The present section discusses the detailed discussions about sorted 51 articles (table-2). The selected articles were categorised into 7 categories. i.e. (i) Modeling of genesis, track and storm surge of TCs’ (ii) Improvements in prediction of cyclone intensity (iii) Techniques utilizing Artificial Neural Network (ANN) (iv) Impact of Data assimilation on TC track and intensity prediction (v) TC rainfall and structure (vi) Impact of planetary boundary layer/convection/dynamics scheme on TC track and Intensity prediction and (vii) Tropical cyclone disaster management respectively.
5.1 Modeling of genesis, track and storm surge of TCs:
Kuroda et al. (2010) numerically simulated the extremely severe cyclonic storm Nargis (2008) over Myanmar oceanic region, using two numerical models (i.e. Japan Meteorological agency’s Nonhydrostatic model (NHM) and Princeton Ocean model). The study concluded that the downscaling of the NHM with a horizontal resolution of ~ 10 km produced more realistic features of the TC in comparison to operational global analysis and global spectral model forecasts. The storm surge could not be reproduced in Princeton Ocean Model (horizontal resolution ~ 3.5 km) experiment using global spectral model forecast data, but the simulation of the ocean model using NHM forecast data predicted a rise in sea level of over 3 metres. In a similar type of study by Singh et al. (2021), the authors found that India’s National Centre for Medium Range Weather Forecasting operational global model NCUM (horizontal resolution ~ 12 km) was highly skilful in predicting the track and intensity of an extremely severe cyclonic storm ‘Fani’ (2019) over the Bay of Bengal. However, the study by Singh et al. (2021) did not analyse the storm surge associated with the TC. The efficacy of NCUM global and regional models was further proved in simulation of a Super Cyclone (SuCS) 'Amphan' over the same ocean basin (i.e. Bay of Bengal; Ashrit et al. 2021). The performance of a non-hydrostatic limited-area atmospheric model “Consortium for Small-scale Modelling (COSMO)” was evaluated and found commendable in detection of early genesis of a very severe cyclonic storm ‘Ockhi’ (2017) over Arabian Sea (Subrahamanyam et al. (2019)). In line with the study by Ashrit et al. (2021), Chen et al. (2022) also evaluated the skills of operational models of ‘China’ for super typhoon ‘Lekima’(2019) over east-China sea region. The main difference in these two studies was that while Chen et al. (2022) evaluated the skill of operational probabilistic (ensemble) model along with deterministic models, Ashrit et al. (2021) evaluated the skill of global and regional models only (i.e. deterministic models) only. Further, the operational forecasts of various leading meteorological agencies were also compared by Chen et al. (2022). The authors further concluded that all the operational deterministic models were efficient enough for predicting landfall point of super typhoon ‘Lekima’, as reported by Ashrit et al. (2021) for Super Cyclone (SuCS) 'Amphan' using NCMRWF’s operational model. The nested finite volume solver (FV-3) based Hurricane Analysis and Forecast System (HAFS) of National Oceanic and atmospheric administration (NOAA) is a hurricane modelling system that is in evolution phase at present. Utilizing global-nested and regional versions of HAFS modeling suite, the real-time simulations were conducted in 2019 creating first baseline for HAFS progressions. Hazelton et al. (2021) showed that despite the fact that the ability to predict rapid intensification was slightly inferior to the operational model ability, the global-nested configuration of the modeling suite was more skilful in predicting track, intensity and structure of hurricanes formed during July-October, 2019.
The Advanced Research version of the WRF (Weather Research and Forecasting) model was used to investigate the rapid intensification of a category-5 hurricane Katrina (2005) before it made landfall in the southern United States (Liu et al., 2017). The findings of the study suggested that the accurate hurricane intensity prediction requires better representation of vertical mixing and surface fluxes in the planetary boundary layer. Zhu et al.(2019) found an improvement in prediction skill of WRF-ARW model for two typhoon rainfall cases (Typhoon ‘Meranti’ (2016) and Typhoon ‘Fitow’ (2013)) over western north pacific ocean, when the “Kain-Fritsch” convection parametrization scheme of the model was optimized using a micro-genetic algorithm. Here, it is worth discussing that in place of convection parametrization scheme, Liu et al., (2017) had compared two popular planetary boundary layer (PBL) schemes (i.e. the Mellor-Yamada-Janjic (MYJ) and the Yonsei University (YSU) schemes) of the WRF-ARW model. As a result, the MYJ scheme performed better, simulating a more accurate track and intensity evolution, as well as a more accurate vortex structure and landfall time and location of the hurricane ‘Katrina’(2005).
