, Mike Pritchard, Stephan Rasp, Gael Reinaudi, and Galen Yacalis. Could machine learning break the convection parameterization deadlock? Geophysical Research Letters, 45(11):5742–5751, 2018

    2018

  8. [8]

    Machine learning for clouds and climate

    Tom Beucler, Imme Ebert-Uphoff, Stephan Rasp, Michael Pritchard, and Pierre Gentine. Machine learning for clouds and climate. Clouds and their climatic impacts: Radiation, circulation, and precipitation, pages 325–345, 2023

    2023

  9. [9]

    A moist physics parameterization based on deep learning

    Yilun Han, Guang J Zhang, Xiaomeng Huang, and Yong Wang. A moist physics parameterization based on deep learning. Journal of Advances in Modeling Earth Systems, 12(9):e2020MS002076, 2020

    2020

  10. [10]

    An ensemble of neural networks for moist physics processes, its generalizability and stable integration

    Yilun Han, Guang J Zhang, and Yong Wang. An ensemble of neural networks for moist physics processes, its generalizability and stable integration. Journal of Advances in Modeling Earth Systems, 15(10):e2022MS003508, 2023

    2023

  11. [11]

    ClimateBench v1

    Duncan Watson-Parris, Yuhan Rao, Dirk Olivié, Øyvind Seland, Peer Nowack, Gustau Camps-Valls, Philip Stier, Shahine Bouabid, Maura Dewey, Emilie Fons, et al. ClimateBench v1. 0: A benchmark for data-driven climate projections. Journal of Advances in Modeling Earth Systems, 14(10):e2021MS002954, 2022

    2022

  12. [12]

    ClimSim: A large multi-scale dataset for hybrid physics-ml climate emulation

    Sungduk Yu, Walter Hannah, Liran Peng, Jerry Lin, Mohamed Aziz Bhouri, Ritwik Gupta, Björn Lütjens, Justus C Will, Gunnar Behrens, Julius Busecke, et al. ClimSim: A large multi-scale dataset for hybrid physics-ml climate emulation. Advances in neural information processing systems, 36:22070–22084, 2023

    2023

  13. [13]

    Energy Exascale Earth System Model (E3SM).https://e3sm.org, 2025

    E3SM Project. Energy Exascale Earth System Model (E3SM).https://e3sm.org, 2025. Accessed: 2025-09-15

    2025

  14. [14]

    Geiss and P.-L

    A. Geiss and P.-L. Ma. Neuralmie (v1.0): an aerosol optics emulator. Geoscientific Model Development, 18 (5):1809–1827, 2025. doi:10.5194/gmd-18-1809-2025. URL https://gmd.copernicus.org/articles/18/ 1809/2025/

    work page doi:10.5194/gmd-18-1809-2025 2025

  15. [15]

    S. J. Silva, P.-L. Ma, J. C. Hardin, and D. Rothenberg. Physically regularized machine learning emulators of aerosol activation. Geoscientific Model Development, 14(5):3067–3077, 2021. doi:10.5194/gmd-14-3067-2021. URLhttps://gmd.copernicus.org/articles/14/3067/2021/

    work page doi:10.5194/gmd-14-3067-2021 2021

  16. [16]

    Estimating submicron aerosol mixing state at the global scale with machine learning and earth system modeling

    Zhonghua Zheng, Jeffrey H Curtis, Yu Yao, Jessica T Gasparik, Valentine G Anantharaj, Lei Zhao, Matthew West, and Nicole Riemer. Estimating submicron aerosol mixing state at the global scale with machine learning and earth system modeling. Earth and Space Science, 8(2):e2020EA001500, 2021

    2021

  17. [17]

    Quantifying the structural uncertainty of the aerosol mixing state representation in a modal model

    Zhonghua Zheng, Matthew West, Lei Zhao, Po-Lun Ma, Xiaohong Liu, and Nicole Riemer. Quantifying the structural uncertainty of the aerosol mixing state representation in a modal model. Atmospheric Chemistry and Physics, 21(23):17727–17741, 2021

    2021

  18. [18]

