Tropical Cyclone Intensity Research Articles

A Systematic Framework for Data Augmentation for Tropical Cyclone Intensity Estimation Using Deep Learning

AbstractGiven the exponential relationship between the intensity and the expected damage of tropical cyclones, accurately estimating their intensity from satellite images is a crucially important area of research. Yet, the imbalance and limited size of available data sets significantly hinder model training and generalization capabilities, especially for state‐of‐the‐art deep learning models. Therefore, it is standard in this field to use data augmentation, a family of transformations to increase the size and variety of a given data set. However, the principles behind the usage of these techniques for the estimation of the intensity of tropical cyclones has been largely unexamined. In this paper, we introduce a framework for establishing how much augmentation to apply and which techniques to use. To determine the ideal amount of augmentation, we use a modified Gini coefficient to understand how the augmentation will affect the imbalance in the data set, and we find that there is an upper bound on how much augmentation should be done (for our data set, 50 times per image). Our results also indicate that data augmentation works best when used to reduce the amount of imbalance in a data set rather than uniformly over the entire data set, as is typically done in the tropical cyclone intensity estimation literature. We then devise a backward elimination feature selection algorithm to determine which augmentation techniques work best. Our findings suggest that all augmentation techniques are effective for the estimation of tropical cyclones intensity, including random erasing, which we were the first to implement in this context.

Case study of high waves in the South Pacific generated by Tropical Cyclone Harold in 2020

This study highlighted a high wave case by severe tropical cyclone Harold and conducted a simulation with a newly developed wave forecasting system for the South Pacific based on the Japan Meteorological Agency third generation wave model (JMA MRI-III) using the National Center for Environment Prediction Global Forecast System (GFS) winds. Harold was a very intense tropical cyclone (TC) and very high waves up to 10 m affected parts of Vanuatu and Fiji. The model results were reasonable and verified against observations of orbital satellites and a wave buoy at Komave in Fiji. The statistical verifications were carefully analysed. The Root Mean Squared Error (RSME), Scatter Index (SI), Bias and R2 are all showing very impressive results. The new wave forecasting system is the first high resolution operational model at Fiji Meteorological Service (FMS), which covers the whole Fiji area. The system will provide guidance to FMS in preparing marine alerts and warning better and more confidence in providing the marine forecast accurately.

Fidelity of global tropical cyclone activity in a new reanalysis dataset (CRA40)

AbstractThis study systematically evaluated global tropical cyclone (TC) activity in a new global atmospheric reanalysis dataset named the “40‐year Global Reanalysis” (CRA40) against the best track data. For comparison, four state‐of‐the‐art reanalyses—ERA5, JRA55, CFSR, and MERRA2—were also assessed. The results showed that there is a general underestimation of global TC genesis frequency and intensity in both CRA40 and other reanalyses. A detailed investigation of spatial distribution, seasonality, interannual variation, and long‐term trend for TC genesis frequency, as well as pressure–wind relationship for TC intensity, revealed similarities and differences among these reanalyses datasets. Overall, CRA40 does not exhibit clear advantages over other reanalyses in these aspects, but its biases are also not more pronounced. However, regarding TC translation speed, CRA40 outpeforms other reanalyses, evident by its high level of consistency with the observation in the zonal average pattern, meridional distribution at peak latitudes, and interannual variation, suggesting its reasonable capability in capturing large‐scale atmospheric characteristics. Our findings indicate that the use of CRA40 is appropriate for conducting TC‐related studies, within the scope of its limitations.

Uncertainties Inherent from Large-Scale Climate Projections in the Statistical Downscaling Projection of North Atlantic Tropical Cyclone Activity

Abstract North Atlantic tropical cyclone (TC) activity under a high-emission scenario is projected using a statistical synthetic storm model coupled with nine phase 6 of the Coupled Model Intercomparison Project (CMIP6) climate models. The ensemble projection shows that the annual frequency of TCs generated in the basin will decrease from 15.91 (1979–2014) to 12.16 (2075–2100), and TC activity will shift poleward and coastward. The mean of lifetime maximum intensity will increase from 66.50 to 75.04 kt (1 kt ≈ 0.51 m s−1). Large discrepancies in TC frequency and intensity projections are found among the nine CMIP6 climate models. The uncertainty in the projection of wind shear is the leading cause of the discrepancies in the TC climatology projection, dominating the uncertainties in the projection of thermodynamic parameters such as potential intensity and saturation deficit. The uncertainty in the projection of wind shear may be related to the different projections of horizontal gradient of vertically integrated temperature in the climate models, which can be induced by different parameterizations of physical processes including surface process, sea ice, and cloud feedback. Informed by the uncertainty analysis, a surrogate model is developed to provide the first-order estimation of TC activity in climate models based on large-scale environmental features.

