Tropical cyclones (TCs) represent the most significant weather phenomenon within the western North Pacific (WNP), profoundly impacting millions of people in the surrounding regions particularly when they make landfall. These powerful storms cause widespread destruction, including severe flooding, loss of life, and significant damage to infrastructure and agriculture1. A reliable and accurate projection of the future occurrence of TC tracks within the WNP is beneficial and indispensable for government planners and businesses to effectively prepare for the devastating effects of these powerful storms.

Over the last few decades, numerous studies have presented observed trends in diverse TC characteristics within the WNP2,3,4,5,6,7,8,9,10,11,12,13,14. For instance, some research have detected a poleward migration in TC genesis ___location, attributed to the poleward expansion of the Hadley cell3,12. Kossin et al. 4 discovered a statistically significant tendency of a poleward migration of the average latitude of TC maximum lifetime intensity on a global scale, encompassing the WNP. Kossin et al. 5 further emphasized that this poleward shift is robust over the WNP and is projected to persist until the end of the century. Moreover, recent decades have witnessed an increase in intense TC number, the TC maximum lifetime intensity, and landfalling TC intensity within the WNP7,8,14. Xu et al. 14 suggested that the intensifying TC intensity within the WNP is partially driven by the warming of Tibetan Plateau.

A distinct spatial pattern of tendency in the frequency of TC occurrences has been identified globally since 1980, with a notable decrease over the WNP, as reported by Murakami et al. 10. Murakami9 further revealed that increased aerosol emissions across South and East Asia have substantially contributed to the decrease in the TC occurrences within the WNP. Additionally, Chen et al. 2 found a northward shift in the landfall positions of TCs over the WNP during the past four decades, which was attributed to a decrease in the proportion of westward TC tracks and an increase in northward TC tracks. The accumulated power dissipation index following landfall displays substantial rising trends, primarily due to the longer mean lifespan of TCs over land and the heightened intensity of TCs at the time of landfall6.

Currently, there exists a lack of a consensus regarding the projections of TC track changes within the WNP. Various modeling studies have presented different perspectives on this issue. Some models indicated a northward migration in TC tracks within the WNP15,16,17. Wu et al. 17 identified a slight trend for the projected TC activity in the WNP to shift poleward under climate change based on GFDL Regional Atmospheric Model. Other models have indicated an eastward migration in TC tracks towards the central North Pacific18,19,20,21,22,23. For example, Li et al. 21 utilized a global high-resolution atmospheric general circulation model to examine the TC frequency over the North Pacific under climate change and found a notable migration in the ___location of TCs from the western to central Pacific, which is linked to the variance of tropical synoptic-scale disturbances. Additionally, some researchers have found a combination of both northward and eastward shifts in TC tracks over the WNP24,25,26. Manganello et al. 27 observed a southeastward migration in TC tracks within the WNP, further adding to the diversity of shifts. In addition, other related works have discussed the dependence of simulations on resolution and tracking algorithms and the effect of SST warming patterns on the future TC tracks over the WNP28,29,30,31,32,33,34,35. For example, Zhao et al. 35 examined the responses of TC track density over the WNP to the two distinct global warming patterns and found that both El Niño-like and La Niña-like warming patterns reduce TC track density over the WNP except that in the latter pattern there are increases in TC track density in the South China Sea and along China coast. Thus, the overall pattern of future changes in TC tracks within the WNP remains uncertain and inconclusive. It is crucial to conduct further research to assess the reliability and statistical significance of TC track changes over the WNP under global warming by using different models.

Recently, an extensive millennium-scale large ensemble simulation, titled the database for Policy Decision making for Future climate change (d4PDF), has been produced utilizing a 60-km atmospheric general circulation model across various warming scenarios36. Prior research has validated the proficiency of the d4PDF simulation to accurately replicate observed interannual and interdecadal fluctuations in large scale atmospheric circulation, which are closely associated with TC activity37,38,39,40,41,42. Yoshida et al. 41 utilized this high-resolution simulation to predict a future decrease of 33% in the global frequency of TCs and a significant decline in category 4 and 5 TCs due to future warming. More recently, Cao et al. 43 analyzed the global pattern of latitudinal shifts in TC genesis under climate change and discovered a consistent poleward migration in TC genesis during active seasons in both hemispheres, which was driven by the weakening of the Hadley circulation. However, there remains a gap in research examining the changes of TC tracks as well as the physical mechanisms over the WNP in this high-resolution simulation in future warming conditions. Therefore, the objective of this present study is to explore the future changes in TC tracks including landfall within the WNP under climate change and the underlying physical mechanisms responsible for these changes in TC tracks utilizing the large ensemble high-resolution simulation.

