Pith. sign in

REVIEW 4 major objections 5 minor 55 references

FuXi-Air: Urban Air Quality Forecasting Based on Emission-Meteorology-Pollutant multimodal Machine Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read FuXi-Air claims that an attention-based model fusing AI weather forecasts, a global emission inventory, and station observations produces 72-hour hourly forecasts for six pollutants across megacity monitoring networks in 25–30 seconds and…

desk verdict Promising integration of AI weather forecasts with emission and station data for urban air quality, but the headline claim against numerical models is undermined by an unfair comparison that gives FuXi-Air a persistent observational memory advantage. read the letter →

arxiv 2506.07616 v1 pith:XQXSZGDG submitted 2025-06-09 cs.LG

Zhixin Geng , Xu Fan , Xiqiao Lu , Yan Zhang , Guangyuan Yu , Cheng Huang , Qian Wang , Yuewu Li
show 4 more authors
Weichun Ma Qi Yu Libo Wu Hao Li
This is my paper
classification cs.LG
keywords airqualityforecastingFuXi-Airmultimodalmachinelearningmeteorology-emissioncouplingself-attentionemissioninventoryquantileurbanmegacities
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

FuXi-Air is a machine-learning system that forecasts urban air quality by fusing three data streams: AI-generated weather forecasts (FuXi-2.0), a monthly global emission inventory (CAMS), and ground-level pollutant observations from city monitoring networks. The paper argues that this fusion lets a single model forecast six pollutants at hourly resolution for 72 hours across all stations in a city in 25–30 seconds, and that in Shanghai the forecasts beat two configurations of the WRF-CMAQ numerical model for every pollutant, with O3 RMSE down 53.95% versus the SA07 configuration and 36.21% versus the CB6 configuration. If true, this matters because operational air-quality forecasting currently leans on numerical simulations that take hours to run and integrate observations poorly, while this approach runs in under a minute and outputs quantile-based uncertainty. The ablation results also show that meteorological inputs contribute more than emission inventories, but the best accuracy comes from using all three data types together.

What carries the argument

The central machinery is the meteorology–emission–pollutant coupling module: a cross-attention layer whose query is the self-attention-enhanced state of the monitoring sites and whose key-value pairs are residual-network features extracted from the gridded FuXi-2.0 weather fields and CAMS emission fields. Around it, an autoregressive base model predicts 6-hour steps and a frame-interpolation model fills the intervening hours, while relative positional encodings from site latitude and longitude and temporal embeddings of hour-of-day and day-of-year inform the attention. This design is what lets the model move information from gridded weather and emissions to irregularly spaced urban stations and produce hourly multi-pollutant forecasts with limited dependence on dense input.

What would settle it

Run the Shanghai 2023 benchmark using only FuXi-2.0 forecasts as meteorological input, aligned to the same initialization times as the WRF-CMAQ runs, and compare O3 RMSE over steps 4-72; the claim collapses if the reported 53.95% and 36.21% reductions over SA07 and CB6, respectively, do not persist.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that a transformer-based, attention-driven model can learn the joint dynamics of pollution, weather, and emissions well enough to match or beat numerical air-quality models for operational urban forecasting. The model combines a self-attention module that encodes relationships among monitoring sites, a residual convolutional network that extracts features from gridded meteorological and emission fields, and a cross-attention module in which site states query those gridded features; a 6-hour autoregressive forecaster produces the long-range trend and a frame-interpolation model refills the hourly steps. A quantile loss gives probabilistic forecasts at different percentiles. Evaluated on 2023 data for Beijing, Shanghai, and Shenzhen, the model keeps mean relative errors for major pollutants below 53% over 72 hours, reaches $R=0.85$ for O3, and in Shanghai reduces RMSE relative to WRF-CMAQ for all six pollutants; it also reports FuXi-2.0-driven forecasts close to ERA5-driven ones.