In recent times, though significant progress have been made in TC track prediction across all the world basins, the prediction of Intensity still haven’t improved much in numerical models. The reason behind this sluggish improvement in TC intensity prediction is that numerical models are unable to resolve ‘complex internal dynamics’ occurring in the core of TCs (Rhome, 2006, Lian Shou, 2012). In this regard, the task of improving intensity forecasts in numerical models mandates the researchers to look into plausible causes on which intensity of TCs depend. In the tropical cyclone secondary circulation, the weaker inertial stability in the outflow layer is found to reduce the energy sink, resulting in a faster intensification to maximum potential intensity of TC (Rappin et al. (2011)). Convective bursts (CBs) occurring inside the upper-level radius of maximum wind, have the potential to influence hurricane intensification by causing compensatory subsidence of lower-stratospheric air (Miller et al. (2015). The decrement in tilt of cyclonic-vortex and ‘symmetric convection’ leads to intensification of weak tropical cyclones (Berrios, 2020).Shieh et al. (2013) proposed that features of upper tropospheric troughs must be considered while forecasting intensity of TCs. Short and Petch (2018) delineated superiority of the convection permitting regional unified model (UM: horizontal resolution ~ 4.4 km) over its global counterpart for prediction of intensities of strong TCs formed in the western north pacific ocean. The TC ‘Dora’ was formed over southwest Indian ocean in the year 2007. The causes of TC’s rapid intensification under its interaction with upper-level trough forcing was investigated by Leroux et al. (2013) utilizing an limited area operational NWP model. The NWP model with 8-kilometer horizontal resolution was capable of reproducing the main characteristics of the eyewall replacement cycle observed for the TC. Potential Vorticity (PV) superposition in conjunction with secondary eyewall formation induced by eddy momentum flux convergence and trough’s vertical velocity forcings were identified as the main mechanisms of the TC vortex intensification. Further, it was found that for stronger cyclonic vortices, intensification of the TC was greater under upper level forcing Leroux et al. (2016).
A thorough examination of high-resolution ensemble predictions of different intensity change (computed from operational ‘HWRF’ and ‘Hurricanes in a Multi-scale Ocean-coupled Non-hydrostatic (HMON) models’) thresholds was carried out by Torn et al. (2021) for the select cyclones formed over Atlantic and eastern pacific oceans during 2017–2019. After the application of a quantile-based bias correction scheme, both the HWRF and HMON ensemble based models provided the skilful forecasting of intensity changes for cases without rapid intensification. The intercomparison of track and intensity forecasts of TCs over North Atlantic Ocean (NAO) basin, which formed in the year 2019, revealed that the newly operationalized regional model ‘Real-Time Hurricane Analysis and Forecast System (HAFS) Stand-Alone Regional (SAR)’outperformed the three operational models i.e. ‘Global Forecast System (GFS)’, ‘Hurricane Weather Research and Forecasting model (HWRF)’, and ‘Hurricanes in a Multi-scale Ocean-coupled Non-hydrostatic model (HMON)’ in track prediction of TCs (Dong et al. (2020). However the HFAS-SAR model was less skilful in intensity prediction of the TCS as compared to HWRF and HMON models. The model showed a little improvement in intensity prediction over the GFS model. The improved convection scheme (i.e. the Finite-Volume cubed sphere (FV3)) may be attributed for the better TC track predicting skill of the HFAS-SAR model.
