pith. sign in

arxiv: 2606.25076 · v1 · pith:FJSYO22Qnew · submitted 2026-06-23 · ⚛️ physics.ao-ph · cs.CY

Machine learning is revolutionizing weather forecasting -- the next step is a change in how we work

Peter Dueben , Peter Bauer , Oliver Fuhrer , Nikolay Koldunov , J{\o}rn Kristiansen This is my paper

Pith reviewed 2026-06-25 21:29 UTC · model grok-4.3

classification ⚛️ physics.ao-ph cs.CY
keywords machine learningweather forecastingvalue chaindigital technologiesagentic software engineeringgenerative methodsverification workflowsclimate services
0
0 comments X

The pith

Machine learning will reshape the entire weather forecasting value chain from model coding to service delivery.

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

The paper argues that machine learning's success in matching traditional forecast skill now requires a shift from output quality to the underlying working practices. It identifies how models are coded and developed, how observations and Earth-system data are exploited, how data and computing are managed, how systems are verified, and how information is turned into services. These changes are driven by six areas such as agentic software engineering, open and compressed data, shared verification workflows, interactive computing, and generative methods. A sympathetic reader would care because the changes promise faster and more accessible modelling and services while the centers must still preserve operational reliability and public-service obligations.

Core claim

Following the success of machine learning in producing weather predictions with competitive skill compared to complex traditional systems, machine learning and recent digital technologies will reshape the forecasting value chain: how models are coded and developed, how observations and Earth-system data are exploited, how data and computing are managed, how systems are verified, and how information is created, evaluated and turned into services. The paper discusses six non-exhaustive areas in which agentic software engineering, open and compressed data, shared verification workflows, interactive computing and generative methods may make modelling, evaluation and service creation faster, more

What carries the argument

Reshaping of the forecasting value chain via six areas (agentic software engineering, open and compressed data, shared verification workflows, interactive computing, and generative methods) that alter coding, data use, management, verification, and service creation.

If this is right

  • Model development will move from traditional coding to agentic software engineering.
  • Observations and Earth-system data will be handled through open and compressed formats.
  • Verification will rely on shared workflows across centres rather than isolated processes.
  • Service creation will incorporate interactive computing and generative methods.
  • Infrastructures, data stewardship, trust frameworks, and skills at centres must adapt.

Where Pith is reading between the lines

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

  • Smaller research groups or national services outside major centres could gain faster access to advanced modelling tools.
  • The shift may accelerate open-science practices by lowering barriers to data and code sharing.
  • Long-term human expertise may evolve toward oversight of automated systems rather than direct model building.

Load-bearing premise

These digital and machine-learning changes can be adopted at weather and climate centres while still maintaining scientific understanding, operational reliability, human expertise, and the public-service role.

What would settle it

Operational adoption of the six areas that results in measurable loss of forecast reliability or scientific insight in at least one major weather centre within five years.

read the original abstract

Following the success of machine learning in producing weather predictions with competitive skill compared to complex traditional systems, this article shifts attention from forecast output to the working practices that make prediction systems possible. We argue that machine learning and recent digital technologies will reshape the forecasting value chain: how models are coded and developed, how observations and Earth-system data are exploited, how data and computing are managed, how systems are verified, and how information is created, evaluated and turned into services. We discuss six non-exhaustive areas in which agentic software engineering, open and compressed data, shared verification workflows, interactive computing and generative methods may make modelling, evaluation and service creation faster, more interactive and more widely accessible. These changes will require weather and climate centres to adapt their infrastructures, data stewardship, trust and quality-assurance frameworks, skills and service delivery while maintaining scientific understanding, operational reliability, human expertise and their public-service role.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 2 minor

Summary. The manuscript is a perspective article arguing that recent successes of machine learning in producing competitive weather predictions will drive changes across the forecasting value chain. It identifies impacts on model coding and development, exploitation of observations and Earth-system data, management of data and computing, system verification, and the creation/evaluation of services. The paper discusses six non-exhaustive areas (agentic software engineering, open and compressed data, shared verification workflows, interactive computing, and generative methods) that could accelerate modelling, evaluation, and service creation while preserving scientific understanding, operational reliability, human expertise, and the public-service role of weather and climate centres.

