Introduction

Tropical cyclones (TCs) rank among the most disastrous weather phenomenon globally. The northern Atlantic Ocean (NATL) is a hot spot in the Northern Hemisphere. The TCs over the NATL impact the Mexico, United States, the Gulf and Atlantic coasts, and the Caribbean islands, inducing significant loss of life and damage of property1,2,3. Consequently, understanding the mechanisms driving the TC genesis and its long-term variation over the NATL is of critical importance4.

Prior research has documented observed trends in TCs worldwide including the NATL using various metrics5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20. Those metrics measure the TC formation, best-track and intensity, the latitude at which TCs reach their peak intensity, as well as the damage associated with TCs3,14,21,22. For instance, Klotzbach et al.14 found a significant upward trend in both short-lived (<2 days) and long-lived (>2 days) tropical storms in the NATL. A distinct spatial pattern of increasing tendency of TC occurrence frequency is also found over the NATL since 198022. This increase is partly due to the decreasing effect of volcanic eruptions and anthropogenic aerosols23,24. Kossin et al.21 have discerned a marked upward trend in the proportion of major hurricane (maximum wind speed > 96 kts) using a satellite-based TC intensity dataset (ADT-HURSAT). Consequently, the TC-related damage in the United States has displayed a tendency of increase3,25.

Future projection studies have explored the potential impacts of anthropogenic global warming on TCs over the NATL9,10,26,27,28,29,30,31. Numerous studies, utilizing diverse climate models, have consistently revealed a decreasing trend in the global number of TC genesis events, including those over the NATL, in response to greenhouse warming29,32,33,34. Recently, employing high-resolution d4PDF simulations and incorporating four distinct warming scenarios, Cao et al.30 uncovered a pronounced poleward displacement of the peak TC genesis area over the NATL, transitioning from 10°-15°N latitudes to 15°-20°N latitudes with the intensification of global warming.

Previous studies have reported a poleward migration of the lifetime maximum intensity (LMI) of TCs at the global scale in the last three decades except for the NATL11,13. Over the NATL, the LMI of TCs shows a trend of southward shift, but the trend is statistically insignificant. Daloz and Camargo20 further investigated whether the poleward displacement of TC LMI is linked to a poleward displacement of TC genesis. They found a poleward shift of TC genesis toward higher latitudes over the Pacific Ocean basins. In contrast, the meridional movement of TC genesis is weak over the NATL. Tennille and Ellis35 asserted that the linear trend of TC LMI within the NATL basin may be highly sensitive to the selection of the analysis period. This raises one question of whether the latitudinal migration of TC genesis over the NATL is dependent on the TC intensity. Our observational analysis demonstrates a robust southward shift in the hurricane genesis location over the NATL in the satellite era during which the data are more reliable. The main objective of this study is to delve into the fundamental physical processes that underpin the southward migration in the genesis location of NATL hurricane during active TC seasons (from June to November, JJASON) from 1979 to 2022.

Results

Southward shift of the NATL hurricane genesis

Previous studies have found a tendency of southward shift in TC genesis over the NATL utilizing IBTrACS dataset, but the trend is insignificant from 1980 to 201320. One issue is whether the latitudinal migration of NATL TC genesis depends on the TC intensity. Thus, we examine the linear tendency of TC genesis location for TCs with different intensities. The regional average position of NATL TCs is determined by computing the mean latitude and longitude of TCs during JJASON period that attain the initial intensities of 34 kts, 64 kts, and LMI (at least hurricane intensity), respectively. Figure 1a shows a linear trend of the JJASON-mean latitude and longitude of all TCs with the maximum wind speed exceeding 34 kts. The linear trend of latitude and longitude of all TCs is insignificant (Fig. 1a, b), which is in line with previous study20. However, when we only consider the TCs that later reach the hurricane intensity (TY), there is a significant southward shift in the genesis location of hurricane, exhibiting a linear trend of 0.114° latitude/year since 1979 (Fig. 1c and Supplementary Fig. 1b). Meantime, the mean longitude of hurricane genesis location displays a significant eastward shift with a linear trend of 0.218°/year (Fig. 1d). Furthermore, the latitudinal locations of TCs when they reach the hurricane intensity and the LMI both display a significant southward shift (Fig. 1e, g), consistent with the hurricane genesis location. However, the longitudinal locations of TCs with hurricane intensity and their LMI are insignificant (Fig. 1f, h). It indicates that the change in the latitude of TCs with at least hurricane intensity shows a stationary feature from genesis to LMI.

