A warm ocean with a low-pressure system is the niche for the formation of TCs. However, the thermal gradient of the NIO consisting of BoB and ARB gives rise to many furious TCs. Several studies have been done to understand the dynamics and thermodynamics of the ocean involved in the TCs’ formation (Webster et al., 2005). In addition, there are studies done in the past to understand the impact of TC on the BoB region. The findings from the studies confirm the rise in the NIO SST (Tiwari et al., 2021). Mishra et al. (2020) reported a strong role of ARB warming on increasing weather extremes. In addition to this, a study by Mohanty et al. (2012) confirms the role of SST in increasing the number of SCS in the ARB. Also, there are recent studies done so far to understand the ocean’s role in TC formation and its activities. For example, Prakash and Pant (2017) studied the dominant role of mixed layer heat budget in TC genesis. The study’s findings confirmed the reduction in the magnitude and diurnal periodicity of the net surface heat fluxes interconnected with the cloud cover during the cyclone. On the other hand, Patwardhan and Bhalme (2001) and Jadhav and Munot (2009) identified a significant decrease in the trend of TC frequency seasonally despite increasing SST in recent decades. Therefore, to simulate the accurate TC obtaining a realistic SST is a big challenge to researchers. However, there are studies conducted to do sensitivity experiments of models to get realistic SST. SST obtained through satellite corresponds to TC (Mandal and Mohanty, 2010; Mohanty et al., 2019). While few contrasting studies (Demaria and Kalpana, 1994) reveal that SST has a relatively less impact on the intensification and propagation of TC after the formation. Bhatla et al. (2020) mentioned that global warming had increased the upper troposphere temperature and also the SST. This rise in SST could be a reason for intense TCs (Chauhan et al., 2021). Another study by Frank and Young (2007) also confirmed the role of SST in the formation and increased intensity of TC.

A 28-years (1990–2017) mean SST in the pre-monsoon (AMJ) and post-monsoon (OND) seasons over the NIO has been shown in Figures 5A,B using the Hadley Centre Global Sea Ice and Sea Surface Temperature (HadISST) data. The pre-monsoon season is associated with boreal summer, and thus the SST in this season was relatively high (Figure 5A). Southeast ARB was warmer than other regions, and most of the intense ARB TCs formed over this region. Due to the Indian Summer Monsoon Rainfall season from June to September, the SST during the post-monsoon becomes low but sufficient to conceive TCs (Figure 5B). Central to south-east BoB was comparatively warm, and most of the intense BoB TCs developed over this region.

Other oceanic subsurface parameters play a vital role in modifying the TCs genesis and other characteristics. Tropical Cyclone Heat Potential (TCHP), currents, and eddies over the warm oceans help in the TC’s rapid intensification (RI). Moreover, a trace of eddies formation near TC tracks could assess the RI phase (Jangir et al., 2020). Lin et al. (2013) also mentioned the role of TCHP and eddy currents in the TC’s RI. A spatial demonstration of TCHP over the BoB during the life of VSCS Titli is shown in Figure 6. A cyclonic system like VSCS Titli that encounter the TCHP and eddy currents are more disastrous during the landfall and further inland progression. Mawren and Reason (2017) and Patnaik et al. (2014) also mentioned the contribution of eddies in the TC intensification over the NIO. In addition, there are sparse studies carried out using the oceanic numerical models to understand and forecast the TC. For example, Das et al. (2014) used Princeton Ocean Model (POM) to study the thermodynamics of oceans over BoB.

