Pith. sign in

REVIEW 3 major objections 5 minor 28 references

MVAR: MultiVariate AutoRegressive Air Pollutants Forecasting Model

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read MVAR claims that an autoregressive transformer using only the past two six-hourly observations can forecast six air pollutants across 75 North China cities for the next 120 hours, beating models that use twenty steps of history.

desk verdict A useful forecasting architecture with a likely normalization leak in Eq. 17 that must be fixed before the headline results can be trusted. read the letter →

arxiv 2507.12023 v1 pith:Y62C7PYH submitted 2025-07-16 cs.CV cs.LG

Xu Fan , Zhihao Wang , Yuetan Lin , Yan Zhang , Yang Xiang , Hao Li This is my paper · ORCID
classification cs.CVcs.LG
keywords airpollutionforecastingmultivariatetimeseriesautoregressivetransformermeteorologicalcouplingcross-attentionNorthChinaFuXi-2.0long-horizon
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

MVAR is a transformer model that claims to forecast six air pollutants across 75 North China cities for the next 120 hours using only the two most recent observation steps as input. The paper introduces a multivariate autoregressive training paradigm that iterates the model on its own outputs with a step-weighted loss, so long-horizon predictions are learned without long input windows. It also introduces a Meteorological Coupled Spatial Transformer block that lets city-level pollutant data query gridded AI weather forecasts, allowing meteorological transport and photochemical information to enter the prediction. The authors report that MVAR outperforms six baselines in RMSE across pollutants and lead times, and that coupling with FuXi-2.0 forecast data yields the largest gains. If correct, the result would mean operational air-quality warnings can start from a very short observational history and directly exploit AI weather models.

What carries the argument

The load-bearing mechanism is the Multivariate Autoregressive Training Paradigm, which iterates the same two-step-input model τ times so the output of step γ becomes part of the input for step γ+1, trained with a Step Weighted loss that assigns gradually decreasing weights to later prediction steps. The second mechanism is the Meteorological Coupled Spatial Transformer block, where pollutant city features serve as queries and gridded meteorological forecast features serve as keys and values in cross-attention, so weather information is retrieved at the locations most relevant to each city's pollutant transport. Together these mechanisms let the model forecast twenty steps from only two input steps while coupling sparse urban observations with dense weather fields.

What would settle it

Recompute the normalization in Eq. (17) using only the 2018–2022 portion of the record, retrain or re-evaluate MVAR on the 2023 test set, and compare RMSE with the paper's Table 2 and Table 3 values; if MVAR's advantage over baselines shrinks or reverses, the full-record normalization is the cause.

Watch

Extended reading notes

Core claim

The central discovery is that an autoregressive transformer with a two-step input window can match or beat models that consume twenty steps of history, when trained with the Multivariate Autoregressive Training Paradigm. The model predicts the change in pollutant concentrations relative to the current step, iterating τ times; the step-weighted loss balances early and late prediction steps. Meteorological coupling makes the pollutant representation act as queries against downsampled gridded weather fields, so transport and photochemical information enters through cross-attention. The paper reports consistent RMSE improvements over STID, AGCRN, STGCN, STNorm, AirFormer, and GAGNN across SO2, NO2, PM2.5, PM10, CO, and O3, with the best results when FuXi-2.0 forecast data is used at test time, enabling a claimed 120-hour forecasting span at 1-hour resolution via a greedy ensemble of single-step lead-time models.

Load-bearing premise

The load-bearing premise is that the city-level normalization in Eq. (17) is computed without using any test-period information from 2023; if the 2023 values enter the mean and standard deviation, the model could benefit from future distributional knowledge and the reported RMSE gains would not hold in a real forecast.

