Hurricane aviation is changing faster than at any point since the WP-3D Orion entered service in the 1970s. The aircraft that fly into tropical cyclones today are aging, the sensors they carry are being overtaken by more capable ones, and entirely new platforms, from expendable drones to global satellite constellations, are moving into operational service. At the same time, artificial intelligence and machine learning are starting to reshape how forecast models digest the data all these platforms collect. Together, these advances point at the most stubborn gap in hurricane science: the inability to reliably predict how strong a storm will get. This article surveys the technologies, platforms, and scientific challenges that will define the next era of hurricane aviation.
Next-Generation Aircraft
NOAA's two WP-3D Orion turboprops, N42RF ("Kermit") and N43RF ("Miss Piggy"), were built in 1975 and 1976. They've been the backbone of civilian hurricane research for nearly five decades, but their age brings growing maintenance challenges and limits how easily modern sensors can be integrated. NOAA has pursued a recapitalization plan for the research fleet since the 2010s, studying replacements that include modified C-130J aircraft and purpose-built research platforms. The aircraft profiles article covers today's fleet in detail.
In 2020 and 2021, NOAA received Congressional funding for acquisition studies to evaluate successor aircraft. The bar is high: any replacement has to match or beat the P-3's ability to carry heavy scientific instrumentation, hold together in the severe turbulence of an eyewall, work at low altitude (1,500 to 10,000 feet, or 460 to 3,000 meters), and stay airborne for 8-to-10-hour missions. It also has to support deploying expendable sensors, dropsondes and small uncrewed aircraft alike.
On the military side, the Air Force Reserve's 53rd Weather Reconnaissance Squadron flies the WC-130J Super Hercules, a much newer platform delivered between 1999 and 2005 to replace the older WC-130H. These aircraft should remain operational for decades, though they too will eventually need succession planning. The WC-130J carries palletized weather instrumentation including the Improved Atmospheric Sounding System (IASS), which processes and transmits dropsonde data in real time.
The Gulfstream IV-SP (N49RF, "Gonzo"), which flies high-altitude synoptic surveillance around hurricanes at 41,000 to 45,000 feet (12,500 to 13,700 meters), also faces eventual replacement. Any future high-altitude platform will need to carry an expanded sensor suite while keeping the G-IV's exceptional endurance and ceiling.
Unmanned Aerial Systems
Uncrewed aerial systems are the most transformative technology in hurricane observation since the dropsonde. They can work where crewed aircraft can't, fly at altitudes and for durations no human crew could sustain, and be thrown away as expendable instruments returning data from otherwise unreachable parts of a storm.
The Coyote UAS
NOAA's Coyote is a small, expendable uncrewed aircraft weighing about 13 pounds (6 kilograms) with a 5-foot (1.5 m) wingspan. It launches from a sonobuoy tube aboard the WP-3D Orion while the aircraft is flying inside the hurricane, then drops into the boundary layer, the critical zone from the ocean surface to roughly 2,000 feet (610 m), where crewed aircraft can't safely operate because of the turbulence and the nearness of the sea.
The Coyote was first demonstrated in a tropical cyclone during Hurricane Edouard in 2014.1 Since then it's flown in multiple storms, returning unprecedented in-situ measurements of wind, temperature, humidity, and pressure in the lowest levels of the vortex. Those measurements matter because the boundary layer is where the ocean feeds energy to the storm through evaporation, and where the strongest surface winds live, the winds that do the damage at landfall.
The ALTIUS-600 is being tested as a next-generation expendable UAS that may eventually complement or succeed the Coyote. Like the Coyote it deploys from crewed aircraft, but it offers more sensor capacity and endurance for sustained boundary-layer work.
NASA Global Hawk
At the far opposite end of the altitude range, NASA's Northrop Grumman RQ-4 Global Hawk can fly above 60,000 feet (18,000 meters) for more than 30 hours at a stretch. It was used heavily in NASA's Hurricane and Severe Storm Sentinel (HS3) campaign from 2012 to 2014, overflying multiple Atlantic storms and dropping sondes from the upper troposphere and lower stratosphere.
HS3, documented by Braun et al. (2016), used the Global Hawk to chase two fundamental questions: what environmental conditions decide whether a disturbance becomes a hurricane, and what controls how fast a hurricane intensifies.2 Flying far above the storm, it delivered measurements of the large-scale environment that no other platform could match for duration and coverage.
Future UAS concepts imagine fleets of autonomous drones working at several altitudes at once, small expendable systems in the boundary layer, medium-altitude platforms inside the storm, and high-altitude long-endurance systems above it, forming a three-dimensional, persistent observing network no combination of crewed aircraft could achieve.
Advanced Sensor Technology
The sensors aboard reconnaissance aircraft keep getting refined. Next-generation dropsondes carry improved humidity sensors for more accurate moisture profiles through the depth of the atmosphere. Since the water vapor available to a hurricane directly sets how much energy it can pull from the ocean, precise humidity is essential for initializing the intensity models.