On the Realism of Tropical Cyclone Intensification in Global Storm‐Resolving Climate Models

AbstractThe physical processes governing a tropical cyclone's lifecycle are largely understood, but key processes occur at scales below those resolved by global climate models. Increased resolution may help simulate realistic tropical cyclone intensification. We examined fully coupled, global storm‐resolving models run at resolutions in the range 28–2.8 km in the atmosphere and 28–5 km in the ocean. Simulated tropical cyclone activity, peak intensity, intensification rate, and horizontal wind structure are all more realistic at a resolution of ∼5 km compared with coarser resolutions. Rapid intensification, which is absent at typical climate model resolutions, is also captured, and exhibits sensitivity to how, and if, deep convection is parameterized. Additionally, the observed decrease in inner‐core horizontal size with increasing intensification rate is captured at storm‐resolving resolution. These findings highlight the importance of global storm‐resolving models for quantifying risk and understanding the role of intense tropical cyclones in the climate system.

Tropical cyclone tracking from geostationary infrared satellite images using deep learning techniques

ABSTRACT Accurate tracking of tropical cyclones (TCs) can provide regions of interest for intelligent forecasting of TC tracks and intensity. There have been few studies on algorithms for automatic TC tracking. This study proposes an effective TC tracking method based on deep learning combined with infrared satellite images. The study first constructed a TC tracking dataset based on the infrared images of the China Fengyun-2D geostationary satellite covering six different TC intensity levels between 2009 and 2012. This included 47 complete cases (video sequences) of TCs from generation to extinction. Based on deep learning, the visual tracking algorithm SiamRPN was used as the model framework. Combining Bi-GRU and TC cloud spatiotemporal evolution characteristics to improve the performance of the SiamRPN network, the SiamTCNet target-tracking model was designed to track TCs automatically. Considering that the shape and scale of TC changes with time, a TC is regarded as a typical non-rigid object with obvious timing characteristics, so the first frame of a TC video sequence is combined with the satellite images of the first three frames of the current frame as inputs to the proposed SiamTCNet model, which then extracts the evolution of the TC’s spatial structure and its bidirectional temporal change information. The experimental results show that the TC tracking of the proposed model is a significant improvement over the original SiamRPN model.

Comparative Analysis of Machine Learning Models for Tropical Cyclone Intensity Estimation

Estimating tropical cyclone (TC) intensity is crucial for disaster reduction and risk management. This study aims to estimate TC intensity using machine learning (ML) models. We utilized eight ML models to predict TC intensity, incorporating factors such as TC location, central pressure, distance to land, landfall in the next six hours, storm speed, storm direction, date, and number from the International Best Track Archive for Climate Stewardship Version 4 (IBTrACS V4). The dataset was divided into four sub-datasets based on the El Niño–Southern Oscillation (ENSO) phases (Neutral, El Niño, and La Niña). Our results highlight that central pressure has the greatest effect on TC intensity estimation, with a maximum root mean square error (RMSE) of 1.289 knots (equivalent to 0.663 m/s). Cubist and Random Forest (RF) models consistently outperformed others, with Cubist showing superior performance in both training and testing datasets. The highest bias was observed in SVM models. Temporal analysis revealed the highest mean error in January and November, and the lowest in February. Errors during the Warm phase of ENSO were notably higher, especially in the South China Sea. Central pressure was identified as the most influential factor for TC intensity estimation, with further exploration of environmental features recommended for model robustness.