Load-bearing premise

The load-bearing premise is that the AI weather forecasts used in real operation are accurate enough to preserve the skill learned from the historical ERA5 reanalysis fields, since the paper's own Table 2 shows a systematic RMSE degradation when the driver switches to FuXi-2.0 and the main Shanghai comparison does not state which driver was used.

Editorial extensions

If this is right

  • Air-quality agencies could run 72-hour forecasts for a megacity in under a minute instead of waiting 2–3 hours for WRF-CMAQ, enabling rapid updates when weather changes.
  • The quantile-loss output gives 10th and 90th percentile bands, so the same model can drive probabilistic pollution warnings rather than point forecasts only.
  • Because the base model needs only two 6-hour-apart observation frames plus weather and emission fields, the architecture tolerates missing or noisy station data better than models that require complete hourly histories.
  • Offline coupling with AI weather forecasts extends the practical horizon: FuXi-2.0-driven accuracy approaches ERA5-driven accuracy, so longer or more frequent operational integrations are feasible.
  • The Shanghai comparison suggests that a data-driven urban model can outperform a physics-based regional model even with coarser emission inputs, which changes the cost-benefit calculus for operational forecasting systems.

Reading between the lines

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

  • A direct test the paper does not run is whether training on FuXi-2.0 fields, or mixing ERA5 and FuXi fields during training, closes the Table 2 degradation gap; if it does, the operational headline numbers would be more directly supported.
  • Because the CAMS inventory is monthly, the emission branch will miss hour-to-hour emission changes; coupling with traffic or industrial activity data should improve primary pollutants like NO2 and CO more than secondary ones like O3.
  • The cross-city success is demonstrated by retraining per city; the paper does not test zero-shot transfer from one city to another, so 'generalization' here means architectural portability rather than transfer without data.
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

4 major / 5 minor

Summary. The paper introduces FuXi-Air, a multimodal, attention-based machine learning model for urban air quality forecasting. It combines FuXi-2.0 AI meteorological forecasts, CAMS emission inventories, and station-level pollutant observations, and couples an autoregressive 6-hour forecasting module with a frame interpolation module to produce hourly 72-hour forecasts for six pollutants at multiple monitoring sites in Beijing, Shanghai, and Shenzhen. The model is evaluated on independent 2023 data; Table 1 reports MRE, MAE, RMSE, and R for all six pollutants across the three cities. Section 3.2.3 compares FuXi-Air with two WRF-CMAQ configurations in Shanghai and reports large RMSE reductions (e.g., O3 reduced by 53.95% vs SA07 and 36.21% vs CB6). Section 3.3.1 presents ablations showing that meteorological inputs contribute more than emission inventories, and Section 3.3.2 compares FuXi-driven and ERA5-driven forecasts. The paper claims operational efficiency (25–30 seconds per 72-hour forecast) and generalizable skill across megacities with different pollution regimes.

Significance. If the comparative claims are supported, the integration of an AI-based meteorological forecast, a dynamic emission inventory, and station observations in a single forecast model is a meaningful contribution to operational air quality forecasting. The evaluation uses independent 2023 observations, which avoids circularity, and the multi-city design with three cities of different pollution regimes is a strength. The explicit ERA5-versus-FuXi comparison in Table 2 is informative and helps quantify the cost of switching from reanalysis to forecast meteorological drivers. The paper also acknowledges limitations of monthly emission inventories and regional transport in the conclusion, which is a useful framing. However, the central comparative claim against numerical models rests on a single city, two WRF-CMAQ configurations, and a comparison that does not control for the model's use of live pollutant observations; the reported improvements for PM2.5, NO2, and PM10 are small and are not accompanied by uncertainty intervals or significance tests.

