Weather Lab users can explore and compare the predictions from various AI and physics-based models. When read together, these predictions may also help weather agencies and emergency service experts higher anticipate a cyclone’s path and intensity. This might help experts and decision-makers higher prepare for various scenarios, share news of risks involved and support decisions to administer a cyclone’s impact.
It is vital to stress that Weather Lab is a research tool. Live predictions shown are generated by models still under development and aren’t official warnings. Please keep this in mind when using the tool, including to support decisions based on predictions generated by Weather Lab. For official weather forecasts and warnings, consult with your local meteorological agency or national weather service.
AI-powered cyclone predictions
In physics-based cyclone prediction, the approximations required to satisfy operational demands mean it’s difficult for a single model to excel at predicting each a cyclone’s track and its intensity. It’s because a cyclone’s track is governed by vast atmospheric steering currents, whereas a cyclone’s intensity will depend on complex turbulent processes inside and around its compact core. Global, low-resolution models perform best at predicting cyclone tracks, but don’t capture the fine-scale processes dictating cyclone intensity, which is why regional, high-resolution models are needed.
Our experimental cyclone model is a single system that overcomes this trade-off, with our internal evaluations showing state-of-the-art accuracy for each cyclone track and intensity. It’s trained to model two distinct varieties of data: an unlimited reanalysis dataset that reconstructs past weather over your complete Earth from hundreds of thousands of observations, and a specialized database containing key information concerning the track, intensity, size and wind radii of nearly 5,000 observed cyclones from the past 45 years.
Modeling the evaluation data and cyclone data together greatly improves cyclone prediction capabilities. For instance, our initial evaluations of NHC’s observed hurricane data, on test years 2023 and 2024, within the North Atlantic and East Pacific basins, showed that our model’s 5-day cyclone track prediction is, on average, 140 km closer to the true cyclone location than ENS — the leading global physics-based ensemble model from ECMWF. That is comparable to the accuracy of ENS’s 3.5-day predictions — a 1.5-day improvement that has typically taken over a decade to attain.
While previous AI weather models have struggled to calculate cyclone intensity, our experimental cyclone model outperformed the typical intensity error of the National Oceanic and Atmospheric Administration (NOAA)’s Hurricane Evaluation and Forecast System (HAFS), a number one regional, high-resolution physics-based model. Preliminary tests also show our model’s predictions of size and wind radii are comparable with physics-based baselines.
Here we visualize track and intensity prediction errors, and show evaluation results of our experimental cyclone model’s average performance as much as five days prematurely, in comparison with ENS and HAFS.