Editorial extensions

If this is right

  • MVAR forecasts six pollutants for 120 hours from only two six-hourly observations, which would cut the data-availability requirement for operational warnings and allow forecasts to start shortly after an observation gap.
  • Coupling with FuXi-2.0 meteorological forecasts improves RMSE for PM2.5, PM10, O3, and NO2 relative to pollutant-only training, suggesting AI weather models can be used directly in air-quality forecasting pipelines.
  • The step-weighted loss and autoregressive iteration reduce the need for long input windows; the paper reports that 8-step training with SW loss matches 20-step training performance while using fewer training samples.
  • The 1-hour-resolution extension with a greedy ensemble of single-step lead-time models produces 120-hour hourly forecasts, the resolution needed for practical pollution alerts.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves implicit that its two-step input advantage likely rests on the strong periodicity of pollutant concentrations; a testable extension is to check whether the advantage persists for cities with irregular pollution events or under climate regimes with weaker diurnal cycles.
  • Because normalization statistics are computed over the complete 2018–2023 record including the 2023 test year, the reported RMSE gains should be re-verified with normalization computed only on the training period; this is a direct extension the paper does not perform.
  • The meteorological cross-attention design could transfer to other sparse-observation forecasting tasks where gridded forecast products exist but station data is scarce, such as river levels or crop stress, though the paper does not claim this.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This paper proposes MVAR, a transformer-based multivariate autoregressive model for forecasting six air pollutants across 75 North China cities. Using only the two most recent concentration observations, MVAR iteratively predicts future steps via a Multivariate Autoregressive Training Paradigm (MATP) with a step-weighted loss, and optionally couples gridded ERA5 or FuXi-2.0 meteorological forecasts through a Meteorological Coupled Spatial Transformer. The authors construct a 2018-2023 benchmark, train on 2018-2022, test on 2023, and report RMSE comparisons against spatial-temporal and air-quality baselines under several input-length settings, including a 120-hour (20-step) forecast. The central claim is that MVAR consistently outperforms all baselines.

Significance. If the evaluation protocol is correct, the paper would offer a compact and computationally efficient alternative to long-window forecasting models, with the practical advantage of 120-hour forecasts from a two-step input. Strengths include the release of a standardized multi-city, multi-pollutant dataset with explicit quality control, alignment of test sets across settings, and clean ablations of the multivariate formulation and the step-weighted loss. The meteorological cross-attention design is sensible, and the FuXi-2.0 variant is operationally plausible. However, the current manuscript does not yet establish the central claim because of a normalization ambiguity that may leak test-period information and because the headline comparisons use unequal training data sizes.

major comments (3)
  1. [Section 4.1, Eq. (17)] The normalization in Eq. (17) is computed from X~i, defined as "the complete historical dataset for the i-th city." Since the dataset spans 2018-2023 and 2023 is the test year, this phrasing implies that the test-year concentrations contribute to the mean and standard deviation used to standardize both training and test inputs. If so, the model is given distributional information about the test period, which is particularly consequential here because MVAR deliberately consumes only two input steps and the normalization is the only pathway for long-term city-level statistics to enter. This affects MVAR, MVARera5, and MVARfuxi equally, so the reported gains over baselines would be inflated and the 120-hour operational conclusion would be unsupported. Please either explicitly restrict the statistics to the 2018-2022 training split or rerun all experiments with train-only normalization and report whether the qualitative ranking changes.
  2. [Section 4.5, Tables 2 and 3] Tables 2 and 3 compare models under unequal training data sizes. MVAR is trained with an 8-step autoregressive unroll (tau=8, Section 4.2), so each training window spans 10 hours, yielding 6464 samples, whereas the 20-step baselines in Table 2 use 40-hour windows (5120 samples) and the 2-step baselines in Table 3 use 22-hour windows (5888 samples). The paper aligns test sets but not training sets, so the consistent advantage of MVAR in these tables confounds model architecture with data quantity. The only experiment with a matched training size (Table 4) covers only 1-48h predictions; therefore the 97-120h superiority claimed in Table 2 is not supported by a fair training-data comparison. Please add a matched-size evaluation for the 20-step setting, or restrict the 120-hour claim to settings where training sizes are aligned.
  3. [Section 4.5] Section 4.5 states that MVARera5 uses ERA5 reanalysis during the testing phase. ERA5 is a reanalysis product and is not available in real time, so MVARera5 is an oracle variant; its results cannot support the operational claim of 120-hour forecasting. The operational claim should be based solely on MVARfuxi (or on a variant using forecast meteorological data), and the comparison against baselines should be reported for that variant separately, not for MVARera5.
minor comments (5)
  1. [Section 4.5] In Section 4.5, "demonstrats" is a typo for "demonstrates."
  2. [Section 4.4] Section 4.4 says "8:00 and 20:00 (UTC-8)"; China Standard Time is UTC+8, so the sign is likely wrong and should be corrected.
  3. [Section 4.5] The abbreviation "SFT" appears in Section 4.5, but the baselines are introduced as "STF" (Spatial-Temporal Forecasting); please harmonize the abbreviation throughout.
  4. [General] The model name is inconsistently written as "MVAR" and "MV AR" in the text and tables; please standardize to a single form.
  5. [Eq. (16)] Equation (16) uses t both as the summation index and as the total number of steps, which is confusing; please rewrite with a different index (e.g., j) and an upper limit tau.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the central claim is an empirical benchmark comparison against external baselines; the only mild flag is a non-load-bearing self-citation to FuXi-2.0, while the normalization ambiguity in Eq. 17 is a correctness/leakage risk rather than circularity.