Results

Future change of TC tracks over the WNP

A comparison between the observed and simulated TC tracks during JASO is presented in Fig. 1. Generally, the spatial distribution of TC track density in observational data closely resembles that of the historical simulation, with two notable maxima observed east of the Philippines and over the South China Sea (Fig. 1a and 1b). However, the magnitude of these maxima, especially the one east of the Philippines, is severely underestimated in the simulation. Under the future 2 K and 4 K warming scenarios, the spatial pattern of TC tracks remains similar to the historical run. However, there is a gradual decrease in the density of TC tracks from the 2 K to the 4 K warming simulations (Fig. 1c and 1d). Figure 1e and 1f compare the TC track density between W2K and HST and between W4K and HST, respectively. Significant decrease occurs south of 40°N and west of 160°E, including the South China Sea, east of the Philippines, south of Japan, South Korea and east and south China coast. Conversely, an increase in TC tracks is seen north of 40°N including northeastern China and Hokkaido of Japan, and regions east of 160°E. As the future warming intensifies, the differences in TC tracks become even more pronounced (Fig. 1e and 1f). To illustrate the robustness of the model simulated changes, we have analyzed the TC track changes in W2K and W4K runs in individual SST warming scenario. The obtained changes in TC track density in each SST warming scenario (Supplementary Figs. 1 and 2) are mostly consistent with those based on the 90-member ensemble mean (Fig. 1e, f) except for the NCAR-CCSM4 and MRI-CGCM3 models in W2K runs (Supplementary Fig. 1c and Supplementary Fig. 2e). While there are differences in the significance levels of changes, the spatial patterns of changes are quite similar. This consistency indicates the robustness of the future projections in TC tracks.

Fig. 1: The climatological mean TC track density in observation and models along with the change in future warming scenarios.
figure 1

The climatological mean TC track density distribution from 1979 to 2010 in the (a) IBTrACS data, (b) HST run, (c) W2K run, and (d) W4K run. The contrast of TC track density between (e) W2K and HST and (f) W4K and HST during JASO over the WNP based on the d4PDF ensemble for last 32 years. The cross-hatched region in (e, f) denotes that the difference in the TC track density is significant at the 95% confidence level.

Full size image

In the following, we focus on the projected outcomes of TC tracks in two key regions that show a significant increase and in eight distinct sub-regions along East Asian coast, as illustrated in Fig. 2. Our discussion is centered on the average percentage changes in the W2K and W4K projections relative to the historical run. In general, the percentage changes indicate an increase north of 40°N and east of 160°E as well as a decrease south of 40°N and west of 160°E, with the higher warming scenario (W4K) exhibiting more pronounced changes. Specifically, TC tracks north of 40°N are projected to rise by 25% under 2 K warming scenario and 65% under 4 K warming scenario (Fig. 2a). TC tracks south of 40°N and east of 160°E are projected to rise by 15% and 36%, respectively, under the two warming scenarios (Fig. 2a). Previous studies18,19 have indicated that the projected frequency of TC occurrence will increase in the central Pacific. Katsube and Inatsu44 used a regional atmospheric model to investigate the response of TC tracks to uniform SST warming in the WNP and found that there is a faster northward movement of TCs in the uniform 2 K warming run.

Fig. 2: The projected percentage change in TC track density.
figure 2

The projected percentage change in TC track density in different regions under different climate change scenarios.