major comments (4)
  1. [Section 3.2.3, Figure 4] The comparison between FuXi-Air and WRF-CMAQ is not an apples-to-apples test of model skill. FuXi-Air consumes the observed pollutant concentrations X_{t-6} and X_t at each autoregressive step (Section 3.1.1), whereas the WRF-CMAQ configurations described in Section 3.2.3 are not reported to include any observational initialization or data assimilation. Because the autoregressive chain begins from X_t, forecasts at lead times 4–6 hours and several subsequent steps still carry strong observation-memory; the restriction to steps 4–72 does not remove this advantage. The reported RMSE reductions therefore conflate the value of live observational initialization with the learned emission–meteorology–pollutant mapping. Please provide lead-time-stratified comparisons (e.g., 25–48 h and 49–72 h only), and either initialize WRF-CMAQ with equivalent observations or clearly frame the comparison as 'operational ML with observational input versus pure numerical forecast' and quantify the information asymmetry.
  2. [Section 3.2.3, Figure 4] No uncertainty intervals or significance tests are provided for the RMSE and R differences. The reported gains are small for several pollutants (e.g., PM2.5 reduced by 11.29% vs SA07 and 5.30% vs CB6; NO2 reduced by 12.09% and 6.29%; PM10 reduced by 6.41% and 2.41%), and without confidence intervals or a significance test it is unclear whether these differences are within sampling noise. The paper also does not report the number of station-hours used in the comparison. Please add confidence intervals, significance tests, or both, and state the effective sample size.
  3. [Table 1 and Section 3.2.1] The captions of Table 1 and the surrounding text do not state which meteorological driver (ERA5 or FuXi-2.0) produced the headline results. Table 2 shows that switching from ERA5 to FuXi increases RMSE for several pollutants (e.g., NO2 49–72 h: 17.30 vs 15.22; PM2.5 49–72 h: 19.10 vs 16.22). The Shanghai values in Table 1 match the FuXi column of Table 2, which suggests that Table 1 is FuXi-driven, but this must be stated explicitly in Sections 2 and 3 and in the abstract. If any headline metric is ERA5-driven, the operational claim would not be directly supported.
  4. [Section 3.2.3 and Abstract] The abstract and conclusion claim that FuXi-Air 'outperforms the mainstream numerical air quality models in operational forecasting work,' but the evidence is limited to Shanghai and two WRF-CMAQ configurations with different chemical mechanisms and nesting resolutions. The claimed generality is broader than the evidence. Please narrow the claim to the tested setting or add at least one additional city comparison or a quantitative benchmark against published operational forecast skill.
minor comments (5)
  1. [Section 3.3.1] The text refers to 'Table 2 in the Dataset Introduction section,' but Table 2 is located in Section 3.3.2; the cross-reference should be fixed.
  2. [Figure 4 and Supplementary Figure 10] The caption says 'steps 4–72' but does not explicitly state that 'steps' are forecast hours from initialization or which initialization cycles are used. Please clarify this in the caption and in the text.
  3. [Supplementary Figures 1–9] The supplementary figures show 10th and 90th percentile forecasts, but no probabilistic verification is provided (e.g., empirical coverage or CRPS). Adding a coverage or CRPS evaluation would strengthen the quantile-loss claim.
  4. [Abstract and Author List] There are several language and formatting issues: 'highly precision' in the abstract, 'for the first time' in Section 3.3.2 should be substantiated or removed, and author names such as 'Y u' and 'Y uewu Li' appear with spacing artifacts from LaTeX.
  5. [Data Availability] Provide a direct access link or contact for the FuXi-2.0 forecast data, similar to the links given for CAMS and ERA5.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: FuXi-Air forecasts are scored against withheld station observations, and the FuXi meteorological input is explicitly checked against ERA5 rather than assumed.