Quality Control of Geostationary Lightning Mapper Observations for Tropical Cyclone Applications

Abstract The Geostationary Lightning Mapper (GLM) has been providing unprecedented observations of total lightning since becoming operational in 2017. The potential for GLM observations to be used for forecasting and analyzing tropical cyclone (TC) structure and intensity has been complicated by inconsistencies in the GLM data from a number of artifacts. The algorithm that processes raw GLM data has improved with time; however, the need for a consistent long-term dataset has motivated the development of quality control (QC) techniques to help remove clear artifacts such as blooming events, spurious false lightning, ‘bar’ effects, and sun glint. Simple QC methods are applied that include scaled maximum energy thresholds and minima in the variance of lightning group area and group energy. QC and anomaly detection methods based on machine learning (ML) are also explored. Each QC method is successfully able to remove artifacts in the GLM observations while maintaining the fidelity of the GLM observations within TCs. As the GLM processing algorithm has improved with time, the amount of QC flagged lightning within 100 km of Atlantic TCs is reduced, from 70% during 2017, to 10% in 2018, to 2% during 2021. These QC methods are relevant to the design of ML-based forecasting techniques which could pick up on artifacts rather than the signal of interest in TCs if QC wasn’t applied beforehand.

Effect of Tropical Cyclone Intensity on the Relationship Between Hydrometeor Distribution and Rapid Intensification by GPM GMI

AbstractThis study analyzes hydrometeor evolution during rapid intensification (RI) and tropical cyclone (TC) intensity dependence using satellite data. Previous studies have suggested ice cloud water or non‐convective precipitation as a predictor of RI from different perspectives. However, few studies have focused on the impact of TC intensity or comprehensive comparisons to identify better indicators. During RI, hydrometeor contents in weak TCs increase over the entire region, whereas they increase mainly in the inner‐core region and decrease in advance in the outer‐core region for strong TCs. Hydrometeor contents in the inner‐core are higher in RI than in slow intensification, and their maxima location is related to TC intensity and intensification rate. Cloud water path (CWP) in the inner‐core region is most correlated with the intensification rate, especially in weak TCs. Therefore, the CWP can serve as a predictor of RI and can be applied to all TC intensities.

The role of time-varying external factors in the intensification of tropical cyclones

Abstract. The role of time-varying external parameters in tropical cyclone (TC) dynamics is explored through a low-order conceptual box model. Specifically, we look at stable-to-stable state transitions which may be linked to tropical cyclone intensification, dissipation, or eyewall replacement cycles (ERCs). To this end, we identify two parameters of interest: the exponent of radial decline and sea-surface temperature. We examine how variations in these parameters affect the stable states of the model and consider the behaviour of the system under different time-dependent forcing profiles for the parameters. By externally forcing the exponent of radial decline and sea-surface temperature, we show the existence of rate-dependent behaviour in the model. These findings are brought together in a case study of Hurricane Irma (2017). The results highlight the role of the radial vorticity gradient in behaviour such as rate-induced tipping and overshoot recovery. They also show that a simple model can be used to explore relatively complex tropical cyclone dynamics.

Precipitation variations associated with tropical cyclone intensity and intensity change in the Northwest Pacific using 21-Yr GPM IMERG product

ABSTRACT Precipitation variations associated with tropical cyclone (TC) intensity and intensity changes in the Northwest Pacific are examined using the Integrated Multi-satellite Retrievals for Global Precipitation Measurement product spanning from June 2000 to December 2020. The initial analysis focuses on the precipitation properties of TCs in a Neutral stage (TC-N). Results indicate a linear increase in both the precipitation area and the maximum rain rate with rising TC intensity, accompanied by a reduced radius of the maximum rain rate. Increased precipitation probability and intense rainfall occur in the down-shear quadrants relative to vertical wind shear. Compared to TC-N, TCs undergoing intensification display larger precipitation areas, a more symmetric spatial distribution, and an increased mean rain rate in the inner core area. The increase in precipitation area and the mean rain rate in the inner core area correlate with the intensification rate. TCs undergoing weakening exhibit a more asymmetric spatial distribution of precipitation. The increase in the asymmetry index is positively correlated with the weakening rate. Findings suggest that the precipitation area and the mean rain rate in the inner core area could serve as potential predictors for forecasting TC intensification rates. Additionally, an asymmetric spatial distribution of precipitation emerges as a preferred condition for predicting TC weakening.

