University of Miami
Assistant Professor of Atmospheric Sciences
My research focuses on tropical cyclone structure and intensity change, particularly the inner-core processes that govern rapid intensification. The work draws on airborne Doppler radar, geostationary satellite observations, and machine learning, with an aim of developing methods that transfer to an operational setting.
About
My appointment at the University of Miami spans two units: I am an Assistant Professor in the Department of Atmospheric Sciences at the Rosenstiel School, and core faculty at the Frost Institute for Data Science and Computing. The joint appointment reflects the structure of the work, which draws on tropical meteorology and on statistical and machine-learning methodology in roughly equal measure. Much of it develops from collaboration with NOAA’s Hurricane Research Division and the airborne radar observations that underpin it.
Operational track forecasts have improved substantially over recent decades; forecasts of intensity have improved far less. My work addresses that disparity. Using observations, numerical simulations, and machine learning, I examine how convection, vortex tilt, and the surrounding environment interact to determine whether a storm undergoes rapid intensification.
A second theme concerns observational coverage. Aircraft reconnaissance samples only a small fraction of global tropical cyclones, almost entirely in the North Atlantic. Recent work asks how much of the storm state can be recovered from satellite-only inputs — models trained on aircraft observations but requiring none at inference — in order to extend structural analysis to basins where reconnaissance is unavailable.
Research
The questions are interdependent: results on vortex tilt inform predictor selection in the statistical models, which in turn supply the data products described below.
Track forecast skill has improved substantially; intensity forecast skill has not. How much of the residual error is intrinsic — set by chaotic, convective-scale variability in the inner core — and how much remains recoverable through improved observations and model formulation?
A vertically misaligned vortex can align by more than one route, and the pathway taken appears to condition the subsequent intensity evolution. Identifying these pathways in observations, and determining what selects among them, is the subject of several recent papers.
Between the synoptic forecast range and the seasonal outlook lies the subseasonal range, where modes such as the MJO and BSISO modulate the timing and location of genesis. The spatial and temporal limits of that predictability remain to be established.
Aircraft sample only a small fraction of storms. TC-SWARM evaluates whether continuously available satellite and reanalysis fields can reconstruct the three-dimensional wind field at latencies compatible with operational use rather than retrospective analysis.
Projects & platforms
Each originated as a research requirement and is now maintained as a public resource: an interactive platform, a modeling framework, and an open observational dataset.
Interactive access to real-time and historical tropical cyclone observations
TC-ATLAS integrates satellite, radar, aircraft, and environmental observations for tropical cyclones within a single interactive platform. The real-time component tracks active storms globally using infrared imagery, passive microwave overlays, scatterometer surface winds, aircraft reconnaissance, and numerical model guidance. A global archive spans the historical best-track record, and the TC-RADAR airborne Doppler database is accessible case by case.
The platform is intended to serve operational situational awareness, research access to the underlying observations, and instructional use, exposing the same data at each level.
Hurricane structure from satellite imagery alone
Aircraft reconnaissance provides the most reliable constraint on inner-core structure, but samples only a minority of storms and, for those, only intermittently. TC-SWARM — the Tropical Cyclone Satellite Wind Analysis and Reconstruction Model — is a hierarchical machine-learning framework that learns the mapping from continuously available geostationary infrared imagery and reanalysis environmental fields to the three-dimensional wind fields measured by airborne Doppler radar.
Its first stage is GHOST, the Geostationary-based Hurricane Objective Strength Technique, which estimates a storm’s scalar state — maximum sustained wind, minimum pressure, and radius of maximum wind — from satellite and environmental data. Both are trained on reconnaissance observations but require none at inference: reconnaissance-informed training, reconnaissance-independent application. The objective is to extend structural analysis to storms that reconnaissance does not sample.
The GHOST manuscript is in preparation for Weather and Forecasting, with the TC-SWARM wind-field reconstruction to follow.
Trained on TC-RADAR airborne Doppler analyses · applied to satellite and reanalysis inputs
Storm-centered airborne Doppler radar analyses, 1997–2025
The Tropical Cyclone Radar Archive of Doppler Analyses with Recentering compiles tail Doppler radar observations from NOAA P-3 reconnaissance flights into a uniform, storm-centered database. Version 4.0 extends the record through the 2025 season: 1,640 individual swath analyses and 474 merged analyses spanning the North Atlantic and eastern and central North Pacific.
Every case carries three-dimensional winds, reflectivity, and derived kinematic fields on a common grid, with multiple center-fix conventions including tilt-relative recentering. The dataset underlies much of the observational work described above and provides the training labels for TC-SWARM.
Publications
Selected from 20 peer-reviewed articles, with manuscripts submitted and in preparation, and invited chapters. The complete list is on Google Scholar and in the CV.
Media
Video features, television interviews, and press coverage of recent research, and commentary on storms as they develop.
1:30
A short feature on using machine learning and three decades of aircraft observations to fill gaps in what we know about hurricane structure and intensity.
Watch on YouTube
Interview on how Hurricane Hunter radar reveals, about a day ahead, that a tilted storm is about to align and strengthen.
Experience & recognition
A summary of appointments, sponsored research, and awards. Full detail is in the curriculum vitae.
Group & teaching
Graduate and undergraduate researchers working on tropical cyclone structure, airborne radar, and machine learning, and the courses I teach at the University of Miami.
Contact
I welcome inquiries from students interested in tropical cyclones, remote sensing, or applied machine learning, and from colleagues working with TC-RADAR, TC-ATLAS, or the satellite-based reconstruction models. A brief note describing your interests and background is a useful starting point.
Send an email