full rationale

The paper's central claim ('MVAR consistently outperforms all baseline models') is an empirical result obtained by training on 2018-2022 and testing on 2023 against external baselines (STGCN, AGCRN, STNorm, STID, AirFormer, GAGNN). Equations (12)-(15) define a residual autoregressive network and iterative rollout; Eq. (16) is a weighted MSE loss; Eq. (17) is standard z-score normalization. None of these makes the predicted quantity equal to an input by construction, and no fitted parameter is renamed as a prediction. The FuXi-2.0 meteorological input is cited to prior work with overlapping authors (Zhong et al., 2024), but it is used only as a data source and does not carry a load-bearing uniqueness or derivation step, so the self-citation is minor and non-circular. The only substantive concern is the ambiguity in Eq. (17): if mean(X~i) and std(X~i) are computed over the full 2018-2023 record while 2023 is the test year, the evaluation protocol leaks test-period distributional information. That is a data-leakage and correctness risk, not a reduction of the result to its own inputs, and it does not by itself establish circularity. Score 2 reflects the presence of a minor, non-load-bearing self-citation; the central empirical comparison remains independent.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The model's performance depends on several hand-set hyperparameters, most notably the SW loss schedule, rollout length, MCST depth/width, and data-quality thresholds. There are no invented physical entities. The domain assumptions about city-level aggregation, Kriging interpolation, and the sufficiency of two input steps are load-bearing for the empirical claims.

free parameters (5)
  • SW loss weight schedule = linearly decreasing from 5 to 0.1
    Eq. 16 weights early/late forecast steps; values are hand-chosen with no sensitivity analysis.
  • Autoregressive unroll length tau = 8
    Section 4.2 sets tau=8 during training; interaction with 20-step inference is not analyzed.
  • MCST layers and hidden dimension = L=3, din=128, heads=4
    Section 4.2; architecture capacity choices not tuned across values.
  • 1-hour model lead times = 1, 3, 6, 24 hours
    Appendix A.2; four single-step models chosen for greedy composition, affects error accumulation.
  • Data quality thresholds = 50% missing station removal; 20% missing time-step exclusion
    Appendix A.1; hand-set thresholds shape the released dataset and may affect generalizability.
assumptions (4)
  • domain assumption City-level pollutant concentration equals the maximum over monitoring stations in that city.
    Appendix A.1 uses station maximum to represent city level; this is a target choice, not validated against alternatives.
  • domain assumption Kriging interpolation adequately replaces missing station values.
    Appendix A.1 second step; bias in interpolation can propagate into forecasts.
  • domain assumption Two six-hourly observations are sufficient to initialize a 120-hour forecast.
    Eq. 14 and Tables 1-3; the central short-window claim assumes strong autocorrelation/periodicity without stationarity analysis.
  • domain assumption FuXi-2.0 forecast fields are accurate enough to act as future meteorology.
    Section 4.1 and Eq. 15; if AI weather forecasts drift, pollutant forecasts inherit errors.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MVAR: MultiVariate AutoRegressive Air Pollutants Forecasting Model." pith.science (2026). https://pith.science/paper/Y62C7PYH

@misc{pith2026250712023,
  author       = {Pith},
  title        = {Pith review of: MVAR: MultiVariate AutoRegressive Air Pollutants Forecasting Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y62C7PYH}},
  note         = {Machine review of arXiv:2507.12023}
}
read the original abstract