5.2 Improvements in prediction of cyclone intensity:
In the present times, researchers are trying to develop easy to use tools for intensity forecast guidance in real time. For the western North Pacific basin (excluding the South China Sea), a model post-processing tool called “TC intensity guidance on rapid intensification (TINT-RI)” has been developed recently (Tam, 2021). Utilizing robust logistic regression technique and the naïve Bayes classifier, the tool is able to give rapid intensification forecasts of TCs well in advance (Upto 48 hr).Similarly, a tool, incorporating probabilistic methodology, was developed for rapid intensification prediction of cyclones forming over the oceans which come under JTWC’s area of observation and responsibility (Knaff et al. (2020)). Vertical wind shear's influence on tropical cyclone intensity change is well documented (DeMaria (1996), Wang (1996), Wong and Chan (2004)). A higher vertical wind shear inhibits the progress and further intensification of cyclones. The warming of the midlevel atmosphere suppresses the convective activity and prevents the development of storms (DeMaria (1996). Hazelton et al. (2020) incorporated NCEP’s FV3GFS to study the rapid intensification of hurricane Michael (2018) in a sheared environment and demonstrated that vortex tilt reduction results in upshear humidification and is, therefore, a driving factor for intensification. The uncertainty in the physical representation of the air-sea fluxes (i.e. momentum and enthalpy fluxes) at high winds in storms puts a constraint in predictability of intensity of TCs (Jaimes et al., (2015)). In the proof-of-concept by Jaimes et al.(2015), Nystrom et al.(2021) investigated the feasibility of estimating model parameters governing air–sea energy transfer processes using ensemble-based data assimilation.
Satellites have become one of the most essential platforms for collecting information about the cyclone and its surrounding environment while monitoring and forecasting. Many traditional forecasting algorithms have struggled to identify variations in very low contrast cloud characteristics associated with cyclones, as well as the integration of data received from numerous sensors and the detection of spectral and spatial patterns (Lee and Liu, 2004, Villmann et al. (2003)). Although there are techniques that offer similar capabilities (e.g., numerical and statistical-dynamical), these traditional procedures are difficult, need high-end machines to run, and are subjected to erroneous initial conditions (Jeffries et al., 1993, 1995). Furthermore, remotely sensed data is frequently noisy and typically quite vast in size. (Hoque et al., 2017, Elsberry 2014, Heming et al., 2019). The data's intricacy, as well as the computational demands of existing forecasting methodologies, has prompted studies and sparked researches into new approaches for processing remotely sensed pictures to anticipate cyclones.
5.3 Techniques utilizing Artificial Neural Network (ANN):
The use of ANNs for data processing is one of these new approaches. Adaptivity, Robustness and power/speed are the characteristics of ANN-based cyclone forecasting algorithms. ANN techniques, being a paramount tool for evaluating satellite data for predicting cyclone track and intensity forecasts because of the advantages of the input from satellite pictures which are generally available and mostly free; is simple to implement on conventional desktop computers. So, these methods are computationally inexpensive and although it takes a long time to make the network adapt using all of the satellite pictures, it only takes a few minutes or fractions of a minute to produce cyclone track and intensity projections after the network is trained. Neural network-based algorithms (Wei et al., 2010) can produce forecasts with a reasonable level of accuracy.
For intensity prediction of TCs, Cloud et al. (2019) developed a feed forward neural network (FFNN) which was significantly skilful in the prediction of rapid intensification of TCs in comparison to operational observation-adjusted HWRF model and probabilistic Statistical Hurricane Intensity Prediction System. Similarly, a multilayer perceptron (MLP) TC intensity prediction model was found more skilful in predicting the rapid intensification of TCs as compared to the four operational statistical-dynamical TC intensity prediction models (Xu et al. (2021)). The MLP model was based on deep-learning. Utilizing a decision-tree-based machine learning algorithm “XGBoost ”, an increased TC intensity forecast skills have been achieved for ECMWF ensemble prediction system model outputs (Chan et al., 2021). Despite their usefulness in intensity prediction of TCs, the machine learning technique provides short lead-time predictions of the cyclones, due to which its utility is yet to be adopted by disaster managers for real-time operations (Chen et al., 2020). Based on evolutionary programming, Schaffer et al. (2020) developed a statistical-dynamical forecast model for intensity prediction of TCs.
5.4 Impact of Data assimilation on TC track and intensity prediction:
Strong TCs have the highest initial errors and uncertainties in terms of TC intensity, since data assimilation (DA) technique struggles to mimic the convective-scale and mesoscale processes that are key to intense TC forecasting. Position (and track) forecast errors are significantly higher for initially weak TCs than for initially strong TCs on average, especially in the short range. Forecasts have a long-standing systematic inaccuracy caused by a slow propagation speed. There are signs that the main source of inaccuracy occurs during propagation into the mid-latitudes, which could have an impact on extra-tropical transitions. Forecasts of westward-moving TCs, particularly weak TCs, frequently drift northward abnormally. Generally, a non-linear forecast error growth is associated with rapidly intensifying TCs (Minamide et al. (2020)). Both, the real observations and axis-symmetric synthetic data assimilation in NWP models are complementary to each other. Their simultaneous ingestion results in a more realistic representation of initial cyclonic vortex and a stronger TC with a longer rapid intensification period (Chang (2015)).
