Diurnal Analysis of Nor’westers over Gangetic West Bengal as Observed from Weather Radar

Abstract

Intense thunderstorms known as Nor’westers develop in the Eastern and North Eastern parts of India and Bangladesh before the monsoon season (March to May). The associated severe weather can cause extensive damage to property and livestock. This study uses the pre-monsoon volumetric data of S-band radar from 2013 to 2018 located in Kolkata to investigate the diurnal variation in the characteristics of the storms over Gangetic West Bengal. The cell initiation, echo top heights, maximum reflectivity, and core convective area are determined by using a flexible feature tracking algorithm (PyFLEXTRKR). The variation of the parameters in diurnal scale is examined from 211,503 individual cell tracks. The distribution of the severe weather phenomena based on radar based thresholds in spatial and temporal scale is also determined. The results show that new cell initiation peaks in the late evening and early morning, displaying bimodal variability. Most of these cells have a short lifespan of 0 to 3 h, with fewer than 5 percent of storms lasting beyond 3 h. The occurrence of hail is much greater in the afternoon due to intense surface heating than at other times. In contrast, the occurrence of lightning is higher in the late evening hours when the cell initiation reaches its peak. The convective rains are generally accompanied by lightning, exhibiting a similar diurnal temporal variability but are more widespread. The findings will assist operational weather forecasters in identifying locations that need targeted observation at certain times of the day to enhance the accuracy of severe weather nowcasting.

1. Introduction

Nor’westers are intense thunderstorms that develop during the pre-monsoon season in Eastern and Northeastern India, Southern Nepal, Bhutan, and Bangladesh. These meteorological phenomena are also called Kalbaisakhis [1]. These storms generate substantial precipitation, electrical discharges, thunder, hailstorms, and occasionally tornadoes, resulting in significant damage to property and loss of life, predominantly in rural areas [2]. The storms are generally initiated by thermal heating over the landmass around midday, leading to convection that intensifies as they progress towards Gangetic West Bengal and Bangladesh, interacting with warm, moist air from the Bay of Bengal [3]. A seasonal low-pressure system in conjunction with thermal lows and southerly winds from the sea contributes to the formation of these storms. Nor’westers are influenced by the passage of troughs in the westerlies at 500 hpa and the intensification of the subtropical westerly jet stream [4]. The storms undergo a continuous development and dissipation cycle throughout the diurnal period [5]. The duration of the convective cell ranges from minutes for isolated cells to several hours for extensive convective squall lines. The regional topography also plays a crucial role in the formation and life cycle of the storm cells [6].

Convective systems typically undergo three distinct stages: (i) Cumulus Stage: In this initial stage, moisture is lifted upwards, which can be triggered by heating, wind convergence, or orographic lifting (when air is forced to rise over mountains). As the moisture ascends, it cools and condenses into liquid droplets, releasing latent heat. This latent heat warms the surrounding air, causing it to rise further and create updrafts. (ii) Mature Stage: As the warm air rises, it eventually reaches a level that can no longer ascend due to warmer surrounding air. At this point, the air spreads out, forming an anvil shape at the top of cumulonimbus clouds. The liquid droplets coalesce into larger drops and transform into ice particles. These ice particles fall as rain when they melt. If the updrafts are strong enough, some ice particles may remain suspended for an extended period and fall as hail. The descent of hydrometeors generates downdrafts within the storm. The simultaneous presence of updrafts and downdrafts indicates a mature storm. This internal turbulence can lead to lightning, strong winds, and potentially tornadoes. (iii) Dissipation Stage: If the vertical wind shear is insufficient, the cool air brought down by downdrafts can impede the inflow of warmer air, leading to the storm’s eventual dissipation. This stage is characterized primarily by downdrafts, which can produce downbursts [7,8,9].