The atmospheric effect of aerosols on future tropical cyclone frequency and precipitation in the Energy Exascale Earth System Model

This study uses experiments from the Energy Exascale Earth System Model (E3SM) to compare the influence on tropical cyclone (TC) activity of: (i) the atmospheric effect of aerosols under specified sea-surface temperatures (SSTs); and (ii) the net effect of greenhouse gases (GhGs) (including changes in SSTs). The experiments were performed using the CMIP6 Shared Socioeconomic Pathway SSP5-8.5 emissions scenario with GhG-induced SST warming specified and atmospheric aerosol effects simulated but without explicit ocean coupling. Insignificant changes in global TC frequency are found in response to the atmospheric effect of future aerosols and GhGs, as significant regional responses in TC frequency counteract each other. Future GhGs contribute to more frequent TCs in the North Atlantic, and reductions over the Northwestern Pacific and Southern Indian Ocean. The atmospheric effect of future aerosols drives more frequent TCs over the Northwestern Pacific and reductions over the Northeast Pacific and North Atlantic. Along with increases in TC intensity, global TC precipitation (TCP) is projected to increase by 52.8% (14.1%/K) due to the combined effect of future aerosols and GhGs. Although both forcings contribute to TCP increases (14.7–19.3% from reduced aerosols alone and 28.1–33.3% from increased GhGs alone), they lead to different responses in the spatial structure of TCP. TCP increases preferentially in the inner-core due to increased GhGs, whereas TCP decreases in the inner-core and increases in the outer-bands in response to the atmospheric effects of decreased aerosols. These changes are distinct from those caused by aerosol-induced SST changes, which have been considered in other studies.

Tropical Cyclone Intensity Estimation by Feature Extraction Techniques using Satellite Imagery

Tropical Cyclone Intensity Estimation by Feature Extraction Techniques using Satellite Imagery

Observational fine-scale evolutionary characteristics of concentric eyewall Typhoon Doksuri (2023)

The variations in the track and intensity of a tropical cyclone (TC) are closely correlated with the fine-scale evolution of its structure. The fine-scale track, intensity, and structural evolution of Typhoon Doksuri (2023) can be comprehensively analyzed by combining multi-source observations. The results herein show that Doksuri (2023) experienced secondary eyewall formation (SEF), concentric eyewall maintenance (CEM), and eyewall replacement cycle (ERC) processes when entering the South China Sea and prior to landfall. These processes can be further delineated into three subsequent stages. In the first stage, the SEF phase, the secondary (outer) eyewall formed, exhibiting features that were non-concentric with the inner eyewall. Concurrently, the track of Doksuri (2023) displayed notable oscillations in both its forward translational direction and speed, accompanied by the emergence of two radial maxima centers of wind speed. Subsequently, during the second stage, the CEM phase, the geometric centers of the inner and outer eyewalls of Doksuri (2023) coincided, initiating a rapid intensification process characterized by an accelerated forward translational speed. Both the inner and outer eyewalls further contracted during this phase. In the third stage, the ERC phase, the asymmetry of the inner eyewall increased, and the outer eyewall gradually contracted while the inner eyewall dissipated until the replacement was completed prior to landfall. Accordingly, Doksuri (2023) experienced rapid weakening. These findings have the potential to enhance our understanding of the physical mechanisms governing the intricate structures of TCs at fine scales, bolstering the forecast accuracy of TC tracks and intensities.

Rapid intensification of tropical cyclones in the Gulf of Mexico is more likely during marine heatwaves

Tropical cyclones can rapidly intensify under favorable oceanic and atmospheric conditions. This phenomenon is complex and difficult to predict, making it a serious challenge for coastal communities. A key contributing factor to the intensification process is the presence of prolonged high sea surface temperatures, also known as marine heatwaves. However, the extent to which marine heatwaves contribute to the potential of rapid intensification events is not yet fully explored. Here, we conduct a probabilistic analysis to assess how the likelihood of rapid intensification changes during marine heatwaves in the Gulf of Mexico and northwestern Caribbean Sea. Approximately 70% of hurricanes that formed between 1950 and 2022 were influenced by marine heatwaves. Notably, rapid intensification is, on average, 50% more likely during marine heatwaves. As marine heatwaves are on the increase due to climate change, our findings indicate that more frequent rapid intensification events can be expected in the warming climate.

