Pith. sign in

REVIEW 4 major objections 4 minor 52 references

PCDCNet: A Surrogate Model for Air Quality Forecasting with Physical-Chemical Dynamics and Constraints

T0 review · 4 major / 4 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 PCDCNet embeds emissions, meteorology, and mass-conservation constraints into a graph-recurrent deep network and reports lower 72-hour PM2.5 and O3 forecast errors than existing baselines in two Chinese regions.

desk verdict PCDCNet probably delivers real forecasting gains, but the mass-conservation loss is not what the paper claims it to be; the physical-consistency story needs substantial repair before the paper is publishable as is. read the letter →

arxiv 2505.19842 v2 pith:STA7GKIV submitted 2025-05-26 cs.LG cs.AI

Shuo Wang , Yun Cheng , Qingye Meng , Olga Saukh , Jiang Zhang , Jingfang Fan , Yuanting Zhang , Xingyuan Yuan
show 1 more author
Lothar Thiele
This is my paper · ORCID
classification cs.LGcs.AI
keywords airqualityforecastingsurrogatemodelgraphneuralnetworkphysicalconstraintsPM2.5ozoneemissioninventoryreal-time
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

The paper claims that a deep surrogate model for air quality forecasting can match the interpretability of numerical chemical transport models while staying fast enough for real-time use. PCDCNet explicitly folds in emissions, forecasted meteorology, and a loss that enforces the atmospheric continuity equation, then reports the lowest mean absolute error for 72-hour PM2.5 and O3 forecasts across two Chinese regions. The reported gains are 19.8% for PM2.5 and 18.4% for O3 over the next best baseline, with errors growing more slowly at long lead times than comparator models. A sympathetic reader would take the contribution as evidence that embedding physical constraints and domain inputs into a graph-recurrent network improves both accuracy and generalization for station-level forecasting.

What carries the argument

The load-bearing mechanism is the interplay of three modules and a residual prediction step. The Local Interaction Dynamics module uses an MLP over concatenated past concentrations, meteorology, and emissions to capture local chemistry; the Spatial Transport Dynamics module applies a graph convolution with the normalized Laplacian of a 200 km geodesic station graph, with a readout that produces per-node transport gradients; and the Temporal Accumulation Dynamics module uses a GRU cell to accumulate past states. Prediction is iterative: at each step the model outputs a concentration change $\Delta \hat{X}_t = \mathrm{Linear}(H_t)$ and updates $\hat{X}_t = \hat{X}_{t-1} + \Delta \hat{X}_t$. The Domain-Informed Constraints loss is designed to enforce the continuity equation by requiring the spatial transport gradients to sum to zero across nodes and to remain stable over time, though the implemented loss (Eq. 16) acts on node-level aggregates rather than extracted pairwise fluxes.

What would settle it

Train PCDCNet and then compute the pairwise transport fluxes $\nabla \hat{X}^t_{M,v\to v'}$ from the STD module's per-edge messages (if they exist); verify numerically whether, for every node and time step, $\sum_{v'} \nabla \hat{X}^t_{M,v\to v'}=0$ and whether the DIC loss in Eq. 16 is actually a sum over these fluxes. If the per-edge fluxes are not extractable or their sum is not zero, the mass-conservation claim is not supported, even though the MAE numbers may remain valid.

Watch

Extended reading notes

Core claim

Stated on its own terms, the paper's central discovery is that a forecasting network structured like a chemical transport model—local interaction MLP, graph convolution for advective-diffusive transport, GRU for temporal accumulation, and a domain-informed loss that penalizes violation of the continuity equation—achieves state-of-the-art 72-hour station-level PM2.5 and O3 forecasts. On the BTHSA and YRD test sets it reports MAE reductions of 19.8% (PM2.5) and 18.4% (O3) relative to the next best method, and it maintains relatively stable accuracy as the horizon extends to 72 hours, where baselines' errors climb. The paper also claims that the physical constraint improves test-set generalization, and that the deployed system provides free real-time forecasts.

Load-bearing premise

The physical-consistency claim rests on the assumption that the spatial transport module's output can be decomposed into conservative pairwise pollutant fluxes between stations, so that the DIC loss genuinely enforces the continuity equation; the implementation, however, only constrains node-level aggregates and never extracts those pairwise fluxes.