    Matrix (multiconfiguration aerosol tracker of mixing state): an aerosol microphysical module for global atmospheric models

    SE Bauer, DL Wright, D Koch, ER Lewis, R McGraw, L-S Chang, SE Schwartz, and R Ruedy. Matrix (multiconfiguration aerosol tracker of mixing state): an aerosol microphysical module for global atmospheric models. Atmospheric Chemistry and Physics, 8(20):6003–6035, 2008

    2008

  19. [19]

    X. Liu, R. C. Easter, S. J. Ghan, R. Zaveri, P. Rasch, X. Shi, J.-F. Lamarque, A. Gettelman, H. Morrison, F. Vitt, A. Conley, S. Park, R. Neale, C. Hannay, A. M. L. Ekman, P. Hess, N. Mahowald, W. Collins, M. J. Iacono, C. S. Bretherton, M. G. Flanner, and D. Mitchell. Toward a minimal representation of aerosols in climate models: description and evaluati...

    work page doi:10.5194/gmd-5-709-2012 2012

  20. [20]

    The global aerosol-climate model echam-ham, version 2: sensitivity to improvements in process representations

    K Zhang, Declan O’Donnell, Jan Kazil, Philip Stier, Stefan Kinne, Ulrike Lohmann, Sylvaine Ferrachat, Betty Croft, Johannes Quaas, H Wan, et al. The global aerosol-climate model echam-ham, version 2: sensitivity to improvements in process representations. Atmospheric Chemistry and Physics, 12(19):8911–8949, 2012

    2012

  21. [21]

    Aerosols in the E3SM version 1: New developments and their impacts on radiative forcing

    Hailong Wang, Richard C Easter, Rudong Zhang, Po-Lun Ma, Balwinder Singh, Kai Zhang, Dilip Ganguly, Philip J Rasch, Susannah M Burrows, Steven J Ghan, et al. Aerosols in the E3SM version 1: New developments and their impacts on radiative forcing. Journal of Advances in Modeling Earth Systems, 12(1):e2019MS001851, 2020

    2020

  22. [22]

    E. R. Whitby, P. McMurry, U. Shankar, and F. Binkowski. Modal aerosol dynamics modeling. Technical report, Computer Sciences Corp., Research Triangle Park, NC (USA), 1991. 14 ML Emulation of Aerosol Microphysics in E3SMv2A PREPRINT

    1991

  23. [23]

    M7: An efficient size-resolved aerosol microphysics module for large-scale aerosol transport models

    Elisabetta Vignati, Julian Wilson, and Philip Stier. M7: An efficient size-resolved aerosol microphysics module for large-scale aerosol transport models. Journal of Geophysical Research: Atmospheres, 109(D22), 2004. doi:https://doi.org/10.1029/2003JD004485. URL https://agupubs.onlinelibrary.wiley.com/doi/abs/ 10.1029/2003JD004485

    work page doi:10.1029/2003jd004485 2004

  24. [24]

    Easter, Steven J

    Richard C. Easter, Steven J. Ghan, Yang Zhang, Rick D. Saylor, Elaine G. Chapman, Nels S. Laulainen, Hayder Abdul-Razzak, L. Ruby Leung, Xindi Bian, and Rahul A. Zaveri. MIRAGE: Model description and evaluation of aerosols and trace gases. Journal of Geophysical Research: Atmospheres, 109(D20), 2004. doi:https://doi.org/10.1029/2004JD004571. URL https://a...

    work page doi:10.1029/2004jd004571 2004

  25. [25]

    Tropospheric aerosol size distributions simulated by three online global aerosol models using the M7 microphysics module

    Kai Zhang, Hui Wan, Bin Wang, Meigen Zhang, J Feichter, and Xiaohong Liu. Tropospheric aerosol size distributions simulated by three online global aerosol models using the M7 microphysics module. Atmospheric Chemistry and Physics, 10(13):6409–6434, 2010

    2010

  26. [26]

    Intercomparison and evalua- tion of global aerosol microphysical properties among AeroCom models of a range of complexity