Significance. If the forward-looking argument holds, the paper provides a timely framework for weather and climate centres to anticipate and plan adaptations in infrastructure, data stewardship, trust frameworks, skills, and service delivery. It explicitly balances technological opportunity with the need to retain core scientific and operational standards, which could help shape institutional responses to ML integration in operational forecasting.

minor comments (2)
  1. [Abstract] The abstract is lengthy and contains multiple clauses that could be streamlined for greater readability while retaining all key points.
  2. [Main text (discussion of areas)] The six areas are presented as non-exhaustive possibilities; adding one or two concrete, published examples of each (with citations) would strengthen the discussion without altering the perspective nature of the piece.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary, assessment of significance, and recommendation to accept the manuscript. No major comments were raised that require point-by-point response.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The manuscript is a forward-looking perspective article with no equations, derivations, fitted parameters, or quantitative claims. Its central argument consists of qualitative predictions about how ML and digital technologies may reshape forecasting practices; these are presented as non-exhaustive possibilities rather than results derived from internal logic or self-referential definitions. No load-bearing steps reduce to self-citations, ansatzes, or renamed known results. The text explicitly notes the need to preserve scientific understanding and reliability, keeping the discussion self-contained as an opinion piece on external technological trends.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a perspective paper with no quantitative models. It introduces no free parameters, mathematical axioms, or new entities. It takes the success of ML in weather forecasting as given without evidence in the abstract.

pith-pipeline@v0.9.1-grok · 5704 in / 1107 out tokens · 40034 ms · 2026-06-25T21:29:37.105982+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

21 extracted references · 9 canonical work pages

  1. [1]

    doi: 10.1109/MCSE.2021.3059437. Mihai Alexe, Eulalie Boucher, Peter Lean, Ewan Pinnington, Patrick Laloyaux, Anthony McNally, Simon Lang, Matthew Chantry, Chris Burrows, MarcinChrust,FlorianPinault,EthelVilleneuve,NielsBormann,andSeanHealy.Graphdop:Towardsskilfuldata-drivenmedium-rangeweather forecasts learnt and initialised directly from observations,

    work page doi:10.1109/mcse.2021.3059437 2021

  2. [2]

    Anna Allen, Stratis Markou, Will Tebbutt, et al

    URLhttps://arxiv.org/abs/2412.15687. Anna Allen, Stratis Markou, Will Tebbutt, et al. End-to-end data-driven weather prediction.Nature, 641:1172–1179,

    arXiv

  3. [3]

    Impact of grib compression on weather forecast data and data-handling applications, 08/2022

    Eugen Betke, Tiago Quintino, Simon Smart, and Tomas Wilhelmsson. Impact of grib compression on weather forecast data and data-handling applications, 08/2022

    2022

  4. [4]

    Accurate Medium-Range Global Weather Forecasting with

    doi: 10.1038/s41586-023-06185-3. Cristian Bodnar, Wessel P Bruinsma, Ana Lucic, Megan Stanley, Johannes Brandstetter, Patrick Garvan, Maik Riechert, Jonathan Weyn, Haiyu Dong, Anna Vaughan, et al. A foundation model for the earth system.Nature, 641:1180–1187,

    work page doi:10.1038/s41586-023-06185-3

  5. [5]

    doi: 10.1038/s41586-025-09005-y. Zied Ben Bouallègue, Mariana C A Clare, Linus Magnusson, Estibaliz Gascón, Michael Maier-Gerber, Martin Janoušek, Mark Rodwell, Florian Pinault, Jesper S Dramsch, Simon T K Lang, Baudouin Raoult, Florence Rabier, Matthieu Chevallier, Irina Sandu, Peter Dueben, Matthew Chantry,andFlorianPappenberger. Theriseofdata-drivenwea...