Numerous studies have been conducted to elucidate the characteristics of these pre-monsoon convective systems. In an early study by the India Meteorological Department, the Nor’westers occurring in this region were broadly classified as A, B, C, and D [10]. Type A Nor’westers are predominant, typically originating in the afternoon in the West Bengal and Chota Nagpur regions of India. The cells originating in the sub-mountainous district of West Bengal and Northern Bangladesh are classified as Type-B, primarily initiating during nighttime and early morning hours. The cells originating in the eastern hills are Type-C, which are infrequent. Type-D cells exhibit similar characteristics to Type-B but originate at the foot of the Khasi Hills [11,12,13]. The origin of different storms is depicted in Figure 1. Thermodynamic studies were conducted to identify the relationships between thermodynamic indices and the forecasting of convective weather development at different locations [14]. Synoptic studies were also undertaken to understand the influence of large-scale and mesoscale environments on the onset of storms. Although the thermodynamic indices and synoptic studies provide information regarding favorable environmental conditions for storm development, they could not provide the spatial and temporal variations of storm characteristics over the region [15,16,17]. Satellites have been utilized to study the mesoscale convective complexes and storm characteristics of Nor’westers. These studies document storm initiation, growth, and movement across the region [18] and has provided significant insights in the understanding the spatial and temporal distributions. Radar-based analysis of storms offer significant advantages over satellite images, mainly because satellite imagery can be limited during nighttime, and infrared (IR) images usually have lower resolution than radar. Additionally, previous studies have used geosynchronous satellites with passive sensors, which restricts their capability to capture the three-dimensional structures of storms. In contrast, ground-based remote sensing devices, like Doppler radar, can provide this crucial data.

Operational forecasters across the world face challenges in nowcasting severe thunderstorms due to its rapidly changing dynamics in a short time scale. The determination of storm characteristics from weather radar data can help in better understanding and nowcasting of the severe weather [19]. The first Doppler weather radar in India was installed in the year 2002 at Chennai followed by radars in Kolkata, Visakhapatnam, and Machilipatanam. The weather radar technology and algorithms have progressed significantly over the past few decades, resulting in numerous studies around the globe which can determine the characteristics of storms and their variability with a weather radar [20,21,22]. Studies utilizing the single polarization data from the Indian radar network using latest techniques and algorithms have been limited, and there is a need to utilize the large data sets archived from the weather radar network in India to understand the different weather phenomena over the country for better operational nowcasting [23,24,25].

The present study has three main objectives:

(a) To utilize the archived single polarization weather radar data from Doppler weather radar network of India to determine long term storm characteristics using the latest feature tracking algorithm.

(b) Examine the characteristics of the storm cells to identify new relations and to compare with other studies undertaken in the region.

(c) Determine the spatial and temporal distribution of the associated weather phenomena to aid the operational forecasters for nowcasting.

This paper is organized as follows. Section 2 describes the methodology used in this study for determining the characteristics of the storms. Section 3 shows the analysis of the Nor’westers and distribution of their associated weather. Section 4 summarizes the conclusions.

2. Methodology

2.1. Study Domain and Datasets

The study domain is in the Gangetic West Bengal of India, 250 km around the radar location at 22.5705° N/88.353° E, as shown in Figure 1. The S-band radar has a frequency of 2.875 GHz with a range resolution of 150 m and a beam width of around 1.0°. The volume scan has 10 elevation angles of 0.2, 1.5, 2.0, 3.0, 4.5, 6.0, 9.0, 12.0, 16.0, and 21.0, with a scan range of 250 km and generates all base moments such as reflectivity, velocity, and spectrum width every 10 min. The radar was in continuous operation between the years of 2013 and 2018, after which the operations were on demand due to technical issues. The volume data from 2013 to 2018 for the pre-monsoon period is used for this study to determine long term characteristics. The feature tracking algorithm used in this study determines the characteristics of the storms by tracking its movement across time scales in a single 2D gridded plane. The volumetric reflectivity data between 0 and 18 km height and a range of 250 km in the polar coordinates is converted into a 2D Cartesian grid with a horizontal resolution of 500 m and a vertical resolution of 1 km. The reflectivity values in the gates are resolved to the Cartesian grid based on the method defined by Doviak and Zrnic [26] assuming standard atmosphere. For each polar radar gate, a radius of influence is calculated based on the distance of the beam from the radar and the beam width. The weighted field values for that gate are added to all grid points within that radius as per Pauley and Wu [27].

