Assertion Supported AI assessment confidence: 95% certainty 4/5 debate potential 1/5

Battaglia: Early ML in weather improved downstream calibration, not base models

Peter Battaglia · How AI is changing weather forecasting · Jan 2, 2026 · at 16:52

Google DeepMind's Peter Battaglia discusses the historical role of statistical methods and machine learning in meteorology.

0:00 / 0:31exact quote · 31.4s

720p mp4 · rendered on demand · StarZero watermark

“So actually, the first advent of like, AI and machine learning in weather forecasting, or at least some of the earliest, was not, like, trying to overhaul the whole weather forecast process itself, but making, you know, using more and more, like, statistical methods and linear regression and non-linear regression and neural networks and other types of Earlier machine learning techniques to improve the, not the base forecast, but like the specific application, so maybe we can calibrate your, you know, chance of rain better if we have a slightly better downstream model.”

quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →

The claim, plainly

The earliest applications of machine learning in weather forecasting focused on downstream post-processing and calibration rather than overhauling the base forecasting process itself.

a quote alone can be ambiguous, so this restates it with the references and context of the conversation around it resolved. AI-written; the tape below the player is the source.

Assessment

Peter Battaglia's historical characterization is fully supported by meteorological and computational literature. Early applications of artificial neural networks and statistical machine learning in weather forecasting during the 1990s and 2000s functioned primarily as extensions of Model Output Statistics (MOS), downstream post-processing, severe weather classification, downscaling, and bias correction (e.g., McCann 1992; Marzban & Stumpf 1996; Marzban 2003). Machine learning methods designed to overhaul or replace the core dynamical equations of numerical weather prediction (NWP) models in an end-to-end data-driven fashion (such as FourCastNet, Pangu-Weather, and GraphCast) did not emerge until 2022–2023.

Early operational and research applications of neural networks in meteorology during the 1990s and 2000s (such as McCann 1992 and Marzban 2003) focused on post-processing dynamical model output and statistical calibration rather than replacing core dynamical models.
Comprehensive reviews in meteorological literature (e.g., Royal Society 2021 reviews and ECMWF technical surveys) confirm that for decades ML was confined to downstream calibration, post-processing, and parameterization emulation.
Full end-to-end machine learning models overhauling the base NWP forecasting process (e.g., FourCastNet, Pangu-Weather, GraphCast) were only introduced starting around 2022.

Checked against

[1] wikipedia.org [2] stanmoreuk.org [3] ametsoc.org [4] weathersats.com [5] medium.com [6] weather.gov [7] arxiv.org [8] washington.edu

"Marzban" "neural network" "postprocessing" OR "MOS" weather 1990s 2000s early applications of machine learning neural networks weather forecasting review 1990s 2000s postprocessing MOS "Rasp" OR "Dueben" "machine learning" history numerical weather prediction post-processing "neural networks" "model output statistics" weather forecasting 1990s 2000s review history machine learning weather forecasting post-processing calibration NWP artificial neural networks weather forecasting history MOS post processing review literature machine learning in numerical weather prediction history postprocessing dynamical models review

assessed as of 2026-09-02

This assessment is an AI reading of a public statement, made against the cited references with the confidence shown. It can be wrong. The quote, the tape and the references are all on this page; if something is off, use report an error below. full disclaimer →

Mentioned

AI and machine learning weather forecasting

⚑ Spot an error? Report it

More from Peter Battaglia

Opinion

Battaglia: Data pipelines matter more than neural network architectures today

“I don't feel that neural network architectures these days tend to be the sort of exciting part. It's usually more of like the training and the sort of data how you handle the data and that kind of thing.”

Peter Battaglia Jan 2, 2026 ▶ 25:22 How AI is changing weather forecasting

Disclosure

DeepMind uses transformers to model spatial interactions in weather, not time

“We actually use transformers not to model the spatial, the interactions in weather over time, like the sequence of text, but in space.”

Peter Battaglia Jan 2, 2026 ▶ 29:54 How AI is changing weather forecasting

Insight

Battaglia: AI weather models leverage macroscopic spatial ranges unlike traditional simulations

“In a traditional model, the way it simulates that is it, in very fine detail, it kind of figures out, like, what's the pressure, and the temperature, and the wind, and the moisture, and what are those things? What's going to happen next is determined strictly …”

Peter Battaglia Jan 2, 2026 ▶ 31:24 How AI is changing weather forecasting

Disclosure

DeepMind primarily uses traditional supervised learning for AI weather models

“What we use in our weather forecasting models, and a lot of folks out in the community are using it as, you know, as this field is advancing and this AI-based weather forecasting is developing, we're still mostly using fairly traditional machine learning, supe…”

Peter Battaglia Jan 2, 2026 ▶ 24:01 How AI is changing weather forecasting

Insight

Battaglia: Precipitation and wind are harder to forecast due to fine-scale dynamics

“So that type of very high resolution, complex, you know, patterns of precipitation, for example, Wind as well. Those are much harder to predict because you're effectively predicting a lot more information. You can't just sort of summarize it by saying, oh, eve…”

Peter Battaglia Jan 2, 2026 ▶ 21:58 How AI is changing weather forecasting

Insight

Battaglia: Weather is Markovian unlike text sequences in language models

“Weather is actually different from text in a fundamental way. In fact, most physical processes are. They are what's called Markov, in that the most recent state of the system determines the subsequent state.”

Peter Battaglia Jan 2, 2026 ▶ 29:17 How AI is changing weather forecasting

Battaglia: Early ML in weather improved downstream calibration, not base models

More from Peter Battaglia

Turn any episode into a week of clips.