Air pollutants pose a significant threat to the environment and human health, thus forecasting accurate pollutant concentrations is essential for pollution warnings and policy-making. Existing studies predominantly focus on single-pollutant forecasting, neglecting the interactions among different pollutants and their diverse spatial responses. To address the practical needs of forecasting multivariate air pollutants, we propose MultiVariate AutoRegressive air pollutants forecasting model (MVAR), which reduces the dependency on long-time-window inputs and boosts the data utilization efficiency. We also design the Multivariate Autoregressive Training Paradigm, enabling MVAR to achieve 120-hour long-term sequential forecasting. Additionally, MVAR develops Meteorological Coupled Spatial Transformer block, enabling the flexible coupling of AI-based meteorological forecasts while learning the interactions among pollutants and their diverse spatial responses. As for the lack of standardized datasets in air pollutants forecasting, we construct a comprehensive dataset covering 6 major pollutants across 75 cities in North China from 2018 to 2023, including ERA5 reanalysis data and FuXi-2.0 forecast data. Experimental results demonstrate that the proposed model outperforms state-of-the-art methods and validate the effectiveness of the proposed architecture.

Figures

Figures reproduced from arXiv: 2507.12023 by the authors.

Figure 1
Figure 1. (a) The overall architecture of the proposed model. (b) Meteorological Coupled Spatial Transformer. (c) Multivariate Autoregressive [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. RMSE trend curve of the effect of Multivariate Prediction. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Relative RMSE trend curve of different training [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: RMSE Trend Curve of MVAR1hour output is Xt+3. During the prediction process, we iteratively invoke the four single-step prediction models, using the prediction result of the previous step as the input for the next step. A greedy algorithm is employed to reduce accumula…
Figure 6
Figure 6. Figure 6: Two pollutants evolution processes. provides a good prediction of the peak concentration timing. The second column presents a comparison of the PM2.5 pollution event. Starting from the forecast initiation, PM2.5 and PM10 concentrations first decrease, gradually rise af…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 25 canonical work pages

  1. [1]

    Adaptive graph convolutional recurrent network for traffic forecasting

    [Bai et al., 2020] Lei Bai, Lina Yao, Can Li, Xianzhi Wang, and Can Wang. Adaptive graph convolutional recurrent network for traffic forecasting. Advances in neural infor- mation processing systems, 33:17804–17815,

    work page 2020

  2. [4]

    However, in real-world scenarios, we aim for a finer temporal resolution of 1 hour to achieve more precise predictions

    A.2 Application - 1-Hour Resolution Air Pollutants Prediction In the experiments conducted in previous sections, we fore- cast air pollutant concentrations at 6-hour intervals, repre- senting a coarse-grained temporal resolution. However, in real-world scenarios, we aim for a finer temporal resolution of 1 hour to achieve more precise predictions. Therefo...

    work page 2023

  3. [7]

    Quantifying the health impact of pm2

    [Kryza et al., 2024] Maciej Kryza, Małgorzata Werner, Michael Holland, and Helen ApSimon. Quantifying the health impact of pm2. 5 with various chemical transport models and for different years- a case study for poland. Environment International, 194:109179,

    work page 2024

  4. [8]

    on May 8, 2023, and the PM 2.5 pollution event is initi- ated for forecasting at 20:00 (UTC-8) on November 22,

    work page 2023

  5. [9]

    Iden- tification of long-range transport pathways and potential source regions of pm2

    [Li et al., 2020] Hanlin Li, Qing He, and Xinchun Liu. Iden- tification of long-range transport pathways and potential source regions of pm2. 5 and pm10 at akedala station, cen- tral asia. Atmosphere, 11(11):1183,

    work page 2020

  6. [10]

    Airformer: Predicting nationwide air quality in china with transformers

    [Liang et al., 2023] Yuxuan Liang, Yutong Xia, Songyu Ke, Yiwei Wang, Qingsong Wen, Junbo Zhang, Yu Zheng, and Roger Zimmermann. Airformer: Predicting nationwide air quality in china with transformers. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 14329–14337,

    work page 2023

  7. [11]

    Probing the capacity of a spatiotem- poral deep learning model for short-term pm2