Editorial extensions

If this is right

  • If the reported gains hold, 72-hour station-level forecasts can be produced cheaply enough for public real-time services, replacing or supplementing expensive CMAQ-type simulations.
  • Explicit inclusion of emissions and forecasted meteorology contributes to accuracy: the paper reports 9.8% (PM2.5) and 3.7% (O3) error reductions over variants without emissions data.
  • The physical constraint improves test-set MAE, suggesting that continuity-equation penalties act as a beneficial regularizer.
  • The architecture's CPU-friendly deployment and hourly update cycle make it practical for operational air quality management.
  • The reported case studies (Spring Festival haze, Beijing pollution, California wildfires) indicate the model captures event-driven pollution dynamics outside its training regions.

Reading between the lines

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

  • A direct ablation that varies DIC strength while holding the input set fixed would separate the benefit of the physics loss from the benefit of the extra exogenous inputs (emissions, forecasted meteorology).
  • The paper's physical-consistency claim is stronger than the implementation supports, because the DIC loss in Eq. 16 is applied to node-level gradients, not to the pairwise station fluxes postulated in Eq. 14; checking whether message passing actually decomposes into conservative pairwise fluxes would test whether the constraint is genuine physics or merely a regularizer.
  • The same framework could be transferred to other advected quantities (dust, pollen, chemical species) by swapping the emission inventory and reaction parameterization, provided the station graph captures the transport scale.
  • The reported competition result for the Guangdong-Hong Kong-Macao region would strengthen the generalization claim if quantified on that region's test data, which is not part of the reported BTHSA/YRD evaluation.
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 / 4 minor

Summary. The paper proposes PCDCNet, a deep-learning surrogate model for 72-hour station-level PM2.5 and O3 forecasting over two regions in China (BTHSA and YRD). The architecture combines a graph-based spatial transport module (STD), a GRU for temporal accumulation (TAD), and an MLP for local interactions (LID). A domain-informed constraint (DIC) loss is introduced with the stated purpose of enforcing mass conservation and the atmospheric continuity equation. The authors report lower MAE and RMSE than several baselines (XGBoost, LightGBM, GC-LSTM, PM2.5-GNN, iTransformer, TimeXer, AirPhyNet) on a 2016–2023 dataset, and describe a deployed real-time forecasting platform.

Significance. If the reported forecasting gains hold, PCDCNet is a practically valuable, computationally efficient surrogate for operational air quality forecasting, with strong societal impact through its deployed service. The paper contains a large real-world evaluation, a reproducible experimental setup, and a clear comparison to relevant baselines. However, the paper's second central claim—that the model enforces mass conservation and physical consistency—is not supported by the implementation as written. The DIC loss in Eq. 13 and Algorithm 1 is a temporal smoothness penalty on node-level STD readouts, not a mass-conservation constraint, and the pairwise flux in Eq. 14 is never defined or extracted from the model. The empirical error reductions are credible from Table 1, but the headline percentages (19.8% and 18.4%) do not match the table's numbers. The physical-consistency narrative requires either a corrected implementation or a substantially weakened claim.