full rationale

The paper's derivation chain is not circular. FuXi-Air is trained with station pollutant observations, CAMS emission inventories, and FuXi/ERA5 meteorological fields, and its 72-hour forecasts are evaluated against independent 2023 station observations in Table 1 and Section 3.2.1. The autoregressive use of the current observation X_t as an initial condition is a standard forecasting setup, not a relabeling of the target as an input. No equation defines an input in terms of the predicted quantity, and no fitted parameter is later presented as a prediction. Section 3.3.2 and Table 2 explicitly compare FuXi-driven and ERA5-driven model variants against observations, providing an external check of the FuXi-2.0 meteorological driver rather than relying on the authority of the authors' prior FuXi papers. The WRF-CMAQ comparison in Section 3.2.3 is restricted to steps 4-72 because of initialization-time differences, and the paper reports the comparison transparently. The only caveat, that FuXi-Air begins from live pollutant observations while the WRF-CMAQ runs are not described as assimilating such observations, is an evaluation-fairness concern and not a circular-reasoning issue. Self-citations to FuXi papers are input-source citations and are not load-bearing for the central accuracy claim.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities or conservation laws. It relies on standard ML components and domain data. The main 'free parameters' are architectural choices, not fitted physical constants. The key domain assumption is the adequacy of FuXi forecasts and the representativeness of station data.

free parameters (3)
  • Temporal input step (X_t-6 and X_t)
    The model uses only two 6-hourly time steps as input, chosen by hand to reduce data dependence. This is a modeling choice rather than a fitted value, but it is central to the design.
  • 6-hour autoregressive interval and interpolation strategy
    The forecast is generated in 6-hour steps then interpolated to 1-hour resolution. This temporal design is a manual architectural choice that could affect accuracy.
  • Quantile levels for quantile loss
    The paper mentions quantile loss but does not specify which quantile levels were used (e.g., 10th, 50th, 90th). These would be tuned choices that affect the reported forecasts.
assumptions (4)
  • standard math Attention mechanisms and cross-attention operations are applied as described, following standard definitions (Eq. 6-14).
    The model relies on standard self-attention and cross-attention, which are accepted mathematical tools.
  • domain assumption Monitoring station pollutant data are representative of city-scale air quality.
    The model is trained and evaluated on data from 11-19 stations per city; this assumes these stations capture the spatial variability relevant for forecasting.
  • domain assumption CAMS-GLOB-ANT emission inventory provides a useful representation of urban emissions despite monthly resolution and possible delays.
    The ablation shows emissions add value, but the inventory is coarse in time; the paper assumes this is adequate.
  • domain assumption FuXi-2.0 meteorological forecasts are accurate enough for air quality forecasting.
    The operational claim depends on FuXi forecast quality; Table 2 shows some degradation relative to ERA5, so this assumption is load-bearing.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FuXi-Air: Urban Air Quality Forecasting Based on Emission-Meteorology-Pollutant multimodal Machine Learning." pith.science (2026). https://pith.science/paper/XQXSZGDG

@misc{pith2026250607616,
  author       = {Pith},
  title        = {Pith review of: FuXi-Air: Urban Air Quality Forecasting Based on Emission-Meteorology-Pollutant multimodal Machine Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XQXSZGDG}},
  note         = {Machine review of arXiv:2506.07616}
}
read the original abstract

Air pollution has emerged as a major public health challenge in megacities. Numerical simulations and single-site machine learning approaches have been widely applied in air quality forecasting tasks. However, these methods face multiple limitations, including high computational costs, low operational efficiency, and limited integration with observational data. With the rapid advancement of artificial intelligence, there is an urgent need to develop a low-cost, efficient air quality forecasting model for smart urban management. An air quality forecasting model, named FuXi-Air, has been constructed in this study based on multimodal data fusion to support high-precision air quality forecasting and operated in typical megacities. The model integrates meteorological forecasts, emission inventories, and pollutant monitoring data under the guidance of air pollution mechanism. By combining an autoregressive prediction framework with a frame interpolation strategy, the model successfully completes 72-hour forecasts for six major air pollutants at an hourly resolution across multiple monitoring sites within 25-30 seconds. In terms of both computational efficiency and forecasting accuracy, it outperforms the mainstream numerical air quality models in operational forecasting work. Ablation experiments concerning key influencing factors show that although meteorological data contribute more to model accuracy than emission inventories do, the integration of multimodal data significantly improves forecasting precision and ensures that reliable predictions are obtained under differing pollution mechanisms across megacities. This study provides both a technical reference and a practical example for applying multimodal data-driven models to air quality forecasting and offers new insights into building hybrid forecasting systems to support air pollution risk warning in smart city management.

