Pith. sign in

REVIEW 2 major objections 4 minor 1 cited by

SolarSeer: Ultrafast and accurate 24-hour solar irradiance forecasts outperforming numerical weather prediction across the USA

T0 review · 2 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read SolarSeer maps six hours of satellite imagery to a 24-hour, 5-km solar irradiance forecast, and the paper reports it beats HRRR by 27.28% in RMSE on reanalysis data and 15.35% at 1,800 stations, while running over 1,500 times faster.

desk verdict SolarSeer is a serious engineering result, but the headline 27.28% improvement over HRRR is not a clean 24h-vs-24h comparison at most initialization hours. read the letter →

arxiv 2508.03590 v2 pith:PCRAD2KC submitted 2025-08-05 cs.LG cs.CE

Mingliang Bai , Zuliang Fang , Shengyu Tao , Siqi Xiang , Jiang Bian , Yanfei Xiang , Pengcheng Zhao , Weixin Jin
show 9 more authors
Jonathan A. Weyn Haiyu Dong Bin Zhang Hongyu Sun Kit Thambiratnam Qi Zhang Hongbin Sun Xuan Zhang Qiuwei Wu
This is my paper · ORCID
classification cs.LGcs.CE
keywords solarirradianceforecastingcloudcoversatelliteimagerynumericalweatherpredictiondeeplearningtransformerrenewableenergyHRRR
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

SolarSeer is an end-to-end AI model that claims to beat the best operational numerical weather prediction in the contiguous United States at forecasting solar irradiance. It does this by mapping just six hours of satellite imagery directly onto a 24-hour cloud cover and irradiance forecast at 5 km resolution, skipping data assimilation and partial differential equation (PDE) based atmosphere simulation entirely. The paper reports a 27.28% lower RMSE than HRRR on ERA5 reanalysis and 15.35% lower RMSE across 1,800 ground stations, while running more than 1,500 times faster, in under 3 seconds. A sympathetic reader should care because day-ahead irradiance accuracy is what lets grid operators trust solar power in electricity markets, and prior AI methods only worked for short lead times.

What carries the argument

The carrying mechanism is a two-block network: a cloud block of Adaptive Fourier Neural Operator (AFNO) Transformer layers, a frequency-domain Transformer that learns cloud motion and evolution, maps the six-hour satellite image sequence to the future 24-hour cloud cover. An irradiance block of Swin Transformer layers, a shifted-window vision Transformer for spatial detail, then combines that cloud forecast with the deterministic clear-sky irradiance from the Ineichen-Perez model, which supplies the Sun-Earth geometry as a physical prior. Bypassing data assimilation is what removes the supercomputer-level cost: the model goes from raw satellite images to forecasts end to end.

What would settle it

Hold out a set of days with rapid convective initiation where the six hours of preceding imagery shows little sign of cloud growth, and compare SolarSeer's 24-hour cloud RMSE against HRRR's; if SolarSeer's advantage disappears or reverses on those days, the satellite-only premise fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that a learned mapping from historical satellite observations to future cloud cover can replace the data-assimilation-plus-PDE pipeline for solar irradiance forecasting and still outperform the operational NWP baseline. SolarSeer outputs both cloud cover and irradiance; its cloud block reduces 24-hour cloud-cover RMSE by 29.13% and MAE by 25.24% against HRRR, with 99.99% of CONUS grid cells improved. The irradiance block then reduces RMSE by 27.28% against ERA5 and 15.35% against 1,800 station observations, and improves forecast accuracy of first-order irradiance differences, meaning it captures ramps in solar output.

Load-bearing premise

The load-bearing premise is that six hours of satellite imagery contains enough information to predict the next 24 hours of cloud cover at 5 km resolution; any cloud development driven by variables invisible in the images, such as temperature and humidity, is something the model can only learn statistically from training data.