major comments (4)
  1. [§3.3.2–§3.3.3, Algorithm 1] The implemented DIC loss is not a mass-conservation constraint. Eq. 13, which is the only DIC term actually computed in Algorithm 1 (line 11), penalizes the temporal difference of node-level STD readouts, ||∇X^t_M − ∇X^{t-1}_M||². This is a smoothness penalty: it can be zero even when ∑_v ∇X̂^t_M is large and growing. The spatial conservation condition stated in Eq. 6, ∑_v ∇X̂^t_M = 0, is never included as a loss term, and the spatial term in Eq. 16 is not implemented as written. Consequently, the claim that PCDCNet 'enforces mass conservation' or 'maintains atmospheric continuity' is unsupported by the training objective actually used.
  2. [§3.3.2, Eq. 14] Eq. 14 postulates pairwise fluxes ∇X̂^t_{M,v→v'} that are 'directly extracted from STD,' but the STD module in Eq. 5 outputs node-level messages M_t, and Eq. 12 computes ∇X̂^t_M = Linear(M_t), also a node-level vector. No mapping from node-level outputs to edge fluxes is defined. A node-level readout of a graph convolution admits infinitely many pairwise decompositions, so Eq. 14 and the spatial mass-conservation constraint built on it are not well-posed. The authors need to either define how edge-level fluxes are computed from the message-passing operation or remove this constraint from the narrative.
  3. [§4.4, Figure 4] The empirical evidence in Figure 4 does not support the physical-consistency claim. The figure shows that the DIC loss decreases even when λ=0, which is consistent with the loss measuring internal smoothness rather than a physical residual. No experiment reports the magnitude of ∑_v ∇X̂^t_M under the trained model, nor any direct test of whether predicted concentrations approximately satisfy the continuity equation. Without such verification, the interpretation that DIC loss 'enforces physical consistency' and improves generalization through physical alignment is not established; the improvement at λ=10 could equally arise from the regularizing effect of a smoothness penalty.
  4. [§4.2, Table 1] The claimed error reductions relative to the next best method do not match Table 1. For PM2.5, the best competing MAE is 20.60 (TimeXer, BTHSA) and 12.86 (iTransformer, YRD); PCDCNet's MAE is 15.46 and 10.97, giving reductions of 24.9% and 14.7%, respectively, not the stated 19.8%. For O3, the next best MAEs are 22.58 and 23.28, while PCDCNet gets 16.73 and 17.27, giving reductions of 25.9% and 25.8%, not 18.4%. Please correct the percentages or clarify which baseline and averaging procedure were used.
minor comments (4)
  1. [§1, §4.1.2] The paper repeatedly emphasizes computational efficiency over CMAQ and WRF-Chem, but no runtime or throughput comparison is provided. A quantitative statement (e.g., inference time for one 72-hour forecast on a CPU) would substantiate the surrogate-model advantage.
  2. [Table 3 and §A.6] Table 3 marks AirPhyNet as 'Phy ✓', but the text in §A.6 says AirPhyNet 'fails to fully integrate exogenous variables or enforce physical constraints.' This is contradictory; please reconcile the table and the text.
  3. [§3.2.2, Eq. 5] The description of the normalized graph Laplacian L̃ = I − D^{-1/2}AD^{-1/2} is standard, but the claim that a single graph convolution 'captures pollutant dispersion as a second-order process' is vague; the link between the graph Laplacian and the diffusion term in Eq. 2 is not made explicit.
  4. [Notation throughout] The notation ∇X̂^t_M is used both for a node-level derivative readout (Eq. 12) and, in Eq. 15, as a quantity whose total derivative is taken; the symbol X̂^t_M is defined in Table 2 but is not consistently distinguished from the concentration prediction X̂^t. Please clarify the distinction between the STD readout and the final predicted concentration.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: PCDCNet's performance claims rest on held-out test evaluation and external data, and its physical-consistency loss is a regularizer, not a fitted target.

full rationale

The central empirical claim (Table 1, Section 4.2) is evaluated on 2022–2023 test data with external inputs (ERA5 meteorology, MEIC emissions, CNEMC observations), so the reported predictions are not constructed from the quantities they are meant to predict. The DIC loss (Eqs. 10, 13, 16) is added to the L1 prediction loss as a regularizer; its strength lambda is tuned on validation data, which is standard model selection rather than circular reasoning. The implementation discrepancy between Eq. 13's temporal smoothness penalty and Eq. 14/16's pairwise flux conservation is a substantive correctness and falsifiability concern for the physical-consistency narrative, but it is not an input–output equivalence by construction. Self-citations do not carry the load: PM2.5-GNN [44] appears only as a baseline and as an example of an AQF model without explicit physical constraints, while the other author-affiliated references are contextual or methodological. No uniqueness theorem, ansatz, or fitted parameter is imported from the authors' prior work as the justification for the model's predictions. Therefore no circular step meeting the required quote-and-reduction standard was found.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The model introduces no new physical entities. It relies on the assumption that a station graph can represent pollutant transport, that emission inventories are accurate, and that the DIC loss actually enforces mass conservation. The latter assumption is the most fragile and is not supported by the implementation as written.