Fig. 1: The temporal evolution of TC genesis and intensity locations over the NATL.
figure 1

The evolution of time series of the JJASON-mean a latitude (°) and b longitude (°) of all TC geneses over the NATL. c, d Same as (a, b) except for genesis locations (°) of those TCs that later reach the hurricane intensity. e, f Same as (c, d) except for locations (°) of those TCs when they reach the hurricane intensity. g, h Same as (e, f) except for the LMI location (°) of those TCs that later reach the hurricane intensity. The red line depicts the linear tendency and the text above indicates the significance level and the slope of linear trend.

Full size image

The trend of TC genesis position may vary depending on the region of TC formation17,36. Cao et al.17 detected a remarkable westward movement in the longitudinal location of TC genesis over the southern Indian Ocean, which is intimately tied to a rise in TC formation number in the southwestern section, coupled with a corresponding fall in the northeastern section. Therefore, we segregate the whole NATL region into two separated regions utilizing the latitude of 20°N as a dividing line. The southward migrating tendency in hurricane genesis over the NATL aligns with a linear upward trend in the number of hurricane genesis in the southern sector (10°-20°N) of the NATL, significant at the 99% confidence level (Supplementary Fig. 1c). The hurricane genesis number in the northern part (20°-50°N) of the NATL displays a decreasing trend, which, however, is not significant (Supplementary Fig. 1e). We repeat the analysis using the latitude of 25°N as a dividing line and obtain similar results (Supplementary Fig. 1d, f). We further examine the meridional shift in hurricane genesis in the TC-active region of 10°-30°N. The result is still valid (Supplementary Fig. 2). This indicates that the southward shift of hurricane genesis location over the NATL is primarily attributed to the upward trend of the hurricane genesis frequency in the southern portion of the NATL.

Physical mechanisms underlying the southward migration of NATL hurricane genesis

Based on the above analysis, one question is what causes the increase in the hurricane genesis number over the southern region (10°-20°N) of the NATL. We embark on an analysis of changes in the large-scale parameters that are intimately intertwined with the TC genesis2,4,37,38,39,40. First, we examine the difference of dynamical and thermodynamical fields during JJASON, including 850-hPa and 200-hPa winds, 500-hPa vertical p-velocity, 200–850-hPa vertical wind shear, sea surface temperature (SST), and 700-hPa specific humidity between 2001–2022 and 1979–2000 (Supplementary Figs. 3, 4). The linear trends of those variables display features similar to the difference fields.

The interdecadal difference of winds shows westerly and easterly in the region of 10°-20°N over the NATL in the lower and upper levels, respectively (Supplementary Fig. 3a, b). This leads to a decrease in total vertical wind shear between 10° and 20°N (Supplementary Fig. 3d), which corresponds to the increase of hurricane genesis in this region. The interdecadal difference and linear trend in total vertical wind shear are predominantly driven by variations in zonal wind components (Supplementary Fig. 4). The easterly wind shear difference is against climatological mean westerly vertical wind shear, which reduces vertical wind shear in the southern part of the NATL. The reduced vertical wind shear over the NATL is beneficial to more hurricane genesis there. The contribution from the mid-level vertical motion change is small for hurricane genesis (Supplementary Fig. 3c). The SST and mid-level specific humidity display a uniform increase over most regions of the NATL, particularly the extratropical region (Supplementary Fig. 3e, f). The ocean heat content shows an increase over most regions of the NATL as well (figure not shown). Therefore, they cannot directly explain why the increase of hurricane genesis number occurs along 10°-20°N. It indicates that the dynamical factor dominates the long-term tendency of the hurricane formation number within the NATL.

