How Hurricane Forecasting Works

Hurricane forecasting has improved dramatically over the past half-century. Five-day track forecasts today are about as accurate as three-day forecasts were a few decades ago, and average track errors have fallen by more than half since 1990, progress that has given coastal communities far more time to prepare and evacuate.¹ That progress rests on a chain of technology that turns scattered observations into actionable forecasts.

Watching the Storm: Satellites

Satellites are the backbone of hurricane monitoring, especially over the open ocean where there are no other observations. Geostationary satellites such as NOAA's GOES series hover over a fixed point and image the Earth every few minutes, tracking a storm's position, structure, and rapid changes in near-real time.² Polar-orbiting satellites add high-resolution passes and microwave data that can peer through clouds to reveal the eye and inner core. When direct measurements aren't available, forecasters estimate a storm's intensity from the imagery alone using the Dvorak technique, which relates cloud patterns to wind speed.¹

Flying Into the Storm: Hurricane Hunters

Satellites can't directly measure the wind and pressure inside a hurricane, so Hurricane Hunter aircraft fly through it. NOAA's WP-3D Orions and the Air Force Reserve's WC-130J aircraft penetrate the eyewall and release dropsondes, instrument packages that radio back temperature, humidity, pressure, and wind as they parachute to the sea.¹

Tail Doppler radar aboard the aircraft maps the storm's three-dimensional wind field. This in-storm data measurably improves forecasts: feeding aircraft observations into the models has been shown to reduce track and intensity errors.³ It's the part of the job that keeps crews flying into the eyewall in the first place, since there's still no way to get those numbers from space.

Predicting the Path: Numerical Models

The heart of modern forecasting is numerical weather prediction: supercomputers solving the physics of the atmosphere. Forecasters consult global models like the American GFS and the European ECMWF, along with hurricane-specific, high-resolution models such as HWRF and its successor HAFS.² Because no single model is perfect, forecasters run ensembles (many slightly different model runs) and compare the "spaghetti" plots of their tracks; agreement signals confidence, and divergence signals uncertainty.¹

Those spaghetti plots are, in our experience, the most misread graphic the public sees. A tight bundle of lines is not a promise, and a single stray line is not nothing: each strand is one plausible future, not a vote, and the value is in the spread, not in any one noodle. We read the whole envelope, and so should you.

The NHC Forecast Process and the Cone

Every six hours during an active storm, the National Hurricane Center synthesizes all of this (satellite, aircraft, buoy, and model data) into an official forecast of the storm's track and intensity out to five days.¹ The familiar "cone of uncertainty" shows where the storm's center is likely to go: it's sized so that the center stays within the cone about two-thirds of the time, based on historical error.¹ The cone depicts only the probable path of the center, not the storm's size or its hazards, which routinely extend far outside it. That distinction is the single most important thing to understand about the graphic, and it's the one most often missed. When a storm threatens land, the NHC issues hurricane watches (conditions possible within 48 hours) and warnings (expected within 36 hours), plus separate storm-surge watches and warnings.¹

Intensity: the Hard Problem

Track forecasting has improved steadily, but intensity forecasting remains harder, and it's the part of the job we trust least. Predicting exactly when and how fast a storm will strengthen depends on small-scale, fast-changing processes in the inner core that the models still struggle to resolve.¹ Rapid intensification, a jump of at least 35 mph (56 km/h) in 24 hours, is the case we lose sleep over: hardest to call and most dangerous when it happens, as storms like Michael (2018) and Ida (2021) showed by strengthening right up to landfall. We can tell you where a storm is going with real confidence days out; we still can't reliably tell you how strong it'll be when it arrives. Closing that gap is a central goal of current research.³

A Track Record of Improvement

The long-term trend is one of steady, hard-won gains: better satellites, more aircraft data, faster computers, and smarter models have together transformed hurricane forecasting from a few hours' warning into a multi-day outlook. The 2017 addition of dedicated storm-surge watches, warnings, and inundation maps extended that progress to the deadliest hurricane hazard.¹ Each increment of lead time translates directly into lives saved.

Sources

1. National Hurricane Center. Definition of the NHC Track Forecast Cone. NOAA. https://www.nhc.noaa.gov/aboutcone.shtml ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹

2. National Oceanic and Atmospheric Administration (2023). NOAA launches new hurricane forecast model as Atlantic season starts strong. https://www.noaa.gov/news-release/noaa-launches-new-hurricane-forecast-model-as-atlantic-season-starts-strong ↩ ↩²

3. NOAA Atlantic Oceanographic and Meteorological Laboratory, Hurricane Research Division. Aircraft reconnaissance and hurricane model improvement. https://www.aoml.noaa.gov/hrd/ ↩ ↩²