free parameters (4)
  • hidden_size = 32
    Selected by sensitivity analysis on BTHSA validation (Fig. 5).
  • DIC weight lambda = 10
    Selected via validation sweep; higher lambda improves test MAE (Fig. 4).
  • graph_edge_threshold = 200 km
    Hand-chosen geodesic distance threshold for station graph connectivity, following prior work [35, 42].
  • historical_window_length_T_prime
    The input window length is not reported in the paper, yet it defines the context for each prediction.
assumptions (6)
  • domain assumption Graph Laplacian smoothing on the station graph models advection and diffusion of pollutants.
    STD uses (I - D^-1/2 A D^-1/2)H and claims it represents second-order dispersion; this is a modeling assumption, not derived from the transport PDE.
  • domain assumption MEIC monthly emissions, downscaled to hourly via the T/CSES 144 methodology, are a valid representation of actual hourly emissions.
    The paper uses this input Q without uncertainty quantification; errors in the inventory propagate to predictions.
  • domain assumption The station graph with 200 km edges captures the relevant pollutant transport pathways.
    Graph construction follows prior work, but the scale of true atmospheric transport depends on wind speed and chemical lifetime.
  • ad hoc to paper The DIC loss on node sums is equivalent to enforcing the continuity equation for pollutant transport.
    Eq. 14 postulates pairwise fluxes that are never concretely defined; the implemented loss regularizes aggregate node outputs, which is not the continuity equation.
  • domain assumption Forecasted meteorology (GFS/ECMWF) and emissions remain sufficiently accurate over 72 hours.
    The deployment uses forecast meteorology, and errors in those forecasts bound the achievable accuracy.
  • standard math The residual update Xhat_t = Xhat_{t-1} + Delta Xhat with GRU accumulation yields stable multi-step forecasts.
    This is a standard autoregressive formulation, but stability over 72 steps is an empirical property, not guaranteed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PCDCNet: A Surrogate Model for Air Quality Forecasting with Physical-Chemical Dynamics and Constraints." pith.science (2026). https://pith.science/paper/STA7GKIV

@misc{pith2026250519842,
  author       = {Pith},
  title        = {Pith review of: PCDCNet: A Surrogate Model for Air Quality Forecasting with Physical-Chemical Dynamics and Constraints},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/STA7GKIV}},
  note         = {Machine review of arXiv:2505.19842}
}
read the original abstract

Air quality forecasting (AQF) is critical for public health and environmental management, yet remains challenging due to the complex interplay of emissions, meteorology, and chemical transformations. Traditional numerical models, such as CMAQ and WRF-Chem, provide physically grounded simulations but are computationally expensive and rely on uncertain emission inventories. Deep learning models, while computationally efficient, often struggle with generalization due to their lack of physical constraints. To bridge this gap, we propose PCDCNet, a surrogate model that integrates numerical modeling principles with deep learning. PCDCNet explicitly incorporates emissions, meteorological influences, and domain-informed constraints to model pollutant formation, transport, and dissipation. By combining graph-based spatial transport modeling, recurrent structures for temporal accumulation, and representation enhancement for local interactions, PCDCNet achieves state-of-the-art (SOTA) performance in 72-hour station-level PM2.5 and O3 forecasting while significantly reducing computational costs. Furthermore, our model is deployed in an online platform, providing free, real-time air quality forecasts, demonstrating its scalability and societal impact. By aligning deep learning with physical consistency, PCDCNet offers a practical and interpretable solution for AQF, enabling informed decision-making for both personal and regulatory applications.

Figures

Figures reproduced from arXiv: 2505.19842 by the authors.

