Njord: A Probabilistic Graph Neural Network for Ensemble Ocean Forecasting

Abstract

Ocean dynamics are inherently chaotic, yet existing machine learning ocean models produce only deterministic forecasts. We introduce Njord, a probabilistic data-driven model for ocean forecasting, applicable to both global and regional domains. Njord combines a deep latent variable framework with a graph neural network architecture, enabling sampling each forecast step in a single forward pass. We apply Njord globally at 0.25° resolution and regionally to the Baltic Sea at 2 km resolution. To scale to these large ocean grids we introduce K-means cluster meshes that adapt to irregular sea surface geometry. Experiments demonstrate strong performance on both domains compared to deterministic machine learning baselines, while also providing uncertainty estimates from the sampled ensemble forecasts. On the global OceanBench benchmark, Njord achieves the lowest errors on average across upper-ocean variables when evaluated against real-world observations, with the largest improvements in surface temperature prediction.

Conditional-Risk Optimization and KKT Geometry of a Regime- and Horizon-Adaptive Probabilistic Ensemble for Local Weather Forecasting
Boris V. Malozyomov, Nikita V. Martyushev, Denis V. Valuev +4
Mathematics13 Oct 2026Peer-reviewedCC BY 4.0Atmospheric science & meteorologyCrossref
SDECast: Probabilistic Weather Forecasting in Continuous Time with Neural SDEs
Maria Marchenko, Martin Andrae, Fredrik Lindsten +1
arXiv2 Oct 2026PreprintarXiv licenceFull textAtmospheric science & meteorology StatisticsarXiv
Graph-based probabilistic earthquake clustering and forecasting in Sumatra
Giuseppe Petrillo, Stefania Gentili, Luca Dal Zilio
Tectonophysics1 Oct 2026Peer-reviewedCC BY 4.0Full textGeophysics & seismologyHAL
Advancing Near-Term Water Quality Forecasting with Explainable Machine Learning and Probabilistic Uncertainty Quantification
Rohit Shukla, Cayelan C. Carey, Adrienne Breef-Pilz +1
ACS ES&T water16 Oct 2026Peer-reviewedCC BYFull textEnvironmental science & pollutionEurope PMC
Predictability-Guided Multiscale Probabilistic Forecasting of Wind Direction under Extreme Shear
Hailong Shu
arXiv15 Sep 2026PreprintCC BY-NC-ND 4.0Full textAtmospheric science & meteorologyarXiv

Cite

BibTeX RIS

BibTeX

@misc{holmberg2026njord,
  title = {{Njord: A Probabilistic Graph Neural Network for Ensemble Ocean Forecasting}},
  author = {Daniel Holmberg and Joel Oskarsson and Erik Wikingsson and Fredrik Lindsten and Teemu Roos},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2605.15470},
  eprint = {2605.15470},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2605.15470},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - Njord: A Probabilistic Graph Neural Network for Ensemble Ocean Forecasting
AU  - Daniel Holmberg
AU  - Joel Oskarsson
AU  - Erik Wikingsson
AU  - Fredrik Lindsten
AU  - Teemu Roos
PY  - 2026
DA  - 2026/10/14
PB  - arXiv
DO  - 10.48550/arxiv.2605.15470
UR  - https://arxiv.org/abs/2605.15470
AB  - Ocean dynamics are inherently chaotic, yet existing machine learning ocean models produce only deterministic forecasts. We introduce Njord, a probabilistic data-driven model for ocean forecasting, applicable to both global and regional domains. Njord combines a deep latent variable framework with a graph neural network architecture, enabling sampling each forecast step in a single forward pass. We apply Njord globally at 0.25° resolution and regionally to the Baltic Sea at 2 km resolution. To scale to these large ocean grids we introduce K-means cluster meshes that adapt to irregular sea surface geometry. Experiments demonstrate strong performance on both domains compared to deterministic machine learning baselines, while also providing uncertainty estimates from the sampled ensemble forecasts. On the global OceanBench benchmark, Njord achieves the lowest errors on average across upper-ocean variables when evaluated against real-world observations, with the largest improvements in surface temperature prediction.
ER  -

Abstract

Related

Cite