2.2. Method of Identification and Classification of Cells

The typical method for identifying convective cells involves using radar reflectivity thresholds, where the cell is identified as convective if it exceeds a specific threshold. This method may not be able to identify an developing cell cluster unless it crosses the threshold reflectivity value. Another method, i.e., the Texture-based method, identifies moderate to weak convective cells that would be missed by a threshold-based method, making it more general in studying convective cloud populations with varying sizes, depths, and intensities. This study uses a modified version of a technique introduced by Steiner et al. [28] to perform a texture-based approach that primarily analyzes radar composite reflectivity (the highest radar reflectivity in a vertical column) and horizontal texture to locate convective cells as demonstrated in PyFLEXTRKR. First, the grid points having reflectivity values greater than 50 dBZ are identified as a convective grid point. Next, for the grid points not identified as convective, the mean background reflectivity for a radius of 11 km of each grid point in a horizontal 2D field is calculated and if the value exceeds 60 dbZ then the grid point is identified as convective. Third, the difference between the reflectivity of each grid point and its background mean reflectivity is calculated. If the difference exceeds a threshold of 10 dB, the grid point is identified as convective. A total of more than 4 square kilometers area from adjoining convective grid points is identified as core convective area. Non core convective area surrounding the convective area is determined based on the Intensity-distance relation of the background mean reflectivity with the grid point. The total area of both core convective and non- core convective areas is considered as a single cell [29].

The parameters to identify the convective cells, such as the background radius, radar reflectivity differences between grid points and the background, and reflectivity thresholds are determined by the data resolution and the characteristics of the convective cells under investigation. Studies have shown that this method of identifying convective cells works well in various conditions, including shallow and isolated cells, intense deep convection, and organized convective clusters embedded in mesoscale convective systems (MCS) with broad stratiform rain areas [30]. The method can distinguish weak and intense cells, preventing contamination from powerful bright bands associated with MCS stratiform rain. The identification of the cells from the radar data can be seen in Figure 2a.

The algorithm can also track the convective cells across the time and provide valuable insights into the cell’s life cycle. The area within the boundary of the convective cell is labeled as a cell mask. A large scale 2D advection is calculated between the two consecutive images by performing a 2D cross-correlation implemented by Padfield [31]. The cell masks from the first image are shifted based on the advection calculated. The advection shifted cell masks from the first image are overlapped on the second image. The overlapping ratio between the cell masks from first and second image are calculated. The cell masks having maximum overlapping ratio are considered as same cell and an tracking number is provided. The process is repeated for consecutive images across the desired time to determine the life of a cell identified by tracking number. The fundamental premise of the overlap technique is that the dataset’s temporal resolution is adequate to discern the spatial movements of the cell and that objects exhibiting considerable overlap between the time stamps can be classified as belonging to the same cell. To enhance tracking, the convective cell masks are progressively dilated outward within a specified radius in increments, commencing with the largest cell in a scene without merging the cells. This enlargement of cells results in larger footprints for the cell masks, enhancing the accuracy of cell tracking. The tracks of different convective cells can be seen in Figure 2b [32].

2.3. Determine the Long Term Storm Characteristics

Various parameters, such as cell initiation times and the lifetime of individual cells, maximum cell convective area, maximum reflectivity in the cells, and echo top heights for 20 dbZ and 50 dbZ, are obtained from the radar data from 2013 to 2018 for the individual cell tracks. A total of 211,503 individual tracks are identified in the six years. The number of tracks that start within each hour of the 24 h (local time) across the day is given in Figure 3a which shows the convection initiation time as a track is identified only if it has a core convective area. The lifetime of all the individual tracks in hours is shown in Figure 3b which is determined by tracking the cell mask across the timeline. The box plot with outliers removed of the cell’s maximum area, cell’s maximum reflectivity, and echo top heights for 20 dBZ and 50 dBZ observed from all the tracks for each of the 24 h a day is shown in Figure 4 and Figure 5 at local time to understand the diurnal variability of the parameters.

For visualization of spatial and temporal distribution of different weather phenomena by the operational meteorologist, the region of radar coverage is classified into 256 sub-domains. The total area, 500 km × 500 km, is divided into 256 blocks, each having a dimension of 30 km × 30 km, with the radar at the center. The number of occurrences of each of the weather phenomena determined based on radar based thresholds in each block for every 3 h period across 24 h is shown in Figure 6, Figure 7 and Figure 8 at local time. It is to be noted that the number of occurrences shown in areas close to the radar [Blocks adjacent to the radar location] are not accurate due to the cone of silence caused by the maximum elevation of 21 degrees in the scan strategy. The uncertainty could not be quantified due to absence of other weather radar in the vicinity.

