HKO finds that AI models such as FengWu tracked 2024 typhoons better

In its 2024 verification, the Hong Kong Observatory found that AI weather models matched or slightly beat traditional models for 10-day patterns and had smaller errors in typhoon tracks.

  • 2 an organization

    HKO found smaller mean errors in tropical cyclone tracks for the AI models than for traditional models in 2024.

Depth runs from 1, a convenience or time saved, to 5, a life or a livelihood preserved.

Status
Reported benefit.
Evidence

foundstudiedreportedclaimed

Only the deployer supports the benefit.

Compared with

Traditional numerical weather prediction models verified by HKO for 2024. Measured.

The system’s part
HKO verified the AI models' own forecasts against traditional models over 2024, which isolates the models' skill. HKO describes the models as deep learning systems trained on decades of meteorological data. HKO runs the models itself, so it is their operator, while the makers are outside HKO. No effect on HKO's official forecasts is measured.
Limits
Not yet checked.
Costs and harms
Not yet checked.
Later
Not yet checked.

Download data Cite Suggest a correction

Benefit type
Improved Task Accuracy
System
AI weather prediction models run in real time by HKO, including FengWu, FuXi, Pangu-Weather and AIFS
Made by
Shanghai Artificial Intelligence Laboratory (FengWu), Fudan University (FuXi), Huawei Cloud (Pangu-Weather) and ECMWF (AIFS)
Used by
Hong Kong Observatory
Where
Hong Kong, China (western North Pacific and South China Sea)
Stage
In routine use
Date
, when the source was published
Kind
Aggregate result
Broad benefit
Estimating the Quantifiable Characteristics of Products, Events, or Information
Last revised
31 October 2026

Sources

  1. Report, by the organization that used it.

    Hong Kong Observatory. Applications of Artificial Intelligence Models in Weather Forecasting. Article by LAI Sin-ki, December 2025.

    According to the Observatory's verification on weather prediction models in 2024, the performance of various AI models in forecasting large-scale weather patterns over the next 10 days has already reached a level similar to or even slightly surpassed that of traditional models.

Details

AI system

Workflow

HKO ran the AI models in real time and turned their output into forecasting products such as charts of tropical cyclone tracks. HKO corrected model biases from past performance with statistical and machine learning methods.

Follow-up

Conditions for replication

Forecasters still weigh several traditional and AI models and adjust for local conditions. Intensity forecasts for mature cyclones come from traditional models, since AI models underestimate them.

Entry details

Sector

Weather forecasting / tropical cyclones.

Purpose

Huawei, the Shanghai Artificial Intelligence Laboratory, Fudan University and ECMWF built Pangu-Weather, FengWu, FuXi and AIFS to forecast global weather with models trained on decades of past weather data.

Not yet checked: model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.

Cite this record

Peter H. Vartanian. (2026). HKO finds that AI models such as FengWu tracked 2024 typhoons better (AIBD-0753). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0753.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)

BibTeX Data Field definitions