    [Liao et al., 2024] Qi Liao, Mingming Zhu, Lin Wu, Dawei Wang, Zixi Wang, Si Zhang, Wudi Cao, Xiaole Pan, Jie Li, Xiao Tang, et al. Probing the capacity of a spatiotem- poral deep learning model for short-term pm2. 5 forecasts in a coastal urban area. Science of The Total Environment, 950:175233,

    work page 2024

  8. [12]

    Forecasting air quality index in yan’an using temporal encoded informer

    [Ma et al., 2024] Shuai Ma, Jinrong He, Jinwei He, Qiang Feng, and Yingzhou Bi. Forecasting air quality index in yan’an using temporal encoded informer. Expert Systems with Applications, 255:124868,

    work page 2024

Show all 28 references
  1. [13]

    Significant impact of urban tree bio- genic emissions on air quality estimated by a bottom-up inventory and chemistry transport modeling

    [Maison et al., 2024] Alice Maison, Lya Lugon, Soo-Jin Park, Alexia Baudic, Christopher Cantrell, Florian Cou- vidat, Barbara d’Anna, Claudia Di Biagio, Aline Gratien, Val´erie Gros, et al. Significant impact of urban tree bio- genic emissions on air quality estimated by a bot...

    2024

  2. [14]

    The long hazy tail: Analysis of the impacts and trends of severe out- door and indoor air pollution in north china

    [Meng et al., 2024] Wenjun Meng, Yafang Cheng, Guofeng Shen, Huizhong Shen, Hang Su, and Shu Tao. The long hazy tail: Analysis of the impacts and trends of severe out- door and indoor air pollution in north china. Environmen- tal Science & Technology, 58(19):8326–8335,

    2024

  3. [15]

    Effect of air pollutants particulate matter (pm2

    [Meo et al., 2024] Sultan Ayoub Meo, Narmeen Shaikh, Metib Alotaibi, Abdullah Abdulziz AlWabel, and Hamid Alqumaidi. Effect of air pollutants particulate matter (pm2. 5, pm10), sulfur dioxide (so2) and ozone (o3) on cognitive health. Scientific Reports, 14(1):19616,

    2024

  4. [16]

    Integrated analysis of air quality-vegetation-health effects of near- future air pollution control strategies

    [Pan et al., 2025] Ruixin Pan, Jiqi Zhu, Dihui Chen, Haichun Cheng, Ling Huang, Yangjun Wang, and Li Li. Integrated analysis of air quality-vegetation-health effects of near- future air pollution control strategies. Environmental Pol- lution, 366:125407,

    2025

  5. [19]

    Spatial-temporal identity: A simple yet effective baseline for multivariate time series forecast- ing

    [Shao et al., 2022] Zezhi Shao, Zhao Zhang, Fei Wang, Wei Wei, and Yongjun Xu. Spatial-temporal identity: A simple yet effective baseline for multivariate time series forecast- ing. In Proceedings of the 31st ACM International Con- ference on Information & Knowledge Management...

    2022

  6. [20]

    Quantitative ev- idence from vocs source apportionment reveals o3 control strategies in northern and southern china

    [Wang et al., 2023] Zhenyu Wang, Xiao Tian, Jie Li, Feng Wang, Weiqing Liang, Huan Zhao, Bo Huang, Zaihua Wang, Yinchang Feng, and Guoliang Shi. Quantitative ev- idence from vocs source apportionment reveals o3 control strategies in northern and southern china. Environment In-...

    2023

  7. [21]

    Multi-step air quality index forecasting based on parallel multi-input transformers

    [Xie et al., 2023] Jie Xie, Jun Li, Mingying Zhu, and Qiong Wang. Multi-step air quality index forecasting based on parallel multi-input transformers. In Asian Conference on Pattern Recognition, pages 52–63. Springer,

    2023

  8. [22]

    Spatio-temporal graph convolutional networks: a deep learning framework for traffic forecasting

    [Yu et al., 2018] Bing Yu, Haoteng Yin, and Zhanxing Zhu. Spatio-temporal graph convolutional networks: a deep learning framework for traffic forecasting. In Proceedings of the 27th International Joint Conference on Artificial In- telligence, pages 3634–3640,

    2018

  9. [24]

    Deep learning for air pollutant concentration predic- tion: A review

    [Zhang et al., 2022] Bo Zhang, Yi Rong, Ruihan Yong, Dongming Qin, Maozhen Li, Guojian Zou, and Jianguo Pan. Deep learning for air pollutant concentration predic- tion: A review. Atmospheric Environment, 290:119347,

