Pathak, Jaideep, Abbas, Mohammad Shoaib, Harrington, Peter, Potvin, Corey, Clark, Adam, Skinner, Patrick, Gupta, Aayush, Ge, Tao, Ravuri, Suman, Adams, Corey, Brenowitz, Noah, Carpentieri, Alberto, Stock, Jason, Leinonen, Jussi, Hu, Zeyuan, Hennigh, Oliver, Geneva, Nicholas, Pritchard, Mike, Durran, Dale · Jan 24, 2026
Skillful Thunderstorm Forecasts Learned Directly from Observations
Markdown · JSON · Original source
The authors introduce StormScope, a system for nowcasting—the forecast of storms over the first few hours after observations are collected. It learns storm-scale evolution from satellite, radar, and lightning observations, without using a numerical model for training or initialization. The authors report that across more than 900 forecasts, StormScope outperformed a forthcoming operational numerical modeling system in deterministic and probabilistic skill through the 0-2-h nowcasting window. They also report stronger skill for strong radar echoes, including at both 30 min and 2 h lead times, and say the system anticipates storms that develop after forecast initialization. That matters because the approach is not limited to projecting existing radar echoes forward, according to the authors. The paper presents learning directly from observations as a way around initialization and microphysics bottlenecks that, it argues, limit numerical models and their emulators. The abstract does not give the evaluation details needed to assess how the reported results translate to operational warnings. How does StormScope perform in operational use, beyond the forecasts evaluated in the paper?
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
physics.ao-ph · Article
Source details
Original source date: Jan 24, 2026 (2026-01-24).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-10T07:09:59.148Z. First observed: 2026-11-03T08:51:16.698Z. Retrieved: 2026-11-03T18:02:42.872Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The abstract reports forecast comparisons but does not provide the evaluation setup, detailed metrics, or operational-use results needed to assess how the reported performance translates to warnings. The paper’s reported performance and claims are the authors’ findings; the supplied evidence is the abstract, not the full evaluation.
Open question
How does StormScope perform in operational use, beyond the forecasts evaluated in the paper?
Source-stated
The authors describe StormScope as learning storm-scale evolution from satellite, radar, and lightning observations, without numerical models in training or initialization.
learns storm-scale evolution directly from satellite, radar, and lightning observations over the continental United States with no numerical model used for training or initialization
Source-stated
The authors report that StormScope exceeded a forthcoming operational numerical modeling system in deterministic and probabilistic skill across more than 900 forecasts in the 0-2-h nowcasting window.
Across more than 900 forecasts, it greatly exceeds the National Oceanic and Atmospheric Administration's state-of-the-art forthcoming operational numerical modeling system in both deterministic and probabilistic skill through the 0-2-h nowcasting window.
Source-stated
The authors report that StormScope anticipates storms developing after forecast initialization, not just the continuation of existing radar echoes.
Crucially, it also anticipates storms that develop after forecast initialization, establishing its utility beyond extrapolation of existing radar echoes.
Source-stated
The authors present learning directly from observations as a route to storm-scale forecasts not constrained by bottlenecks they identify in numerical models and their emulators.
a route toward skillful, well-calibrated storm-scale forecasts that is unconstrained by the initialization and microphysics bottlenecks limiting numerical models and their emulators.
Source and contributor context
Pathak, Jaideep, Abbas, Mohammad Shoaib, Harrington, Peter, Potvin, Corey, Clark, Adam, Skinner, Patrick, Gupta, Aayush, Ge, Tao, Ravuri, Suman, Adams, Corey, Brenowitz, Noah, Carpentieri, Alberto, Stock, Jason, Leinonen, Jussi, Hu, Zeyuan, Hennigh, Oliver, Geneva, Nicholas, Pritchard, Mike, Durran, Dale · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: 80ec1f62-3eb1-4f55-8313-3bcb78a95575.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-10T07:09:59.148Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 5f969bdc5edba8fc9c1ee41aa83d2651a204f136a5f99ae73802210f16ea11bd. Source text hash: 456cb6333a4f6700f37b2a4300e1da4213b9cf0ed262897556540e2d12e45883.
Revision and source-check history
Version 1 · 2026-11-10T07:09:59.148Z
Skillful Thunderstorm Forecasts Learned Directly from Observations
The authors introduce StormScope, a system for nowcasting—the forecast of storms over the first few hours after observations are collected. It learns storm-scale evolution from satellite, radar, and lightning observations, without using a numerical model for training or initialization. The authors report that across more than 900 forecasts, StormScope outperformed a forthcoming operational numerical modeling system in deterministic and probabilistic skill through the 0-2-h nowcasting window. They also report stronger skill for strong radar echoes, including at both 30 min and 2 h lead times, and say the system anticipates storms that develop after forecast initialization. That matters because the approach is not limited to projecting existing radar echoes forward, according to the authors. The paper presents learning directly from observations as a way around initialization and microphysics bottlenecks that, it argues, limit numerical models and their emulators. The abstract does not give the evaluation details needed to assess how the reported results translate to operational warnings. How does StormScope perform in operational use, beyond the forecasts evaluated in the paper?
