Shifting seasonality of cyclones and western boundary current interactions in Bay of Bengal as observed during Amphan and Fani

In recent years, the seasonal patterns of Tropical Cyclones (TC) in the Bay of Bengal have been shifting. While tropical depressions have been common in March–May (spring), they typically have been relatively weaker than the TCs during October–December. Here we show that the spatial pattern of recent warming trends during the last two decades in the southwestern Bay has allowed for stronger springtime pre-monsoon cyclones such as Amphan (May 2020, Super Cyclone) and Fani (April–May 2019, Extremely Severe Cyclone). The tracks of the pre-monsoon cyclones shifted westward, concurrent with an increasing rate of warming. This shift allowed both Fani and Amphan tracks to cross the northeastward warm Western Boundary Current (WBC) and associated warm anti-cyclonic eddies, while the weaker Viyaru (April 2013, Cyclonic Storm) did not interact with the WBC. A quantitative model linking the available along-track heat potential to cyclone’s intensity is developed to understand the impact of the WBC on cyclone intensification. The influence of the warming WBC and associated anti-cyclonic eddies will likely result in much stronger springtime TCs becoming relatively common in the future.

Tropical Cyclones are one of the most devastating natural disasters, especially over coastal regions, due to impacts on densely populated low-lying areas and conditions over shallow continental shelves that strengthen Tropical Cyclone intensity 1 . In recent years, the increasing trend in the intensification rates of Tropical Cyclones has been observed on a global scale 2 .
The Bay of Bengal, a unique tropical ocean basin in the northern Indian Ocean, is a potentially active region for the genesis of Tropical Cyclones. Formation of Tropical Cyclones over this region is seasonal with a primary peak formation during the post-monsoon season (October-December) and a secondary peak formation during the pre-monsoon season (March-May) 3 . The role of atmospheric parameters and sea surface temperature (SST) in cyclone formation and intensification are well established from the satellite and available in-situ observations during the different cyclonic events [4][5][6][7][8][9][10] . The SST, the dynamic topography, and the presence of oceanic eddies at the surface all have a significant contribution to Tropical Cyclone intensification 5 . Surface eddies are strongly related to the variability of winds and sea surface height anomalies (SSHA) in the Bay of Bengal 11,12 . Warm-core eddies (with positive SSHA) have a deeper thermocline, which supplies significant heat for cyclone intensification. In contrast, cold-core eddies (with negative SSHA) weaken cyclone intensity due to a shallower thermocline and significantly less heat content 11,12 .
The oceanic circulation in the spring or pre-monsoon season is dominated by the northward flowing warm western boundary current (WBC) and its eddies 13,14 . Formation of the WBC is due to the anti-cyclonic wind gyre that develops in November and continues through April-May by the integrated wind stress curl or Ekman pumping 15,16 . The WBC propagates northward from 10 to 17° N along the edge of the continental shelf and then separates at around 18° N and flows eastward 13 with multiple eddies (Fig. 1a). This springtime WBC is observed 50-100 km away from the coast with a width of 200-300 km and extending down to depth of 250-450 m 17,18 Shifting seasonality (stronger pre-monsoon cyclones due to warming SST trends). During 1982-2020, a total of 77 cyclones formed over the Bay of Bengal with landfall in India and the Bangladesh coast (source: IMD cyclone e-atlas 32 ). Of these, 19 (25%) formed in the pre-monsoon (March-May) season and 58 in the post-monsoon (October-December) season. The formation locations and subsequent tracks of the 8 premonsoon cyclones prior to 2000 and 10 pre-monsoon cyclones after 2000 are shown in Fig. 1b,c. Note that one pre-monsoon cyclone in 1989 could not be considered due to unavailability of its data. The IMD-recognized sub-regions are also shown in Figs. 1b-d. During 1982During -2000, most of the pre-monsoon cyclones formed over the east-central (ECB) and southeastern (SEB) Bay of Bengal (Fig. 1b). Six out of eight tracks passed through the ECB and northern Bay of Bengal. Therefore, they did not interact with the western boundary current, which extends to 88 • E in the northern bay between 18 and 20° N and is limited to 85 • E between 8 and 10 • N . In this study, we consider 0.15 m of SSHA to represent the western edge of the WBC (Fig. 1a and subsequent figures). All of these cyclones made landfall to the north of 20 • N, except one in 1990, which made landfall at around 16 • N and crossed the WBC, and became a super cyclone.
