How Do Forecasters Predict a Tropical Cyclone's Track and Intensity?

Short answer: cyclone forecasters combine satellites, aircraft, radar, ocean measurements, global and regional models, statistical guidance, ensemble probabilities, and expert judgment. Track depends mainly on large-scale steering winds; intensity depends on smaller processes in the storm core and ocean, which are harder to observe and resolve. A forecast therefore includes uncertainty, not one guaranteed line and wind speed.

What must be predicted?

A tropical-cyclone forecast is several linked forecasts:

• Genesis: whether a disturbance will organize into a cyclone
• Track: the path of the circulation centre
• Intensity: commonly the maximum sustained near-surface wind and minimum central pressure
• Size: how far tropical-storm and hurricane-force winds extend
• Hazards: storm surge, waves, rainfall, flooding, tornadoes, and landslides

A good centre-track forecast can coexist with a damaging rainfall or surge forecast error. The centre line is not a map of total risk.

How forecasters establish the current storm

Prediction begins with an initial state. Geostationary satellites follow cloud motion and structure frequently. Polar-orbiting satellites measure temperature, moisture, winds, rain, sea surface, and microwave signatures through clouds. Scatterometers infer ocean-surface winds.

Where available, reconnaissance aircraft measure pressure, wind, temperature, humidity, and the storm centre, while expendable instruments fall through the core. Coastal radar samples rain and wind as a storm approaches land. Buoys, ships, floats, and gliders describe the ocean and surface conditions.

Every observation has coverage and error limits. Data assimilation combines these measurements with a previous forecast to produce a physically consistent three-dimensional estimate.

Why track is usually easier than intensity

A cyclone is carried by broad atmospheric flow, often described as steering winds. Ridges, troughs, monsoons, and nearby weather systems shape the path over hundreds or thousands of kilometres. Global models represent these large patterns increasingly well.

Intensity depends on the eyewall, rainbands, convection, internal vortices, wind shear, dry-air intrusion, ocean heat beneath the surface, and exchanges across the air-sea boundary. Some operate at scales near or below model-grid spacing and change rapidly.

How numerical models make a forecast

1. Initialize: assimilate observations into atmospheric and ocean states.
2. Integrate physics: calculate winds, pressure, heat, moisture, clouds, radiation, and surface exchange forward in time.
3. Nest high resolution: regional hurricane models place a finer grid around the storm to represent the core.
4. Couple the ocean: simulate how winds mix cooler water upward and how ocean heat feeds the storm.
5. Post-process: correct systematic biases and derive track, wind radii, intensity, rain, and hazard guidance.
6. Compare guidance: forecasters assess multiple models, ensembles, observations, and recent trends.

What ensembles add

An ensemble forecast runs many plausible versions with small differences in initial conditions or model formulation. The spread shows sensitivity. A tight cluster supports higher track confidence; divergent branches reveal competing scenarios.

The “spaghetti plot” is not itself a probability map because members may not be independent or calibrated. Official forecasts use verified ensembles, model skill, known biases, and expert interpretation.

What the forecast cone means

A track cone represents the probable range of the cyclone centre based on historical forecast errors at each lead time. It normally widens farther into the future. It does not show storm size, wind coverage, rainfall, surge, or a guaranteed containment boundary.

Hazards can occur well outside the cone, and the centre can move outside it. Users should follow watches, warnings, surge maps, rain forecasts, and local emergency instructions rather than focus only on the line.

Rapid intensification

Rapid intensification is commonly defined as a large rise in maximum sustained wind within 24 hours, often at least 30 knots in Atlantic practice. Warm deep water, moist air, weak shear, and a favorable core raise the chance, but internal convective evolution remains hard to time.

Satellites, aircraft, ocean profiling, rapid-update models, and machine learning improve probabilities. A probabilistic warning can be valuable even when the exact wind increase is uncertain.

Why the ocean below the surface matters

A thin warm surface layer may cool quickly when cyclone winds mix colder water upward. A deep reservoir of warm water supplies more energy. Ocean heat content, eddies, salinity layers, currents, and the storm's speed therefore affect intensity.

Slow storms churn the same water longer but can also produce extreme rainfall. Fast storms cross ocean features quickly and bring hazards to land sooner.

AI and machine-learning models

AI models learn atmospheric evolution from historical analyses and can generate global forecasts rapidly. Specialized cyclone systems can predict tracks, intensity, and wind extent with large ensembles at lower computational cost. They may complement physics-based models and free resources for uncertainty sampling.

They still depend on the quality of training and real-time observations. Rare extremes, climate shifts, observation changes, and physically inconsistent outputs require evaluation and safeguards.

A measured AI forecast gain

A Nature study of WeatherNext Cyclones evaluated storms from 2023 to 2025 and live operation during 2025. It reported roughly one additional day of useful forecast skill on average across track, intensity, and wind extent compared with selected operational baselines. Read An AI Cyclone Model Delivers a Day More Forecasting Skill, on Average.

“One extra day” is an average equivalence in verification scores, not a promise that every five-day forecast is now as accurate as every four-day forecast. Gains vary by basin, lead time, storm, metric, and comparator. The model supports, rather than replaces, official centres and local warnings.

From model output to an official forecast

Human forecasters check initialization, recent motion, structural changes, model biases, discontinuities, and consistency across forecast cycles. They may avoid following one sudden model jump until evidence supports it, while communicating low-probability high-impact alternatives.

Official centres coordinate with national services and emergency managers. Local rainfall, terrain, river response, buildings, exposure, and evacuation time determine impact beyond meteorology.

How forecast skill is measured

• Track error: distance between forecast and observed centre
• Intensity error: difference in maximum sustained wind or pressure
• Wind-radii error: accuracy of storm-size contours
• Probability skill: calibration and discrimination for events such as rapid intensification
• Hazard verification: rain, surge, waves, and warning performance
• Consistency and timeliness: whether guidance arrives early and avoids unnecessary jumps

Average error can hide rare catastrophic misses. Verification should be stratified by basin, lead time, storm strength, observation quality, and landfall context.

What users should do with uncertainty

Use the latest official forecast, not a single raw model shared online. Recheck as the storm approaches because new observations narrow possibilities. Prepare for hazards across the full warning area and act according to local authorities, especially when evacuation lead time is long.

The mental model

Track forecasting asks which broad atmospheric river will carry the cyclone. Intensity forecasting asks how a turbulent heat engine will reorganize while interacting with ocean and air. Models propose many paths, observations update the starting point, and expert centres turn evolving probabilities into warnings tied to local hazards.