Figures

Figures reproduced from arXiv: 2506.07616 by the authors.

Figure 1
Figure 1. Schematic diagram of the proposed model structure. 2.2 Study Area and Selected Cities Three representative cities in China—Beijing, Shanghai, and Shenzhen—are selected as the target cities in this study ( [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Study area and site distribution map. 2.3 Dataset Construction Three main data sources are utilized in this study for the cities of Beijing, Shanghai, and Shenzhen: (1) national air quality monitoring station data concerning pollutant concentrations, (2) atmospheric pollutant monitoring and emission inventory data from the Copernicus Atmosphere Monitoring Service (CAMS), and (3) meteorological forecast data from the… view at source ↗
Figure 3
Figure 3. MRE evaluation metric curves produced by the FuXi-Air model for six pollutants over a 72-hour forecasting horizon in Shanghai, Beijing, and Shenzhen. 8/37 [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Comparison among the 72-hour forecasts of the FuXi-Air model and two numerical models in Shanghai. The SA07 configuration (CMAQ–SA07) adopted a 3-domain nested structure (36–12 km resolution). The CB06 configuration (CMAQ–CB06) applied a finer 4-domain nesting (27–9–3 …
Figure 5
Figure 5. Figure 5: R curves derived from ablation experiments involving Beijing, Shanghai, and Shenzhen. 13/37 [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Time series forecasts produced for pollution episodes in Shanghai by different model configurations (timestamps in China Standard Time, CST) 15/37 [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Differences among the evaluation metrics produced across different ablation experimental groups. Panels (a) and (b): DEEMS vs. STN_ONLY; (c)(d): ALL vs. DEMET; (e)(f): DEMET vs. STN_ONLY; and (g)(h): ALL vs. DEEMS. 18/37 [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

55 extracted references · 51 canonical work pages

  1. [1]

    & Chen, H.-L

    Chen, S.-L., Chang, S.-W., Chen, Y .-J. & Chen, H.-L. Possible warming effect of fine particulate matter in the atmosphere. Commun. Earth & Environ. 2, 208 (2021)

    work page 2021

  2. [2]

    & Long, Y

    Xu, X., Huang, L., Yao, L., Yoshida, Y . & Long, Y . Rising socio-economic costs of pm2.5 pollution and medical service mismatching. Nat. Sustain. 1–11 (2025)

    work page 2025

  3. [3]

    Long-term exposure to ambient pm 2.5, particulate constituents and hospital admissions from non-respiratory infection

    Feng, Y .et al. Long-term exposure to ambient pm 2.5, particulate constituents and hospital admissions from non-respiratory infection. Nat. Commun. 15, 1518 (2024)

    work page 2024

  4. [4]

    Environmental Protection Agency

    U.S. Environmental Protection Agency. Integrated science assessment (isa) for particulate matter (final report, dec 2019). Tech. Rep. EPA/600/R-19/188, U.S. Environmental Protection Agency, Washington, DC (2019). Final Report

    work page 2019

  5. [5]

    Liang, L., Wang, Z. & Li, J. The effect of urbanization on environmental pollution in rapidly developing urban agglomerations. J. Clean. Prod. 237, 117649 (2019)

    work page 2019

  6. [6]

    Wang, J. et al. Anthropogenic emissions and urbanization increase risk of compound hot extremes in cities. Nat. Clim. Chang. 11, 1084–1089 (2021)

    work page 2021

  7. [7]

    Amann, M. et al. Cost-effective control of air quality and greenhouse gases in europe: Modeling and policy applications. Environ. Model. & Softw. 26, 1489–1501 (2011)

    work page 2011

  8. [8]