Editorial extensions

If this is right

  • With 24-hour irradiance forecasts produced in seconds, day-ahead market participants could update solar positions far more frequently than NWP currently allows.
  • The satellite-only design means operational solar forecasting no longer must wait for an NWP initialization cycle; a fresh satellite image can trigger a forecast at any hour.
  • Because SolarSeer improves cloud-cover forecasts at all 24 lead times and across nearly all CONUS regions, the irradiance error reductions follow in the same areas, with the largest gains in the western and southern United States.
  • Better first-order irradiance difference accuracy means rapid ramp events, a main threat to grid stability, are tracked more faithfully than by HRRR.
  • SolarSeer produces 24-hour forecasts at every initialization hour of the day, whereas HRRR only offers 24-hour forecasts at four initialization times, extending forecast availability for grid operators.

Reading between the lines

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

  • Editorial inference: If the satellite-only premise holds under weather regimes absent from training, such as wildfire smoke or rapid convection, the same architecture could be retrained for any region with geostationary satellite coverage, effectively exporting skill without a local high-resolution NWP.
  • Editorial inference: A natural stress test is to hold out entire seasons or extreme-event years; if skill drops sharply on rare cloud regimes, the model may be partly memorizing climatological cloud patterns rather than learning cloud processes from the images.
  • Editorial inference: The six-hour input window is a tunable design choice; varying it or adding coarse atmospheric state fields would reveal how much of the predictive signal actually comes from the imagery alone.
  • Editorial inference: Because the clear-sky prior fixes the irradiance geometry, the model's irradiance errors should concentrate where cloud attribution is wrong; separating errors on cloudy versus clear grid cells could identify whether the cloud block or the irradiance block is the limiting component.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper introduces SolarSeer, an end-to-end AI model that maps six hours of satellite imagery to 24-hour cloud-cover and solar-irradiance forecasts over the contiguous United States at roughly 5-km resolution, bypassing data assimilation and PDE solving. Against NOAA's operational HRRR NWP, the authors report large average RMSE reductions (29.13% for cloud cover, 27.28% for irradiance on ERA5 ground truth, and 15.35% across 1,800 stations) and claim inference under 3 seconds, more than 1,500 times faster than HRRR. The central claim is that SolarSeer is the first end-to-end AI model to outperform state-of-the-art NWP for 24-hour solar irradiance forecasting.

Significance. If substantiated, the result is significant for solar energy forecasting: a fast, end-to-end, satellite-based model that beats an operational NWP at kilometer scale would be practically valuable for day-ahead electricity markets and would support the broader trend of learned weather emulators. The paper deserves credit for benchmarking against a real operational model (HRRR), for reporting station-based validation in addition to reanalysis, and for emphasizing inference speed. However, the headline comparison currently mixes forecast horizons across initialization hours, so the quantitative claims need to be re-established on matched lead times before the significance can be assessed.