Most of the subsequent pre-monsoon cyclone tracks have exhibited a distinct westward shift during 2001-2020 (Fig. 1c) compared to their 1982-2000 path. In the recent decades, six out of ten cyclones moved through the west-central bay B) and seven through northwestern bay (NWB). Therefore, the chance of their interaction with the warm WBC associated with positive sea surface height has increased. The mean track during 1982-2000 starts near the central part of SEB and makes landfall over Bangladesh (Fig. 1c). The dashed black line shows the mean track while including the anomalous 1990 cyclone, which crossed over the WBC region and became a super cyclone. The mean longitude and latitude of the genesis location in the earlier decades (1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000) was 89.6 • E, 11.6 • N, and in the recent decades (2001-2020), it is 87.9 • E, 13.9 • N. Comparison of the mean tracks shows that the westward shifting is maximum in the central Bay of Bengal. Additionally, this westward shifting of tracks is more than one standard deviation (green dash line, Fig. 1d) of the 1982-2000 path in the southern and central Bay of Bengal. The current intensity (CI), a measure of the instantaneous cyclone strength (see "Methods" section) showed a downward trend from 1982 to 2000 and an upward trend after 2000 (green bars, Fig. 1e). The CI varies from 1.5 to 7 as provided by IMD best track data 33,34 (see "Methods" section). Together with the comparison of mean formation region and subsequent mean tracks in the two periods, it is www.nature.com/scientificreports/ evident that there has been a significant westward shift in the genesis location and the post-2000 pre-monsoon cyclones are moving through a more westward path which increases their chances of interaction with the warm springtime WBC in the Bay of Bengal. The possible impact of SST and its trends have been shown to affect the intensity of cyclones in recent studies in the Indo-Pacific region 35,36 . To understand this relationship between the formation regions and the tracks with the SST, the trends of SST over the Bay of Bengal were computed for different seasons using the Hadley SST 37 (see "Methods" section). The SST in the WBC region is generally 1-2 °C warmer than the central basin associated with the positive SSHA (Fig. 1a) and anticyclonic eddies 13,14 . The SST trends in April and May are positive in the east-central (during 1982-2000) and southern regions (during 2001-2020) of the Bay of Bengal. A student's two-tail t-test was performed which identified the SST trends with significance level more than 90%  36 (see their Fig. 7) for the Indo-Pacific tropical region. The formation regions of Tropical Cyclones in the last two decades are aligned with the maximum SST trend region (0.02-0.04 °C per year, 95% significant) depicted by the southeast-to-northwest elongated core structure (4-12° N, Fig. 2) in the southeastern Bay of Bengal, where the recent Tropical Cyclones Viyaru, Fani and Amphan were all formed. Interestingly, the northern part of this core is only observed in the May trend (Fig. 2) in contrast to the April-May trend (Fig. 1c). Note that the Amphan cyclone rapidly intensified into a super cyclone while passing over the northern part of this core. In 24 h (17th May 0600 UTC to 18th May 0600), it became a super cyclonic storm (13.4 • N , 86.2 • E ) from severe cyclonic storm (11.5 • N , 86 • E ). On the other hand, cyclone Fani, which formed in late April, did not get the benefit of passing through the northern core structure, and was limited to a weaker ESCS status (albeit, similarly devastating to Amphan due to its inshore track) before landfall. The SST trend in the southern Bay of Bengal (6-11 • N , 88-93 • E ) also showed a remarkable warming trend of 0.033 • C per year from 2001 to 2020, which was almost negligible from 1982 to 2000. The changes in the SST in last two decades also match with the increasing changes of the CI of the cyclones (Fig. 1d).