Phased-array radar is a significant step up from the mechanically scanning radars on today's research aircraft. Phased-array systems can scan the storm's precipitation and wind structure volumetrically, far faster than conventional radar, catching features like eyewall mesovortices and rainband convection on time scales of seconds instead of minutes. That speed is exactly what's needed to watch the fast-moving processes behind rapid intensification.
Compact wind lidar is being developed for airborne use. Unlike radar, which infers wind from the motion of precipitation, lidar measures wind directly from the drift of aerosol particles, which lets it work in clear air, including inside the eye and in the rain-free surroundings where radar wind retrieval fails.
The Stepped Frequency Microwave Radiometer (SFMR), which estimates surface wind from the microwave emission of the wind-roughened ocean, is being refined too, with neural-network approaches to optimize its retrieval algorithm and sharpen the surface-wind estimates that set a storm's Saffir-Simpson category.
Satellite Constellation Advances
Where aircraft deliver high-resolution data in and near a storm, satellites supply the global context that track forecasting needs, and the coverage for basins where reconnaissance aircraft don't routinely fly. Several recent and planned missions are sharply improving how often, and how well, storms are observed from space.
COSMIC-2
The Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC-2), launched in June 2019, is six microsatellites that measure temperature and moisture profiles by GPS radio occultation. By tracking how GPS signals bend passing through the atmosphere, COSMIC-2 returns accurate vertical profiles even through the heavy cloud and rain that degrade other satellite techniques, and those profiles are especially good for characterizing a storm's warm core, a direct index of intensity.
TROPICS
NASA's Time-Resolved Observations of Precipitation structure and storm Intensity with a Constellation of Smallsats (TROPICS), launched in 2023, is a constellation of small satellites in low Earth orbit. Its big advantage is revisit time: it can observe a given storm roughly every hour, against the six-hour revisit of traditional polar orbiters. That cadence lets forecasters watch a hurricane's precipitation structure, warm-core intensity, and surrounding moisture evolve in near-real time from space.
SWOT and GOES-19
The Surface Water and Ocean Topography (SWOT) satellite, launched in December 2022, carries a wide-swath altimeter that measures ocean-surface topography at unprecedented resolution. For hurricanes, SWOT reveals the distribution of ocean heat content, the storm's fuel supply, by mapping the sea-surface-height anomalies tied to warm eddies and the depth of the thermocline. Reading the ocean's thermal structure beneath a storm is central to predicting whether it strengthens or weakens as it crosses different water.
GOES-U, launched in June 2024 and now designated GOES-19, carries an improved Geostationary Lightning Mapper (GLM) alongside the Advanced Baseline Imager. Research has shown that shifts in lightning inside a hurricane's inner core can flag imminent intensity change, which makes the GLM a useful real-time watch on rapid intensification from geostationary orbit.
Artificial Intelligence and Machine Learning
AI and machine learning are emerging as powerful ways to pull information out of the enormous data volumes that aircraft, satellites, and models generate. Several applications already show measurable skill gains over traditional methods.3
AI-based rapid-intensification prediction has shown improved skill at flagging which storms are likely to rapidly intensify, defined as a wind-speed jump of 35 mph (56 km/h; 30 kt) or more in 24 hours. Traditional statistical models struggle with RI because it hinges on processes interacting across scales, from large-scale patterns down to eyewall dynamics, and machine learning can find the complex, nonlinear relationships those methods miss.
ML is also being applied to satellite imagery for automated center-fixing. Pinning a hurricane's exact center from satellite imagery takes expert interpretation, and analysts don't always agree, which introduces inconsistency. Algorithms trained on thousands of images can do it with a consistency and speed that supplements the human analysts.
Deep-learning models are being built for dropsonde quality control, automatically flagging bad data points in the vertical profiles. With hundreds of dropsondes deployed in a single mission, automated QC is essential to keep only good data flowing into the assimilation pipeline.
NOAA is also exploring AI for real-time mission planning, using algorithms to pick the best flight patterns and sensor drops during a mission based on the storm's current structure and the specific forecast questions on the table, so scientists can wring the most value out of every flight hour.
Here's the opinion we'll put on record, because the hype runs ahead of the reality: AI is not magic here. You cannot machine-learn your way out of a data gap, and the hurricane boundary layer is a data gap. The algorithms are improving fast, but the binding constraint on intensity forecasting is still observations, not cleverness. That's why, of everything on this page, the boundary-layer drones may matter most.
Improved Numerical Models
The numerical models that ingest reconnaissance and produce the forecasts are themselves being transformed. The Hurricane Analysis and Forecast System (HAFS) is replacing the legacy HWRF (Hurricane Weather Research and Forecasting) and HMON (Hurricanes in a Multi-scale Ocean-coupled Non-hydrostatic) as NOAA's primary hurricane model. HAFS is a genuine shift in philosophy, embedding a high-resolution moving nest inside a global framework.