Impacts of free tropospheric turbulence parametrisation on a sheared tropical cyclone

AbstractThe turbulent transport of momentum, heat, and moisture can impact tropical cyclone intensity. However, representing subgrid‐scale turbulence accurately in numerical weather prediction models is challenging due to a lack of observational data. To address this issue, a case study of Hurricane Maria was conducted to analyse the influence of different free tropospheric turbulence parametrisations on sheared tropical cyclones. The study used the current Met Office Unified Model (MetUM) parametrisation, as well as a parametrisation scheme with significantly reduced free tropospheric mixing length. Convection‐permitting ensemble simulations were performed for both mixing schemes at two initialisation times (four 18‐member ensembles in total), revealing an improvement in the intensity forecasts of Hurricane Maria when the mixing length was decreased in the free troposphere. By implementing this change, the less diffuse simulations presented a drier mid‐level. The resolved downward transport of drier air from the mid‐levels into the inflow layer (so‐called “downdraft ventilation”) was thus more effective in reducing the storm's intensity. In contrast to earlier studies, where decreasing the diffusivity in the boundary layer intensified the storm, we show that decreasing the free tropospheric diffusivity can weaken the storm by enhancing shear‐related weakening processes. While this study was performed using the MetUM, the findings highlight the general importance of considering turbulence parametrisation, and show that changes in diffusivity can have different impacts on storm intensity depending on the environment and where the changes are applied.

Unveiling the Dominant Factors Controlling the Long‐Term Changes in Northwest Pacific Tropical Cyclone Intensification Rates

AbstractTropical cyclones (TCs), especially intense TCs, pose serious threats to life and property particularly in the affected coastal regions. Understanding the factors determining the TC intensification rate (IR) remains a great challenge. This study identifies the dominant factors responsible for the observed significant increase in TC IR over the western North Pacific in recent decades using the energetically based dynamical system model of TC intensification. It is found that the environmental dynamical efficiency mainly contributed by vertical wind shear and upper‐level divergence is responsible for the long‐term changes in TC IR during the strong TC stage, but it played a secondary role in the long‐term changes in IR during the weak TC stage. The latter were primarily contributed by the maximum potential intensity, which is primarily determined by sea surface temperature. Results also strongly suggest that global warming is the primary driver of the long‐term changes in TC IR.

Effects of tropical cyclone intensity on spatial footprints of storm surges: an idealized numerical experiment

Abstract Storm surges caused by tropical cyclones (TCs) have long ranked first among all types of marine disasters in casualties and economic losses, and can lead to further regional exacerbation of consequences stemming from these losses along different coastlines. Understanding the spatial footprints of storm surges is thus highly important for developing effective risk management and protection plans. To this end, we designed an ideal storm surge model based on FVCOM to explore the relationship between TC intensity and the spatial footprint of storm surges, and its intrinsic mechanism. The spatial footprints of both positive and negative storm surges were positively correlated with TC intensity; however, the latter was more sensitive to the intensity when the TC intensity is weaker than CAT3 TC’s. The average positive storm surge footprint of CAT1 was 574 km, with CAT3 and CAT5 increasing by 6% and 25%, respectively, compared to CAT1. The average spatial footprint of the negative storm surge of CAT1 was 1,407 km, with CAT3 and CAT5 increasing by 18% and 29%, respectively, compared to CAT1. The decomposition and mechanism analysis of the storm surge show that the main contributing component of the total surge at the south end of the storm’s landfall and during the time of the forerunner was the Ekman surge, whereas the contribution of the normal surge component to the north and during the time of the main surge and resurgence was dominant. In addition, not all the spatial footprints of the storm surge components increased with the TC intensity, as the total surge did, similar to the Ekman surge. These quantitative analyses and intrinsic mechanisms provide a theoretical basis for predicting and evaluating storm surge risks.