Quantification of the impact of the WBC and its associated eddies on the cyclone intensity. While the answer to the first question appears to be affirmative (SST warming trend is partially and significantly responsible for relatively westward springtime cyclone genesis in recent two decades), the answer www.nature.com/scientificreports/ to the second question lies in the possibility of the cyclone's path crossing the WBC and its anticyclonic eddies, thereby gathering energy. This interaction has happened for Fani and Amphan, and not for Viyaru. A simple quantitative model of computing the "along-Track Available Potential Heat" (TRAPH) or the heat potential under the footprint of the evolving cyclone's stronger wind region is developed first. Relating this TRAPH with the current intensity (CI) will then be used to examine the relationship between the ocean heat content and the evolution of the cyclone intensity. The TRAPH is calculated in four distinct steps. First, wind speed around the center of the cyclone is analyzed to obtain a typical radius of influence for the cyclone's energy. It is typically considered as the distance from the center where the wind speed reaches its maximum and then decreases. For all three cyclones, this distance was determined from the zonal and meridional variation of the wind speed along the radials from the center of the cyclone. However, as the cyclone evolves along its path, the wind speeds are not radially symmetric. There are patches of stronger winds, or footprints of the high-energy bands of the cyclone. These are caused by complex non-linear processes of multi-layer atmospheric flow, driven by advection, convection, precipitation and interaction with the ocean 39,40 . Thus, in the second step, we choose a threshold wind speed, where the maximum sustainable winds occur. The threshold values are considered close to the 90th percentile of the winds over the Bay of Bengal for three cyclones. See "Methods" section for details on threshold calculation. This threshold wind speed was chosen as 15 m s −1 for the SuCS (Amphan), and 12 m s −1 for ESCS (Fani) cyclones. For Viyaru, which reached only to the stage of cyclonic storm, this speed was chosen as 10 m s −1 . In the third step, the region within the radius of influence that had wind speeds more than the threshold was demarcated as the "energy gathering footprint" of the cyclone. This is the effective region within the cyclone where the winds gather energy from the ocean. In the fourth and final step, the TRAPH is determined as the oceanic heat content within the active footprint of the cyclone using the SST distribution from the first day of the cyclone initiation as follows: where C p is the specific heat (4.2 × 10 3 J kg −1 K −1 ), is the sea water density (1025 kg m −3 ), H is the mixed-layer depth, and T x, y is the SST distribution. H is assumed to be constant at 50 m 41-43 for this simple calculation. Available Argo profile (WMO: 2902283) in central Bay of Bengal around 84.3 • E, 14.85° N during May 2020 also confirmed a mixed-layer depth of 50-60 m. The area integral is performed for the footprint region only. TRAPH is interpolated at 6-hourly intervals from the daily SST and wind fields obtained from satellites (see "Methods" section for sources). probably losing some of its heat. The tropical cyclone Viyaru veered to the right and missed interacting with the WBC eddies to its west. Also note that the WBC and its eddies were situated further to the east during 2019 and 2020 (88 • E) than in 2013 (90 • E). This allowed the recent cyclones (Fani and Amphan) to gather more heat and strengthen while passing over the eddies. On the other hand, Viyaru did not interact with the WBC and its eddies in 2013 and could not strengthen beyond the stage of CS.
Eddy passage by a Tropical Cyclone is one of the most crucial parameters determining the evolution of the storm intensity. Cyclones intensify after passing over an anticyclonic (warm) eddy with positive SSHA, and dissipate when it is crossing over a cyclonic (cold) eddy with negative SSHA 4,5 . Climatologically, during the pre-monsoon season (April-May), a strong anti-cyclonic gyre dominates over the western Bay of Bengal and extends from 7 to 18° N 12,17 . Both Amphan (Fig. 3a-e) and Fani (Fig. 4a-e) passed close to strong positive SSHA anomalies and had generally northward tracks. However, Viyaru (Fig. 5a-e) curved to the northeast near 14° N, moving the Tropical Cyclone well away from the strong positive SSHA near the continental shelf. The proximity to the WBC and anti-cyclonic eddy likely boosted the intensity of Amphan and Fani, while the northeastward track of Viyaru likely was the primary factor limiting its intensity to CS.
It is evident from the three bottom-right panels of Figs. 3, 4 and 5 that the TRAPH follows and leads the current intensity (CI) reasonably well for all three cyclones. For Fani and Amphan, this quantitative pattern provides support for the hypothesis that indeed these two severe cyclones reached their peak after interacting with the WBC and its associated anticyclonic warm eddies. Figure 5 also indicates that since Viyaru did not interact with the WBC and its eddies, it did not gather strength after becoming a CS in the first 2-3 days and maintained its stage of CS throughout its journey.