    2022

  10. [25]

    Fuxi-2.0: Advancing machine learning weather forecasting model for practical applications

    [Zhong et al., 2024] Xiaohui Zhong, Lei Chen, Xu Fan, Wenxu Qian, Jun Liu, and Hao Li. Fuxi-2.0: Advancing machine learning weather forecasting model for practical applications. arXiv preprint arXiv:2409.07188,

    2024 arXiv

  11. [28]

    NO 2, which participates in atmospheric photochemical reactions with O3, is also well captured by MV AR1hour in terms of concentration changes

    MV AR1hour effectively captures the diurnal variation of ozone, with strong consistency in pre- dicting the timing of O 3 concentration peaks. NO 2, which participates in atmospheric photochemical reactions with O3, is also well captured by MV AR1hour in terms of concentration...

    2023

  12. [2007]

    St-norm: Spatial and tem- poral normalization for multi-variate time series forecast- ing

    [Deng et al., 2021] Jinliang Deng, Xiusi Chen, Renhe Jiang, Xuan Song, and Ivor W Tsang. St-norm: Spatial and tem- poral normalization for multi-variate time series forecast- ing. In Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining , pages 269–278,

    2021

  13. [2016]

    Spatio-attention embedded recurrent neural network for air quality prediction

    [Huang et al., 2021] Yu Huang, Josh Jia-Ching Ying, and Vincent S Tseng. Spatio-attention embedded recurrent neural network for air quality prediction. Knowledge- Based Systems, 233:107416,

    2021

  14. [2018]

    Multi-step ahead forecasting of re- gional air quality using spatial-temporal deep neural net- works: a case study of huaihai economic zone

    [Zhang et al., 2020] Kefei Zhang, Jesse Th ´e, Guangyuan Xie, and Hesheng Yu. Multi-step ahead forecasting of re- gional air quality using spatial-temporal deep neural net- works: a case study of huaihai economic zone. Journal of Cleaner Production, 277:123231,

    2020

  15. [2020]

    Accurate medium- range global weather forecasting with 3d neural networks

    [Bi et al., 2023] Kaifeng Bi, Lingxi Xie, Hengheng Zhang, Xin Chen, Xiaotao Gu, and Qi Tian. Accurate medium- range global weather forecasting with 3d neural networks. Nature, 619(7970):533–538,

    2023

  16. [2021]

    Gaussian error linear units (gelus)

    [Hendrycks and Gimpel, 2016] Dan Hendrycks and Kevin Gimpel. Gaussian error linear units (gelus). arXiv preprint arXiv:1606.08415,

    2016 arXiv

  17. [2022]

    Half the world’s population are exposed to increasing air pollution

    [Shaddick et al., 2020] Gavin Shaddick, Matthew L Thomas, Pierpaolo Mudu, Giulia Ruggeri, and Sophie Gumy. Half the world’s population are exposed to increasing air pollution. NPJ Climate and Atmospheric Science, 3(1):1–5,

    2020

  18. [2023]

    Air pollution modeling–an overview

    [Daly and Zannetti, 2007] Aaron Daly and Paolo Zannetti. Air pollution modeling–an overview

    2007

  19. [2024]

    Learning skillful medium-range global weather forecasting

    [Lam et al., 2023] Remi Lam, Alvaro Sanchez-Gonzalez, Matthew Willson, Peter Wirnsberger, Meire Fortunato, Ferran Alet, Suman Ravuri, Timo Ewalds, Zach Eaton- Rosen, Weihua Hu, et al. Learning skillful medium-range global weather forecasting. Science, 382(6677):1416– 1421,

    2023

  20. [2025]

    Fourcast- net: A global data-driven high-resolution weather model using adaptive fourier neural operators

    [Pathak et al., 2022] Jaideep Pathak, Shashank Subrama- nian, Peter Harrington, Sanjeev Raja, Ashesh Chattopad- hyay, Morteza Mardani, Thorsten Kurth, David Hall, Zongyi Li, Kamyar Azizzadenesheli, et al. Fourcast- net: A global data-driven high-resolution weather model using ...

    2022 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.