TC is always a concern for meteorologists because it results from synoptic and mesoscale interactions accompanied by other natural phenomena like strong winds, heavy rainfall, and storm surge. Likewise, ocean components other essential factors of the weather system that contribute to the formation of TCs are Coriolis force, vorticity, low vertical wind shear, moisture and moist convection. Moreover, in tropical regions, moist convection dominates the process of transporting mass, energy, and momentum through the atmosphere (Kumar et al., 2017). Albert et al. (2022) reported a 0.58 correlation between power dissipation index (PDI) and SCS category TCs from 1979 to 2019. They also found that relative humidity at 600 hPa, positive relative vorticity at 850 hPa, and reduced outgoing longwave radiation at 500 hPa have significantly contributed to the NIO’s increased TC frequency. Balaji et al. (2018) used the accumulated cyclone energy (ACE) metric to analyze TC activities over the NIO from 1981 to 2014 and observed an increasing trend in recent years. SST, UOHC, atmospheric water vapour, and the genesis potential index (GPI) were all strongly associated with the growth and variability of ACE. According to Duan et al. (2021), the lower and middle troposphere’s relative humidity and vertical wind shear are the two key contributors to TC genesis. The combination of these two variables drives the bimodal seasonal cycle of TCs’ in both the ARB and the BoB. Tiwari et al. (2021) investigated the characteristics of post-monsoon season TCs’ over the BoB from 1979 to 2018 using various metrices such as ACE and PDI. They also used correlation and principal component analysis on the VSCS, SCS, and CS category TCs, taking into account SST, vertical wind shear, MSW, minimum sea level pressure (MSLP), relative vorticity, and specific rainwater content, among other parameters and revealed that relative vorticity is the most influential among all the controlling elements that impact the final intensity of the BoB TCs’. A study by Chan (2005) explained the importance of the baroclinic process in the TC. However, the energy and heat transfer through convection modifies the temperature gradient and vertical wind shear (Osuri et al., 2010). Again, the representation of dynamical changes in the NWP model is a big challenge. However, this has been further improved with satellite products that help to develop the better initial vortex position and structure of TC in the NWP models.

Further to confirm the data dependencies for position and structure of TC, Courtney et al. (2019) used different best-track datasets from IMD and Joint Typhoon Warning Center (JTWC) to analyze the intensity of VSCS Hudhud. They found different results with both the dataset as underestimation in one and overestimation in other due to different criteria of sources for MSW. Mohanty et al. (2020) portray the impact of dry air intrusion on the ESCS Fani caused by asymmetric wind. This dry air significantly impacts the vortex initialization that affects the structure, landfall, and TC intensity. However, few studies focused on improving the vortex initialization by using better satellite and reanalysis data (Nadimpalli, 2020a; Nadimpalli, 2020b). In addition to SST, atmospheric high temperature and diabatic heating also plays a dominant role in the evolution of the intensity and determining the track of TC (Singh and Bhaskaran, 2020). Henceforth, many studies confirmed the direct correlation between seasonal changes and TC formation. Therefore, the NIO favours the formation of more intense BoB TCs in the post-monsoon season and less frequent but intense ARB TCs in the pre-monsoon season (Nadimpalli et al., 2021).

The land is a sink for TCs, and they get energetic when crossing the lands (Pattanayak and Mohanty, 2010). Advancements in data assimilation techniques have improved the forecast skills of the NWP models in wide ranges during pre-and post-monsoon seasons. Also, land surface features have helped to enhance the mesoscale features such as drying, precipitation, and deep convection. It further helps in better representation of Land Surface Models (LSM), Land use and Land Cover (LULC), soil moisture and other parameters (Osuri et al., 2017). Raju et al. (2011) progressively showed the importance of land data assimilation in improving the TC’s intensity prediction through better representation of boundary layer flux exchange. Rajesh et al. (2017) confirmed the same conclusion of the role of the land surface in mesoscale moist convection. Scholarly articles indicate the inland role in a TC, whereas the “Brown Ocean” concept was also introduced, indicating the role of wetland in mimicking the ocean to fuel moisture to the TC intensification over the many regions across the world (Andersen and Shepherd, 2017). Nair et al. (2019) reported the role of LULC change, soil moisture, heat flux in the intensification of TC over land. Mohanty et al. (2001) mentioned in the Indian Ocean Experiment (INDOEX) that deep offshore plume-like structure resulting from diurnal variability and topography heterogeneity plays a vital role in modulating the local and large-scale circulation patterns. Chang et al. (2009) have signified the role of soil condition before the storm in predicting the landfalling storm. Kishtawal et al. (2012) analyzed the role of change in soil bulk density and change in land features over the decay of the TCs and post-landfall intensity changes. However, there are very few studies to evaluate the role of land in TC’s intensification and weakening. Also, the forecast of accurate landfall and post-landfall intensification is a question of debate. Henceforth, many studies suggested the use of improved land force parameters for better prediction of TCs (Mohanty et al., 2001; Corsaro and Toumi, 2017; Rajesh et al., 2017).