Figure 1
Figure 1. Air quality (PM2.5 and O3) is shaped by complex interactions between meteorology (e.g., UV radiation, wind) and emissions (e.g., NO𝑥 , VOC). Capturing these spatiotempo￾ral dynamics, including pollutant transport and secondary formation, requires integrating emissions data with meteo￾rology, which poses significant modeling challenges. 1 Introduction Air pollution remains a pressing global challenge, disproportion￾a… view at source ↗
Figure 2
Figure 2. The framework of PCDCNet for air quality forecasting (AQF), comprising three stages: Data Pre-processing (graph [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. MAE trends for PM2.5 in BTHSA and O3 in YRD over a 72-hour prediction horizon. illustrates their input-output frameworks based on their original designs. • Machine Learning Models: These include XGBoost [10] and LightGBM [23], which directly fit air quality predictions to meteorological and emissions data using gradient boosting but lack temporal and spatial modeling. • General Time-Series Forecasting Models: These … view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Analysis of Domain-Informed Constraints (DIC). Temporal and spatial DIC loss components confirm the necessity [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Performance evaluation in BTHSA. (Left) Sensitiv [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Research regions and constructed spatial graphs for [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Comparison of input-output frameworks in air qual [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 9
Figure 9. Figure 9: Model prediction (left) and real-world observation [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 8
Figure 8. Figure 8: Nationwide air quality during New Year’s Eve and [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 13
Figure 13. Figure 13: Temporal variation of NO𝑥 emissions at four air quality monitoring stations: 1001A (Beijing city center), 1010A (Beijing suburban), 1147A (Shanghai city center), and 3272A (Shanghai suburban). The data, converted to UTC+0, highlights distinct morning and evening peaks…
Figure 14
Figure 14. Figure 14: Spatial distribution of emissions over the study [PITH_FULL_IMAGE:figures/full_fig_p016_14.png]
Figure 15
Figure 15. Figure 15: MAE comparison for BTHSA PM2.5 and YRD O3 predictions using XGBoost and PCDCNet, with and without emissions data. The spatial distribution of emissions ( [PITH_FULL_IMAGE:figures/full_fig_p017_15.png]
Figure 16
Figure 16. Figure 16: Online Real-Time Accuracy Monitoring of AQF [PITH_FULL_IMAGE:figures/full_fig_p018_16.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 46 canonical work pages

  1. [1]

    Muhammad Naveed Anwar, Muneeba Shabbir, Eza Tahir, Mahnoor Iftikhar, Hira Saif, Ajwa Tahir, Malik Ashir Murtaza, Muhammad Fahim Khokhar, Mohammad Rehan, Mortaza Aghbashlo, et al. 2021. Emerging challenges of air pollution and particulate matter in China, India, and Pakistan and mitigating solutions. Journal of Hazardous Materials 416 (2021), 125851

    work page 2021

  2. [2]

    K Wyat Appel, Sergey L Napelenok, Kristen M Foley, Havala OT Pye, Christian Hogrefe, Deborah J Luecken, Jesse O Bash, Shawn J Roselle, Jonathan E Pleim, Hosein Foroutan, et al . 2017. Description and evaluation of the Community Multiscale Air Quality (CMAQ) modeling system version 5.1. Geoscientific model development 10, 4 (2017), 1703–1732

    work page 2017

  3. [3]

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

    2023

  4. [4]

    Cristian Bodnar, Wessel P Bruinsma, Ana Lucic, Megan Stanley, Johannes Brand- stetter, Patrick Garvan, Maik Riechert, Jonathan Weyn, Haiyu Dong, Anna Vaughan, et al. 2024. Aurora: A foundation model of the atmosphere. arXiv preprint arXiv:2405.13063 (2024)

    arXiv 2024

  5. [5]

    Michael Brauer, Greg Freedman, Joseph Frostad, Aaron Van Donkelaar, Randall V Martin, Frank Dentener, Rita van Dingenen, Kara Estep, Heresh Amini, Joshua S Apte, et al. 2016. Ambient air pollution exposure estimation for the global burden of disease 2013. Environmental science & technology 50, 1 (2016), 79–88

    work page 2016

  6. [6]

    Marshall Burke, Marissa L Childs, Brandon de la Cuesta, Minghao Qiu, Jessica Li, Carlos F Gould, Sam Heft-Neal, and Michael Wara. 2023. The contribution of wildfire to PM2. 5 trends in the USA. Nature 622, 7984 (2023), 761–766

    work page 2023

  7. [7]

    Baozhang Chen, Sheng Zhong, Nicholas AS Hamm, Hong Liao, Tong Zhu, Huifang Zhang, Lifeng Guo, Kun Hou, et al. 2024. Region-oriented simultane- ously joint two-pollutant control strategies are required to substantially reduce deaths attributed to both PM2. 5 and ozone pollution in China. Atmospheric Environment 334 (2024), 120708

    work page 2024

  8. [8]