The synoptic distribution of GHRSST (Global High Resolution Sea Surface Temperature) for all three cyclones are shown in the color background field of Figs. 3, 4 and 5. During the passage of Fani in April-May 2019, the northward propagating WBC was strong, extending northward to 18 • N , and then bifurcating to form the cyclonic eddy near 18 • N over the northern Bay of Bengal (Fig. 4). On 30th April 2019, the center of Fani was over a warm temperature region (> 30.8 °C) in the west-central Bay of Bengal (Fig. 4b). While temperatures generally decreased to the north, the presence of an anti-cyclonic eddy kept temperatures warmer in the western Bay of Bengal (Fig. 4a-e). Note that the area of the higher SST (for SST > 30.8 ℃) is much larger in case of Amphan than that of Fani. In contrast, during the passage of Viyaru in May 2013, the WBC only extended to 16 • N before moving northeastward away from the coast (Fig. 5; see contour of SSHA = 0.15 m). www.nature.com/scientificreports/   Fig. 3a- www.nature.com/scientificreports/ Similarly, for Viyaru, initially, SST was higher in the central Bay of Bengal (Fig. 5a,b) and SST decreased with the northward movement of the cyclone center (Fig. 5c-e). However, because of the much more southerly detachment point of the western boundary current, Viyaru encountered substantially cooler waters much further south than Fani encountered. Note that the decline of TRAPH for all three cyclones nearing the end of the time-series in Figs. 3f, 4f and 5f is a signature of the post-storm cooling, which is a function of the rapidly diminishing CI of the cyclone 44,45 .
Historically , and dissipated after landfall as a tropical cyclone. In this study, we showed that all three cyclones were formed in the southeastern Bay, which is rare during 1982-2020. This is related to an increase of a significant SST trend in May (0.02-0.04 °C per year) in that region, possibly due to excessive heat storage during winter and spring in recent years (which needs to be investigated in the future). Favorable atmospheric conditions interacting with the strong northward WBC and its eddies during April-May 2019 and May 2020 might also have helped to intensify both Fani and Amphan 31 , which will require further investigation including coupled numerical model experiments. A quantitative measure of along-Track Available Heat Potential (TRAPH) shows the peaking of cyclone intensity along the path over the WBC and its eddies for Fani and Amphan, which was missing for Viyaru. Our analysis suggests that more frequent occurrence of such high-intensity pre-monsoon cyclones (Fig. 1d) in the Bay of Bengal might be the new normal in the Bay of Bengal in a warming future. This shifting seasonality obviously has deep implications for public safety and hazards, agriculture, and maritime activities in the future.
Finally, it is worth mentioning a couple of possible worldwide implications of our findings. Similar processes can lead to future intensification of north Atlantic storms passing over the Gulf Stream and its warm core rings. Recent observational studies have shown that the warm core ring formation over the Gulf Stream region (75-55° W, 35-45° N) increased from an average of 18 per year during (1980)(1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999) to 33 per year during (2000-2017) 47,48 . The formation of these warm core rings reaches its peak in August. The implication of such abundance of warm rings at the beginning of the Atlantic hurricane season is unknown currently. Similar implications may apply to the Pacific typhoons crossing the Kuroshio. In addition, Yang et al. 49 documented poleward shifts of Western Boundary Currents and projected northward shifts and increased poleward heat transport in 2050 and beyond from CMIP6 downscaling experiments 50 . Our findings (and the quantitative approach presented herein), supported by additional refinement with high-resolution data and coupled ocean-atmosphere modeling systems could contribute towards better predictability of the propagation paths of these cyclones and their subsequent amplitude manifestation along-track. In 2021, the VSCS cyclone Yaas formed in the northern Bay of Bengal (16.3° N 89.7° E) in the warm pool region in the northern Bay of Bengal near 18° N (Fig. 2). The formation location  www.nature.com/scientificreports/ was exceptionally warm (by about 2 °C above climatological value), and the path of Yaas went over a large warm eddy area. However, according to IMD report on Yaas 51 , this cyclone never went through rapid intensification during its lifetime, indicating that atmospheric parameters, mainly the wind shear, might have hindered the rapid intensification of this cyclone despite very favorable SSTs. Further studies on this latest cyclone and such competition between atmospheric and oceanic roles in cyclone intensification are needed. To conclude, the evolving extreme severity of hurricanes, typhoons and cyclones can be aided by the warming Western Boundary Currents and their warm eddies in the future climate.

