Novel Methods for Exploring the Limits of Tropical Cyclone Forecasting
Open Access
- Author:
- Kowaleski, Alexander Michael
- Graduate Program:
- Meteorology
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- August 03, 2017
- Committee Members:
- Jenni Evans, Dissertation Advisor/Co-Advisor
Jenni Evans, Committee Chair/Co-Chair
George Spencer Young, Committee Member
Fuqing Zhang, Committee Member
Francesca Chiaromonte, Outside Member - Keywords:
- tropical cyclones
potential intensity
ensemble forecasts
clustering - Abstract:
- Novel methods are employed to explore the limits of tropical cyclone (TC) intensity and to improve TC track and structure forecasting. Observations from 88 hurricane passages are used to construct radial thermodynamic storm profiles. These profiles are compared to the idealized boundary layer in Potential Intensity (PI) theory. Data from 85 passages are used to calculate ocean-air enthalpy fluxes. Temperature decreases with decreasing radius, while moist entropy begins to increase with decreasing radius beyond the radius of maximum winds, especially for major hurricanes. Total enthalpy fluxes calculated using composite observed conditions differ substantially from fluxes calculated using PI theory idealizations, though the sign of the difference depends on the flux calculation method used. Modifying the (PI) calculation to account for energy production beyond the eyewall produces higher maximum intensities. The radial shape of the wind profile modulates the maximum intensity increase. For outer limits of energy production at 2-2.5 times the radius of maximum winds, PI can increase by more than 10 ms-1 over its standard value. These results provide a potential explanation for why individual TCs can exceed their potential intensity. Regression mixture-model clustering of TC forecast tracks from 30 120-hour forecasts of the ECMWF Integrated Forecast System ensemble produces interpretable partitions by direction and speed of motion across initialization times. Clustering synthesizes the forecast spread within the ensemble into a small number of representative trajectories. Clustering of track and Cyclone Phase Space (CPS) forecasts of Hurricane Sandy from four ensemble prediction systems also produces meaningful track and CPS partitions. Rand Index and Adjusted Rand Index calculations demonstrate a substantial relationship between track and CPS cluster membership. Storm-centered composites show that both track and CPS clustering yield substantial variations in structural evolution among clusters. Ensemble forecasts from the ECWMF and GEFS ensemble prediction systems are used to initialize an ensemble of regional simulations. Tracks and CPS evolutions from these simulations are then clustered using regression mixture models. Among this ensemble, there is a substantial relationship between track and CPS evolution of Sandy. This is due to extratropical transition (ET) timing and lower-tropospheric thickness asymmetry during ET. In the four most populous clusters Sandy develops a warm seclusion structure just prior to landfall; however clustering reveals subtle differences in the stage of storm-trough interaction at landfall.