Pith. sign in

REVIEW 9 cited by

Data driven weather forecasts trained and initialised directly from observations

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2407.15586 v1 pith:ZXUHLWRI submitted 2024-07-22 physics.ao-ph cs.LG

classification physics.ao-phcs.LG
keywords observationsweatherdataforecastfutureapproachbeenforecasts
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

Skilful Machine Learned weather forecasts have challenged our approach to numerical weather prediction, demonstrating competitive performance compared to traditional physics-based approaches. Data-driven systems have been trained to forecast future weather by learning from long historical records of past weather such as the ECMWF ERA5. These datasets have been made freely available to the wider research community, including the commercial sector, which has been a major factor in the rapid rise of ML forecast systems and the levels of accuracy they have achieved. However, historical reanalyses used for training and real-time analyses used for initial conditions are produced by data assimilation, an optimal blending of observations with a physics-based forecast model. As such, many ML forecast systems have an implicit and unquantified dependence on the physics-based models they seek to challenge. Here we propose a new approach, training a neural network to predict future weather purely from historical observations with no dependence on reanalyses. We use raw observations to initialise a model of the atmosphere (in observation space) learned directly from the observations themselves. Forecasts of crucial weather parameters (such as surface temperature and wind) are obtained by predicting weather parameter observations (e.g. SYNOP surface data) at future times and arbitrary locations. We present preliminary results on forecasting observations 12-hours into the future. These already demonstrate successful learning of time evolutions of the physical processes captured in real observations. We argue that this new approach, by staying purely in observation space, avoids many of the challenges of traditional data assimilation, can exploit a wider range of observations and is readily expanded to simultaneous forecasting of the full Earth system (atmosphere, land, ocean and composition).

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 9 Pith papers

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

  1. Global reanalysis from observations alone with machine learning

    physics.ao-ph 2026-07 conditional novelty 7.0 of 10

    Observation-only machine learning can generate multi-decade global atmospheric reanalyses with large-scale skill near ERA5 and surface errors between ERA-Interim and ERA5, in a single day of compute.

  2. Prithvi-Precip: Integrating Satellite Observations into an Atmospheric AI Foundation Model for Precipitation Forecasting

    physics.ao-ph 2026-08 conditional novelty 6.0 of 10

    Training on IMERG satellite precipitation and directly ingesting satellite observations improves medium-range AI precipitation forecasts, with the largest gains at lead times under 40 hours.

  3. Learning from nature: insights into GraphDOP's representations of the Earth System

    physics.ao-ph 2025-08 conditional novelty 6.0 of 10

    GraphDOP, trained purely on meteorological observations, spontaneously forms unified latent representations of Earth system state, viewing effects, and synoptic dynamics.

  4. CERA: A Framework for Improved Generalization of Machine Learning Models to Changed Climates

    physics.ao-ph 2025-08 conditional novelty 6.0 of 10

    CERA uses an autoencoder with latent-space alignment to learn climate-invariant representations, improving generalization of moist-physics parameterizations to a +4K climate without warmer-climate labels.

  5. Generative Lagrangian data assimilation for ocean dynamics under extreme sparsity

    physics.ao-ph 2025-07 conditional novelty 6.0 of 10

    A neural operator-conditioned diffusion model reconstructs ocean surface states with high-wavenumber fidelity from 99% to 99.9% sparse observations, outperforming standard UNET and FNO baselines.

  6. Diffusion-LAM: Probabilistic Limited Area Weather Forecasting with Diffusion

    cs.LG 2025-02 conditional novelty 6.0 of 10

    A diffusion-based limited area weather model that uses future boundary conditions from a global model improves short-lead forecast accuracy and boundary consistency.

  7. OMG-HD: A High-Resolution AI Weather Model for End-to-End Forecasts from Observations

    physics.ao-ph 2024-12 conditional novelty 6.0 of 10

    An observation-only AI model trained on the RTMA analysis reports lower 0-12h surface forecast errors than operational HRRR and IFS-HRES over the contiguous US.

  8. GraphDOP: Towards skilful data-driven medium-range weather forecasts learnt and initialised directly from observations

    physics.ao-ph 2024-12 conditional novelty 6.0 of 10

    A graph-neural-network weather model trained only on raw observations produces skillful global forecasts out to five days, with tropical 2-meter temperature forecasts competitive with the operational IFS.

  9. Satellite Observations Guided Diffusion Model for Accurate Meteorological States at Arbitrary Resolution

    cs.LG 2025-02 reject novelty 5.0 of 10

    A satellite-conditioned diffusion model with station-guided sampling is claimed to downscale ERA5 weather fields to 6.25 km more accurately than existing methods, but the evaluation is circular.

Pith tools