    Graham W Mann, Kenneth S Carslaw, Carly L Reddington, Kirsty J Pringle, Michael Schulz, Ari Asmi, Do- minick V Spracklen, David A Ridley, Matthew T Woodhouse, Lindsay A Lee, et al. Intercomparison and evalua- tion of global aerosol microphysical properties among AeroCom models of a range of complexity. Atmospheric chemistry and physics, 14(9):4679–4713, 2014

    2014

  27. [27]

    Emulating aerosol microphysics with machine learning

    Paula Harder, Duncan Watson-Parris, Dominik Strassel, Nicolas Gauger, Philip Stier, and Janis Keuper. Emulating aerosol microphysics with machine learning. arXiv preprint arXiv:2109.10593, 2021

    work page arXiv 2021

  28. [28]

    The aerosol-climate model ECHAM5-HAM

    Philip Stier, Johann Feichter, Stefan Kinne, Silvia Kloster, Elisabetta Vignati, Julian Wilson, Laurence Ganzeveld, Ina Tegen, Martin Werner, Yves Balkanski, et al. The aerosol-climate model ECHAM5-HAM. Atmospheric Chemistry and Physics, 5(4):1125–1156, 2005

    2005

  29. [29]

    Physics-informed learning of aerosol microphysics

    Paula Harder, Duncan Watson-Parris, Philip Stier, Dominik Strassel, Nicolas R Gauger, and Janis Keuper. Physics-informed learning of aerosol microphysics. Environmental Data Science, 1:e20, 2022

    2022

  30. [30]

    Data-driven emulation of modal aerosol microphysics via neural operator-based modeling

    Zhe Bai and Damian Rouson. Data-driven emulation of modal aerosol microphysics via neural operator-based modeling. Scientific Reports, 2025

    2025

  31. [31]

    Easter, Rudong Zhang, Po-Lun Ma, Balwinder Singh, Kai Zhang, Dilip Ganguly, Philip J

    Hailong Wang, Richard C. Easter, Rudong Zhang, Po-Lun Ma, Balwinder Singh, Kai Zhang, Dilip Ganguly, Philip J. Rasch, Susannah M. Burrows, Steven J. Ghan, Sijia Lou, Yun Qian, Yang Yang, Yan Feng, Mark Flanner, L. Ruby Leung, Xiaohong Liu, Manish Shrivastava, Jian Sun, Qi Tang, Shaocheng Xie, and Jin-Ho Yoon. Aerosols in the E3SM version 1: New developmen...

    work page doi:10.1029/2019ms001851 2020

  32. [32]

    Liu, P.-L

    X. Liu, P.-L. Ma, H. Wang, S. Tilmes, B. Singh, R. C. Easter, S. J. Ghan, and P. J. Rasch. Description and evaluation of a new four-mode version of the modal aerosol module (MAM4) within version 5.3 of the Community Atmosphere Model. Geoscientific Model Development, 9(2):505–522, 2016. doi:10.5194/gmd-9-505-2016. URL https://www.geosci-model-dev.net/9/505/2016/

    work page doi:10.5194/gmd-9-505-2016 2016

  33. [33]

    S. J. Ghan and R. C. Easter. Impact of cloud-borne aerosol representation on aerosol direct and indirect effects. Atmospheric Chemistry and Physics, 6(12):4163–4174, 2006. ISSN 1680-7316. URL http://www. atmos-chem-phys.net/6/4163/2006/

    2006

  34. [34]

    Description and evaluation of glomap-mode: A modal global aerosol microphysics model for the UKCA composition-climate model

    Graham W Mann, KS Carslaw, DV Spracklen, DA Ridley, PT Manktelow, MP Chipperfield, SJ Pickering, and CE Johnson. Description and evaluation of glomap-mode: A modal global aerosol microphysics model for the UKCA composition-climate model. Geoscientific Model Development, 3(2):519–551, 2010

    2010

  35. [35]

    H. Wan, P. J. Rasch, K. Zhang, J. Kazil, and L. R. Leung. Numerical issues associated with compensating and competing processes in climate models: an example from echam-ham. Geoscientific Model Development, 6 (3):861–874, 2013. doi:10.5194/gmd-6-861-2013. URL https://gmd.copernicus.org/articles/6/861/ 2013/

    work page doi:10.5194/gmd-6-861-2013 2013

  36. [36]

    Terai, Hailong Wang, Qi Tang, Jiwen Fan, Susannah Burrows, Wuyin Lin, Mingxuan Wu, Xiaoliang Song, Yuying Zhang, Mark A

    Shaocheng Xie, Christopher R. Terai, Hailong Wang, Qi Tang, Jiwen Fan, Susannah Burrows, Wuyin Lin, Mingxuan Wu, Xiaoliang Song, Yuying Zhang, Mark A. Taylor, Jean-Christophe Golaz, James J. Benedict, Chih- Chieh-Jack Chen, Yan Feng, Walter M. Hannah, Ziming Ke, Yunpeng Shan, Vincent E. Larson, Xiaohong Liu, Michael J. Prather, Jadwiga H. Richter, Manish ...

    work page doi:10.1029/2025ms005120 2025

  37. [37]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

    2016

  38. [38]

    Identity mappings in deep residual networks

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Identity mappings in deep residual networks. In European conference on computer vision, pages 630–645. Springer, 2016

    2016

  39. [39]

    Data-driven medium-range weather prediction with a resnet pretrained on climate simulations: A new model for weatherbench

    Stephan Rasp and Nils Thuerey. Data-driven medium-range weather prediction with a resnet pretrained on climate simulations: A new model for weatherbench. Journal of Advances in Modeling Earth Systems, 13(2): e2020MS002405, 2021

    2021

  40. [40]

    On deep learning-based bias correction and downscaling of multiple climate models simulations

    Fang Wang and Di Tian. On deep learning-based bias correction and downscaling of multiple climate models simulations. Climate Dynamics, 59(11):3451–3468, 2022

    2022

  41. [41]

    Hannah, Andrew M

    Walter M. Hannah, Andrew M. Bradley, Oksana Guba, Qi Tang, Jean-Christophe Golaz, and Jon Wolfe. Separating physics and dynamics grids for improved computational efficiency in spectral ele- ment earth system models. Journal of Advances in Modeling Earth Systems, 13(7):e2020MS002419, 2021. doi:https://doi.org/10.1029/2020MS002419. URL https://agupubs.onlin...

    work page doi:10.1029/2020ms002419 2021

  42. [42]

    P. J. Rasch, S. Xie, P.-L. Ma, W. Lin, H. Wang, Q. Tang, S. M. Burrows, P. Caldwell, K. Zhang, R. C. Easter, P. Cameron-Smith, B. Singh, H. Wan, J.-C. Golaz, B. E. Harrop, E. Roesler, J. Bacmeister, V . E. Larson, K. J. Evans, Y . Qian, M. Taylor, L. R. Leung, Y . Zhang, L. Brent, M. Branstetter, C. Hannay, S. Mahajan, A. Mametjanov, R. Neale, J. H. Richt...

  43. [43]

    URL https://agupubs.onlinelibrary.wiley.com/ doi/abs/10.1029/2019MS001629

    doi:https://doi.org/10.1029/2019MS001629. URL https://agupubs.onlinelibrary.wiley.com/ doi/abs/10.1029/2019MS001629

    work page doi:10.1029/2019ms001629

  44. [44]

    Rasch, Po-Lun Ma, Richard Neale, Vincent E

    Shaocheng Xie, Wuyin Lin, Philip J. Rasch, Po-Lun Ma, Richard Neale, Vincent E. Larson, Yun Qian, Peter A. Bogenschutz, Peter Caldwell, Philip Cameron-Smith, Jean-Christophe Golaz, Salil Mahajan, Balwinder Singh, Qi Tang, Hailong Wang, Jin-Ho Yoon, Kai Zhang, and Yuying Zhang. Understanding cloud and convective characteristics in version 1 of the E3SM atm...

    work page doi:10.1029/2018ms001350 2018