Thermodynamic Pathways of Vertical Wind Shear Impacting Tropical Cyclone Intensity Change: Energetics and Lagrangian Analysis

Abstract This study analyzes the variations in the thermodynamic cycle and energy of a tropical cyclone (TC) under the influence of vertical wind shear (VWS), exploring the possible thermodynamic pathways through which VWS affects TC intensity. The maximum energy harnessed by the TC diminishes alongside a decrease in storm intensity in the presence of VWS. In the sheared TC, the ascending branch of the thermodynamic cycles of TC shifts toward lower entropy, which is related to the reduction in entropy in the eyewall and/or the increase in entropy and enhanced upward motion outside the eyewall. Moreover, the descending leg shifts toward higher entropy due to the increase in entropy and weakening of downward motion in both the ambient environment and the upper troposphere. These changes in the ascending and descending branches could reduce the work done by the heat engine cycle, with the former playing a primary role in the presence of VWS. Given that the ascending branch is influenced by the eyewall and the rainbands outside the eyewall under VWS, the thermodynamic pathways could be categorized into inner ventilation and outer ventilation based on the location of their roles. The pathways associated with inner ventilation primarily reduce the entropy in the eyewall. In addition to the conventional low- and midlevel ventilation, the inner ventilation also encompasses new pathways entering the midlevel eyewall after descending from the upper level and ascending from the boundary layer. Conversely, the pathways of outer ventilation are related to the increase in the entropy outside the eyewall. These include the ascent of high-entropy air to the middle and upper troposphere related to the inner and outer rainbands, the outward advection of high-entropy air from the eyewall in the midlevel and upper level, and air warming by the descending draft from the upper- to midlevel troposphere. These insights contribute to a nuanced understanding of the sophisticated interactions within TCs and their response to VWS. Significance Statement Vertical wind shear (VWS) is known to modify the thermodynamic structure of tropical cyclones (TCs), thereby affecting their development. The incorporation of multiple thermodynamic mechanisms amplifies the complexity of related studies and predictions. Within the context of a unified heat engine framework, this study aims to paint a relatively comprehensive picture of the key thermodynamic pathways through which VWS impacts the intensity of TCs. Contrary to previous studies, we emphasize the increase in entropy and the intensified upward motion outside the eyewall, which play pivotal roles in diminishing the intensity of TCs in the presence of VWS. Gaining an understanding of these critical mechanisms and structural changes offers insights and guidance that can enhance the prediction of sheared TCs.

The Role of Turbulence in an Intense Tropical Cyclone: Momentum Diffusion, Eddy Viscosities, and Mixing Lengths

Abstract Improved representation of turbulent processes in numerical models of tropical cyclones (TCs) is expected to improve intensity forecasts. To this end, the authors use a large-eddy simulation (with 31-m horizontal grid spacing) of an idealized category 5 TC to understand the role of turbulent processes in the inner core of TCs and their role on the mean intensity. Azimuthally and temporally averaged budgets of the momentum fields show that TC turbulence acts to weaken the maximum tangential velocity, diminish the strength of radial inflow into the eye, and suppress the magnitude of the mean eyewall updraft. Turbulent flux divergences in both the vertical and radial directions are shown to influence the TC mean wind field, with the vertical being dominant in most of the inflowing boundary layer and the eyewall (analogous to traditional atmospheric boundary layer flows), while the radial becomes important only in the eyewall. The validity of the downgradient eddy viscosity hypothesis is largely confirmed for mean velocity fields, except in narrow regions which generally correspond to weak gradients of the mean fields, as well as a narrow region in the eye. This study also provides guidance for values of effective eddy viscosities and vertical mixing length in the most turbulent regions of intense TCs, which have rarely been measured observationally. A generalized formulation of effective eddy viscosity (including the Reynolds normal stresses) is presented. Significance Statement This study uses a turbulence-resolving simulation of a category 5 tropical cyclone to understand the role of turbulence in intense storms. Results show that turbulence clearly modulates storm structure and intensity. This study provides guidance for the values of turbulent quantities (which are usually parameterized in comparatively coarse operational TC forecast models) in scarcely observed regions of intense storms. Furthermore, a complete formulation of the effective eddy viscosities is proposed, incorporating contributions from typically ignored Reynolds normal stress terms.