This research identifies large-scale synoptic controls that are relevant for rapid intensification (RI) in the Atlantic basin. Spatial statistical analysis techniques were performed on NASA MERRA data from 1979–2009. Rotated principal component analysis (RPCA) was performed, looking for common patterns in the datasets. The RPC’s were grouped using hierarchical clustering techniques, allowing for finding events similar in synoptic structure. The clustered events, representing the total RI and non-RI composites, were averaged yielding composite maps for different scenarios. To verify the results, a permutation test was done to show which variables are good distinguishers of RI and non-RI cases. These variables were used as input in two prediction schemes: logistic regression and support vector machine classification. The prediction scheme was a slight improvement in forecasting RI when using the synoptic variables mid-level vorticity, vertical velocity, low-level potential temperature and specific humidity, as the most significant in predicting RI.

An analysis to determine physical and spatial patterns of the surface latent heat flux (LHF) and near surface (5m) salinity (NSS) beneath tropical cyclones (TCs) in the North Atlantic and eastern North Pacific basins during the first 24 hours of rapid intensification (RI) was conducted using empirical orthogonal function (EOF) analysis. To determine if these patterns were unique to RI, TC RI cases were compared to three non-RI intensification thresholds, 10 kt, 15 kt and 20 kt, for both LHF and NSS. Though similarities exist between non-RI and RI cases physical and spatial patterns unique to the RI cases did exist. Sea surface temperatures associated with statistically identified TC groups were assessed for their potential influence on RI. While inconclusive in the eastern North Pacific, NSS in the Atlantic may play a role for RI TCs in areas affected by river discharge from South America.

Forecasting rapid intensification (RI) of tropical cyclones (TCs) is considered one of the most challenging problems for the TC operational and research communities and remains a top priority for the National Hurricane Center. Upon landfall, these systems can have detrimental impacts to life and property. To support continued improvement of TC RI forecasts, this study investigated large-scale TC environments undergoing RI in the North Atlantic basin, specifically identifying important diagnostic variables in three-dimensional space. These results were subsequently used in the development of prognostic machine learning algorithms designed to predict RI 24 hours prior to occurrence. Using three RI definitions, this study evaluated base-state and derived meteorological parameters through S-mode and T-mode rotated principal component analysis, hierarchical compositing analysis, and hypothesis testing. Additionally, nine blended intelligence ensembles were developed using three RI definitions trained on data from the Statistical Hurricane Intensity Prediction Scheme- Rapid Intensification Index, Global Ensemble Forecast System Reforecast, and Final Operational Global Analysis. Performance metrics for the intelligence ensembles were compared against traditional linear methods. Additionally, a tenth ensemble was created using forecast data generated from Weather Research and Forecasting model simulations of TC RI events in the open North Atlantic and compared against linear methods. Results revealed modest classification ability of machine learning algorithms in predicting the onset of RI 24 hours in advance by including TC environmental spatial information of temperature and moisture variables, as well as variables indicative of ambient environmental interactions.

The climatology of precipitation asymmetries in Tropical Cyclones (TCs) and their relationship to TC intensity changes using 16 years of data from the Tropical Rainfall Measuring Mission (TRMM) satellite. TC Inner core precipitation asymmetries were quantified using the Fourier wavenumber decomposition method upon the pixel level data of 3,542 TRMM TMI overpasses. Composites of wavenumber–1 and wavenumber 1–6 total precipitation asymmetries were constructed to show the distribution pattern under different storm motion speed, vertical wind shear and the combined effects of varying vertical wind shear and storm motion. Results indicate that motion–relative total precipitation asymmetry is located down–motion. The phase of motion–relative maximum asymmetry shifts cyclonically by adding the wavenumber–2–6 asymmetry to wavenumber–1. Shear is more dominant than motion on the distribution of precipitation asymmetry. The analysis of combined effects of motion and shear shows when shear is weak, and shear is to the left of motion, the precipitation asymmetry is explained more by storm motion. The main contributor to the general asymmetry pattern is from the moderate and heavy precipitation. The wavenumber 2–6 energy localizes the maximum heavy precipitation asymmetry. The quantified wavenumber 1–6 asymmetries is also applied to differentiate between different intensity change categories and the asymmetry evolution of a rapidly intensifying storm. The precipitation asymmetry properties of rapid intensification (RI) and non–RI storms are examined. The dataset of 2,186 global tropical storms through category 2 hurricanes is divided by future 24–h intensity change and includes storms with at least moderately favorable environmental conditions. The normalized wavenumber 1–6 asymmetries, indicates quantitatively that the lower asymmetry of precipitation is most strongly correlated with future intensity change. The precipitation field of non–RI storms are more asymmetric than RI storms. The 595 sampled overpasses are classified into 14 categories in the timeline of an RI event from 48 hours before RI until RI ends. The decrease of normalized wavenumber 1–6 asymmetries in the inner core region of all four types of precipitation several hours before RI onset was quantitatively demonstrated to be critical for TC RI.