    Han Chen, Weihang Zhang, and Lifang Sheng. 2025. Canadian record-breaking wildfires in 2023 and their impact on US air quality. Atmospheric Environment 342 (2025), 120941

    work page 2025

Show all 52 references
  1. [9]

    Ricky TQ Chen, Yulia Rubanova, Jesse Bettencourt, and David K Duvenaud. 2018. Neural ordinary differential equations. In NeurIPS

    2018

  2. [10]

    Tianqi Chen and Carlos Guestrin. 2016. Xgboost: A scalable tree boosting system. In ACM SIGKDD. 785–794

    2016

  3. [11]

    Ziyue Chen, Danlu Chen, Chuanfeng Zhao, Mei-po Kwan, Jun Cai, Yan Zhuang, Bo Zhao, Xiaoyan Wang, Bin Chen, Jing Yang, et al. 2020. Influence of meteorolog- ical conditions on PM2. 5 concentrations across China: A review of methodology and mechanism. Environment international 1...

    2020

  4. [12]

    Ziyue Chen, Xiaoming Xie, Jun Cai, Danlu Chen, Bingbo Gao, Bin He, Nianliang Cheng, and Bing Xu. 2018. Understanding meteorological influences on PM 2.5 concentrations across China: a temporal and spatial perspective. Atmospheric Chemistry and Physics 18, 8 (2018), 5343–5358

    2018

  5. [13]

    Johanna Einsiedler, Yun Cheng, Franz Papst, and Olga Saukh. 2021. Interpretable and Transferable Models to Understand the Impact of Lockdown Measures on Local Air Quality. arXiv preprint arXiv:2011.10144 (2021)

    2021 arXiv

  6. [14]

    Christopher Emery, Kirk Baker, Gary Wilson, and Greg Yarwood. 2024. Com- prehensive Air Quality Model with Extensions: Formulation and Evaluation for Ozone and Particulate Matter over the US. Atmosphere 15, 10 (2024)

    2024

  7. [15]

    Matthias Fey and Jan Eric Lenssen. 2019. Fast graph representation learning with PyTorch Geometric. In ICLR Workshop

    2019

  8. [16]

    Chinese Society for Environmental Sciences. 2024. Technical guideline on city- level air pollutant emission inventory development. T/CSES 144—2024. https: //www.ttbz.org.cn/StandardManage/Detail/108461/

    2024

  9. [17]

    Zhaoqi Gao and Xuehua Zhou. 2024. A review of the CAMx, CMAQ, WRF- Chem and NAQPMS models: Application, evaluation and uncertainty factors. Environmental Pollution 343 (2024), 123183

    2024

  10. [18]

    Guannan Geng, Yuxi Liu, Yang Liu, Shigan Liu, Jing Cheng, Liu Yan, Nana Wu, Hanwen Hu, Dan Tong, Bo Zheng, et al. 2024. Efficacy of China’s clean air actions to tackle PM2.5 pollution between 2013 and 2020. Nature Geoscience (2024), 1–8

    2024

  11. [19]

    Kethmi Hirushini Hettige, Jiahao Ji, Shili Xiang, Cheng Long, Gao Cong, and Jingyuan Wang. 2024. AirPhyNet: Harnessing Physics-Guided Neural Networks for Air Quality Prediction. In ICLR

    2024

  12. [20]

    Peter V Hobbs. 2000. Introduction to atmospheric chemistry. Cambridge University Press

    2000

  13. [21]

    Tangyan Hou, Shaocai Yu, Yaping Jiang, Xue Chen, Yibo Zhang, Mengying Li, Zhen Li, Zhe Song, Pengfei Li, Jianmin Chen, et al. 2022. Impacts of chemical initial conditions in the WRF-CMAQ model on the ozone forecasts in eastern China. Aerosol and Air Quality Research 22, 7 (202...