major comments (2)
  1. [Results: Cloud cover forecast results / Solar irradiance forecast results] The evaluation protocol is explicitly mismatched for most initialization hours. The text states that at non-synoptic hours the authors evaluate "the 24-hour forecasts of SolarSeer and the 18-hour forecasts of HRRR-NWP," while at UTC 00/06/12/18 they compare 24-hour forecasts of both. Because forecast error generally grows with lead time, comparing a 24-hour SolarSeer forecast against an 18-hour HRRR forecast at 20 of 24 initialization hours systematically inflates SolarSeer's apparent advantage. The reported CONUS averages (29.13% cloud-cover RMSE, 27.28% irradiance RMSE, 36.08% irradiance MAE) therefore do not establish the claim that SolarSeer outperforms HRRR at the same 24-hour lead time. The authors should recompute all aggregate metrics using only matched lead times (e.g., lead times 1-18 h for all hours and 24 h only for the four synoptic hours) and report the 24-hour-vs-24-hour comparison separately. Without this, the central claim of "outperforming NWP for 24-hour forecasts" is not supported by the current numbers.
  2. [Results: Solar irradiance forecast results] The repeated use of "significantly" (e.g., "SolarSeer significantly outperforms HRRR-NWP," "significantly reduces the root mean squared error") is not accompanied by any uncertainty intervals or statistical significance tests. Given the strong spatial and temporal correlations in the data, the very large domain fractions (99.99% of CONUS) do not by themselves establish that the differences are beyond sampling noise. The authors should provide confidence intervals (for example, via block bootstrap over forecast initialization dates) for the headline RMSE/MAE reductions and for the 15.35% station-level result, or otherwise state the uncertainty associated with each point estimate.
minor comments (4)
  1. [Solar irradiance forecast results] Typos: "start-of-the-art" should be "state-of-the-art" in the sentence describing HRRR-NWP.
  2. [Abstract] The abstract claims that SolarSeer "significantly enhances the first-order irradiance difference forecasting accuracy," but no results or metrics for this claim appear in the main text; please add the corresponding analysis or remove the claim from the abstract.
  3. [Abstract / Results] The station-based evaluation across 1,800 stations is mentioned in the abstract and results overview, but the main text does not describe the station data source, the evaluation period, or whether lead times are matched there; please provide these details.
  4. [SolarSeer overview] The statement that SolarSeer produces forecasts "in under 3 seconds" would benefit from specifying the hardware and whether the time includes input/output and preprocessing; if this is given in the Methods, please cite the relevant subsection here.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SolarSeer's forecasts are a learned mapping evaluated against external benchmarks; the HRRR 18-hour comparison is an evaluation fairness concern, not a circular derivation.

full rationale

SolarSeer is presented as an end-to-end learned mapping from six hours of satellite imagery to 24-hour cloud cover and irradiance forecasts, with no analytic derivation chain whose outputs are equivalent to its inputs by construction. The headline 27.28% RMSE reduction on ERA5 is an empirical comparison against an external reanalysis product; the text does not state that ERA5 was the training target, and the independent 1,800-station comparison provides an external benchmark. The mismatch between SolarSeer's 24-hour forecasts and HRRR's 18-hour forecasts at non-synoptic initialization times is a benchmark-protocol fairness issue (forecast error generally grows with lead time), not a circularity: the forecast is not defined in terms of the HRRR baseline, and no fitted parameter is renamed as a prediction. No load-bearing self-citation, uniqueness theorem, or ansatz-smuggling-via-citation is present. Accordingly, no circular step can be exhibited from the paper's own text, and the appropriate finding is no significant circularity.

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

The method rests on the assumption that cloud evolution is learnable from satellite images alone, and on the use of reanalysis and analysis products as evaluation truth. The model's learned weights are many fitted parameters; the paper does not disclose their values or the hyperparameters, which limits independent reconstruction.

free parameters (3)
  • Learned weights of cloud block (AFNO Transformer layers) = Not disclosed; set by training on five years of satellite data
    The cloud cover forecast is produced by these learned weights, so all forecast skill depends on this fitted function.
  • Learned weights of irradiance block (Swin Transformer layers) = Not disclosed; set by training on five years of irradiance and cloud data
    The irradiance forecast is a learned function of cloud forecasts and clear-sky irradiance; the reported error reductions depend on these fitted weights.
  • Architecture hyperparameters (number of layers, heads, embedding dimension, patch size) = Not disclosed in reviewed text
    These are chosen by the authors and materially affect forecast accuracy; without them the model cannot be reconstructed.
assumptions (3)
  • domain assumption Six hours of satellite imagery contains sufficient information to predict 24 hours of cloud cover at 5 km resolution.
    The model has no dynamical equations or additional atmospheric inputs; this assumption is the basis for bypassing data assimilation and PDE solving.
  • domain assumption ERA5 reanalysis is a valid ground truth for solar irradiance forecast errors.
    ERA5 is used as the reference for the 27.28% RMSE improvement; reanalysis is not a direct surface observation and has its own biases.
  • domain assumption RTMA is a valid ground truth for total cloud cover.
    Cloud-cover improvements are measured against RTMA, an analysis product, not raw observations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SolarSeer: Ultrafast and accurate 24-hour solar irradiance forecasts outperforming numerical weather prediction across the USA." pith.science (2026). https://pith.science/paper/PCRAD2KC