To quantitatively assess the influences of various environmental factors on the hurricane genesis variations over the NATL, the dynamic genesis potential index (DGPI) proposed by Wang and Murakami41 is utilized for a further analysis. Figure 2 displays the contrasts in the JJASON-mean DGPI between 2001–2022 and 1979–2000 and the contributions of individual factors. Remarkably, positive variations in the DGPI are discernible in the southern portion of the NATL, which aligns with the enhanced hurricane genesis observed in that area (Fig. 2a and Supplementary Fig. 1c). The cumulative sum of the four terms on the right–hand side of Eq. (1) is consistent with the total value on its left-hand side (Fig. 2a, b). This underscores the credibility of the linear decomposition method in estimating the relative contributions of environmental parameters to TC formation. The dominant term promoting the rising tendency of hurricane genesis in the southern portion of the NATL is vertical wind shear (Fig. 2f). The secondary contribution is from the meridional shear of mid-level zonal wind (Fig. 2d). The remaining two terms, including absolute vorticity and mid-level vertical motion, make a relatively insignificant contribution to the variations in hurricane genesis within the southern sector of the NATL (Fig. 2c, e). The linear trends of DGPI and the individual terms display results consistent with those from the composite analysis (Supplementary Fig. 5).

Fig. 2: Contrasts of DGPI and its contributing factors.
figure 2

Contrasts of a DGPI, b the cumulative sum from all terms, and the individual contributions from c 850-hPa absolute vorticity (ABS, s−1), d meridional shear of 500-hPa zonal wind (UY, s−1), e vertical pressure velocity at 500 hPa (OMEGA, Pa s−1), and f vertical wind shear (US, m s−1) during JJASON between 2001–2022 and 1979–2000. The blue box denotes the major region of NATL. The black dots indicate statistically significant differences at the 95% confidence level.

Full size image

In addition to the DGPI, we have also analyzed the genesis potential index (GPI) change introduced by Murakami and Wang27 for comparison (Supplementary Text 1). The GPI difference is positive over most regions of the NATL (Supplementary Fig. 6a), which does not match with the hurricane genesis change in some regions. It indicates that the DGPI displays a better skill in capturing the long-term change of hurricane genesis in the NATL. This finding is in harmony with the conclusion drawn by Wang et al.42 that concluded that DGPI offers a superior representation of the decadal variability in TC formation within the western North Pacific, in contrast to the other conventional GPI formula.

The aforementioned analysis underscores the crucial impact of vertical zonal wind shear on the observed tendency in hurricane genesis frequency within the southern part of the NATL. The change in vertical zonal wind shear is closely linked to the change in the meridional temperature gradient, as dictated by the thermal wind balance. Thus, subsequently we delve into the trends of 200–850-hPa averaged tropospheric temperature (Fig. 3). Notably, a ubiquitous warming trend is observed across the entire NATL region, with the magnitude increasing poleward from the equatorial region (Fig. 3a). The north-to-south temperature gradient in the main development zone of NATL, characterized by the difference in mean temperature between the tropical belt (10°S-5°N, 10°-80°W) and the subtropical belt (15°N-30°N, 10°-80°W) averaged between 850 hPa and 200 hPa, displays a significant downward trend (Fig. 3b), in line with the negative tendency of vertical zonal wind shear in the tropical region of the NATL based on the thermal wind balance (Supplementary Fig. 4e).