3. Results and Discussion

3.1. Analysis of Cell Characteristics

3.1.1. Cell Initiation and Duration

Earlier studies using geostationary satellite data have identified peaks in thunderstorm activity during the afternoon/evening hours and early morning in the pre-monsoon season [18,33]. The present analysis also reveals a bimodal diurnal variability in cell initiation; however, the timing of the peaks differs. As observed from weather radar data, peak activity in cell initiation occurs during the early morning and late evening hours as shown in Figure 3a. The number of initiating cells begins to increase from around 15:00 IST, reaching a maximum in the late evening at approximately 22:00 IST. Notably, the frequency of cell initiation is considerably higher during the late evening compared to the afternoon period. Thunderstorm activity decreases slightly after midnight but increases again in the early morning, reaching a peak around 06:00 IST.

A study on the diurnal variability of summer thunderstorms based on manual synoptic observations also reported peak activity late night and early morning over eastern India Sharma et al. [34]. The initiation of convective cells during the night and early morning hours is influenced by valley winds and the presence of hills in the region, as described by Sahu et al. [35]. Since weather radar is a surface-based remote sensing instrument, it can detect developing thunderstorms, including those without precipitation which is similar to what is reported in manual observations. In contrast, satellite-based observations relying on passive sensors are more likely to capture thunderstorms once they have developed sufficiently to produce colder cloud tops, a stage more commonly reached during the afternoon than at other times of the day.

Although Nor’westers tend to originate somewhat randomly, their genesis and propagation can be generalized to a certain extent based on spatial and temporal patterns. These systems are capable of traveling hundreds of kilometers within just a few hours. However, since individual cumulonimbus cells are typically very short-lived, it is not a single storm cell that traverses the entire region. Instead, a sequence of thunderstorms forms successively, with each storm initiating the next in a preferred direction, continuing until the system encounters a stable atmospheric environment. Most convective cells are short-lived; 95 percent have a lifetime of less than three hours, as shown in Figure 3b. Accurately identifying the lifespan of such storms requires instruments with high temporal resolution and appropriate tracking algorithms. While there are limited studies on the lifetime characteristics of Nor’westers using remote sensing data, the findings of the current study are consistent with those reported in the earlier comprehensive satellite-based analysis by Tyagi et al. [18].

3.1.2. Cell Core Convective Area

The core convective area is a critical parameter for assessing the intensity of thunderstorms and their potential to produce severe weather phenomena. The evolution of this core region can serve as an effective indicator for identifying events such as tornadoes and severe lightning activity [36,37,38]. However, studies focusing on the characterization of convective core areas, particularly in the Gangetic West Bengal region, are less. Figure 4a presents a box plot illustrating the diurnal variation in the core convective area of Nor’westers. The maximum area of individual cells ranges from approximately 20 km² to 190 km². The median value throughout the 24 h period remains on the lower end, at around 35 km², reflecting the predominance of smaller storm cells relative to the relatively fewer large ones. This distribution is further evidenced by the mean values, which show a notable decline in core area between 06:00 IST and 09:00 IST coinciding with a reduction in the number of cell initiations.

Larger core areas tend to form when the frequency of cell initiation increases, as smaller cells merge to produce more expansive convective systems. Conversely, the core convective area diminishes in tandem with a decline in new cell formation. During periods of heightened initiation particularly in the early morning and late evening the maximum core area occasionally exceeds 175 km², although the mean values during these times typically remain within the range of 100–125 km². When non-core convective areas are included, the total spatial extent of storm cells can surpass several hundred square kilometers. Importantly, an increase in the maximum core convective area is also associated with enhanced lightning activity, indicating a strong linkage between storm structure and electrical intensity.

3.1.3. Maximum Reflectivity

The maximum reflectivity recorded during the life cycle of a storm cell is one of the most critical parameters for nowcasting. Most studies on the detection of severe weather using weather radar rely heavily on reflectivity as a primary diagnostic tool [20]. This parameter serves as a reliable indicator of the strength of a convective cell during its evolution. A cumulative analysis of maximum reflectivity values from all storm cells over a 24 h period provides insights into the potential peak storm intensities at different times of day, as well as the atmospheric factors influencing them.