    H., Apeh, C

    Ekeh, A. H., Apeh, C. E., Odionu, C. S. & Austin-Gabriel, B. Leveraging machine learning for environmental policy innovation: Advances in data analytics to address urban and ecological challenges. Gulf J. Adv. Bus. Res. 3, 456–482 (2025)

    work page 2025

Show all 55 references
  1. [9]

    & Hao, Y

    Yang, W., Wang, J., Zhang, K. & Hao, Y . A novel air pollution forecasting, health effects, and economic cost assessment system for environmental management: From a new perspective of the district-level. J. Clean. Prod. 417, 138027 (2023). 20/37

    2023

  2. [10]

    & Wang, Z

    Ding, Z., Chen, H., Zhou, L. & Wang, Z. A forecasting system for deterministic and uncertain prediction of air pollution data. Expert. Syst. with Appl. 208, 118123 (2022)

    2022

  3. [11]

    Liu, Y ., Zhou, Y . & Lu, J. Exploring the relationship between air pollution and meteorological conditions in china under environmental governance. Sci. Reports 10, 14518 (2020)

    2020

  4. [12]

    pollution reduction and carbon reduction synergy

    Wen, W. et al. Assessing aerosol-radiation interaction with wrf-chem-solar: Case study on the impact of the “pollution reduction and carbon reduction synergy” policy. Atmospheric Res. 308, 107537 (2024)

    2024

  5. [13]

    Mai, Z. et al. Convolutional neural networks facilitate process understanding of megacity ozone temporal variability. Environ. Sci. & Technol. 58, 15691–15701 (2024)

    2024

  6. [14]

    & Mauzerall, D

    Zhou, M., Xie, Y ., Wang, C., Shen, L. & Mauzerall, D. L. Impacts of current and climate induced changes in atmospheric stagnation on indian surface pm2.5 pollution. Nat. Commun. 15, 7448 (2024)

    2024

  7. [15]

    & Yuan, H

    Han, H., Liu, J., Shu, L., Wang, T. & Yuan, H. Local and synoptic meteorological influences on daily variability in summertime surface ozone in eastern china. Atmospheric Chem. Phys. 20, 203–222 (2020)

    2020

  8. [16]

    Schnell, J. L.et al. Exploring the relationship between surface pm2.5 and meteorology in northern india.Atmospheric Chem. Phys. 18, 10157–10175 (2018)

    2018

  9. [17]

    L., Dayan, U., Levy, I

    Yuval, Y . L., Dayan, U., Levy, I. & Broday, D. M. On the association between characteristics of the atmospheric boundary layer and air pollution concentrations. Atmospheric Res. 231, 104675 (2020)

    2020

  10. [18]

    & Liu, S

    Miao, Y ., Che, H., Zhang, X. & Liu, S. Relationship between summertime concurring pm2.5 and o3 pollution and boundary layer height differs between beijing and shanghai, china. Environ. Pollut. 268, 115775 (2021)

    2021

  11. [19]

    Evolution of ozone formation sensitivity during a persistent regional ozone episode in northeastern china and its implication for a control strategy

    Zhang, Y .et al. Evolution of ozone formation sensitivity during a persistent regional ozone episode in northeastern china and its implication for a control strategy. Environ. Sci. & Technol. 58, 617–627 (2023)

    2023

  12. [20]

    Sulaymon, I. D. et al. Using the covid-19 lockdown to identify atmospheric processes and meteorology influences on regional pm2.5 pollution episodes in the beijing-tianjin-hebei, china. Atmospheric Res. 294, 106940 (2023)

    2023

  13. [21]

    & Ying, Z

    Tie, X., Brasseur, G. & Ying, Z. Impact of model resolution on chemical ozone formation in mexico city: application of the wrf-chem model. Atmospheric Chem. Phys. 10, 8983–8995 (2010)

    2010

  14. [22]

    Wang, L. et al. Causal-inference machine learning reveals the drivers of china’s 2022 ozone rebound. Environ. Sci. Ecotechnology 24, 100524 (2025)

    2025

  15. [23]