Fig. 3: Trends and evolutions of tropospheric temperature in the tropical and subtropical regions.
figure 3

a The linear trend of 200–850-hPa averaged tropospheric temperature (K year−1) during JJASON season from 1979 to 2022. b The evolution of time series of 200–850-hPa averaged temperature contrast (K) between the tropical (purple box, 10°S-5°N, 10°-80°W) and the subtropical (purple box, 15°N-30°N, 10°-80°W) regions as depicted in (a). The red line in (b) signifies the linear trend, while the text above it indicates the significance level and the slope of linear trend. Areas highlighted with black dots in (a) represent statistically significant trend at the 95% confidence level.

Full size image

To elucidate the heightened warming pattern observed in the subtropical troposphere above the NATL (Fig. 3a), we conduct a heat budget analysis according to the Eq. (2). The assessment, focusing on the 200–850-hPa layer, reveals insights into this phenomenon, as depicted in Fig. 4. The positive temperature tendency is larger in the subtropics than in the tropics (Fig. 4a). The positive temperature tendency in the subtropics is predominantly attributed to the term associated with static stability (Fig. 4a, c). The advection of temperature is negligible and the diabatic heating has negative effect to the temperature tendency in the subtropics, respectively (Fig. 4b, d).

Fig. 4: Heat budget analysis according to the equation (2).
figure 4

The heat budget according to the Eq. (2) during JJASON from 1979 to 2022. The top left a represents the local time derivative of temperature (K s−1). The top right b represents the term associated with the advection of temperature (K s−1). The bottom left c represents the term associated with static stability (K s−1) and the bottom right d represents the term associated with the diabatic heating (K s−1). The purple box is the same as that in Fig. 3a.

Full size image

To delve deeper into the mechanisms underpinning this tropospheric warming associated with the static stability, Fig. 5 illustrates the linear trends in potential temperature at both lower and upper levels along with the vertical profiles of potential temperature trend across subtropical and tropical zones. A notable finding is that the subtropical region experiences a more pronounced warming trend in the upper troposphere compared to the lower troposphere (Fig. 5a, b). Furthermore, compared to the tropical region, the subtropical zone exhibits a more substantial vertical potential temperature gradient between the upper and lower troposphere (Fig. 5c), indicative of a more stable convective stratification. This observation aligns with previous studies that showed a significant increase in temperature trends with altitude across various observational datasets43,44,45. In particular, Steiner et al.44 posited that the intensified warming in the upper troposphere, in comparison to surface trends, is largely consistent with the moist adiabatic lapse rate theory44,46. This indicates that the moist adiabatic ascent of air masses over the ocean amplifies warming, resulting in a warming peak at ~200 hPa, called tropospheric amplification.

Fig. 5: Linear trends of potential temperature at different pressure levels and regions.
figure 5

The linear trend of potential temperature (K year−1) at a 200 hPa and b 850 hPa. c The vertical profile of the linear trend of potential temperature (K year−1) averaged at the tropical region (10°S-5°N, 10°-80°W) (dashed blue line) and the subtropical region (15°N-30°N, 10°-80°W) (solid red line) (purple boxes of (a, b)). The purple boxes are the same as those in Fig. 3a.

Full size image

One issue is what are the relative roles of natural variability and external forcing in vertical zonal wind shear tendency. To address this issue we examine the ensemble-mean vertical zonal wind shear trend from the 39 Coupled Model Intercomparison Project phase 6 (CMIP6) historical and pre-industrial runs (Fig. 6). Figure 6a shows significant negative tendency of vertical zonal wind shear in the tropical region of the NATL from CMIP6 historical runs, which is similar to the observations (Supplementary Fig. 4e). In comparison, this negative tendency is slightly weaker in the historical runs. The tendency of vertical zonal wind shear is weak and insignificant in the pre-industrial runs (Fig. 6b). It indicates that the significant tendency of vertical zonal wind shear over the tropical region of the NATL is primarily contributed by climate change, not natural variability.