Initial observations from Figure 4b reveal no pronounced diurnal variation in the maximum reflectivity across the 24 h cycle. Notably, reflectivity values exceeding 60 dBZ are observed at various times throughout the day, suggesting that intense convection can occur at any hour. The persistence of high reflectivity is maintained by two primary factors: evening and nighttime cell initiation, and deep convection typically occurring in the afternoon. These processes either contribute to the expansion of the core convective area or lead to an increase in storm intensity.

The mean and median values of maximum reflectivity remain relatively stable throughout the diurnal cycle, ranging between approximately 37 dBZ and 43 dBZ. A marginal decline in reflectivity is noted between 09:00 IST and 12:00 IST, corresponding to a temporary decrease in cell initiation, although this is partially offset by surface heating during this time. Overall, these observations underscore the potential for severe convective thunderstorms and associated hazardous weather phenomena to occur at any time of day, even though their frequency and spatial distribution may vary temporally.

3.1.4. Echo Top Height

Echo top height is typically defined as the highest altitude in a radar volume scan where the reflectivity exceeds a specified threshold. When the radar scan strategy is properly configured, this measurement can also provide an estimate of cloud-top height. Higher echo top heights are indicative of stronger updrafts capable of lifting hydrometeors to upper levels of the troposphere. Severe weather phenomena such as hail, turbulence, and lightning are closely associated with elevated echo top heights. In particular, echo tops corresponding to 50 dBZ reflectivity are considered reliable indicators of hail-producing storms [39].

While several studies have explored the variability of echo top heights over the Indian region using space-borne radar, similar analyses utilizing surface-based weather radars remain relatively limited [40]. In this study, the maximum echo top height of each convective cell during its lifetime, along with the hour at which this height was reached, was derived from cell tracking data. Figure 5a presents the distribution of maximum echo top heights for the 20 dBZ threshold across a 24 h period.

The analysis reveals an increase in echo top height between 10:00 and 15:00 IST, coinciding with enhanced surface heating due to solar insolation. This heating supports the development of deeper convection, especially under pre-existing favorable synoptic conditions. Interestingly, even during periods of reduced cell initiation, echo top heights tend to increase, highlighting the role of surface-driven convective enhancement. In contrast, the early morning, evening, and nighttime periods exhibit a marked reduction in echo top height, correlating with diminished solar heating.

During the daytime period from 09:00 to 17:00 IST, the 20 dBZ echo tops frequently reach altitudes between 12 and 18 km, with a mean value of approximately 8 km. Outside these hours, maximum echo tops generally remain below 12 km, with mean values near the melting level, around 4–5 km. The 50 dBZ echo top heights, shown in Figure 5b, follow a similar diurnal trend, exhibiting a mean height of 5 km, with some cells reaching 10–12 km in the afternoon. Given their strong association with hail formation, 50 dBZ echo tops serve as effective thresholds for hailstorm detection and warning.

3.2. Spatial and Temporal Distribution of Severe Weather

3.2.1. Hail

Hail detection using radar echo top heights involves evaluating the vertical extent of radar reflectivity to infer the potential presence of hail within convective systems. Severe thunderstorms are often characterized by strong updrafts that lift water droplets and ice particles to considerable altitudes. Weather radars detect the resulting echoes, and by measuring the altitude at which specific reflectivity thresholds are exceeded, meteorologists can estimate the likelihood of hail occurrence. Echo top heights associated with high reflectivity values serve as effective indicators of hail-producing storms.

Numerous studies have sought to establish quantitative thresholds for hail detection based on reflectivity and echo top height criteria [41,42,43]. In the present study, spatial and temporal variability in hail occurrence is assessed across a 24 h period using a reflectivity threshold of 50 dBZ at the −10 °C height, as illustrated in Figure 6. Radiosonde data were utilized to identify the altitude corresponding to the −10 °C isotherm in the vertical temperature profile during the pre-monsoon season, which is critical for applying this threshold. The radio sonde observations were taken at 05:30 IST and the average altitude of the −10 °C isotherm for the season was taken for the analysis.

The analysis reveals that hailstorm activity is markedly reduced during the morning hours, from 05:30 IST to 14:30 IST, and is largely confined to localized areas influenced by topographic features. Although the first peak in cell initiation, shown in Figure 4, occurs between 05:30 IST and 08:30 IST, widespread hail occurrences are not observed during this period. Hail formation in the morning is limited to regions where orographic lifting enhances convection sufficiently to reach hail-producing altitudes.