@misc{pith2026250803590,
  author       = {Pith},
  title        = {Pith review of: SolarSeer: Ultrafast and accurate 24-hour solar irradiance forecasts outperforming numerical weather prediction across the USA},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PCRAD2KC}},
  note         = {Machine review of arXiv:2508.03590}
}
read the original abstract

Accurate 24-hour solar irradiance forecasting is essential for the safe and economic operation of solar photovoltaic systems. Traditional numerical weather prediction (NWP) models represent the state-of-the-art in forecasting performance but rely on computationally costly data assimilation and solving complicated partial differential equations (PDEs) that simulate atmospheric physics. Here, we introduce SolarSeer, an end-to-end large artificial intelligence (AI) model for solar irradiance forecasting across the Contiguous United States (CONUS). SolarSeer is designed to directly map the historical satellite observations to future forecasts, eliminating the computational overhead of data assimilation and PDEs solving. This efficiency allows SolarSeer to operate over 1,500 times faster than traditional NWP, generating 24-hour cloud cover and solar irradiance forecasts for the CONUS at 5-kilometer resolution in under 3 seconds. Compared with the state-of-the-art NWP in the CONUS, i.e., High-Resolution Rapid Refresh (HRRR), SolarSeer significantly reduces the root mean squared error of solar irradiance forecasting by 27.28% in reanalysis data and 15.35% across 1,800 stations. SolarSeer also effectively captures solar irradiance fluctuations and significantly enhances the first-order irradiance difference forecasting accuracy. SolarSeer's ultrafast, accurate 24-hour solar irradiance forecasts provide strong support for the transition to sustainable, net-zero energy systems.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Data-driven solar forecasting enables near-optimal economic decisions

    physics.geo-ph 2025-09 conditional novelty 6.0 of 10

    SunCastNet combines AI weather forecasting with reinforcement-learning battery control to turn high-resolution solar forecasts into large regret reductions and more profitable industrial solar projects.

Reference graph

Works this paper leans on

41 extracted references · 37 canonical work pages · cited by 1 Pith paper

  1. [1]

    title World Energy Outlook 2023

    author International Energy Agency . title World Energy Outlook 2023 . type Tech. Rep. , institution IEA ( year 2023 )

    work page 2023

  2. [2]

    author Zhang, Z. et al. journal title A machine learning model for hub-height short-term wind speed prediction . Nature Communications volume 16 , pages 3195 ( year 2025 )

    work page 2025

  3. [3]

    title Global Market Outlook for Solar Power 2024-2028

    author SolarPower Europe . title Global Market Outlook for Solar Power 2024-2028 . type Tech. Rep. , institution SolarPower Europe ( year 2024 )

    work page 2024

  4. [4]

    author Nijsse, F. J. et al. journal title The momentum of the solar energy transition . Nature Communications volume 14 , pages 6542 ( year 2023 )

    work page 2023

  5. [5]

    author Wang, J. et al. journal title Inherent spatiotemporal uncertainty of renewable power in china . Nature Communications volume 14 , pages 5379 ( year 2023 )

    work page 2023

  6. [6]

    author Xia, P. et al. journal title Accurate nowcasting of cloud cover at solar photovoltaic plants using geostationary satellite images . Nature Communications volume 15 , pages 510 ( year 2024 )

    work page 2024

  7. [7]

    author Gandhi, O. et al. journal title The value of solar forecasts and the cost of their errors: A review . Renewable and Sustainable Energy Reviews volume 189 , pages 113915 ( year 2024 )

    work page 2024

  8. [8]