Fig. 6: Ensemble-mean linear trend of vertical zonal wind shear from CMIP6 simulations.
figure 6

The ensemble-men linear trend of JJASON vertical zonal wind shear (m s−1 year−1) between 200 hPa and 850 hPa from CMIP6 a historical and b pre-industrial runs during 1979–2014. The blue box denotes the 10-20°N region of NATL. The areas marked with dots denote statistically significant linear trend at the 95% confidence level.

Full size image

Discussion

Previous studies have found a tendency of a southward movement of TC genesis and TC LMI over the NATL, but the trend is insignificant11,20. Present study reveals that the trend is dependent on the TC intensity. A significant southward movement in the JJASON hurricane genesis within the NATL region has been detected after 1979. Meantime, the latitudinal location of TCs that reach the hurricane intensity and LMI also feature a consistent southward shift. Our analysis shows that the southward shift of the hurricane genesis is closely linked to an increase in the hurricane genesis frequency over 10°-20°N of the NATL.

The physical processes for the increase of hurricane genesis frequency over 10°-20°N of the NATL are schematically shown in Supplementary Fig. 7. Analysis according to the DGPI decomposition reveals that the vertical wind shear serves as a pivotal factor in the shifting of location of hurricane genesis. Remarkably, a pronounced negative tendency in vertical shear of zonal wind is identified south of 20°N, which is inherently tied to the weakening meridional temperature gradient based on the principle of the thermal wind balance. This reduced meridional temperature gradient is a direct outcome of enhanced warming within the subtropical troposphere. Consequently, the negative trend in the vertical shear of zonal winds leads to a decrease in vertical wind shear, thereby creating favorable conditions for more hurricane genesis in the southern realm of NATL. Furthermore, the greater warming in the subtropical troposphere is attributable to the static stability rise in the extratropical region. CMIP6 models indicate that the predominant factor driving the vertical zonal wind shear trend over the NATL is climate change, rather than natural variability. These physical processes are similar to Cao et al.17 that detected a robust westward movement of TC genesis within the southern Indian Ocean. It underscores the paramount importance of subtropical tropospheric warming as a universal factor driving the observed latitudinal and longitudinal shifts in TC or hurricane genesis across various ocean basins.

Previous studies have indicated a poleward shift of TC genesis location around the globe20,47. What causes the equatorward shift in hurricane genesis within the NATL? To address this question, we conduct an additional analysis for four basins during their respective TC peak seasons, including the western North Pacific (WNP, 100°-180°E) and the eastern North Pacific (ENP, 140°-85°W) during JJASON and the southern Indian Ocean (SIO, 30°-115°E) and the southern Pacific Ocean (SPO, 135°E-130°W) from November in the preceding year to April (NDJFMA) in the succeeding year.

The larger tropospheric temperature increase in the subtropics than in the tropics is mainly observed in the NATL, ENP, and the SIO (Fig. 7c and Fig. 8c). As such, a consistent negative trend is evident across these three basins (Figs. 7b and 8b). However, the climatological vertical zonal wind shear differs among the three basins. Climatologically, easterly vertical wind shear is prevalent near the equatorial regions over the ENP and the SIO (Figs. 7a and 8a), whereas westerly vertical wind shear is observed over the equatorial NATL (Fig. 7a). The superposition of this negative trend with climatological mean results in the enhanced vertical wind shear near the equatorial regions over the ENP and the SIO, which is unfavorable for the equatorward shift of TC genesis. Conversely, over the equatorial NATL, the negative trend opposes climatological mean westerly vertical wind shear, thereby reducing the vertical wind shear near the equatorial NATL. This creates a favorable condition for TC genesis to shift equatorward in the NATL.