Utilizing the Ensemble Kalman Filter data assimilation technique, the assimilation of the geostationary Multifunctional Transport Satellite (MTSAT) obtained atmospheric motion vectors (AMVs) in WRF model improved the track, intensity, and structure analyses of Typhoon ‘Sinlaku’ (2008) over the pacific ocean (Wu et al. (2014)). On a similar note, the assimilation of these atmospheric vectors obtained from geostationary satellite ‘Himawari-8’ enhanced the efficacy of operational HWRF model in predicting track and intensity of three TCs namely ‘Nepartak’ (2016), ‘Meranti’(2016) and ‘Megi’(2016), all the three formed over western-north pacific ocean (Sawada et al., (2019)). Large improvements in track, minimum sea level pressure and maximum wind speed (vmax) have been found by assimilating Cyclone Global Navigation Satellite System (CYGNSS) derived winds in the regional HWRF model (Mueller et al. (2021)).
In terms of magnitude of wind, the hurricane ‘Patricia (2015)’ was the strongest tropical cyclone of category-5 which formed over eastern pacific ocean. Utilizing a very high resolution WRF model, the hurricane’s intensity evolution was further analysed by Fox and Judt (2018). The authors attributed to ‘vortex initialization’, ‘high resolution (~ 1km)’ and ‘parametrization of dissipative heating’ for successful prediction of intensity of hurricane ‘Patricia’. When Doppler radar ‘radial velocity’ observations were assimilated in WRF-EnKF (Ensemble Kalman Filter) data assimilation system, the intensity forecast of the hurricane Patricia (2015) was significantly improved in comparison to assimilation of only conventional observations in the NWP suite (Nystrom et al. (2019)). Adopting same type of methodology, in ensemble-variational (EnVar) data assimilation technique of HWRF model, the assimilation of ‘upper-level wind observations’ derived from dropsondes have been found improving the numerical analysis and prediction of Hurricane Patricia (2015) during its rapid intensification phase (Feng et al., 2019). Further, In a relatively more innovative approach, Lu et al. (2019) did vortex modification and produced more-realistic three dimensional analyses of hurricane Patricia (2015) by assimilating enhanced atmospheric motion vectors along with inner core observations obtained from different field campaigns. The authors employed the GSI-based, continuously cycled, dual-resolution hybrid ensemble–variational data assimilation (DA) system for in the HWRF Model for the vortex modification. In cycling EnKF DA scheme, the assimilation of all-sky infrared radiances obtained from the Advanced Baseline Imager on GOES-16 satellite potentially improved the intensity forecast of TC Harvey (2017; Minamide (2020)) .Taylor et al. (2021) performed an observing system simulation experiment(OSSE) on a West Pacific tropical cyclone to see if geostationary satellite-based Precipitation Radar reflectivity observations can be used for numerical weather forecasting of TCs. The OSSE proved the utility of the assimilating the radar reflectivity observations in the regional cloud-resolving ‘SCALE-RM model version 5.0.0’ by reducing the intensity errors.
5.5 TC rainfall and structure:
The landfalling TCs are generally associated with various structure and intensity changes i.e. spiral rain band, mesoscale vortices and remote rain bands etc. (Shou et al. (2012)).During the landfall process, weakening or sudden strengthening of a TC over or near the coast can significantly change rainfall characteristics associated with cyclonic storm (Ray et al. (2022). Four predictors associated with TC rainfall and structure features were incorporated in Japan Meteorological Agency version of Statistical Hurricane Intensity Prediction Scheme (SHIPS), which improved the intensity forecasts (both central pressure (Pmin) and maximum wind speed (Vmax)) of TCs (Shimada et al. (2018). Similarly, incorporating several new predictors in SHIPS-RII (rapid intensification index) model the intensity prediction of TCs have been found improved operationally, for both the Atlantic and eastern North Pacific basins (Kaplan et al. (2015).The India Meteorological department has operationally implemented a state-of-the-art dynamical-statistical cyclone prediction model (Kotal et al., (2021)) which has been proved highly efficient in prediction of genesis, structure, track, intensity and the rainfall associated with TCs forming over North Indian Ocean (NIO) region (i.e. Bay of Bengal and Arabian Sea). Ren et al. (2018) developed an objective ‘TC track similarity area index (TSAI)’ for precipitation forecasting of landfalling TCs over south China by employing a combined ‘dynamical–statistical’ approach. Based on this index, a ‘landfalling TC precipitation dynamical–statistical ensemble forecast (LTP_DSEF) model’ was developed which was found more skilful in comparison to three operational models (i.e. ECMWF, GFS, and T639/China) for higher threshold values of intense precipitation associated with TCs.