Methods
Data (cyclone, SST, SSH, winds). For the current study, the cyclone tracks, intensity, maximum sustained winds speed, estimated central pressure, and central pressure drops for all cyclones were obtained from the India Meteorological Department (IMD) datasets (https:// rsmcn ewdel hi. imd. gov. in/). The inter-annual trends of SST were computed from the UK Met Office Hadley Centre's Sea ice and SST data set 37 (source: https:// clima tedat aguide. ucar. edu/ clima te-data/ sst-data-hadis st-v11). Trend (°C/year) is calculated at each grid point over the domain and average is calculated. We have performed the student's two-tailed t-test and identified the regions with a significance level of more than 95% (90%) with p-value less than 0.05 (0.01).
Daily SST data during cyclones were collected from the Global Higher Resolution Sea Surface Temperature (GHRSST) with spatial resolution 1/20° during the season of the cyclones (source: http:// apdrc. soest. hawaii. edu/ las/ v6/ const rain? var= 12679). Altimeter-derived daily SSHA fields of 1/4° spatial resolution, were extracted from the satellite-derived Archiving Validation and Interpretation of Satellite Oceanographic (AVISO) datasets (source: http:// marine. coper nicus. eu/). Scatterometer-derived 25-km winds were obtained from SCATSat-1 for  (Fig. 1) Calculation of threshold wind speed for TRAPH. The wind speed distribution within a cyclone is not radially symmetric. Typically, the winds within a cyclone are thought to reach maximum speed some distance away from the center and then weaken out. However, there are patches of stronger winds, or footprints of the high-energy bands within the cyclone before reaching the maximum wind and afterwards and these patches are distributed within the cyclone area depending upon the ocean-atmosphere feedback and energy conversion. It is thus important to identify this 'energy gathering footprint' of the cyclone. This is the effective region within the cyclone where the winds gather energy from the ocean. We designed a 'threshold wind speed' for a cyclone from the basin-wide wind distribution during the cyclone passage and calculated the area covered by the winds bound by the threshold on either side of the maximum wind speed to determine the 'energy gathering footprint' of the cyclone. The threshold value also helps capture the relative strength of the cyclone and its growth compared to its associated surrounding large scale wind field over the Bay of Bengal, which differs for different cyclones. www.nature.com/scientificreports/ The spatial distributions of winds for 19th May 2020 (Amphan), 1st May 2019 (Fani), and 14th May 2013 (Fani) show the asymmetries in the winds around the center of the cyclone (Fig. 6a-c). Amphan showed the maximum wind speed on the southwest side of the center (Fig. 6a), whereas it is on the southeast for Fani (Fig. 6b). The winds are much weaker for Viyaru, however, the maximum winds are seen on the southeast of the center. Therefore, the histogram of the winds over the whole of Bay of Bengal is analyzed for identifying the threshold value of the winds with higher magnitude. A single-peak modal distribution was found for Amphan with magnitude more 10 m s −1 in the whole Bay of Bengal, indicating the impact on the larger area. For Fani, there are two peaks. The peak with lower wind magnitude around 5 m s −1 indicates the ambient seasonal winds. The other peak with higher magnitude of 10 m s −1 showed the influence of cyclone. In case of Viyaru, there is also a single peak with wind magnitude of 7 m s −1 showed the impact area is much less and mostly influenced by the seasonal winds.
The distribution of winds over the Bay of Bengal for the three cyclones on selected days are shown in Fig. 6d-f. Our approach was to choose the threshold wind speed by using a percentile categorization of the observed wind distribution. A sensitivity test was carried out with different percentile limits for determining the threshold. It was clear that the threshold defined by the 90th percentile wind speed captured most of the energetic footprint around the center similar to Figs. 3, 4 and 5 and as shown in Fig. 6a-c. The footprint signature with wind speeds given by the 80th or 70th percentile captured more wind patches outside and away from the cyclone area, which was rendered undesirable. Thus, we chose the 90th percentile values of the wind speeds (Amphan 15.7 m s −1 , Fani 11.8 m s −1 and Viyaru 10.6 m s −1 ) as representative of the high wind impact areas for all three cyclones. Therefore, wind thresholds are taken as 15, 12 and 10 m s −1 , respectively. The grids with wind speeds over the threshold values are identified, and the SST values at those grids are used for the calculation of the TRAPH.
Finally, note that this approach has at least three important attributes: (i) it appreciates the importance of stronger patchiness within the cyclone; (ii) it eliminates the weaker winds near the center low; and (iii) it captures the areas of stronger winds beyond the maximum wind radius as well.

Data availability
The datasets generated during and/or analyzed during the current study are also available from the corresponding author on reasonable request.