    2022

  14. [22]

    Mark Z Jacobson. 1999. Fundamentals of atmospheric modeling . Cambridge university press

    1999

  15. [23]

    Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. 2017. Lightgbm: A highly efficient gradient boosting decision tree. In NeurIPS

    2017

  16. [24]

    Thomas N Kipf and Max Welling. 2017. Semi-supervised classification with graph convolutional networks. In ICLR

    2017

  17. [25]

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

    2023

  18. [26]

    Jiali Li, Shaocai Yu, Xue Chen, Yibo Zhang, Mengying Li, Zhen Li, Zhe Song, Weiping Liu, Pengfei Li, Min Xie, et al . 2022. Evaluation of the WRF-CMAQ model performances on air quality in China with the impacts of the observation nudging on meteorology. Aerosol and Air Quality...

    2022

  19. [27]

    Ke Li, Daniel J Jacob, Hong Liao, Jia Zhu, Viral Shah, Lu Shen, Kelvin H Bates, Qiang Zhang, and Shixian Zhai. 2019. A two-pollutant strategy for improving ozone and particulate air quality in China. Nature Geoscience 12, 11 (2019), 906–910

    2019

  20. [28]

    Lianfa Li, Jinfeng Wang, Meredith Franklin, Qian Yin, Jiajie Wu, Gustau Camps- Valls, Zhiping Zhu, Chengyi Wang, Yong Ge, and Markus Reichstein. 2023. Im- proving air quality assessment using physics-inspired deep graph learning. npj Climate and Atmospheric Science 6, 1 (2023), 152

    2023

  21. [29]

    Meng Li, Huan Liu, Guannan Geng, Chaopeng Hong, Fei Liu, Yu Song, Dan Tong, Bo Zheng, Hongyang Cui, Hanyang Man, et al. 2017. Anthropogenic emission inventories in China: a review. National Science Review 4, 6 (2017), 834–866

    2017

  22. [30]

    Yasong Li, Tijian Wang, Mengmeng Li, Yawei Qu, Hao Wu, Min Xie, et al. 2024. Exploring the role of aerosol-ozone interactions on O3 surge and PM2. 5 decline during the clean air action period in Eastern China 2014–2020. Atmospheric Research 302 (2024), 107294

    2024

  23. [31]

    Yong Liu, Tengge Hu, Haoran Zhang, Haixu Wu, Shiyu Wang, Lintao Ma, and Mingsheng Long. 2023. iTransformer: Inverted Transformers Are Effective for Time Series Forecasting. In ICLR

    2023

  24. [32]

    Jing Lu, Yuhu Zhang, Mingxing Chen, Lu Wang, Shaohua Zhao, Xiao Pu, and Xuegang Chen. 2021. Estimation of monthly 1 km resolution PM2. 5 concentra- tions using a random forest model over “2+26” cities, China. Urban Climate 35 (2021), 100734

    2021

  25. [33]

    Xiao Lu, Lin Zhang, Xiaolin Wang, Meng Gao, Ke Li, Yuzhong Zhang, Xu Yue, and Yuanhang Zhang. 2020. Rapid increases in warm-season surface ozone and resulting health impact in China since 2013. Environmental Science & Technology Letters 7, 4 (2020), 240–247

    2020

  26. [34]

    Soon-Young Park, Uzzal Kumar Dash, Jinhyeok Yu, Keiya Yumimoto, Itsushi Uno, and Chul Han Song. 2021. Implementation of an ensemble Kalman filter in the community multiscale air quality model (CMAQ model v5. 1) for data assimilation of ground-level PM 2.5. Geoscientific Model ...

    2021

  27. [35]

    Yanlin Qi, Qi Li, Hamed Karimian, and Di Liu. 2019. A hybrid model for spa- tiotemporal forecasting of PM2.5 based on graph convolutional neural network and long short-term memory. Science of the Total Environment 664 (2019), 1–10

    2019

  28. [36]

    Yawei Qu, Tijian Wang, Cheng Yuan, Hao Wu, Libo Gao, Congwu Huang, Yasong Li, Mengmeng Li, and Min Xie. 2023. The underlying mechanisms of PM2. 5 and O3 synergistic pollution in East China: Photochemical and heterogeneous interactions. Science of the Total Environment 873 (202...