In numerical models of the earth’s climate system and mesoscale weather events, the atmosphere, ocean, land, and cryosphere, among other elements, are mathematically represented. One of the best examples of such models is NWP models, which have been the mainstay of operational weather prediction for the last 2 decades or so. As new modelling algorithms, parameterization schemes, and faster computing resources become available, NWP is a complex and specialized field that is constantly evolving. The NWP models cover both hydrostatic and non-hydrostatic assumptions. The conservation of momentum, conservation of mass, conservation of energy, the first law of thermodynamics, equation of state, and the relationship among pressure, temperature, and density are all addressed in the fundamental governing equations of NWP models, and some of them are shown below: d V d t + f k × V =   − ∇ Ø +   σ P s ∂ Ø ∂ σ   ∇ P s + F (1) ∇ . ( P s   V ) +   ∂ ∂ σ   ( P s   σ ) +   ∂ P s ∂ t = 0 (2) ∂ Ø ∂ σ = − R T σ (3)

Eq. 1, 2, and 3 are the horizontal momentum equation, continuity equation, and hydrostatic equation, respectively. Here, V is the horizontal wind velocity, Ps is the surface pressure, T is the temperature, and σ ,   Ø are the diagnostic fields.

Thus, a set of partial differential equations that explain the dynamic and thermodynamic processes in the earth’s atmosphere are used to generate the prediction or forecast. The equation set and model performance are highly influenced by the horizontal and vertical grid structure, initial and boundary conditions, and model domain.

There are various numerical models with different resolutions for predicting and projecting TCs worldwide, including NWP models, regional climate models (RCMs), and global climate models (GCMs) (Table 2). However, due to inaccurate vortex initialization of TCs, incomplete representation of complex physical processes, error in parameterization, and coarse resolution of the models, there are biases in the predictions of TC intensity, genesis and landfall that are still a challenge for forecasters (Chen et al., 2020). Osuri et al. (2012) performed numerical experiments to improve the TC intensity prediction by assimilating satellite-derived wind data. Further research improves uncertainty in the NWP system in predicting heavy rainfall associated with TC through Doppler Weather Radar (DWR) data assimilation (Mohanty et al., 2014). An example of TC simulation using the WRF model is provided in Figure 7. Simulated MSLP (Figure 7A) and surface wind at 10-m height (Figure 7B) during a VSCS Titli (2018) over the BoB was captured well by the model that shows the model’s potential to predict the TC’s over the region.

The prediction skill of the NWP model is highly reliable on the exactness of the initial state of the atmosphere. The conventional measurements like Sonde, Pilot, Profiler, Airep, Buoy, Ship, etc., have benefited the weather predictions, including the tropical belt but to a lesser extent (Guerbette et al., 2016). However, space-borne sensors provide continuous data at a high spatiotemporal resolution over sparse areas, mainly oceans. TCs are one of the deadliest natural hazards over the tropical regions to predict in terms of location and intensity to mitigate the devastation (Kumar and Shukla, 2019). The forecasting of severe weather events highly depends on moisture distribution and pre-convection environment transportation (Sieglaff et al., 2009). Though various conventional and satellite observations are routinely assimilated in the NWP model to produce a precise estimate of the initial model state (Montmerle et al., 2007), the use of satellite radiances perform an important role in the present operational data assimilation system (Kelly and Thépaut, 2007; Zupanski, 2013).

Due to a notable diversity in the observations, the Earth system modelling and its evaluation using satellite data have become challenging. Previous studies suggested that assimilation of clear-sky or cloud-cleared radiance assimilation from Microwave (MW) and Infrared (IR) sensors have improved the temperature and moisture analysis (Madhulatha et al., 2018). Zou et al. (2013) assimilated the Advanced Technology Microwave Sounder (ATMS) radiances in the HWRF model to analyze its influence on track and intensity forecasts. They suggested a consistently positive impact on model predictions. The accuracy in predicting severe weather events by numerical models can be further improved by incorporating precipitation and cloud affected radiances from MW sensors (Madhulatha et al., 2018). In practice, it is difficult to separate the effect of cloud and precipitation in the temperature and moisture observations, therefore the assimilation of such data in the NWP models is still a significant challenge (Kumar and Shukla, 2019).