Fig. 7: Linear trends of vertical zonal wind shear and tropospheric temperature as well as climatological mean vertical zonal wind shear over the Northern Hemisphere.
figure 7

a Climatological mean 850-hPa winds (vector, m s−1) and vertical zonal wind shear between 200 hPa and 850 hPa (shaded, m s−1) during JJASON of 1979–2022. The linear trend of JJASON (b) vertical zonal wind shear (m s−1 year−1) between 200 hPa and 850 hPa and (c) 200–850-hPa averaged tropospheric temperature (K year−1) from 1979 to 2022 over the Northern Hemisphere. The blue boxes denote the western North Pacific, eastern North Pacific, and northern Atlantic Ocean. The areas marked with black dot denote statistically significant linear trend at the 95% confidence level.

Full size image
Fig. 8: Linear trends of vertical zonal wind shear and tropospheric temperature as well as climatological mean vertical zonal wind shear over the Southern Hemisphere.
figure 8

a Climatological mean 850-hPa winds (vector, m s−1) and vertical zonal wind shear between 200  and 850 hPa (shaded, m s−1) during NDJFMA. The linear trend of NDJFMA b vertical zonal wind shear (m s−1 year−1) between 200  and 850 hPa and c 200–850-hPa averaged tropospheric temperature (K year−1) from 1979 to 2022 over the Southern Hemisphere. The blue boxes denote the southern Indian Ocean and southern Pacific Ocean. The areas marked with black dot denote statistically significant linear trend at the 95% confidence level.

Full size image

For the WNP and SPO, the difference in tropospheric temperature trends between the subtropics and the tropics is weak (Figs. 7c and 8c). As such, the trend in vertical zonal wind shear over the WNP and SPO is insignificant (Figs. 7b and 8b). Consequently, the poleward shift of TC genesis over the WNP and SPO is not linked to the vertical wind shear, but rather to mid-level vertical motion (figure not shown). Our analysis reveals that the meridional shift of TC genesis around the globe is dependent on both spatial distribution of tropospheric temperature changes and climatological vertical zonal wind shear.

The aforementioned finding reveals a notable southward migration of hurricane genesis locations since 1979, a phenomenon intimately tied to environmental dynamic parameters. However, the overall latitudinal movement of the TC geneses over the NATL is modest, suggesting that the inclusion of weak TCs that fail to attain the hurricane strength make a contamination on the linear trend of all the TC migration. Meantime, we have examined the sensitivity of our result to the TC duration. When we exclude short-lived (<2 days) TCs, the linear trend of TC genesis location is still insignificant compared with all the TCs (figure not shown). It indicates that the significance level of TC genesis location trend over the NATL is primarily dependent on the TC intensity. This may be due to the dependence of the large-scale environmental influence on the size of TCs. Tropical storms tend to have smaller sizes and may be more susceptible to small-scale and synoptic-scale weather systems. In contrast, hurricanes have larger sizes and are more likely to be influenced by large-scale circulation patterns. The lack of a notable southward shift among those weak TCs necessitates a deeper analysis. Our current finding highlights that hurricane activity is more intimately controlled by background environmental factors compared to the weak TC activity over the NATL. Meantime, we also examine whether this tendency of southward shift of hurricane genesis will persist under global warming by using CMIP6 HighResMIP models48,49. We select three high-resolution models, including CMCC-CM2-VHR, CNRM-CM6-1-HR, and HadGEM3-GC31-HM. We employ the atmosphere-only experiments of three models, which include historical runs named highresSST-present from 1950 to 2014 and future warming runs named highresSST-future from 2015 to 2050 under the SSP5-8.5 scenario. However, no significant trend of the southward shift is detected in hurricane genesis in the multi-model ensemble mean (figures not shown). The projection of models with higher resolution and more members will need to be analyzed in detail in subsequent research endeavors.

Methods

Data

The TC best track dataset in the NATL is sourced from the IBTrACS50. This dataset meticulously records the six-hourly latitude, longitude, and the maximum sustained wind speed of TCs. We define TC genesis as the moment and location when the wind speed of TCs initially exceeds 34 kts (roughly 17 m s−1) over the NATL region17,30,42. The hurricane strength is identified when the maximum wind speed exceeds 64 kts (roughly 33 m s−1). The active TC season over the NATL is from June to November (JJASON)2,14.