    Li, L. et al. Ozone source apportionment over the yangtze river delta region, china: Investigation of regional transport, sectoral contributions and seasonal differences. Atmospheric Environ. 202, 269–280 (2019)

    2019

  16. [24]

    Zhang, C. et al. Source assessment of atmospheric fine particulate matter in a chinese megacity: Insights from long-term, high-time resolution chemical composition measurements from shanghai flagship monitoring supersite. Chemosphere 251, 126598 (2020)

    2020

  17. [25]

    Yu, S.et al. Eta-cmaq air quality forecasts for o3 and related species using three different photochemical mechanisms (cb4, cb05, saprc-99): comparisons with measurements during the 2004 icartt study. Atmospheric Chem. Phys. 10, 3001–3025 (2010)

    2010

  18. [26]

    Chen, L. et al. Fuxi: A cascade machine learning forecasting system for 15-day global weather forecast. npj climate atmospheric science 6, 190 (2023)

    2023

  19. [27]

    Lam, R. et al. Learning skillful medium-range global weather forecasting. Science 382, 1416–1421 (2023)

    2023

  20. [28]

    Pathak, J. et al. Fourcastnet: A global data-driven high-resolution weather model using adaptive fourier neural operators. arXiv preprint arXiv:2202.11214 (2022)

    2022 arXiv

  21. [29]

    Wu, C.-l. et al. A hybrid deep learning model for regional o 3 and no 2 concentrations prediction based on spatiotemporal dependencies in air quality monitoring network. Environ. pollution 320, 121075 (2023)

    2023

  22. [30]

    & Huang, X.-F

    Cheng, Y ., He, L.-Y . & Huang, X.-F. Development of a high-performance machine learning model to predict ground ozone pollution in typical cities of china. J. Environ. Manag. 299, 113670 (2021)

    2021

  23. [31]

    & Liu, D

    Qi, Y ., Li, Q., Karimian, H. & Liu, D. A hybrid model for spatiotemporal forecasting of pm2.5 based on graph convolutional neural network and long short-term memory. Sci. The Total. Environ. 664, 1–10 (2019)

    2019

  24. [32]

    Bodnar, C. et al. A foundation model for the earth system. Nature (2025). 21/37

    2025

  25. [33]

    & Liu, J

    Zhu, J., Niu, L., Zheng, P., Chen, H. & Liu, J. A hybrid pm2.5 interval concentration prediction framework based on multi-factor interval decomposition reconstruction strategy and attention mechanism. Atmospheric Environ. 335, 120730 (2024)

    2024

  26. [34]

    & Thakur, M

    Mandal, S. & Thakur, M. A city-based pm2.5 forecasting framework using spatially attentive cluster-based graph neural network model. J. Clean. Prod. 405, 137036 (2023)

    2023

  27. [35]

    & Shen, W

    Zhang, C., Wang, S., Wu, Y ., Zhu, X. & Shen, W. A long-term prediction method for pm2.5 concentration based on spatiotemporal graph attention recurrent neural network and grey wolf optimization algorithm. J. Environ. Chem. Eng. 12, 111716 (2024)

    2024

  28. [36]

    Prediction of air pollutant concentration based on self-attention mechanism lstm model

    Liu, H. Prediction of air pollutant concentration based on self-attention mechanism lstm model. In International Conference on High Performance Computing and Communication (HPCCE 2021) , vol. 12162, 325–329 (SPIE, 2022)

    2021

  29. [37]

    & Qian, W

    Guyu, Z., Xiaoyuan, Y ., Jiansen, S., Hongdou, H. & Qian, W. A pm2.5 spatiotemporal prediction model based on mixed graph convolutional gru and self-attention network. Environ. Pollut. 125748 (2025)

    2025

  30. [38]

    Zhang, B. et al. Air pollutant diffusion trend prediction based on deep learning for targeted season—north china as an example. Expert. Syst. with Appl. 232, 120718 (2023)

    2023

  31. [39]

    & Yue, S

    Wang, Z., Yang, Y . & Yue, S. Air quality classification and measurement based on double output vision transformer. IEEE Internet Things J. 9, 20975–20984 (2022)