In the assimilation of clear-sky radiances, a large amount of information is discarded due to the non-linearity of the system, spatial and temporal discontinuity of the clouds and precipitation, and limitations in the model dynamics. Also, most of the data assimilation techniques forcefully transform the nonlinear processes into the linearized form (Ohring and Bauer, 2011; Kumar and Shukla, 2019). To avoid these difficulties to some extent, many agencies use the clear-sky IR/MW radiance in operational forecasts (Madhulatha et al., 2018). As a result, the clear sky radiance from different MW sensors has shown a positive impact on the NWP model forecasts.

Ice and precipitation-affected radiances have a massive potential to improve accuracy (Ohring and Bauer, 2011; Zupanski, 2013; Kumar and Shukla, 2019). Modification in observation thinning and quality control allowed Zhu et al. (2016) to assimilate all-sky satellite radiances in the Grid-point Statistical Interpolation (GSI) analysis system and evaluate the outputs over the clear-sky approach. The all-sky assimilation system provided more realistic brightness temperature (TB) results and cloud water analysis increments. Yang et al. (2016) analyzed the impact of clear-sky and all-sky AMSR2 radiances in predicting Hurricane Sandy. All-sky AMSR2 assimilation experiment showed the improved forecast of MSLP, cloud distribution, and warm-core structure compared to clear-sky radiance because more precipitation/cloud-affected data was assimilated through all-sky radiance around hurricane core areas. Assimilation of MW radiometers measured radiances in cloudy and rainy areas has shown recent progress in the humidity analysis (Yang et al., 2016; Zhu et al., 2016). Previously, Guerbette et al. (2016) showed the sensitivity of the Sondeur Atmosphérique du Profil d’Humidité Intertropicale par Radiométrie (SAPHIR) sounder within cloud systems to solid precipitating hydrometeors. Singh et al. (2013) compared the NWP model simulated radiances from SAPHIR using input profiles retrieved from Atmospheric Infra-Red Sounder (AIRS) data and radiosonde profiles. Singh et al. (2013) found good agreement between SAPHIR and Microwave Humidity Sounder (MHS) TB. A representation of SAPHIR TB during the cyclone Ockhi (2017) over the NIO is also demonstrated in Figure 8.

It is noticed that atmosphere, ocean, and land all together have a role in the TC genesis through the exchange of fluxes that ultimately affect TC intensification and track changes. Moreover, it is confirmed by Mohanty et al. (2001) that land-air-sea interaction has a significant impact on the regional circulation pattern. There is a developed strong feedback teleconnection, interaction and feedback between the regional and global scale features. Recent research mentioned the importance of the air-sea interface for intensity prediction. Following the same, there are many studies over the influence of air-sea interaction on TC physical activity and its simulation. It is well established that air-sea moisture difference and surface entropy flux increase the intensity and hence destructiveness. Single standalone models used for the TC forecasts are more biased and inaccurate due to the unrealistic feedback in the model. Such static SST is provided in the NWP models, with a high probability of biases in the results (Bender and Ginis, 2000; Bender et al., 2007). Henceforth, coupled models emerged as a helping tool to the modelling community in improving the predictability skills of the model. Many studies have been done so far to test the predicting skills of coupled ocean-atmosphere models over NIO and other oceans. For example, the sensitivity experiment of the HWRF-POM/HYCOM coupled ocean-atmosphere model was conducted to demonstrate the predicting skills of these widely used regional models (Mohanty et al., 2019). Another study using WRF-ROMS demonstrated the benefits of a coupled atmosphere-ocean model for obtaining realistic simulations of atmospheric and oceanic parameters under the extreme weather conditions associated with a very intense cyclone such as TC Phailin (Prakash and Pant, 2017). There are more studies observed using coupling between ocean-atmosphere but three ways coupled models are significantly less explored. However, few studies demonstrated the usefulness of the inclusion of realistic land surface features has the potential to improve the predicting skills of models.