    , author Thorpe, A

    author Bauer, P. , author Thorpe, A. & author Brunet, G. journal title The quiet revolution of numerical weather prediction . Nature volume 525 , pages 47--55 ( year 2015 )

    work page 2015

Show all 41 references
  1. [9]

    author Zhou, W. et al. journal title Progress and future prospects of decadal prediction and data assimilation: a review . Atmospheric and Oceanic Science Letters volume 17 , pages 100441 ( year 2024 )

    2024

  2. [10]

    & author Kalnay, E

    author Pu, Z. & author Kalnay, E. journal title Numerical weather prediction basics: Models, numerical methods, and data assimilation . Handbook of hydrometeorological ensemble forecasting pages 67--97 ( year 2019 )

    2019

  3. [11]

    , author Lin, C

    author Chen, Y. , author Lin, C. , author Liu, J. & author Yu, D. journal title One-hour-ahead solar irradiance forecast based on real-time k-means++ clustering on the input side and cnn-lstm . Journal of Atmospheric and Solar-Terrestrial Physics pages 106405 ( year 2024 )

    2024

  4. [12]

    & author Toshniwal, D

    author Kumari, P. & author Toshniwal, D. journal title Deep learning models for solar irradiance forecasting: A comprehensive review . Journal of Cleaner Production volume 318 , pages 128566 ( year 2021 )

    2021

  5. [13]

    , author Huang, X

    author Zhu, L. , author Huang, X. , author Zhang, Z. , author Li, C. & author Tai, Y. journal title A novel u-lstm-aft model for hourly solar irradiance forecasting . Renewable Energy volume 238 , pages 121955 ( year 2025 )

    2025

  6. [14]

    author Yang, D. et al. journal title A review of solar forecasting, its dependence on atmospheric sciences and implications for grid integration: Towards carbon neutrality . Renewable and Sustainable Energy Reviews volume 161 , pages 112348 ( year 2022 )

    2022

  7. [15]

    author Srivastava, A. K. , author Kamalasadan, S. , author Patel, D. , author Sankar, S. & author Al-Olimat, K. S. journal title Electricity markets: an overview and comparative study . International Journal of Energy Sector Management volume 5 , pages 169--200 ( year 2011 )

    2011

  8. [16]

    author Bi, K. et al. journal title Accurate medium-range global weather forecasting with 3d neural networks . Nature volume 619 , pages 533--538 ( year 2023 )

    2023

  9. [17]

    author Lam, R. et al. journal title Learning skillful medium-range global weather forecasting . Science volume 382 , pages 1416--1421 ( year 2023 )

    2023

  10. [18]

    author Price, I. et al. journal title Probabilistic weather forecasting with machine learning . Nature volume 637 , pages 84--90 ( year 2025 )

    2025

  11. [19]

    author Chen, L. et al. journal title Fuxi: A cascade machine learning forecasting system for 15-day global weather forecast . npj Climate and Atmospheric Science volume 6 , pages 190 ( year 2023 )

    2023

  12. [20]

    author Bodnar, C. et al. journal title A foundation model for the earth system . Nature pages 1--8 ( year 2025 )

    2025

  13. [21]

    author Espeholt, L. et al. journal title Deep learning for twelve hour precipitation forecasts . Nature communications volume 13 , pages 1--10 ( year 2022 )

    2022

  14. [22]

    , author Nocedal, J

    author Fisher, M. , author Nocedal, J. , author Tr \'e molet, Y. & author Wright, S. J. journal title Data assimilation in weather forecasting: a case study in pde-constrained optimization . Optimization and Engineering volume 10 , pages 409--426 ( year 2009 )

    2009

  15. [23]

    author Schultz, M. G. et al. journal title Can deep learning beat numerical weather prediction? Philosophical Transactions of the Royal Society A volume 379 , pages 20200097 ( year 2021 )

    2021

  16. [24]

    author Dowell, D. C. et al. journal title The high-resolution rapid refresh (hrrr): An hourly updating convection-allowing forecast model. part i: Motivation and system description . Weather and Forecasting volume 37 , pages 1371--1395 ( year 2022 )