Hail occurrence shows a substantial increase after 14:30 IST, corresponding with enhanced surface heating, which promotes higher echo top heights and more vigorous convective development. During this period, the south western and Northern sectors of the radar coverage area exhibit the most active hail regions. The increase in numbers are due to local activity as well as the movement of convective cells from the western region to the eastern region. As the evening progresses, hail signatures begin to reduce around 20:30 IST, with less activity during the night time. The analysis also reveals a notably low frequency of hail over the southern and southeastern oceanic regions, likely due to weaker convective activity over water surfaces.

The spatial and temporal patterns identified in this study are consistent with findings from previous research. The study by Sharma and Sen Roy [44], based on synoptic observations, also identifies the same regions as prone to hail during the summer season. Similarly, the study by Sama et al. [23], using 32 radar-observed events from Kolkata, report that peak hail activity occurs between 14:30 IST and 17:30 IST.

The present study builds upon these earlier findings by leveraging a long-term radar dataset to identify not only the timing but also the specific spatial locations of hail activity across different hours of the day. This provides a more granular understanding of hailstorm dynamics in the region, thereby enhancing the potential for effective nowcasting and localized hazard preparedness.

3.2.2. Lightning

Lightning detection using weather radar is typically indirect, relying on radar-derived thresholds associated with electrification processes within convective storms. The occurrence of hail and lightning is governed by closely linked microphysical processes, particularly involving strong updrafts that lift liquid and ice hydrometeors to elevated levels within the storm. While dual-polarization radar provides enhanced capability for classifying hydrometeor types and identifying potential lightning zones, single-polarization radars primarily depend on parameters such as reflectivity and echo top height for inferring lightning activity.

In this study, a reflectivity threshold of 40 dBZ at the −10 °C level is used to estimate lightning occurrence in the spatial domain, as illustrated in Figure 7 [45,46,47]. The altitude of the −10 °C isotherm was determined using radiosonde observations, as previously described in Section 3.2.1 on hail detection.

Lightning activity is observed to be relatively low between 08:30 IST and 14:30 IST, after which it increases significantly, particularly over the Chota-Nagpur Plateau. The lightning becomes progressively more widespread, especially across the western and northern sectors of the radar coverage area. The intensity of lightning peaks during the evening hours, notably in the storm initiation zones and along the propagation paths of convective cells. The formation of new cells plays a critical role in modulating lightning activity, with peak lightning occurrences coinciding with periods of enhanced convective initiation. In particular, early morning lightning is more pronounced in the northern and North Eastern regions, associated with storms triggered near the foothills of the Himalayas, which subsequently moved into the radar domain.

These findings are consistent with the results reported by Mishra et al. [48], who analyzed two decades of lightning flash cluster data over West Bengal, and with the study by Midya et al. [49], both of which demonstrate that lightning flash frequency is highest between late afternoon and early evening, and lowest from morning to midday. The present study complements previous findings by identifying an additional early morning peak in lightning activity, which is associated with enhanced cell initiation in the northern parts of the region.

3.2.3. Convective Rain

The spatial distribution of convective rainfall in this study is determined using a composite reflectivity threshold of greater than 45 dBZ, a commonly adopted indicator of intense convective precipitation. The observed spatial pattern closely mirrors that of lightning occurrence, suggesting that convective rainfall is frequently accompanied by lightning, except in isolated instances where precipitation occurs in the absence of electrical activity. Notably, the extent of convective rainfall distribution is broader than that of hail and lightning, which can be attributed to the differing sensitivities of the thresholds used. While the thresholds for hail and lightning are more indicative of core convective regions, the threshold applied for convective rain captures both core and non-core convective areas, leading to a more extensive spatial coverage.

Rainfall activity exhibits a distinct bimodal distribution, with prominent peaks during the early morning hours (05:30 IST–08:30 IST) and late evening (20:30 IST–23:30 IST). In contrast, rainfall is relatively suppressed during the mid-morning to early afternoon period (08:30 IST–14:30 IST) across most parts of the domain. Beginning in the afternoon hours (14:30 IST–17:30 IST), rainfall activity intensifies over the southwestern sector, gradually expanding across the region during the evening hours (17:30 IST–20:30 IST), and reaching a maximum in the late evening, as illustrated in Figure 8. During the nighttime hours, convective rainfall becomes relatively subdued when compared to the preceding late evening peak and the subsequent early morning resurgence.