Monthly atmospheric variables are sourced from the ERA5 dataset. This dataset includes zonal wind, meridional wind, vertical pressure velocity and specific humidity, with a 2.5° latitude and 2.5° longitude51. The total vertical wind shear is defined as \(\sqrt{{({u}_{200}-{u}_{850})}^{2}+{({v}_{200}-{v}_{850})}^{2}}\) in which \({u}_{200}-{u}_{850}\) and \({v}_{200}-{v}_{850}\) are the difference of zonal and meridional wind between 200 hPa and 850 hPa, respectively. The Extended Reconstructed Sea Surface Temperature (ERSST) version 5 from NOAA offers a monthly sea surface temperature (SST) dataset, boasting a grid resolution of 2° × 2° and a historical span that extends back to 185452.

In order to examine the relative role of natural variability and external forcing in the vertical wind shear tendency over the NATL, the present study uses monthly mean outputs of 39 historical and pre-industrial runs that participate in the Coupled Model Intercomparison Project phase 6 (CMIP6) archive53. Detailed information of models is presented in Table 1. All the model outputs are converted to a grid resolution of 2.5° × 2.5° for consistency. The CMIP6 historical simulations are forced by historical anthropogenic forcing and natural forcing spanning the time from 1979 to 2014. The CMIP6 pre-industrial simulations correspond to the period prior to the start of large-scale industrialization, with 1850 being the reference year. The pre-industrial simulations are used to study the unforced internal variability of the climate system54. We only choose the last 36 years for a fair comparison with historical simulations. Most models only have one realization, thus we employ the first run in each model.

Table 1 Information of 39 CMIP6 models employed in this study
Full size table

Genesis potential index

The DGPI is widely used to investigate the influence of large-scale environments on TC formation on distinct time scales41. The DGPI index is defined as follows:

$${DGPI}={\left(2+0.1{V}_{{shear}}\right)}^{-1.7}\times {\left(5.5-\frac{\partial u}{\partial y}{10}^{5}\right)}^{2.3}\times {\left(5-20\omega \right)}^{3.4}\times {\left(5.5+\left|{10}^{5}\eta \right|\right)}^{2.4}\times {e}^{-11.8}-1$$
(1)

where Vshear (US) represents the vertical wind shear between 850 hPa and 200 hPa, \(\frac{\partial u}{\partial y}\) (UY) represents the meridional gradient of zonal wind at 500 hPa, ω (OMEGA) represents the vertical pressure velocity at 500 hPa, and η (ABS) represents the absolute vorticity at 850 hPa. The higher DGPI indicates the higher likelihood of TC genesis.

Heat budget

To comprehend the temperature variations associated with TC shift, we calculated the heat budget in accordance with the prior studies55,56. At each constant pressure level, the temperature tendency arises from the combined influence of horizontal temperature advection, adiabatic processes linked to vertical motion, and the atmospheric apparent heat source (Q1). The equation for temperature variation can be formulated as:

$$\frac{\partial T}{\partial t}=-V\,\cdot\, \nabla T+\omega \frac{{RT}}{{{pC}}_{p}}-\omega \frac{\partial T}{\partial p}+\frac{{Q}_{1}}{{C}_{p}}$$
(2)

where T represents temperature, t represents time, V denotes horizontal velocity vector, ∇ indicates the horizontal gradient operator, ω is the vertical pressure velocity, R, p, Cp and Q1 represent the gas constant, pressure, the specific heat at constant pressure, and the apparent heat source, respectively. Note that \(\frac{{RT}}{{{pC}}_{p}}-\frac{\partial T}{\partial p}\) also represents the static stability.

Significance test

The long-term trend is evaluated utilizing the least-square linear regression method. The statistical significance of the linear trend is estimated based on the nonparametric Mann-Kendall test. Furthermore, to assess the significance of the composite analysis, we employ the Student’s t test.