    2022

  32. [40]

    Baruah, A. et al. A novel spatiotemporal prediction approach to fill air pollution data gaps using mobile sensors, machine learning and citizen science techniques. npj Clim. Atmospheric Sci. 7, 310 (2024)

    2024

  33. [41]

    & Wang, P

    Li, B. & Wang, P. A multi-task stations cooperative air quality prediction system for sustainable development. Humanit. Soc. Sci. Commun. 11, 1–11 (2024)

    2024

  34. [42]

    & Chen, J

    Liu, B., Lai, M., Zeng, P. & Chen, J. Air pollutant prediction based on a attention mechanism model of the yangtze river delta region in frequent heatwaves. Atmospheric Res. 311, 107701 (2024)

    2024

  35. [43]

    Zhang, H. et al. Air pollution and control action in beijing. J. Clean. Prod. 112, 1519–1527 (2016)

    2016

  36. [44]

    Feng, X. et al. Formation and dominant factors of haze pollution over beijing and its peripheral areas in winter. Atmospheric Pollut. Res. 5, 528–538 (2014)

    2014

  37. [45]

    & Gao, W

    Chang, L., He, F., Tie, X., Xu, J. & Gao, W. Meteorology driving the highest ozone level occurred during mid-spring to early summer in shanghai, china. Sci. Total. Environ. 785, 147253 (2021)

    2021

  38. [46]

    A., Ojha, N

    Coates, J., Mar, K. A., Ojha, N. & Butler, T. M. The influence of temperature on ozone production under varying nox conditions–a modelling study. Atmospheric Chem. Phys. 16, 11601–11615 (2016)

    2016

  39. [47]

    Seinfeld, J. H. & Pandis, S. N. Atmospheric chemistry and physics: from air pollution to climate change (John Wiley & Sons, 2016)

    2016

  40. [48]

    & Wang, Y

    Kong, L., Xin, J., Zhang, W. & Wang, Y . The empirical correlations between pm2.5, pm10 and aod in the beijing metropolitan region and the pm2.5, pm10 distributions retrieved by modis. Environ. pollution 216, 350–360 (2016)

    2016

  41. [49]

    Observation of regional air pollutant transport between the megacity beijing and the north china plain

    Li, Y .et al. Observation of regional air pollutant transport between the megacity beijing and the north china plain. Atmospheric Chem. Phys. 16, 14265–14283 (2016)

    2016

  42. [50]

    Yin, L. et al. Regional-specific trends of pm2. 5 and o3 temperature sensitivity in the united states. npj Clim. Atmospheric Sci. 8, 12 (2025)

    2025

  43. [51]

    Granier, C. et al. The Copernicus atmosphere monitoring service global and regional emissions (April 2019 version). Ph.D. thesis, Copernicus Atmosphere Monitoring Service (2019)

    2019

  44. [52]

    Soulie, A. et al. Global anthropogenic emissions (cams-glob-ant) for the copernicus atmosphere monitoring service simulations of air quality forecasts and reanalyses. Earth Syst. Sci. Data Discuss. 2023, 1–45 (2023)

    2023

  45. [53]

    Hersbach, H. et al. The era5 global reanalysis. Q. journal royal meteorological society 146, 1999–2049 (2020)

    2020

  46. [54]

    Zhong, X. et al. Fuxi-2.0: Advancing machine learning weather forecasting model for practical applications. arXiv preprint arXiv:2409.07188 (2024)

    2024 arXiv

  47. [55]

    Technical Specifications for Ambient Air Quality Monitoring

    Sun, X. et al. Fuxi weather: An end-to-end machine learning weather data assimilation and forecasting system. arXiv preprint arXiv:2408.05472 (2024). 22/37 Supplementary materials Supplementary materials A. Datasets Introduction ♦ CAMS-GLOB-ANT: CAMS-GLOB-ANT (Copernicus Atmos...

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.