5.6 Impact of planetary boundary layer/convection/dynamics scheme on TC track and Intensity prediction:
The momentum and enthalpy fluxes from the earth’s surface affect its lower atmosphere. The interaction of these fluxes with atmosphere is determined by parametrization of planetary boundary layer (PBL) in NWP models (Stensurd et al. (2009)). The ratio of momentum to heat and moisture exchange in the PBL is governed by a number known as ‘Prandtl number (Pr)’. Kalina et al. (2021) suggested that the value of ‘Prandtl number (Pr)’ significantly affects the intensification and enthalpy fluxes of the rapidly intensifying TC. So, consideration of the ‘Pr’ number, while changing the planetary boundary scheme of Global Forecast System–Eddy Diffusivity Mass Flux (GFS-EDMF), may improve the skill of HWRF model. In the low ‘vertical eddy diffusivity’ (Km) forecasts of HWRF model, lowering Km in the planetary boundary layer increases both inflow and convergence in a TC, resulting in a stronger and more symmetric deep convection (Zhang et al. (2019). Similarly, Singh et al. (2017) conducted numerical simulations to evaluate the impact of five PBL and six cumulus convection schemes of WRF-ARW model on track and intensity predictions of seven landfalling TCs formed over Bay of Bengal. The model was the most skilful in TC track and intensity prediction with the Yonsei University (YSU) PBL and the old simplified Arakawa-Schubert cumulus parametrization scheme. The two microphysical parametrization schemes (i.e. the older and newer versions) of the ‘Coupled Ocean–Atmosphere Mesoscale Prediction System–Tropical Cyclone (COAMPS-TC) model’ were compared for prediction of 15 Atlantic cyclonic storms (Jin et al. (2014)). The newer version of microphysical parametrization used a hybrid approach of double moment in ‘cloud’ ‘ice’ and ‘rain’ in contrast to single-moment scheme for the three hydrometeor species employed in the older scheme. The newer scheme significantly reduced rightward cross-track errors and positive intensity bias in control forecasts. For 2005′s category-5 Hurricane ‘Rita’ Histova et al. (2021) showed that different microphysical parameter combinations in the cloud-permitting community Weather Research and Forecasting model (WRF) produced significantly different microwave signatures for wind and precipitation observations. A detailed analysis of thermodynamic variables of the vortex core of TC Vicente (2012) using a column integrated moist static energy budget revealed that before rapid intensification, the core of the TC was sufficiently humid (Chen et al. (2019).
5.7 Tropical cyclone disaster management:
Hazards associated with tropical cyclone (TC) landfall pose increasing risks as coastal populations grow, yet accurate forecasts of the timing, location, and impacts of TC landfall remain an ongoing challenge. In recent times, the use of ‘remote sensing’ and ‘spatial analysis’ techniques has significantly increased to manage the on-ground impacts of these disasters with rapid advances in a wide range of data availability and processing techniques. Although TC frequency has stayed approximately constant over recent decades, there are growing evidences that the proportion of TCs that became major hurricanes, has increased significantly. Out of 51 studies selected for the detailed review, the two studies by Hoque et al. (Part 1 and 2, 2017) discuss the possible management strategies associated with mitigating the challenges in predicting the track and intensity of cyclones. Both the studies delineates the application of remote sensing and spatial analysis in the context of response, recovery, prevention/reduction, and preparedness. The substantial coastal mitigation and adaptation strategies would be required in the near future due to the global incidence of rapid intensification of TCs due to anthropogenic reasons.