Studies show that using a statistical approach is a challenge. The statistical models and methods have limitations in TC prediction beyond 24 h over a heterogeneous environment like NIO (Mohanty and Gupta, 1997; Gupta, 2006). However, these methods are not far behind in contributing to the improvement in TC forecast. Statistical approaches are used to correct the data, fill the gaps for data, bias correction, downscaling, and many more ways to explore the correlation between environmental drivers and TC activity. A study was conducted by Lee et al. (2020) using the statistical downscaling approach to investigate the TC’s frequency, and the result indicates that future TCs will be more devastating (more intense). Data used in NWP models is always a hindrance in the TC forecast. Statistical approaches help improve the quality of input data for initial and boundary conditions and post-processing.

Orthodox statistical models often fail to address the nonlinear and complex relationship between TC predictors (Demaria et al., 2005; Lee et al., 2015; Wang et al., 2015). To solve such challenges, the concept of machine learning is being used to explore the observational datasets to improve the TCs forecast skills (Chen et al., 2020). It also helps to improve the uncertainties in the NWP models by enhancing the pre-processing, i.e., refining the initial condition of a model through data assimilation techniques. Artificial intelligence, a part of machine learning, is a new development to the current century. According to their applications, we can divide the machine learning algorithms into three categories: feature selection, clustering, and regression (Chen et al., 2020). Richman et al. (2017) applied the support vector regression (SVR) method to an initial predictor pool to reduce the TC seasonal prediction errors. Alemany et al. (2019) proposed a fully connected recurrent neural network to predict the TCs trajectory. Their method successfully predicted the hurricane’s track up to 120 h with reduced error.

Once the TC approaches the land, the automatic weather station (AWS) and conventional observations such as sonde, pilot, profiler, airep, buoy, ship, and doppler weather radar data are relevant for TC prediction. Because conventional observations over the ocean are unavailable, satellite data are being used and have shown to be significant for studying and understanding TC characteristics (Jaiswal et al., 2017). Microwave scatterometers have been effective in studying cyclogenesis in its early phases (Sharp et al., 2002). The use of a sea-winds scatterometer has been valuable in understanding tropical disturbances (Sharp et al., 2002; Li et al., 2003). Jaiswal and Kishtawal (2011) and Jaiswal et al. (2013) applied the scatterometer derived surface wind. They used a wind pattern matching based technique that relied on the availability of a cyclonic disturbance during the satellite overpass.

Apart from conventional observations, the reanalysis products from the European Centre for Medium-Range Weather Forecasts (ERA-I, ERA-40, and ERA-5, and others), National Centers for Environmental Prediction—National Center for Atmospheric Research, Modern-Era Retrospective Analysis for Research and Applications, etc., are widely used in many kinds of literature to study the TCs, such as Evan et al. (2006); Balaji et al. (2018); Duan et al. (2021); Tiwari et al. (2021); Ranji et al. (2022); etc.

The numerical modelling approach is demarcated as best suited for operational forecasting; however, the coupled models are best to perform the cyclone related research (Table 3).

The simulation of cyclone Phailin induced storm surge and inundation was carried out by means of ADvanced CIRCulation (ADCIRC) model by Kumar et al. (2015). Phailin made landfall on 12 October 2013 at Odisha (east coast of India). Within a few days of landfall, the model-simulated inundation extent corresponded closely with field surveys at Ganjam, Odisha. Furthermore, the model nicely simulated the temporal evolution of the surge residual based on measurements from a tide gauge in Paradip. However, the model slightly underestimated the magnitude compared to observations, which might be attributed to the model’s paucity of wave setup and uncertainty in wind and pressure parameters. More or less a similar kind of study was performed by Murty et al. (2014) by implementing a coupled wave + surge hydrodynamic modeling system to simulate storm surge caused by cyclone Phailin. The coupled model setup provides a realistic representation of the dynamic interaction of currents, waves, tides, and wind, which is important for operational forecasts.