    work page doi:10.1038/s41586-025-09005-y

  6. [6]

    The rise of data-driven weather forecasting: A first statistical assessment of machine learning-based weather forecasts in an operational-like context

    doi: 10.1175/BAMS-D-23-0162.1. URL https://journals.ametsoc.org/view/journals/bams/aop/BAMS-D-23-0162.1/BAMS-D-23-0162.1.xml. Noah D Brenowitz, Tao Ge, Akshay Subramaniam, Peter Manshausen, Aayush Gupta, David M Hall, Morteza Mardani, Arash Vahdat, Karthik Kashinath, and Michael S Pritchard. Climate in a bottle: Towards a generative foundation model for t...

    work page doi:10.1175/bams-d-23-0162.1

  7. [7]

    Artificial intelligence for modeling and understanding extreme weather and climate events

    Gustau Camps-Valls, Miguel-Ángel Fernández-Torres, Kai-Hendrik Cohrs, Adrian Höhl, Andrea Castelletti, Aytac Pacal, Claire Robin, Francesco Martinuzzi, Ioannis Papoutsis, Ioannis Prapas, et al. Artificial intelligence for modeling and understanding extreme weather and climate events. Nature Communications, 16(1):1919,

    1919

  8. [8]

    Hans Hersbach, Bill Bell, Paul Berrisford, Shoji Hirahara, András Horányi, Joaquín Muñoz-Sabater, Julien Nicolas, Carole Peubey, Raluca Radu, Dinand Schepers, et al

    doi: 10.1038/s43247-025-02903-z. Hans Hersbach, Bill Bell, Paul Berrisford, Shoji Hirahara, András Horányi, Joaquín Muñoz-Sabater, Julien Nicolas, Carole Peubey, Raluca Radu, Dinand Schepers, et al. The era5 global reanalysis.Quarterly Journal of the Royal Meteorological Society, 146(730):1999–2049,

    work page doi:10.1038/s43247-025-02903-z 1999

  9. [9]

    Carlos E

    URLhttps://arxiv.org/abs/2510.22265. Carlos E. Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik Narasimhan. SWE-bench: Can language models resolve real-world GitHub issues? InInternational Conference on Learning Representations,

    arXiv

  10. [10]

    Nature , volume =

    doi: 10.1038/s41586-024-07744-y. Nikolay Koldunov and Thomas Jung. Local climate services for all, courtesy of large language models.Communications Earth & Environment, 5 (1):13,

    work page doi:10.1038/s41586-024-07744-y

  11. [11]

    Koldunov,Suvarchal K.Cheedela, Sergey Danilov,Dmitry Sidorenko,Sebastian Beyer,and ThomasJung

    NikolayV. Koldunov,Suvarchal K.Cheedela, Sergey Danilov,Dmitry Sidorenko,Sebastian Beyer,and ThomasJung. An oceanmodel portedby a largelanguagemodel:Experienceandlessonsfromfesom2(fortrantoctoc++/kokkos),2026. URLhttps://arxiv.org/abs/2606.11356. Patrick Laloyaux, Mihai Alexe, Eulalie Boucher, Peter Lean, Ewan Pinnington, Simon Lang, Tobias Necker, and An...

    Pith/arXiv arXiv 2026

  12. [12]

    Mathilde Leuridan, James Hawkes, Simon Smart, Emanuele Danovaro, Martin Schultz, and Tiago Quintino

    URLhttps://arxiv.org/abs/2308.13280. Mathilde Leuridan, James Hawkes, Simon Smart, Emanuele Danovaro, Martin Schultz, and Tiago Quintino. Polytope: an algorithm for efficient feature extraction on hypercubes.Journal of Big Data, 12(243),

    arXiv

  13. [13]