    2022

  17. [25]

    author James, E. P. et al. journal title The high-resolution rapid refresh (hrrr): an hourly updating convection-allowing forecast model. part ii: Forecast performance . Weather and Forecasting volume 37 , pages 1397--1417 ( year 2022 )

    2022

  18. [26]

    author De Pondeca, M. S. et al. journal title The real-time mesoscale analysis at noaa’s national centers for environmental prediction: current status and development . Weather and Forecasting volume 26 , pages 593--612 ( year 2011 )

    2011

  19. [27]

    , author Benjamin, S

    author Marquis, M. , author Benjamin, S. , author James, E. , author Molling, C. et al. title A public-private-academic partnership to advance solar power forecasting . type Tech. Rep. , institution NOAA ESRL ( year 2015 )

    2015

  20. [28]

    author James, E. P. , author Benjamin, S. G. & author Marquis, M. journal title A unified high-resolution wind and solar dataset from a rapidly updating numerical weather prediction model . Renewable Energy volume 102 , pages 390--405 ( year 2017 )

    2017

  21. [29]

    , author Yang, D

    author Zhang, G. , author Yang, D. , author Galanis, G. & author Androulakis, E. journal title Solar forecasting with hourly updated numerical weather prediction . Renewable and Sustainable Energy Reviews volume 154 , pages 111768 ( year 2022 )

    2022

  22. [30]

    author Hersbach, H. et al. journal title The era5 global reanalysis . Quarterly Journal of the Royal Meteorological Society volume 146 , pages 1999--2049 ( year 2020 )

    1999

  23. [31]

    author Kamadinata, J. O. , author Ken, T. L. & author Suwa, T. journal title Solar irradiance fluctuation prediction methodology using artificial neural networks . Journal of Solar Energy Engineering volume 142 , pages 031003 ( year 2020 )

    2020

  24. [32]

    title GOES - Geostationary Operational Environmental Satellite

    author National Oceanic and Atmospheric Administration . title GOES - Geostationary Operational Environmental Satellite . howpublished https://www.goes.noaa.gov/index.html

  25. [33]

    author Stein, J. S. , author Hansen, C. W. & author Reno, M. J. title Global horizontal irradiance clear sky models: implementation and analysis. type Tech. Rep. , institution Sandia National Laboratories (SNL), Albuquerque, NM, and Livermore, CA ( year 2012 )

    2012

  26. [34]

    & author Perez, R

    author Ineichen, P. & author Perez, R. journal title A new airmass independent formulation for the linke turbidity coefficient . Solar energy volume 73 , pages 151--157 ( year 2002 )

    2002

  27. [35]

    author Holmgren, W. F. , author Hansen, C. W. & author Mikofski, M. A. journal title pvlib python: A python package for modeling solar energy systems . Journal of Open Source Software volume 3 , pages 884 ( year 2018 )

    2018

  28. [36]

    journal title Choice of clear-sky model in solar forecasting

    author Yang, D. journal title Choice of clear-sky model in solar forecasting . Journal of Renewable and Sustainable Energy volume 12 ( year 2020 )

    2020

  29. [37]

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

    2022 arXiv

  30. [38]

    author Liu, Z. et al. title Swin transformer v2: Scaling up capacity and resolution . In booktitle Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 12009--12019 ( year 2022 )

    2022

  31. [39]

    , " * write output.state after.block = add.period write newline

    ENTRY address archive author booktitle chapter edition editor eprint howpublished institution journal key month note number organization pages publisher school series title type url doi volume year label INTEGERS output.state before.all mid.sentence after.sentence after.block ...

  32. [40]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  33. [41]

    write newline

    " write newline "" before.all 'output.state := FUNCTION string.to.integer 't := t text.length 'k := #1 'char.num := t char.num #1 substring 's := s is.num s "." = or char.num k = not and char.num #1 + 'char.num := while char.num #1 - 'char.num := t #1 char.num substring FUNCTI...

Pith tools

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