Saxby et al. (2021) compared the performance of a regional convection-permitting atmosphere-ocean coupled model, i.e., the Met Office Unified Model atmosphere-only configuration and three-dimensional dynamical ocean model for the simulations of BoB TCs. Results showed that the atmosphere-only configuration produced stronger TCs than the coupled configuration; however, both model configurations reasonably represented the TC dynamics. Sahoo et al. (2019) evaluated the impact of cyclone Phailin’s strong winds and heavy rainfall on the power distribution network. The data from the tailored WRF model was found to be highly effective in managing power distribution and transmission networks in electric power grids. A thorough inspection of the results reveals that the atmospheric model fared well to capture power line tripping time. This study’s overall knowledge offers a greater scope for developing a framework for efficient power network planning operations, resource allocation, and disaster preparedness. Kumar et al. (2017) evaluated the WRF model with nested domains (horizontal resolutions of 27 and 9 km for outer and inner domains, respectively) to simulate cyclone Phailin. Compared to the domain-1 simulation, the cyclone’s intensity was well simulated in domain-2. In the domain-2 simulation, the average root mean square error and standard deviations of surface wind and MSLP were significantly lower. The track of cyclone Phailin was also well demonstrated by domain-2 simulation with respect to IMD best-track observation.

For NIO cyclones real-time prediction, the IMD in New Delhi introduced some in-house global and limited area NWP models (Mohapatra et al., 2014, Mohapatra et al., 2013; Mohanty et al., 2015; Nadimpalli et al., 2020a). IMD employed the Quasi-Lagrangian model as a TC operational model with a coarser resolution and only 16 vertical levels in the twenty-first century’s first decade. Under the auspices of the ‘Forecast Demonstration Project of Landfalling TCs,’ some prestigious research centres and academic institutes in India also provide real-time TC forecasts to IMD for official usage (Mohapatra et al., 2011; Nadimpalli et al., 2020a). Since 2007, some operational centres have adopted the WRF model for real-time TC prediction. Adapting such mesoscale models like WRF and HWRF for the NIO region has been a significant development in recent years. Some works of literature such as Kanase and Salvekar (2015); Mahala et al. (2021); Osuri et al. (2012); Pattanayak et al. (2012a), and Raju et al. (2011) conducted dedicated simulations to determine the optimal parameterization scheme set up within the WRF model for the NIO region. The WRF model was used by Osuri et al. (2013) to simulate 17 NIO TCs. The TCs movement had an eastward and sluggish bias. Track errors ranged from 113 to 375 km with a model resolution of 27 km. Pattanayak et al. (2012b) and Mohanty et al. (2013) found better results with the HWRF model for TCs’ track prediction and improved intensity prediction over the BoB. Nadimpalli et al. (2020a) evaluated the forecasts from WRF and HWRF models in a quasi-operational setup for 10 BoB TCs from 2013 to 2017. Both models performed well for forecasts up to 30 h; however, HWRF generated more accurate results for longer forecasts.

Vinodhkumar et al. (2022) studied the climatology of NIO cyclone RI behaviour from 1990 to 2019. If the MSW speed of a cyclone increases by 15.4 m/s or greater in 24 h, it is classified as RI. During the time period given above, 46 NIO cyclones had the RI feature, with a considerable increase beginning in the year 2000. The post-monsoon season proved more favourable for RI TCs. While around 70% of RI TCs migrated north-westward or westward, eastern states of India (Tamil Nadu, and Andhra Pradesh) are more vulnerable to RI TCs. A possible cause behind the increase in the RI TCs over the NIO could be the higher moisture and SST over the cyclogenesis locations than the non-RI TCs (Nadimpalli et al., 2021). Munsi et al. (2021) used the WRF model to perform three-dimensional variational-Ensemble Kalman Filter data assimilation for three cyclones over the NIO: Fani, Ockhi, and Luban. To investigate the RI of these TCs, upper-air observations, radiometer wind, and radiance data were assimilated. In all cases, the model framework correctly simulated the RI.

Osuri et al. (2013) reported the WRF model to be more efficient in predicting the landfall location than the other attributes. Govindankutty et al. (2010) showed a positive impact of assimilating the conventional and Doppler weather radar radial wind assimilation on the intensity and rainfall distribution of NIO TCs. Sandeep et al. (2017) assimilated the Doppler weather radar wind profiles in the 3DVar and Hybrid 3DVar mode and found a minimal improvement in the TC’s zonal and meridional winds simulation. Gopalakrishnan and Chandrasekar (2018) performed the assimilation of satellite-derived winds, satellite radiance, and conventional observations with 3DVar and four-dimensional variational (4DVar) mode for the first time over the NIO for 4 TCs. The 4DVar experiments revealed better results than 3DVar in terms of track and intensity and rainfall prediction.