Original source
physics.ao-ph · Article
Source details
Original source date: Jan 24, 2026 (2026-01-24).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-10T07:09:59.148Z. First observed: 2026-11-03T08:51:16.698Z. Retrieved: 2026-11-03T18:02:42.872Z.
Limits
The abstract reports forecast comparisons but does not provide the evaluation setup, detailed metrics, or operational-use results needed to assess how the reported performance translates to warnings. The paper’s reported performance and claims are the authors’ findings; the supplied evidence is the abstract, not the full evaluation.
Open question
How does StormScope perform in operational use, beyond the forecasts evaluated in the paper?
Source-stated
The authors describe StormScope as learning storm-scale evolution from satellite, radar, and lightning observations, without numerical models in training or initialization.
learns storm-scale evolution directly from satellite, radar, and lightning observations over the continental United States with no numerical model used for training or initialization
Source-stated
The authors report that StormScope exceeded a forthcoming operational numerical modeling system in deterministic and probabilistic skill across more than 900 forecasts in the 0-2-h nowcasting window.
Across more than 900 forecasts, it greatly exceeds the National Oceanic and Atmospheric Administration's state-of-the-art forthcoming operational numerical modeling system in both deterministic and probabilistic skill through the 0-2-h nowcasting window.
Source-stated
The authors report that StormScope anticipates storms developing after forecast initialization, not just the continuation of existing radar echoes.
Crucially, it also anticipates storms that develop after forecast initialization, establishing its utility beyond extrapolation of existing radar echoes.
Source-stated
The authors present learning directly from observations as a route to storm-scale forecasts not constrained by bottlenecks they identify in numerical models and their emulators.
a route toward skillful, well-calibrated storm-scale forecasts that is unconstrained by the initialization and microphysics bottlenecks limiting numerical models and their emulators.
Source and contributor context
Pathak, Jaideep, Abbas, Mohammad Shoaib, Harrington, Peter, Potvin, Corey, Clark, Adam, Skinner, Patrick, Gupta, Aayush, Ge, Tao, Ravuri, Suman, Adams, Corey, Brenowitz, Noah, Carpentieri, Alberto, Stock, Jason, Leinonen, Jussi, Hu, Zeyuan, Hennigh, Oliver, Geneva, Nicholas, Pritchard, Mike, Durran, Dale · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: 80ec1f62-3eb1-4f55-8313-3bcb78a95575.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-10T07:09:59.148Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 5f969bdc5edba8fc9c1ee41aa83d2651a204f136a5f99ae73802210f16ea11bd. Source text hash: 456cb6333a4f6700f37b2a4300e1da4213b9cf0ed262897556540e2d12e45883.
uncertain · 2026-11-10T07:09:59.148Z · anthropic/claude-opus-5.5
Publication: 80ec1f62-3eb1-4f55-8313-3bcb78a95575
attribution — pass: The reading consistently attributes StormScope, its reported results and the bottleneck argument to the paper's authors. Phrases such as 'The authors report' and 'it argues' keep these as source claims, not established facts.
Here we introduce StormScope, a computationally efficient generative forecasting system
summary — uncertain: The summary leaves out context a general reader needs and weakens the source's force. (1) It never says StormScope is a generative AI/machine-learning model. (2) It drops NOAA as the operator of the comparison system. (3) It turns 'greatly exceeds' into 'outperformed' and omits the reported ~75% and 45% gains and the comparison with ML gains in global medium-range forecasting. (4) It omits that StormScope also beat a leading AI nowcasting model. (5) Central terms are unexplained: deterministic/probabilistic skill, radar echoes, initialization, microphysics and emulators. Smallest repair: name it a generative AI system, name NOAA, restore the figures, and add tooltips.
StormScope's skill for strong radar echoes exceeds the operational state of the art by approximately 75% and 45% at lead times of 30 min and 2 h, a relative improvement several times larger than machine learning has delivered for global medium-range forecasting.
date — pass: The originalPublished date of 2026-01-24 matches the v1 submission date. It does not use the 30 Oct 2026 v2 revision or the November feed date.
Submitted on 24 Jan 2026 (v1)
support — pass: Each claim in the draft has a verbatim excerpt from the abstract and is framed as the authors' report. The 'not limited to projecting existing echoes' point is attributed to the authors. The question about operational use follows from the evidence without assuming an answer. The limitations accurately note that only the abstract page was supplied.
Crucially, it also anticipates storms that develop after forecast initialization, establishing its utility beyond extrapolation of existing radar echoes.
media — pass: The selected media is the verified direct arXiv abstract page for this paper, which is the most direct way to reach the source.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXiv