Evaluating the impact of topography on the initiation of Nor’westers over eastern India

Abstract

Pre-monsoon thunderstorms are catastrophic and frequent over eastern and north-eastern India. The Chhota Nagpur plateau (CNP) believes to be the triggering platform for thunderstorms over east India. This study attempted to determine the influence of the CNP topography on thunderstorm formation and propagation by observing the changes in the values of thermodynamic indices over eastern India. The work attempted ten cases of thunderstorms in multiple simulations, i.e. each case simulated three times with varying elevations of CNP topography (in total 30 simulations) by Advanced Weather Research and Forecasting (WRF–ARW) model. Our analysis revealed that when the elevation of topography of the CNP changes (increases/decreases), the characteristics of several thermodynamic indices alter considerably (with values changing from negative to the positive side for some indices). We found the Cross Total Index (CTI) shows considerable differences in all the thunderstorm simulations, and the values are higher in the increased CNP topography (∼ −4 to 6 °C). However, the Humidity Index (HI) (∼ −10 to 15 °C) and Convective INhibition (CIN) (∼ −200 to 200 J/Kg) values depict considerable changes in decreased topography. The results are crucial to understanding and quantifying the role of CNP in pre-monsoon thunderstorms over east India.

Keywords:

• Thunderstorms
• topography
• numerical simulation
• thermodynamic indices
• CNP

1. Introduction

Thunderstorms are ferocious short-term weather events accompanying lightning, thunder, dense clouds, substantial rain or hail, and strong gusty winds (Harel and Price 2020). Gradually, it develops into a warm and wet updraft that ascends to the cooler parts of the atmosphere. At any moment, around 1800–2000, thunderstorms occur across the globe, with 45,000 thunderstorms per day, leading to 16 million thunderstorms in a year (Dudhia 1996). In India, we experience severe thunderstorms in the Pre-monsoon season, i.e. from March–May over the eastern (states of Odisha, West Bengal, Jharkhand, Bihar, Chhattisgarh) and northeastern (states of Assam, Tripura, Sikkim, Arunachal Pradesh, Meghalaya, Manipur, Mizoram) regions. These storms are commonly called ‘Kal-Baishakhi’ or Nor’westers. These storms, with increased severity, may sometimes lead to the formation of Tornadoes which causes severe destruction to the infrastructures, aviation industry, loss of life, and damage to the natural ecosystem (Tyagi et al. 2022). The solution to lessen the deadly effect is to predict their occurrence and propagation in advance.

Predicting the origin and evolution of these thunderstorm occurrences is difficult due to their small spatial and temporal dimensions (Ghosh et al. 2008; Madala et al. 2013). Various mesoscale models (WRF-NMM, WRF-ARW) are adopted to predict thunderstorm occurrence and propagation accurately (Litta and Mohanty 2008; Litta et al. 2012; Madala et al. 2013). Adequate preliminary information about the atmospheric and topographic conditions is a prerequisite for the model. Due to scale interactions and diverse meteorological circumstances, researchers still cannot fully comprehend the specific initiating agent of thunderstorms (Rodriguez et al. 2010). Over the years, thermodynamic indices have gained immense popularity and shown efficient results in the successful nowcasting of thunderstorm activities over the globe (Peppier 1988; Ravi et al. 1999; Haklander and Delden 2003; Kunz 2007; Tyagi et al. 2011; Sahu and Tyagi 2022; Sahu et al. 2020, 2022a, 2022b, 2022c; Samanta et al. 2020). These thermodynamic indices have been well explored using numerical model simulation to forecast the morphology of thunderstorms (Litta and Mohanty 2008; Latha and Murthy 2011; Robinson et al. 2013; Nayak et al. 2021).

The model data provides higher temporal and spatial resolution. It allows the forecaster to analyse the thunderstorm occurrences with much greater depth and helps to understand the structure and evolution of the thunderstorm after comparing it with actual observations. Litta and Mohanty (2008) discussed the competence of the Weather Research and Forecasting (WRF) model to predict the thunderstorm with a 24–h lead time. Litta et al. (2012) investigated the sensitivity of the microphysics scheme in the WRF model of the nonhydrostatic mesoscale model core (NMM) across West Bengal. Topography substantially impacts weather processes by transferring momentum and energy among mesoscale and large-scale weather systems (Rotach et al. 2015; Malardel and Wedi 2016). Wang et al. (2013) explain the impact of topography on China’s yellow mountain (the Huangshan) region by removing the yellow mountain and changing its land cover categories. The results suggested that changing the yellow mountain’s topography impacts the latent heating and accumulated water vapour, weakening vertical lifting, and reducing precipitation.

It has been reported that most thunderstorms formed in eastern India are initiated at the Chhota Nagpur plateau (CNP) (STORM Science Plan 2005; Ghosh et al. 2008) and later transform into multicell cluster thunderstorms. After mixing with the hot and humid air, these thunderstorms propagate towards the northwest to southeast directions and become more vigorous. According to reports (IMD TN 10 1944), eastern India receives Type-A thunderstorms where CNP provides the triggering. The CNP is well-known in India for its diverse landscape. The region receives an annual average rainfall of roughly 1400 mm, which is lower than India’s rain-forested areas, and more than 80% of the arable land (around 2,287,260 hectares) is rainfed (Dey and Sarkar 2011).

Understanding the impacts of CNP to provide the triggering mechanism to thunderstorms in other associated regions is crucial. Based on the available literatures we have found that no study has been found studying the impact of CNP topography over the eastern Indian region. The current study emphasises finding the effect of the CNP in initiating the thunderstorm and propagating towards eastern India, using the Advanced Research Weather Research and Forecasting model (WRF–ARW). Herein, the prime focus is to see the impact of topographical changes in the CNP (height) on thunderstorm activities. The article is organised as follows: Section 2 describes the study area, whereas Section 3 depicts the data and methodology. Section 4 discusses the results and discussions, and Section 5 provides a research Summary.

2. Study area

The study regions over eastern India includes the states of Jharkhand, Bihar, Odisha, and West Bengal. These states are located along the Indo-Gangetic plain (Bihar, West Bengal), the Eastern Ghats and Deccan plateau (Odisha), and the CNP (Jharkhand). The eastern Indian region experiences a humid subtropical climate, with blistering summers from March–June, monsoons from July–October, and moderate winters from November–February. The region’s interior has a dry climate that is significantly more severe, specifically in the winters and summers, although the entire area experiences intense, persistent precipitation in the monsoon months. Dry deciduous woodlands and tropical and subtropical dry broadleaf forests dominate the vegetation (Champion and Seth 1968).

Agriculture dominates the biotic province, accounting for around 62.10% of the CNP, followed by towns, orchards, and water features. Increased agriculture and other anthropogenic pressure lead to provincial fragmentation and disruption. Recent increases in disturbance profiles also indicate the influence of anthropogenic pressure; about 29.11% of all vegetative areas in the CNP are classified as having the highest disturbance condition (Roy et al. 2012). Figure 1 shows the study area (states of Jharkhand, Odisha, Bihar and West Bengal) with CNP highlighted by red colour.

Figure 1. Study area showing the CNP region (highlighted with red colour).

3. Data and methodology

3.1. Cases adopted in the present study

Based on the data availability to compare with the observation, a total of ten thunderstorm events over the Kolkata region and some parts of West Bengal observed during the pre-monsoon season (March–May) from 2016 to 2019 from the Meteorological & Oceanographic Satellite Data Archival Centre (MOSDAC). The details of the events are shown in Table 1.

Table 1. Details of the thunderstorm events.

Cases	Thunderstorm event date	Details of the thunderstorm activity observed were obtained from STROM reports of IMD
Case 1	03 May 2016	A thunderstorm with lightning was noticed at the Dum Dum area between 22:30 to 23:30 IST, with the wind blowing from the North–West with a maximum wind speed of 33 km/h.
Case 2	17 May 2017	The event was observed with lightning over Dum Dum between 16:30 to 18:00 and 21:00 to 23:30 IST with a wind speed of 24 km/h.
Case 3	07 April 2018	Thunderstorm developed over the Dum Dum, Kolkata from 16:30 to 20:30 IST with light rain and wind speed of 44 km/h.
Case 4	17 April 2018	A strong thunderstorm occurred over Kolkata and associated regions of Howrah, Hooghly, and North 24 Parganas from 19:30 to 21:30 IST with heavy squalls and rainfall of 29.6 mm with wind speed 52 km/h, gusting to 90 km/h.
Case 5	13 May 2018	An echo of a strong thunderstorm with lightning occurred over the Dum Dum area, Kolkata, from 09:00 to 09:30 IST and 13:30 to 18:30 IST with light to moderate rain and a wind speed of more than 33 km/h.
Case 6	22 May 2018	A strong thunderstorm with lightning was spotted over the Dum Dum region, Kolkata, at 16:30 to 20:00 IST with a wind speed of more than 20 km/h with moderate rainfall.
Case 7	17 March 2019	The event was observed over Alipore, Dum Dum area over Kolkata with lightning at 17:40 to17:50 IST in Dum Dum and thunder squall at 19:25 to 19:27 IST in Alipore is observed with wind blows from the NW direction with a wind speed of more than 72 km/h with rainfall 26.2 mm/h (STORM Project Report 2019).
Case 8	06 April 2019	A strong thunderstorm occurred between 18:45 to 24:00 IST over Dum Dum with 30.5 mm rain and 19:00 to 21:20 IST over Alipore with 25.1 mm rain, with a wind drift through NW direction wind speed > 64 km/h.
Case 9	22 May 2019	A strong echo was noticed over the Malda area of West Bengal with frequent lightning between 18:45 to 21:00 and 00:20 to 01:10 IST, with a rainfall of 21.2 mm recorded on this day.
Case 10	25 May 2019	The thunderstorm with thunder-squall observed over Alipore (18:20 to 19:50 IST; 78 km/h wind speed; 8.7 mm rainfall) and Dum Dum region of Kolkata (18:24 to 19:40 IST; 56 km/h wind speed; 23.2 mm rainfall). The wind blows from NW directions.

3.2. Configuration of WRF model and methodology

We have used the WRF-ARW model (Version 3.9), a nonhydrostatic and fully compressible Advanced Research version model, for the present study. The model was run for 30 h, beginning at 00:00 UTC, for all the ten thunderstorm cases, as shown in Table 1. The final global analysis for every 6-h interval at 1°× 1° grids developed by NCEP’s GDAS dataset was used to determine the initial conditions. A triple-nested domain setup was employed for this investigation. The first domain (d01) has a resolution of 27 km, while the second (d02) and third domains (d03) have resolutions of 9 km and 3 km, respectively. The model domain has simulated 51 unequally spaced sigma pressure levels, with the top level having a pressure of 50 hPa. The first domain (d01) covers 9.97°N to 34.11°N and 73.237°E to 103.48°E and (116 × 101) grids. The second domain (d02) has (187 × 181) grids, and the third domain (d03) has (370 × 301) grids. These three domains are shown in Figure 2 with the elevation and are obtained from Global Multiresolution Terrain Elevation Data 2010 (GMTED, 2010) in km provided by the US Geological Survey.

Figure 2. WRF-ARW Domain for the model simulations. d01 represents domain 1 with 27 km, d02 represents domain 2 with 9 km, and d03 represents domain 3 with 3 km spatial resolution for eastern India.

The model physics and parameterisation schemes have been adopted from the previous studies over the Kolkata and Kharagpur regions which are crucial schemes for better forecasting thunderstorms (Madala et al. 2013; Prasad et al. 2014; Nayak et al. 2021). The model utilised Mellor–Yamada–Janjic (MYJ) for Planetary Boundary Layer (PBL) scheme, WRF single–moment 6–class (WSM6) for the microphysics scheme, Betts–Miller–Janjic (BMJ) scheme for cumulus parameterisation. No convection method is employed in this domain (d03) to see if WRF can explicitly simulate convection. For the land surface, the Noah Land surface model was utilised, and for the microphysics choices, the Rapid Radiation Transfer Model (RRTM) for longwave radiation (Mlawer 1997) and the Dudhia scheme (Dudhia 1989) for shortwave radiation. The details of all these schemes are provided in Table 2. For the current study, the following datasets were used for studying the thunderstorm occurrence:

Table 2. WRF-ARW (Version 3.9) model configuration used in this study.

Dynamics	Nonhydrostatics
Datasets	NCEP-FNL
Data interval	6 h
Grid size	d01: (116 × 101) × 51, d02: (187 × 181) × 51 d03: (370 × 301) × 51
Resolution	d01: 27 km ×27 km, d02: 9 km × 9 km d03: 3 km × 3 km
Covered area	(9.9767 − 34.1126 ^°N) and (73.2376 − 103.4824 ^°E)
Map projection	Mercator
Horizontal grid system	Arakawa-C grid
Integration time step	60 s
Vertical co-ordinates	Terrain–following hydrostatic pressure vertical co-ordinate with 51 vertical levels
Time integration scheme	Third-order Runga–Kutta Scheme
Spatial differencing scheme	Sixth-order center differencing
Microphysics	WRF Single–moment 6–class Scheme (WSM6)
Cumulus parameterization	Betts–Miller–Janjic Scheme (BMJ)
Planetary boundary layer (PBL) scheme	Mellor–Yamada–Janjic Scheme (MYJ)
Surface layer parameterisation	Noah land Surface Scheme
Short wave radiation	Dudhia scheme
Longwave radiation Land surface Surface layer physics	RRTM scheme Unified Noah Land Surface Scheme Eta Similarity Scheme

• Upper atmosphere radiosonde data from the University of Wyoming are gathered at 00 and 12 UTC, which includes temperature (in °C), pressure (in hPa), relative humidity (in %), wind speed (in m/s) and direction (in degrees).

• Surface meteorological characteristics such as pressure (in hPa), temperature (in °C), relative humidity (in %), and wind speed (in m/s) were acquired from MOSDAC (Meteorological & Oceanographic Satellite Data Archival Centre) from the Indian Space Research Organization (ISRO).

• NCEP/NCAR FNL (final) global analysis data with 1°× 1° grid resolution is used for initial and boundary conditions every 6–h.

• WRF-ARW model output datasets are used to compare vertical structures with radiosonde datasets for different time steps and the time-series comparison of temperature and relative humidity with MOSDAC datasets.

In this work, we have tried to see the impact of increasing and decreasing elevation of topography over the CNP, having latitude and longitude extent of (22.4 − 23.7°N and 82.8 − 85.7°E) and Figure 3 shows the elevation of topography change over the CNP region.

Figure 3. Topography change over CNP region marked in the white box.

Three model simulations have been done for each thunderstorm case and the topography was changed before the WRF run in the WRF preprocessing system in geogrid data:

1. without changing the topography of CNP (hereafter CNTL),

2. decreasing the height of the topography of CNP by 1/4th time (hereafter Exp.1), and

3. increasing the height of the topography to double CNP (hereafter Exp.2) for each thunderstorm case.

The model simulation has been run to examine how thermodynamic indices alter when the elevation of topography of CNP is modified (increased/decreased) versus when it is left intact and how the change affects thunderstorm initiation and propagation. The thermodynamic indices Convective Available Potential Energy (CAPE), Convective INhibition (CIN), Cross Total Index (CTI), Vertical Total Index (VTI), Totals Total Index (TTI), Humidity Index (HI), K-Index (KI), and Severe Weather Threat Index (SWEAT) were incorporated in this study to compare the topographical changes over CNP and the associated eastern India region. CAPE is the vertically integrated, positive buoyancy force of an adiabatically rising air parcel (Moncrieff and Miller 1976). Higher CAPE values often indicate stronger convection, and they are used to examine the conditional instability of the atmosphere (Williams and Renno 1993). Especially in the eastern Indian region, operational forecasting can certainly benefit from the use of CAPE, which serves as a reliable warning of the onset of convection (Roy Bhowmik et al. 2008). To what extent buoyant energy and moist instability are present in the atmosphere is shown by CAPE (Neelin 1997). Higher CAPE values represent higher convective activities over any region. The CIN index is helpful because it measures the amount of negative buoyant energy applied to an air parcel, which slows down rising air and reduces the likelihood of convection (Colby 1984). In general, the likelihood of convective activity decreases (increases) as CIN values increases (decreases), which is proportional to the magnitude of the change. The K Index (KI) is another helpful predictor of convective commencement (George 1960). By combining the difference in temperature between 850 and 500 hPa with the dew point or moisture between 850 and 700 hPa, this is a metric used to identify the air mass that is producing thunderstorms (George 1960). By measuring the KI, one can infer about the amount of moisture present in the lower atmosphere and the height to which it extends. Total Totals Index (TTI) is an important empirical index for predicting thunderstorms and other convective occurrences. It can withstand a variety of storm intensities. Nonetheless, it did not necessitate latent instability at pressures below 850 hPa (Miller 1967). It incorporates both CTI and VTI. This index considers the lapse rate and the moisture level in the lower atmosphere to determine the parameters for static stability. TTI values ranging ≥55 results in strong thunderstorm, it is also used for forecasting severe local thunderstorms (Reap and Foster 1979).

Difference of dry-bulb and dew-point temperatures at 850 and 500 hPa pressure is the metric of CTI (Miller 1967). Low-level moisture (850 hPa) and comparatively colder air at higher altitudes (500 hPa) both contribute to a higher CTI. Increasing CTI values results severe thunderstorms, the CTI values ≥30 results scattered severe thunderstorms (Miller 1967). The VTI is the difference between the dry-bulb temperatures at 850 and 500 hPa (Miller 1972). Although it measures conditional instability between 850 and 500 hPa, the VTI (Miller 1967) does not take moisture into account. Increasing values of VTI results in forming cluster moderate thunderstorms with the VTI values ranging ≥30. HI is the sum of the differences of dry bulb temperature and the dew point temperature at altitudes of 850, 700, and 500 hPa (Litynska et al. 1976). Lesser the value of HI results severe thunderstorm. Litynska et al. (1976) established ≤30 as the threshold for HI. The SWEAT index (Miller 1972) was developed to evaluate the likelihood of extreme weather events, such as severe thunderstorms and tornadoes, as compared to more common thunderstorms. Particularly destructive convective weather is being targeted with the inclusion of TTI and the 850 hPa dew point temperature. The index is used for accessing the severity of thunderstorm by weather forecasters. The SWEAT values ranging ≥250 is suitable for severe thunderstorm occurrence (David and Smith 1971)

4. Results and discussion

4.1. Model evaluation for simulating the meteorological parameters

A total of ten thunderstorm cases were analysed for this present work. Meteorological parameters play a significant role in developing thunderstorm activities (Asnani 2005). Model simulated surface temperature (in °C) and Relative humidity (RH) (in %) have been compared with ERA5 reanalysis datasets and MOSDAC observations to assess the model performance. We have compared the model output with the ERA5 data due to two reasons: (i) ERA5 has proven its utility to simulate the mesoscale and synoptic scale process like thunderstorm and Tropical Cyclones in the convective environment over the globe (Sahu et al. 2022a, 2022c), and (ii) There are some limitation/gaps in MOSDAC datasets. So, to compare the model output, we used ERA5 data along with in situ observations (MOSDAC).

4.1.1. Surface meteorological parameters variations for thunderstorm days

Figure 4 shows the intercomparison of surface meteorological parameters (temperature and RH) of observation with model-simulated and ERA5 data for the 17 May 2017 thunderstorm event. Based on the data availability and severity of thunderstorms experienced on the event day, this case has been chosen. The comparison was made by choosing the point location as in this case is Kolkata with latitude and longitude [22.57°N and 88.36°E]. We are showing the results simulated the inner domain with horizontal grid resolution of 3 km for this study. From Figure 4a, we can see a sudden drop in temperature with observation, i.e. 36.7 °C to 30.1 °C between 8 UTC and 11 UTC. Similarly, a rise in RH from 53.2% to 77.6% between 8 UTC to 10 UTC can be inferred from Figure 4b. Again, a sharp drop in temperature was observed between 15 UTC and 17 UTC, from 29.6 °C to 24.2 °C (Figure 4a), and a continuous rise in RH was observed between 11 UTC and 15 UTC, i.e. from 58.7% to ∼92%. These variations in the meteorological parameters are associated with the occurrence of thunderstorm events during this time. When comparing the observations with the model simulation, a sudden temperature drop is well captured by the model at 8 UTC. The model overestimated the temperature values, and ERA5 data underestimated the temperature with a lead time of 1 h. The model RH is overestimated with a lag of 1 h, while in ERA5, RH shows an underestimation (Figure 4b). The sudden rise in RH during the thunderstorm can be attributed to the moist air incursion and associated rainfall. The drop in temperature during the event is associated with precipitation.

Figure 4. Comparison of Model simulated surface meteorological parameters (a) Temperature (in °C) (b) Relative humidity (in %), with MOSDAC and ERA5 data over Kolkata from 00 UTC to 06 UTC on 17 to 18 May 2017.

Figure 5 displays the intercomparison of surface temperature and RH of observation with model-simulated and ERA5 data for the 7 April 2018 thunderstorm event. Figure 5a highlights a sudden drop in observed temperature, i.e. 34.2 °C to 28.7 °C between 9 UTC and 13 UTC and a rise in RH from 52.5% to 91.1% between 9 UTC to 15 UTC (Figure 5b). Further, a strident drop in temperature was observed between 14 UTC and 16 UTC, from 28.3 °C to 22.9 °C (Figure 5a). These meteorological conditions indicate the occurrence of a thunderstorm event during that time. After comparing the observational variations with model simulation, the model captures an abrupt drop in temperature at 8 UTC with 1 h lag from the observation. Still, the temperature values of the model are underestimated compared to the observation. ERA5 data underestimates the temperature compared to the observation, but it harmonises temporally with the observational drop. The model underestimated RH compared to the observation with a lead of 1 h, whereas ERA5 RH shows an underestimation with a lag of 1 h (Figure 5b). Similar variation has been observed for all the cases considered in the study. The MOSDAC data is missing for some hours during the study period, making it discontinuous for the Figures 4 and 5.

Figure 5. Comparison of model-simulated surface meteorological parameters (a) Temperature (in ^oC) (b) Relative humidity (in %), with MOSDAC observation and ERA5 reanalysis data over Kolkata from 00 UTC to 06 UTC on 7 to 8 April 2018.

4.1.2. Vertical variations of meteorological parameters for thunderstorm days

For analysing the vertical variability captured by model simulations, we have analysed the vertical variation of temperature, RH, and Wind Speed (WS) at 00 UTC and 12 UTC. All ten cases of the study have been analysed. We have adopted the latitude and longitude of thunderstorm occurrence over the region from the India Meteorological Department STORM report. Figure 6 shows the vertical variation of temperature (°C), RH (%) and WS (m/s) for both 00 and 12 UTC for the thunderstorm event on 17 May 2017. From Figure 6a, we can see the high temperature at 1000 hPa level to 800 hPa, the temperature keeps decreasing, and at 500 hPa, the temperature falls near −5 °C. At 00 UTC, the model simulated temperature profile is well matched with observation and ERA5.

Figure 6b shows the RH variations; more humidity is present in the lower level of the atmosphere up to 800 hPa, which is the region of the planetary boundary layer (PBL) responsible for moisture transfer to the upper levels of the atmosphere. On close observation, one can notice the presence of convectional instability up to 800 hPa. The model captures RH well till 800 hPa; after that, there is a lag in the model profile, overestimating the radiosonde observations and ERA5 values. Figure 6c shows the WS in a range of 0–4 m/s below 800 hPa levels. This WS helps rapidly mix and transfer moisture to the upper levels. The WS is well simulated by the model and matches the ERA5 and the observation. However, between 700 hPa to 600 hPa, the model WS overestimates the observation; after 600 hPa, the model simulation underestimates the observation, whereas, in the ERA5 profile, a lag is observed after 650 hPa. Figure 6d–f represents the intercomparison of the model, observation and ERA5 data of temperature, RH and WS vertical profile for the 6 April 2019 thunderstorm event over Kolkata. In Figure 6d, there is a smaller magnitude in the overestimation of temperature observed in the PBL region.

Figure 6. Vertical variation of model-simulated profiles of temperature (in ^oC), Relative humidity (in %) and wind speed (m/s) with Radiosonde observation and ERA5 reanalysis data over Kolkata region for 00 UTC on 17 May 2017 (Upper Panel) and 6 April 2019 (Lower panel). The first column represents temperature variations (a and d), second column represents RH (b and e), and third column represents WS (c and f).

In Figure 6e, the RH shows dynamic variations in the model simulated profile. The model overestimated RH (compared to both observations and ERA5) from 1000 hPa to 900 hPa and underestimated RH from 900 to 700 hPa, again overestimating RH from 700 hPa to 500 hPa. Clear capture of WS is highlighted in Figure 6f, where the model and the ERA5 profiles are well-matched along with the observation, except 950 hPa to 850 hPa and 600 hPa to 500 hPa, where a lag of WS is observed.

Figure 7 shows the vertical intercomparison of temperature, RH, WS for 12 UTC during the thunderstorm events on 3 May 2016 (Figure 7a–c) and 17 April 2018 (Figure 7d–f). In Figure 7a, the model temperature captures the ERA5 and observation profile except from 900 hPa to 800 hPa where the model and ERA5 temperature overestimated with the observation. The model RH shows underestimation in the surface level up to 950 hPa with the observation after that it is overestimated up to 900 hPa. From 900 hPa onwards, the model RH underestimated the observed values till 570 hPa, afterwards overestimated observation, as shown in Figure 7b. For RH, ERA5 shows underestimation from surface to 800 hPa, then, overestimated up to 650 hPa after that underestimated with the observation. Figure 7c shows the vertical variation of WS where up to 900 hPa, model and ERA5 overestimated WS. The model WS follows overestimation up to 740 hPa thereafter underestimate up to 650 hPa, after that follows with the observed WS. ERA5 underestimated WS up to 575 hPa. Similarly, Figure 7d shows the temperature variation; in which the model as well as the ERA5 temperature profiles captures the observed profile, except in the boundary layer, i.e. up to 850 hPa where it is slightly underestimated the observations. Figure 7e shows that model underestimated RH except from 780 hPa to 680 hPa, whereas ERA5 overestimates RH from 1000 hPa to 900 hPa, 780 hPa to 640 hPa, and 570 hPa onwards, but underestimate RH from 900 hPa to 780 hPa and 640 hPa to 570 hPa (Figure 7e). The model and ERA5 WS profiles show an overestimation for most of the pressure levels below 500 hPa, however, 550 hPa onwards pressure levels show underestimation of WS with observation (Figure 7f).

Figure 7. Vertical variation of Model simulated profiles of temperature (in °C), Relative humidity (in %), and wind speed (m/s) with Radiosonde observation and ERA-5 reanalysis data over Kolkata region for 12 UTC on 3 May 2016 (Upper panel) and 17 April 2018 (Lower panel). The first column represents temperature variations (a and d), second column represents RH (b and e), and third column represents WS (c and f).

4.2. Statistical analysis of model performance analysis for thunderstorm days

Figures 8a and 9a show the statistical performance evaluation between the model and in situ observation for surface meteorological parameters (temperature, RH and WS) for the ten thunderstorm cases over eastern India. The statistical analysis is based on the Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Correlation Coefficient (CC), Mean Bias Error (MBE), and Standard Deviation (SD) (Wilks 2011; Franses 2016), to analyse the model’s performance.

Figure 8. Taylor plots of temperature (in °C), Relative humidity (in %), and wind speed (m/s) (a) Surface variation, (b) Vertical 00 UTC variation, and (c) Vertical 12 UTC variation over Kolkata region for ten thunderstorm cases.

Figure 9. Statistical analysis of temperature (in °C), Relative humidity (in %), and wind speed (m/s) over Kolkata region for Surface and Vertical (00 and 12 UTC) variation for ten thunderstorm cases. The intensity of black shades represents RMSE for respective meteorological parameters, similarly blue for MAE and red for MBE.

4.2.1. Root mean square error

The RMSE is the average square root of all the errors. RMSE is widely used because it is a useful all-around error metric for numerical forecasts. $$\mathrm{RMSE}=\sqrt{\frac{1}{n}{\sum }_{i=1}^n{(S_I-O_I)}^2}$$ where O_I = observations, S_I = forecasted values of a variable, and n = total number of observations. Since RMSE is scale-dependent, RMSE is a useful measure of accuracy only when comparing the forecasting mistakes of multiple models or model configurations for a single variable (Christie and Neill 2021).

4.2.2. Mean absolute error

MAE is a straightforward and intuitive way to estimate average error, and it serves as a useful indicator of model robustness. $$\mathrm{MAE}=\frac{1}{n}\mathrm{ }\sum _{i=1}^n{|O}_I-P_I|$$ where O_I = observations, P_I = forecasted values of a variable, and n = total number of observations.

4.2.3. Correlation coefficient (r)

The coefficient of determination (or Pearson correlation) quantifies the amount of (linear) prediction error that can be attributed to variations in the observed values. $$r=\mathrm{ }\frac{\sum {(O}_I-\overline{O_I}){(S}_I-\overline{S_I})}{\sqrt{\sum {{(O}_I-\overline{O_I})}^2\sum {{(S}_I-\overline{S_I})}^2}}$$

4.2.4. Mean bias error

The main purposes of mean bias error are to assess the overall bias in the model and to determine whether or not the model bias needs to be corrected. The average error in a forecast is measured by the MBE. $$\mathrm{MBE}=\frac{1}{n}\mathrm{ }\sum _{i=1}^n{(P}_I-O_I)$$

4.2.5. Standard deviation

The SD is a measure of dispersion in a dataset in relation to the mean, and it is computed by taking the square root of the variance. It is also denoted as ‘σ’. $$SD=\mathrm{\sigma }=\frac{1}{n}\mathrm{ }\sum _{i=1}^n{{(x}_I-\overline{x})}^2$$ where $x_I$ = Value of ith point of dataset $\overline{x}$ = Mean of the observation sample, n = Total number of sample observation.

The surface temperature is better correlated with observation than surface RH and WS (Figure 8a). The surface temperature for the 17 May 2017, 7 April 2018, and 6 April 2019 cases are highly correlated with the observation with a CC ≥ 0.90. With less SD (0.13 °C), the best-correlated case is 7 April 2018, with CC = 0.92. Similarly, for surface RH, 17 May 2017, 7 April 2018, and 17 April 2018 cases are highly correlated with the observation, i.e. the CC > 0.7 for these cases with less SD. The best-correlated case is 17 April 2018, with (CC = 0.94) the observation with the minimum SD of 0.64%. For surface WS, most thunderstorm cases show higher deviations, but the highly correlated case is 22 May 2018 with (CC = 0.71) and SD of 0.06 m/s. Figure 9a shows the statistical differences in surface meteorological parameters (Temperature, RH, and WS) in the form of RMSE, MAE, and MBE. Here, all the meteorological parameters are shown in different shades of black, red and blue (BRB) in colour. From higher to lower intensity, colour represents temperature, RH, and WS. From Figure 9a, we can observe that the case 7 April 2018 has minimum RMSE, MBE, and MAE apart from other cases (For temperature, MAE: 1.24, RMSE:1.63, MBE: −0.24, for RH, MAE: 6.77, RMSE: 7.76, MBE: −1.62, and for WS, MAE: 2.01, RMSE:2.85, MBE: 1.70) with higher correlations and minimum standard error. In the case of 17 March 2019, these statistical differences are higher, and values are not significant with higher RH: MBE (–28.06%), RMSE (31.44%), and MAE (28.06%). Similarly, the other eight cases are also more or less nonsignificant with the observation.

Figures 8b,c (Taylor plot) and 9b,c (bar plot) show the vertical statistical variance between model and radiosonde observations for temperature, RH and WS for both 00 and 12 UTC for the 10 cases. The Taylor diagram can be used to evaluate and contrast the strengths of several models or to monitor how an existing model’s performance evolves over time. The diagram’s geometric link between the statistical plots offers some insight into how to correctly weight different measurements of pattern correspondence when computing a skill score (Taylor 2001). The vertical temperature at 00 UTC is the best correlated with a minimum bias for all the thunderstorm cases. For vertical RH at 00 UTC, most of the model-simulated thunderstorm case statistics are well correlated, with a correlation between 0.81 and 0.99. The best cases with minimum SD and maximum correlations are 3 May 2016, 7 April 2018, 17 April 2018, 13 May 2018, 17 March 2019, and 22 May 2019. For the vertical variation of WS at 00 UTC; five cases (17 May 2017, 7 April 2018, 17 March 2019, 22 May 2019, and 25 May 2019) are highly correlated with observation with minimum SD. Figure 9b shows the statistical differences of vertical meteorological parameters at 00 UTC (temperature, RH, and WS) in the form of RMSE, MAE, and MBE. From Figure 9b, we can witness that the cases 3 May 2016 (For temperature, MAE: 0.53, RMSE:0.57, MBE: −0.30, for RH, MAE: 3.33, RMSE: 3.68, MBE: −0.65, and for WS, MAE: 2.13, RMSE: 2.25, MBE: −0.15) and 22 May 2019 (For temperature, MAE: 0.31, RMSE:0.36, MBE: −0.31, for RH, MAE: 2.21, RMSE: 2.63, MBE: 0.17 and for WS, MAE: 1.04, RMSE: 1.23, MBE: −0.07) has minimum RMSE, MBE, SD, and MAE and high correlations, apart from other cases with higher correlations and minimum SD. In 00 UTC, the cases with higher statistical differences with high values of MAE, RMSE, and MBE were observed on 13 May 2018, 22 May 2018, 6 April 2019, and 25 May 2019 thunderstorm cases for vertical RH.

Figure 8c shows the Taylor plot of the vertical meteorological parameters (temperature, RH, WS) at 12 UTC. In 12 UTC, the vertical temperature is best correlated with a minimum bias for ten thunderstorm cases. For vertical RH at 12 UTC, the model simulated RH is not well correlated with more SD for most thunderstorm cases. However, the case 13 May 2018 and 25 May 2019 shows some significance and slightly less SD with high correlation than the other thunderstorm simulations. For the vertical variation of WS at 12 UTC, a few thunderstorms simulation cases (3 May 2016, 7 April 2018, 17 April 2018, and 25 May 2019) are highly correlated with observation with minimum SD values as observed in the Taylor diagram. Figure 9c represents the statistical differences considering RMSE, MAE, and MBE for vertical meteorological parameters at 12 UTC. From this figure, we have noticed that the case 25 May 2019 (for temperature, MAE: 1.41, RMSE:2.22, MBE: −0.81, for RH, MAE: 5.70, RMSE: 8.41, MBE: −5.05 and for WS, MAE: 3.79, RMSE: 4.19, MBE: 3.79) has minimum RMSE, MBE, and MAE apart from other cases with higher correlations and minimum standard error concerning the observation. For most cases, RH values show higher MAE, MBE, and RMSE than the WS and temperature for the 12 UTC period.

Figure 10 represents the mean statistical variations of RMSE, MAE, and MBE of ten thunderstorm cases of Surface and Vertical (00 and 12 UTC) temperature, RH and WS. The mean values of horizontal meteorological variables are (for temperature, MAE: 2.19, RMSE:2.72, MBE:0.52, for RH, MAE: 11.56, RMSE: 13.71, MBE: 4.0, and for WS, MAE: 3.35, RMSE: 3.92, MBE: 3.16), vertical at 00 UTC (for temperature MAE: 0.56, RMSE:0.67, MBE:0.09, for RH, MAE: 6.97, RMSE: 8.49, MBE: 2.09, for WS, MAE: 1.49, RMSE: 1.76, MBE: −0.23) and at 12 UTC are (for temperature MAE: 2.62, RMSE:3.42, MBE:1.29, for RH, MAE: 19.36, RMSE: 23.76, MBE: 1.20, for WS, MAE: 5.63, RMSE: 6.35, MBE: 2.55).

Figure 10. Mean RMSE, MAE, and MBE of temperature (in °C), Relative humidity (in %), and wind speed (m/s) over Kolkata region for Surface and Vertical (00 and 12 UTC) variation for ten thunderstorm cases.

4.3. Experiments with changed topography and its impact on precipitation and reflectivity

This section evaluates the relationship between precipitation (PCP), maximum reflectivity and CNP topography variation to comprehend the significance of the CNP in modulating thunderstorms’ movement, location and intensity over eastern India. Topographical triggering plays a vital role in initiating a thunderstorm. For the present work, ten thunderstorm cases over eastern India are simulated by changing the topography of CNP. The unchanged and changed topographical simulations have been used for calculating different thermodynamic indices. The work uses CAPE, CIN, CTI, TTI, VTI, HI, SWEAT, and KI indices. The simulations have been done three times for each thunderstorm case:

1. with decreasing elevation of topography by 0.25 times (1/4th, Exp. 1),

2. with increasing the elevation of topography to 2 times from the actual topography (2.0, Exp. 2), and

3. with no change in the topography (CNTL).

The study showed PCP variations for the 6 April 2019 case and discussed all other cases. Figure 11 shows the model accumulated PCP differences for 30 h on thunderstorm event day, i.e. 6 April 2019 (from 6 April 2019 00:00 UTC to 7 April 2019 06:00 UTC) for Exp. 1 and Exp. 2. The difference was made by subtracting the decreased elevation of topography with CNTL simulation (Exp. 1) and the increased elevation of topography with the CNTL simulation (Exp. 2). The accumulated PCP difference ranges from − 40 to 40 mm. The WB region shows higher differences than other eastern Indian states, with southern parts showing less difference; western, eastern, and central WB showing apparent differences between Exp. 1 and Exp. 2. In Exp. 1, over the Jharkhand region, the PCP pattern decreases compared to Exp. 2. Over Odisha, the difference in PCP is similar for both Exp. 1 and Exp. 2. The PCP difference in Exp. 1 is decreasing than Exp. 2 in the CNP region. For 3 May 2016 thunderstorm simulation, the CNP region shows a clear difference in PCP in Exp. 1 and Exp. 2, with the higher intensity in Exp. 2. The PCP differences from CTNL in both experiments ranged between −40 and 40 mm. WB shows increased PCP (>20 mm) over the region in Exp. 2. Higher differences in PCP for Exp. 2 are also observed for the thunderstorm events on 17 May 2017, 7 April 2018, 13 May 2018, 22 May 2018, 17 March 2019, 22 May 2019, and 25 May 2019. The PCP differences are more prominent over WB regions than Odisha and Jharkhand for the thunderstorm cases. However, the northern Odisha and north/north-eastern Jharkhand regions show significant PCP differences for Exp. 1 and Exp. 2. For the thunderstorm case on 17 April 2018, the PCP differences in Exp. 1 and Exp. 2 (from CTNL) are less, ranging between ∼ −15 mm and 15 mm, with increased PCP differences in Exp. 2.

Figure 11. Accumulated PCP difference (from CNTL) with changing topography for thunderstorm case on 6 April 2019. Left panel shows difference of Exp. 1 from CNTL and right panel shows difference of Exp. 2 from CNTL.

The model simulated reflectivity differences of Exp. 1 and Exp. 2 (from CNTL) are shown in Figure 12 for the thunderstorm case on 6 April 2019. Here we have shown the hourly reflectivity difference from 12 to 15 UTC, as the thunderstorm was recorded over Kolkata between 14 and 15 UTC. Thunderstorm cells travel from the northwest to the southeast, i.e. from the CNP to the Kolkata region (Figure 12). The reflectivity differences are visible in both experiments with higher values (>50 dBZ) in Exp. 2 than Exp. 1. The simulated reflectivity for all other cases (not shown here) suggests that the storm generated at the CNP and propagated in the southeast direction to Kolkata, WB region. The model reflectivity difference exists in all the thunderstorm cases between Exp. 1 and Exp. 2, with the prominent increased/decreased intensity observed for the cases on 17 May 2017, 7 April 2018 (highest among all thunderstorm cases), 13 May 2018 and 17 March 2019. Simulated reflectivity also shows that Exp. 2 is shifting the thunderstorm cell more towards the southeast direction than Exp. 1.

Figure 12. Reflectivity difference (from CNTL) with changing topography for thunderstorm case on 6 April 2019. Top row shows difference of Exp. 1 from CNTL, and bottom row shows difference of Exp. 2 from CNTL for 12–15 UTC. The circle represents the distance of the thunderstorm location in kilometres, with each circle at a distance of 100 km from the thunderstorm location.

4.4. Experiments with changed topography and its impact on thermodynamic indices

According to previous studies related to Nor’westers over eastern Indian region, the severe thunderstorm originated from the CNP and then travel from north– west to south– west direction (STORM Science Plan 2005; Ghosh et al. 2008; Nayak and Mandal 2014; Chatterjee et al. 2015). The CNP region is marked in the Figures in a bigger box format, while the small box represents the Kolkata region. We presented the results of modified experiments with the CNTL simulations of one thunderstorm case on 6 April 2019 and discussed the other cases. The thunderstorm event occurred between 18:45 and 24:00 IST (13:15 to 18:30 UTC), and the indices variations at 15 UTC have been explored to understand the impacts of changing topography.

Figure 13 shows the difference in CTI values at 15 UTC with the changing CNP topography elevation. In Exp. 1, the CTI values decrease slightly from the Exp. 2 simulations in CNP and the other parts of eastern India. In Exp. 2, the CTI values are increasing over most parts of Kolkata and the West Bengal (WB) region. The changes in CTI values are visible in the CNP region (∼ −4 to 6 °C) and other regions (∼ −6 to 10 °C) like some parts of coastal Odisha, WB and the interior Jharkhand region. A significant change in CTI values is observed in the WB region, specifically in Kolkata; the values are more in the Exp. 2. The increase in elevation of topography leads to an increase in the CTI values and vice-versa.

Figure 13. CTI difference (from CNTL) at 15 UTC with changing topography for thunderstorm case on 6 April 2019. Left panel shows difference of Exp. 1 from CNTL and right panel shows difference of Exp. 2 from CNTL.

Figure 14 shows the VTI difference at 15 UTC with changing elevation of topography of CNP. Although overall patterns are similar, Exp. 1 shows slightly increasing VTI values (∼ −4 to 5 °C), whereas Exp. 2 shows a decrease in the VTI values (∼ −5 to 5 °C). In the western and northern parts of Odisha and southern parts of Jharkhand, Exp. 2 shows slightly higher values than Exp. 1. The VTI values are high all over the interior of WB, Jharkhand, and some regions of Odisha. Similarly, TTI values have higher magnitudes in Exp. 1 (∼ −5 to 10 °C), as shown in Figure 15. TTI values are higher over the CNP region (Exp. 1) (∼ −4 to 6 °C) than the Exp. 2 (∼ −4 to 3 °C). The coastal WB regions, mostly Sunderban areas, have higher TTI values in Exp. 2 than in Exp. 1. A clear difference is observed in the WB and some parts of the northern Odisha region for all the simulations.

Figure 14. VTI difference (from CNTL) at 15 UTC with changing topography for thunderstorm case on 6 April 2019. Left panel shows difference of Exp. 1 from CNTL and right panel shows difference of Exp. 2 from CNTL.

Figure 15. TTI difference (from CNTL) at 15 UTC with changing topography for thunderstorm case on 6 April 2019. Left panel shows difference of Exp. 1 from CNTL and right panel shows difference of Exp. 2 from CNTL.

The dewpoint and dry bulb temperatures are used in the computation of the humidity index, which includes saturation at 850, 700, and 500 hPa levels (Litynska et al. 1976). Figure 16 shows the difference in HI values for Exp. 1 and Exp. 2 from the CTNL at 15 UTC on 6 April 2019. HI values differentiate Exp.1 and Exp. 2, with most regions in Exp. 1 showing considerable changes in HI values (∼ −15 to 15 °C). CNP regions show higher/lower HI values for Exp. 1/Exp. 2 (∼ −5 to 15 °C/∼ −15 to 15 °C). In the WB region, Exp. 1 shows higher HI values than Exp. 2. Higher CAPE values signify a more unstable atmosphere. Figure 17 shows the difference in CAPE values for Exp. 1 and Exp. 2 from the CTNL at 15 UTC on 6 April 2019. The CAPE differences are higher for Exp.2 (∼ −1000 to 1000 J/Kg), with notable changes in the coastal regions and parts of the WB with >1000 J/Kg. A clear difference in CAPE was also observed in the CNP region (∼ −600 to 600 J/Kg).

Figure 16. HI difference (from CNTL) at 15 UTC with changing topography for thunderstorm case on 6 April 2019. Left panel shows difference of Exp. 1 from CNTL and right panel shows difference of Exp. 2 from CNTL.

Figure 17. CAPE difference (from CNTL) at 15 UTC with changing topography for thunderstorm case on 6 April 2019. Left panel shows difference of Exp. 1 from CNTL and right panel shows difference of Exp. 2 from CNTL.

CIN is an important index which indicates the energy needed to supply overcoming the inhibition of convection at a place. Figure 18 shows the difference in CIN values for Exp. 1 and Exp. 2 from the CTNL at 15 UTC on 6 April 2019. The CIN values decreased in Exp. 2 and increased in Exp. 1. The highest values of CIN (>400 J/Kg) were observed in the CNP and its adjacent regions in Exp. 1. The results indicate that the topographical reduction of the CNP region increases the CIN values of CNP and adjacent regions.

Figure 18. CIN difference (from CNTL) at 15 UTC with changing topography for thunderstorm case on 6 April 2019. Left panel shows difference of Exp. 1 from CNTL and right panel shows difference of Exp. 2 from CNTL.

SWEAT indicates the convective environment and low-level moisture, with wind speed and direction variations. The higher SWEAT values signify more severity of thunderstorms. Figure 19 shows the difference in SWEAT values for Exp. 1 and Exp. 2 from the CTNL at 15 UTC on 6 April 2019. The results show that though the magnitude was higher in Exp. 2, SWEAT patterns are similar in both Exp. 1 and Exp. 2. The WB region showed a higher SWEAT variation than other eastern Indian regions. The differences in SWEAT (>100) were observed over the central and northern parts of WB, some parts of north Odisha and CNP regions. The SWEAT values are minimum in coastal regions of Odisha and some parts of Jharkhand in both Exp. 1 and Exp. 2.

Figure 19. SWEAT difference (from CNTL) at 15 UTC with changing topography for thunderstorm case on 6 April 2019. Left panel shows difference of Exp. 1 from CNTL and right panel shows difference of Exp. 2 from CNTL.

KI nowcasts the thunderstorm occurrence based on vertical temperature lapse rate and low-level moisture. Figure 20 shows the difference in KI values for Exp. 1 and Exp. 2 from the CTNL at 15 UTC on 6 April 2019. The changes in KI values in Exp. 2 are higher (∼ −10 to 10 °C) than in Exp. 1 (∼ −10 to 8 °C) over most parts of WB and some of the western Odisha regions. The CNP region also differentiates KI patterns for Exp. 1 and Exp. 2.

Figure 20. KI difference (from CNTL) at 15 UTC with changing topography for thunderstorm case on 6 April 2019. Left panel shows difference of Exp. 1 from CNTL and right panel shows difference of Exp. 2 from CNTL.

Similarly, the other nine thunderstorm cases were simulated and discussed hereafter. On 3 May 2016, a thunderstorm occurred over the Kolkata region at 22:30 to 23:30 IST (17:00 to 18:00 UTC). The study analysed the results for 18 UTC of the model and observed that CTI, TTI, and HI significantly changed after changing the CNP region’s topography. CTI differences are more (∼ −8 to 8 °C) when the topography of the CNP area increases (Exp. 2) and less (∼ −10 to 6 °C) when the topography decreases (Exp. 1). The difference is visible more prominently over the CNP area. HI values increased in Exp. 1 (∼ −10 to 20 °C) and decreased in Exp. 2 (∼ −15 to 15 °C), with prominent variations over the western part of Odisha and southern Jharkhand. Similarly, for TTI, the values are increased in Exp. 2 (∼ −4 to 6 °C) over the CNP region.

On 17 May 2017, Kolkata experienced a thunderstorm event at 1630 to 1800 IST and 21:00 to 23:30 IST (11:00 to 12:30 UTC and 15:30 to 18:00). The indices variation at 15 UTC for this case showed that CTI, HI, KI, TTI, CAPE, CIN, and VTI considerably changes with changed CNP topography. CTI differences are higher in the CNP region in Exp. 2 (∼ −10 to 8 °C) and lower in Exp. 1 (∼ −10 to 6 °C). CTI values are higher in Exp. 2 for Kolkata and associated WB regions (∼ −2 to 8 °C). The HI differences are more in Exp. 1 (∼ −20 to 25 °C) than Exp. 2 (∼ −25 to 20 °C), specifically over the northern Odisha and coastal WB regions (>20 °C). In contrast, the north Jharkhand region follows the same patterns in Exp. 1 and Exp. 2, with marginally higher magnitudes in Exp. 1. KI values differentiate the CNP region, with higher differences in Exp. 1 (∼ −10 to 10 °C). The CAPE values are more consistent in Exp. 2 than Exp. 1, and the difference was more prominent over the coastal WB region in Exp. 2 (∼100 to 2000 J/Kg). The CIN differences are higher for Exp. 1 in the CNP and adjacent regions (∼ −100 to 400 J/Kg). TTI values follow a similar pattern in both the Exp. With a higher magnitude in Exp. 2 (∼ −8 to 10 °C). VTI values are also higher in CNP and adjacent regions in Exp. 1 (∼ −4 to 5 °C). Over the WB region, the VTI differences are higher in Exp. 1 (∼ −3 to 5 °C); however, over Kolkata, the differences are almost similar. Similarly, SWEAT values also show a slight variation of their values with Exp. 2, with marginally higher (∼ −100 to 200) differences than Exp. 1.

On 7 April 2018, two thunderstorm events happened between 16:30 to 20:30 IST (11:00 to 15:00 UTC) and 20:30 to 23:30 IST (15:00 to 18:00 UTC) over Kolkata. CTI differences are more in Exp. 2 (∼ −8 to 10 °C) and less in Exp. 1 (∼ −8 to 8 °C). Overall, the same pattern of CTI values was observed over some parts of Kolkata and the WB region. CIN differences are considerably decreased in Exp. 2 and increased in Exp. 1. HI and KI differences are higher in Exp. 1 (∼ −15 to 15 °C and ∼ −10 to 10 °C). The CAPE differences are similar in most parts of the eastern India region, while a clear difference was observed in the CNP region. SWEAT differences are higher in magnitude in the Exp. 2 (∼ −100 to 100). VTI differences was high over Odisha, parts of northern Jharkhand and the CNP region of Exp. 1 (∼ −5 to 5 °C).

On 17 April 2018, a strong thunderstorm occurred over Kolkata and other adjacent regions between 19:30 and 21:30 IST (14:00 to 16:00 UTC) with heavy wind and rainfall. Except for CIN, all other indices showed higher differences in the Exp. 1. VTI, TTI, and SWEAT showed apparent differences in CNP and the Kolkata regions. CAPE differences are minimal in Exp. 1 and Exp. 2, but the difference was prominent in the coastal parts of WB and some parts of northern and northeastern Odisha. CTI differences are marginally high in Exp. 2 (∼ −10 to 10 °C) in the CNP and adjacent regions, while over the Kolkata region, Exp.1 (∼ −2 to 5 °C) showed higher differences than Exp. 2 for both 15 and 18 UTC. HI differences were higher in Exp. 1 (∼ −6 to 10 °C) in the Kolkata and CNP regions than in Exp. 2. KI differences are insignificant in both Exp. 1 and Exp. 2. TTI and VTI differences were also higher in Exp. 1 than Exp. 2 (∼ −5 to 5 °C and ∼ −2 to 3 °C), with clear distinction over the Kolkata region (>3 °C) for both 15 and 18 UTC.

For the thunderstorm activity on 13 May 2018 from 13:30 to 18:30 IST (08:00 to 13:00 UTC) over Kolkata, CTI (∼ −6 to 6 °C), CAPE (∼ −1000 to 1000 J/Kg), KI (∼ −10 to 10 °C), SWEAT (∼ −100 to 100), and TTI (∼ −10 to 10 °C) showed higher differences in Exp.2 than Exp.1. HI (∼ −10 to 20 °C), CIN (∼ −200 to 200 J/Kg) showed marginally higher differences in Exp. 1. For Kolkata region, the VTI differences were more in the Exp. 2 (>4 °C). The CNP and Kolkata regions showed considerable changes in all the thermodynamic indices. A prominent difference in CIN (>200 J/Kg) was observed over the CNP region in Exp. 1.

On 22 May 2018, over Kolkata, thunderstorm activities happened between 16:30 to 20:00 IST (11:00 to 14:30 UTC). For this case, CAPE (∼ −1000 to 1000 J/Kg), CTI (∼ −5 to 5 °C), KI (∼ −10 to 10 °C), SWEAT (∼ −80 to 100), and TTI (∼ −3 to 5 °C) showed considerable differences in Exp.2. In contrast, HI (∼ −20 to 20 °C), CIN (∼ −200 to 200 J/Kg), and VTI (∼ −3 to 4 °C) showed marked differences in Exp. 1. The differences are prominent over the CNP region compared to other parts of eastern India.

The thunderstorm on 17 March 2019 occurred between 17:10 to 22:50 IST (11:40 to 17:20 UTC) over Kolkata. Except HI (∼ −20 to 20 °C), all the thermodynamic indices showed higher differences in Exp. 1 than Exp. 2. The differences are more prominent over CNP and adjacent parts and Kolkata regions of eastern India.

On 22 May 2019, the Kolkata region experienced continuous thunderstorm activity from 18:45 to 21:00 IST (13:15 to 15:30 UTC). The model simulations showed that CAPE (∼ −1000 to 1000 J/Kg), CTI (∼ −4 to 5 °C), TTI (∼ −5 to 5 °C), KI (∼ −4 to 6 °C), and SWEAT (∼ −100 to 100) were higher in Exp. 2. At the same time, the HI (∼ −10 to 15 °C), VTI (∼ −3 to 4 °C), and CIN (∼ −200 to 200 J/Kg) values are higher in Exp. 1. Prominent differences were observed in the CNP region for all the thermodynamic indices.

On 25 May 2019, a thunderstorm happened from 18:24 to 19:40 IST (12:54 to 14:10 UTC). During this time, a heavy thunder squall was observed over Dum Dum and the Alipore region of Kolkata. In this case, all the thermodynamic indices showed marginally higher differences in Exp. 1 than Exp. 2, with HI (∼ −15 to 20 °C) and CAPE (∼ −1000 to 1000 J/Kg) clearly distinguishing the differences in Exp. 1 with higher magnitudes.

5. Conclusions

The present work discussed the role of changing the elevation of topography of the CNP region to bring out atmospheric thermodynamic changes over eastern India as a triggering platform for initiating thunderstorms. The study simulated ten thunderstorm cases with the WRF-ARW model with varying topography conditions. The results can be summarised as follows:

• It has been found that model temperatures are well correlated with observed temperature values than RH and WS in both surface and vertical profiles. However, in some thunderstorm cases, the simulated parameters vary drastically at some levels. The WRF-ARW model adequately describes the present study’s surface and vertical meteorological conditions.

• With increased topography, the study observed the differences of both PCP and reflectivity are higher in Exp. 2, with the PCP and storm intensity slightly shifting towards the right, i.e. towards the southeastern side. The precipitation and maximum reflectivity analysis suggests that the increased topography favours higher PCP and the storm intensity higher in most thunderstorm cases.

• Thermodynamic indices showed a significant change in their values when we changed the topography of CNP. Besides all indices used in this study, CTI, HI, CIN, and SWEAT show considerable changes in their values with altered topography. The CNP regions and parts of WB, Odisha, and Jharkhand show substantial changes in the thermodynamic indices, which may impact the triggering of thunderstorm activity over the study region.

Although the results are clearly indicating the associated changes in thermodynamic indices with topography, we have limited number of observational datasets associated with the Kolkata site. It will be better to consider such simulations with other topographical features known to trigger the convective activities over a region before reaching to a conclusion.

Acknowledgments

The authors are thankful to India Meteorology Department for providing the thunderstorm information for the present study. Mr. Rajesh Kumar Sahu wants to acknowledge the National Institute of Technology Rourkela for providing research facilities.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

The datasets utilised in this article are freely accessible through the following sites: http://weather.uwyo.edu/upperair/sounding.html, https://mosdac.gov.in/ and https://rda.ucar.edu/datasets/ds083.2/#!description. The numerical model used in the study is available freely at https://www.mmm.ucar.edu/models/wrf. All other analysed datasets generated for this study are included in the article.

Additional information

Funding

Authors want to acknowledge the Science and Engineering Research Board (SERB), Department of Science and Technology, Govt. of India for providing the funding [project-funding code: DST/SERB/ECR/2017/001361].