The movement of thunderstorms (Nor’westers) and their activity throughout the day over the Gangetic West Bengal region for the pre-monsoon season of 2014 is verified with the LIS/OTD climatology datasets. The combined flash rate in LIS/OTD 2.5° Low-Resolution Diurnal Climatology (LRDC) data contains gridded climatologies of hourly average rate of total lightning flash rates obtained from two lightning detection sensors—the space-borne Optical Transient Detector (OTD) on Orbview-1 and the Lightning Imaging Sensor (LIS) onboard the Tropical Rainfall Measuring Mission (TRMM) satellite [50].

The climatology data set has a resolution of 2.5° × 2.5°, while the spatial resolution used in this study for analysis is approximately 0.25° × 0.25°. The temporal resolution of the climatology dataset is hourly and the resolution used in this study is every 10 min of radar scanning. As the data sets cannot be compared directly for verification due to significant differences in the spatial and temporal resolution, the general aspects of thunderstorm activity throughout the day can be verified by the combined flash rates in the LIS/OTD datasets as shown in the Figure 9. The diurnal climatology shows significant thunderstorm activity around 15:30 IST, which peaks at 19:30 IST and gradually reduces around midnight. The thunderstorm activity is minimal in the early morning to noon time. The intensity is higher in west and south west regions in the afternoon and the eastern region in late evening and night times. The findings in this study also indicate a similar pattern that the possibility of severe weather is at the maximum between afternoon and midnight distributed among Hail, lightning and convective rain throughout the region as the storm moves from west to east. The high resolution data sets from radar used in this study has provided a in depth view of the spatial and temporal variations which resembles the general pattern of variations observed from an independent climatological dataset.

4. Summary and Conclusions

The archived Doppler weather radar of Kolkata from 2013 to 2018 is used to determine the storm characteristics using latest feature tracking algorithm (PyFLEXTRKR) and different radar observed parameters were obtained from 211,503 individual cell tracks. This large dataset could be used for further studies as well as machine learning applications.

The examination of the characteristics from the parameters reveal a bimodal diurnal pattern in storm initiation, with peak activity occurring in the early morning and late evening. A large majority of the convective cells (95 percent) have a lifetime of less than 3 h. Though the general diurnal behavior of the cells has been studied earlier, the studies on the radar observed characteristics have been very limited over the Indian region. The present study fills the gap by identifying their variability.

The size of the maximum core convective area is closely linked to the number of cell initiations. As cell initiations increase, the core area expands due to the merging of smaller cells, mirroring the diurnal cycle of cell initiations. Further findings indicate that maximum reflectivity fluctuates along with cell initiation rates during morning, evening, and nighttime, but afternoon convection driven by surface heating can obscure this relationship. Notably, the height of echo tops is more significantly influenced by deep convection during midday than by the number of cell initiations, as maximum heights are often observed when cell initiations are at their lowest.

Hail occurrences are also predominantly driven by deep convection related to afternoon surface heating rather than early morning or late evening cell initiations. Additionally, lightning events rise markedly during peak cell initiation periods in late evening and early morning while still being affected by afternoon insolation. Convective rain generally coincides with lightning but has a widespread presence than lightning. Morning to early afternoon hours have relative less occurrence of severe weather associated with Nor’westers than at other time of the day. The spatial distribution of the severe weather activity also varies based on the weather phenomena and time of the day.

There is one significant finding in the study which needs to be further investigated. The peak convective activity observed in this study during the early morning hours were identified using satellite observations by Tyagi et al. [18] mentioning that the Type-B Nor’westers is a major contributor to the thunderstorm activity in the morning. Study by Karmakar [51] finds that copious rainfall without hailstorms is associated with Type-B early morning Nor’westers. The present study also notes that there is a peak convective activity in the early morning with significant rainfall and reduced hail activity. However, the increased lightning activity during early morning is not reflected in the climatology datasets. More investigation is needed to characterize the severe weather caused by peak activity of the Nor’westers in the early morning as synoptic observations also support thunderstorm activity over the region during the time as per the study by Sharma et al. [34]. Hydrometeor classification studies of these storms using a dual polarization weather radar can provide valuable insights on the behavior of the Nor’westers.