    2023

  29. [37]

    S Trivikrama Rao, Huiying Luo, Marina Astitha, Christian Hogrefe, Valerie Garcia, and Rohit Mathur. 2020. On the limit to the accuracy of regional-scale air quality models. Atmospheric chemistry and physics 20, 3 (2020), 1627–1639

    2020

  30. [38]

    John H Seinfeld and Spyros N Pandis. 2016. Atmospheric chemistry and physics: from air pollution to climate change . John Wiley & Sons

    2016

  31. [39]

    Bhupal Shrestha, Jerald A Brotzge, and Junhong Wang. 2022. Observations and impacts of long-range transported wildfire smoke on air quality across New York state during July 2021. Geophysical Research Letters 49, 19 (2022), e2022GL100216

    2022

  32. [40]

    Pierre Sicard, Paola Crippa, Alessandra De Marco, Stefano Castruccio, Paolo Giani, Juan Cuesta, Elena Paoletti, Zhaozhong Feng, and Alessandro Anav. 2021. High spatial resolution WRF-Chem model over Asia: Physics and chemistry evaluation. Atmospheric Environment 244 (2021), 118004

    2021

  33. [41]

    Yongzhao Sun and Xiaoyan Wang. 2022. Meteorological factor contributions to the seesaw concentration pattern between PM2. 5 and O3 in Shanghai. Frontiers in Environmental Science 10 (2022), 1015723

    2022

  34. [42]

    Chunyang Wang, Yanmin Zhu, Tianzi Zang, Haobing Liu, and Jiadi Yu. 2021. Modeling inter-station relationships with attentive temporal graph convolutional network for air quality prediction. In ACM WSDM. 616–634

    2021

  35. [43]

    Pengfei Wang, Hao Guo, Jianlin Hu, Sri Harsha Kota, Qi Ying, and Hongliang Zhang. 2019. Responses of PM2. 5 and O3 concentrations to changes of mete- orology and emissions in China. Science of the Total Environment 662 (2019), 297–306

    2019

  36. [44]

    Shuo Wang, Yanran Li, Jiang Zhang, Qingye Meng, Lingwei Meng, and Fei Gao

  37. [45]

    Yuxuan Wang, Haixu Wu, Jiaxiang Dong, Guo Qin, Haoran Zhang, Yong Liu, Yunzhong Qiu, Jianmin Wang, and Mingsheng Long. 2024. Timexer: Empowering transformers for time series forecasting with exogenous variables. In NeurIPS

    2024

  38. [46]

    Qingyang Xiao, Guannan Geng, Tao Xue, Shigan Liu, Cilan Cai, Kebin He, and Qiang Zhang. 2021. Tracking PM2. 5 and O3 pollution and the related health PCDCNet: A Surrogate Model for Air Quality Forecasting with Physical-Chemical Dynamics and Constraints Conference acronym ’XX, ...

    2021

  39. [47]

    Hui Yang, Xinyuan Huang, Daniel M Westervelt, Larry Horowitz, and Wei Peng

  40. [48]

    Huanbi Yue, Chunyang He, Qingxu Huang, Dan Yin, and Brett A Bryan. 2020. Stronger policy required to substantially reduce deaths from PM2. 5 pollution in China. Nature Communications 11, 1 (2020), 1462

    2020

  41. [49]

    Biao Zhang and Rico Sennrich. 2019. Root mean square layer normalization. In NeurIPS

    2019

  42. [50]

    Air Quality Forecasting in the Guangdong- Hong Kong-Macao Region 2024

    Bo Zheng, Dan Tong, Meng Li, Fei Liu, Chaopeng Hong, Guannan Geng, Haiyan Li, Xin Li, Liqun Peng, Ji Qi, et al. 2018. Trends in China’s anthropogenic emissions since 2010 as the consequence of clean air actions. Atmospheric Chemistry and Physics 18, 19 (2018), 14095–14111. Con...

    2018

  43. [2020]

    In ACM SIGSPATIAL

    PM2.5-GNN: A domain knowledge enhanced graph neural network for pm2.5 forecasting. In ACM SIGSPATIAL. 163–166

  44. [2023]

    Nature Sustainability 6, 1 (2023), 58–68

    Socio-demographic factors shaping the future global health burden from air pollution. Nature Sustainability 6, 1 (2023), 58–68

    2023

Pith tools

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