<div>A tropical cyclone (TC) vortex is an immense, coherent, organized-convective system. Beneath this large-scale organization, is a litany of azimuthally asymmetric convective motions that exist on a spectrum of scales. These asymmetries are especially dominant during periods when the vortex undergoes critical transitions in its intensity and structure. However, the precise nature of influence of the organization of asymmetries on TC intensity change remains an enigma. The inherent difficulty in predicting their behavior is because asymmetries may arise due to different external or intrinsic sources and occur at different spatial and temporal scales while several complex mechanisms act near-simultaneously to dictate their evolution in time. As a result, multiple pathways are possible for a TC vortex that is influenced by these asymmetries. Our preliminary investigations using numerical models made it apparent that there wasn't a single, unifying way to address this problem. In this thesis, I outline multiple novel techniques of diagnosing and predicting which of the many pathways are likely for a TC vortex that is influenced by azimuthal asymmetries. </div><div> </div><div> First, using three-dimensional numerical simulations of a pair of sheared and non-sheared vortices, I demonstrate the diagnostic potential of the juxtaposition in the azimuthal phasing of: </div><div>(i) the asymmetrically distributed vertical eddy flux of moist-entropy across the top of the boundary layer, and the radial eddy flux of moist-entropy within the boundary layer; and (ii) eddy relative vorticity, eddy moist-entropy, and vertical velocity throughout the depth of the vortex. </div><div> </div><div> Second, I introduce an energetics-based diagnostic framework that computes the energy transactions occurring at asymmetries across various length-scales in the wavenumber domain. By applying it to select cases, this thesis uncovers the relative importance of all the energy pathways that support or disrupt the growth of asymmetries within the vortex. Contrary to the traditional explanations of convective aggregation/disaggregation and axi/asymmetrization through barotropic mean-eddy transactions, my thesis reveals that the growth or disruption of asymmetries are predominantly due to (i) the baroclinic conversion from available potential to kinetic energy at individual scales of asymmetries and (ii) the transactions of kinetic energy across asymmetries of different length scales. </div><div> </div><div>Finally, this thesis introduces two further diagnostic frameworks targeted at tackling the problem of real-time forecasting of TC rapid intensity changes. The first is an empirical framework which examines symmetric and asymmetric convection and other state variables within the vortex, and in the environment across a suite of TCs and identifies a set of `important' variables that are significantly different during time periods that precede a rapid intensification as opposed to a rapid weakening. My framework then ranks the variables identified based on how significantly they influence a rapid intensity change in a TC and the amplification factor of any associated variability. We recommend that future observational, and consequent TC modeling and data assimilation efforts prioritize the highest ranked variables identified here. </div><div><br></div><div>The second is a stochastic model wherein a scale-specific stochastic term is added to the equations describing the energy transactions within the TC vortex. By simulating a stochastic forcing that may arise from any scale, I compute the probability of the vortex transitioning into a rapidly intensifying or a rapidly weakening configuration across an ensemble of scenarios. </div><div><br></div><div>In summary, this thesis introduces and applies a variety of diagnostic techniques that help determine the impact of azimuthal asymmetries on TC intensity evolution.</div>

The climatology of overshooting convection in tropical cyclones (TCs) is examined using Tropical Rainfall Measuring Mission (TRMM) Precipitation Radar (PR). The percentage of TC convective systems with overshooting convection is highest over the North Indian Ocean basin, while the northwest Pacific basin contains the highest population of both TC convective systems and convection with overshooting tops. Convective systems in the inner core region are more capable of penetrating 14 km and the associated overshooting convection are featured with much stronger overshooting properties compared with those in the inner rainband and outer rainband regions. In the inner core region of TCs, convection associated with precipitating systems of higher intensity and intensification rates has a larger probability of containing overshooting tops. To identify the relative importance of shallow/moderate versus deep/very deep convection in the rapid intensification (RI) of TCs, four types of precipitation-convection are defined based on the 20 dBZ radar echo height (Z20dBZ). Distributions of four types of precipitation-convection, and their contributions to total volumetric rain and total latent heating are quantified. It is shown that RI is closely associated with increased and widespread shallow precipitation around the storm center, while moderately deep and very deep convection (or overshooting convection) does not increase until in the middle of RI. This is further confirmed by the study of rainfall and convection evolution with respect to the timeline of RI events. Statistically, the onset of RI follows a significant increase in the areal coverage of rainfall, shallow precipitation, and cyan of 37 GHz color composites upshear-left, which in turn could be used as potential parameters to forecast RI. Very deep convection is most frequent 12-24 hours before RI onset and concentrates upshear-left, but it quickly decreases in the following 24 hours. The percent occurrence of very deep convection is less than 1% for RI storms. The tilt of vortex is large prior to, and near the RI onset, but rapidly decreases in the middle of RI, suggesting that the vertical alignment is a result instead of a trigger of RI.