The modeling community is moving toward global convection-permitting models with grid spacings under 1.9 miles (3 kilometers). At that resolution, models can explicitly resolve the eyewall updrafts and downdrafts that drive intensity change, instead of leaning on parameterizations that only approximate them. The transition is computationally brutal, needing exascale resources, but it promises to capture the small-scale physics current operational models miss.
Coupled ocean-atmosphere models are becoming standard. They simulate the storm and the ocean together in real time, capturing the negative feedback when hurricane winds churn cold water to the surface (upwelling) and the positive feedback when a storm crosses a deep warm eddy. The ocean's thermal structure, increasingly well observed by satellites like SWOT and by underwater gliders, is a critical factor in intensity.
Track forecasts have improved roughly 50 percent since 2000,4 on the back of better global modeling, satellite data assimilation, and ensembles.5 Intensity forecasts have crept up far more slowly. The goal is reliable intensity forecasts at 5-day lead times, matching what track forecasts already deliver, and reaching it means closing the observational gaps that limit model initialization, especially in the boundary layer and in the ocean beneath the storm.
The Boundary Layer Problem
The single greatest challenge in intensity forecasting is the boundary layer problem. The hurricane boundary layer, the lowest slice of the atmosphere, from the sea surface to roughly 1,600 feet (500 meters), is the engine room of the storm. It's where the ocean hands heat and moisture to the atmosphere through evaporation, where the strongest winds occur, and where the frictional inflow spiraling toward the eyewall is concentrated. The processes in this thin layer largely decide how strong a hurricane becomes.
And yet it's the least-observed part of the storm. Crewed reconnaissance typically flies at 10,000 feet (about 3,000 meters), far above the boundary layer. Dropsondes profile it on the way down, but each gives only a single snapshot along its fall, and the layer's structure varies so much in space and time that sparse sonde sampling can't capture it.
This is exactly the gap small UAS like the Coyote are built to fill. Flying sustained missions inside the boundary layer, these drones can measure wind, temperature, and humidity continuously at the altitudes where the most important energy exchanges happen. Cione et al. (2020) documented the Coyote observing boundary-layer conditions during Hurricane Maria (2017), resolving structures and gradients that dropsondes alone could not.
I'll be direct about why this section is the honest heart of the article: the boundary layer is the part of the storm our models represent worst, and it's the part that decides intensity. That mismatch, worst-modeled where it matters most, is the central frustration of the field, and closing it is less about a breakthrough than about finally getting sustained measurements from a place that has always been too dangerous to sit in. The processes that set surface wind speed, turbulent mixing, sea-spray evaporation, roll vortices, operate at scales of meters to hundreds of meters, far below what forecast models resolve, and only real observations from that layer will let us build and validate the parameterizations that stand in for them.
Future campaigns picture multiple UAS deployed at once across different sectors of a storm, the eyewall, the eye, the outer rainbands, to capture how the boundary layer varies from place to place. Combined with ocean buoys, underwater gliders, and satellite sea-surface temperatures, that multi-platform approach could finally deliver the comprehensive lower-boundary observations hurricane forecasting has lacked for decades. The future of hurricane aviation isn't one technology; it's the convergence of many, uncrewed systems filling the gaps, AI mining the data, satellite constellations watching continuously, and coupled models simulating the whole ocean-atmosphere system, all aimed at the intensity problem that has eluded us for half a century. If it's solved, a good part of the credit will belong to a throwaway drone flying where no one can follow.
Sources
-
Cione, J. J., et al. (2020). Eye of the storm: observing hurricanes with a small unmanned aircraft system. Bulletin of the American Meteorological Society, 101(2), E186–E205. https://doi.org/10.1175/BAMS-D-19-0169.1 ↩
-
Braun, S. A., et al. (2016). NASA's Hurricane and Severe Storm Sentinel (HS3) investigation. Bulletin of the American Meteorological Society, 97(11), 2085–2102. https://doi.org/10.1175/BAMS-D-15-00186.1 ↩
-
Ko, M.-C., Chen, X., Kubat, M., & Gopalakrishnan, S. (2023). The development of a consensus machine learning model for hurricane rapid intensification forecasts with Hurricane Weather Research and Forecasting (HWRF) data. Weather and Forecasting, 38(8), 1253–1270. https://doi.org/10.1175/WAF-D-22-0217.1 — see also NOAA AOML, "Paper on using machine learning to improve rapid intensification forecasts released online in Weather and Forecasting," which notes the model "has better forecast skill than other models used by specialists at the National Hurricane Center." https://www.aoml.noaa.gov/hurricane_blog/paper-on-using-machine-learning-to-improve-rapid-intensification-forecasts-released-online-in-weather-and-forecasting/ ↩
-
NOAA National Hurricane Center. (2024). NHC Forecast Verification: Trends in Track Forecast Error. NOAA. https://www.nhc.noaa.gov/verification/verify5.shtml ↩
-
Torn, R. D., & Hakim, G. J. (2009). Ensemble data assimilation applied to RAINEX observations of Hurricane Katrina (2005). Monthly Weather Review, 137(9), 2817–2829. https://doi.org/10.1175/2009MWR2656.1 ↩