    Israel Leyva-Mayorga, Marc Martinez-Gost, Marco Moretti, Ana Pérez-Neira, Miguel Ángel Vázquez, Petar Popovski, and Beatriz Soret

    doi: 10.1186/s40537-025-01306-3. Israel Leyva-Mayorga, Marc Martinez-Gost, Marco Moretti, Ana Pérez-Neira, Miguel Ángel Vázquez, Petar Popovski, and Beatriz Soret. Satellite edge computing for real-time and very-high resolution earth observation.IEEE Transactions on Communications, 71(10):6180–6194,

    work page doi:10.1186/s40537-025-01306-3

  14. [14]

    Data driven weather forecasts trained and initialised directly from observations.arXiv preprint arXiv:2407.15586,

    Anthony McNally, Christian Lessig, Peter Lean, Eulalie Boucher, Mihai Alexe, Ewan Pinnington, Matthew Chantry, Simon Lang, Chris Burrows, Marcin Chrust, et al. Data driven weather forecasts trained and initialised directly from observations.arXiv preprint arXiv:2407.15586,

    arXiv

  15. [15]

    Climatelearn: Benchmarking machine learning for weather and climatemodeling

    Tung Nguyen, Jason Jewik, Hritik Bansal, Prakhar Sharma, and Aditya Grover. Climatelearn: Benchmarking machine learning for weather and climatemodeling. InA.Oh,T.Naumann,A.Globerson,K.Saenko,M.Hardt,andS.Levine,editors,AdvancesinNeuralInformationProcessing Systems,volume36,pages75009–75025.CurranAssociates,Inc.,2023. URLhttps://proceedings.neurips.cc/pape...

    2023

  16. [16]

    Do Data-Driven Models Beat Numerical Models in Forecasting Weather Extremes?

    doi: 10.5194/gmd-17-7915-2024. Open Geospatial Consortium. CoverageJSON Community Standard,

    work page doi:10.5194/gmd-17-7915-2024 2024

  17. [17]

    Enrique G Paredes, Linus Groner, Stefano Ubbiali, Hannes Vogt, Alberto Madonna, Kean Mariotti, Felipe Cruz, Lucas Benedicic, Mauro Bianco, JoostVandeVondele,etal.Gt4py:Highperformancestencilsforweatherandclimateapplicationsusingpython.arXivpreprintarXiv:2311.08322,

    arXiv

  18. [18]

    Weatherbench 2: A benchmark for the next generation of data-driven global weather models.arXiv preprint arXiv:2308.15560,

    Stephan Rasp, Stephan Hoyer, Alexander Merose, Ian Langmore, Peter Battaglia, Tyler Russel, Alvaro Sanchez-Gonzalez, Vivian Yang, Rob Carver, Shreya Agrawal, et al. Weatherbench 2: A benchmark for the next generation of data-driven global weather models.arXiv preprint arXiv:2308.15560,

    arXiv

  19. [19]

    Advancing earth observation through machine learning: A torchgeo tutorial.arXiv preprint arXiv:2603.02386,

    Caleb Robinson, Nils Lehmann, Adam J Stewart, Burak Ekim, Heng Fang, Isaac A Corley, and Mauricio Cordeiro. Advancing earth observation through machine learning: A torchgeo tutorial.arXiv preprint arXiv:2603.02386,

    arXiv

  20. [20]

    Petteri Taalas

    doi: 10.1038/s41467-025-62024-1. Petteri Taalas. Free global access to climate and weather data must continue,

    work page doi:10.1038/s41467-025-62024-1

  21. [21]

    High-level,high-resolutionoceanmodelingatallscaleswithoceananigans.arXivpreprintarXiv:2502.14148,

    Gregory L Wagner, Simone Silvestri, Navid C Constantinou, Ali Ramadhan, Jean-Michel Campin, Chris Hill, Tomas Chor, Jago Strong-Wright, XinKaiLee,FrancisPoulin,etal. High-level,high-resolutionoceanmodelingatallscaleswithoceananigans.arXivpreprintarXiv:2502.14148,

    arXiv