Full size image

We further focus on the percentage change of projected TC landfalling along the East Asian coast under two warming scenarios (Fig. 2b). Under the 2K warming scenario, a median increase of 12.8% in TC tracks is projected for the Hokkaido of Japan. The increase is escalated to 28.4% under the 4 K warming scenario. Similarly, northeastern China is expected to witness significant increases, with rates of 33.7% and 72.8%, respectively, under the 2 K and 4 K warming scenarios. South of 40°N, the magnitude of decreasing percentage changes in TC tracks increases with the decreasing latitude (Fig. 2b). Over the Korean Peninsula and southern Japan, only slight decreases in TC tracks are projected, with rates of 4.3% and 10.8% under the 2 K warming scenario. Under the 4 K warming scenario, the rate increases to approximately 24% in both regions. In both East China and South China, the percentage changes in TC tracks are comparable, with decreases of −19.5% and −22.5% under the 2 K warming scenario, and −44% and −47.8% under the 4 K warming scenario, respectively. The most significant changes are observed in lower latitudes, including the Indo-China Peninsula and the Philippines (Fig. 2b). Under the 2K warming scenario, a significant decrease of 28.4% in TC tracks is projected over the Indo-China Peninsula. It escalates to 53.5% under the 4K warming scenario. Similarly, in the Philippines, decreases of 27.8% and 60% are projected under the 2K and 4K warming scenarios, respectively.

Our projections are not the same as previous studies15,18,19,45,46. For instance, utilizing the CMIP5 models, Bell et al. 45 obtained a significant reduction in TC track exposure in southern China and the Philippines, accompanied by a significant increase for the Korean peninsula and Japan. More recently, Liu et al. 15 studied the potential future changes in TC landfalling activity along the East Asian coast under varying global warming conditions by using CMIP6 model. Their study revealed that the northern part of the East Asian coastline, encompassing the East China, Japan and Korean Peninsula, is likely to witness an increase in TC track frequency by a range from 17% to 60%, while the southern part may experience a decline of 14–27%. The notable diversity is found to be located in high-latitude regions, including Korean peninsula, Japan and northeast China. Our study underscores that Korean peninsula and Japan do not exhibit a uniform pattern, with a rise north of 40°N, but a decline south of 40°N. In contrast, previous studies consistently reported a significant increase in TC track frequency across Korean peninsula and Japan. Meantime, TC track changes over northeast China are rarely discussed in previous studies. Our study projects a substantial increase in TC tracks over northeast China under future warming scenarios, thereby contributing novel insights to this less studied region.

Possible mechanisms underlying the future change of TC tracks

The projected shift in TC tracks is believed to stem from the changes in TC genesis ___location and steering flows16,18,47. Firstly, we examine the ability of model simulation to reproduce TC genesis in historical run with the observational data, as well as the differences between future warming runs and historical run (Fig. 3). Figure 3a-d depict the TC genesis density during JASO in HST, W2K and W4K as well as the climatological mean pattern from the IBTrACS dataset. In general, the climatological mean in TC genesis number during JASO in historical run (18.4) is rather close to the observational value (17.8) (Fig. 3a and 3b). It indicates that the TC tracking method is reliable to detect the TCs in the simulation. The ensemble-mean result indicates a progressive decrease in the frequency of WNP TC genesis as global warming intensifies. Specifically, the TC genesis count stands at 15.2 and 11.0 in W2K and W4K runs, respectively (Fig. 3c and 3d). Compared to the climatological mean value in HST, the mean TC genesis number is decreased by 17% and 40% in the W2K and W4K runs, respectively.

Fig. 3: The climatological mean TC genesis density in observation and models along with the change in future warming scenarios.
figure 3

The climatological mean TC genesis frequency distribution (count/10 years) from 1979 to 2010 in the (a) IBTrACS data, (b) HST run, (c) W2K run, and (d) W4K run. The number on the top right denotes the JASO-mean TC genesis number in the respective scenarios. The contrast of TC genesis density between (e) W2K and HST and (f) W4K and HST during JASO over the WNP based on the d4PDF ensemble for last 32 years. The cross-hatched region in (e, f) denotes that the difference in the TC genesis frequency is significant at the 95% confidence level.

Full size image

The future changes in TC genesis density are further illustrated in Fig. 3e, f. Significant differences in TC genesis locations are observed between W2K and HST runs as well as between W4K and HST runs. Evidently, the TC genesis ___location shifts northward and eastward from HST to W2K and W4K runs, leading to a notable increase in TC genesis north of 25°N and east of 160°E (Fig. 3e-f). The significant reduction in TC genesis south of 25°N and west of 160°E indicates a decline in the TC number approaching the coasts of South China and East China, which seems to align with the future change in TC track changes (Figs. 1 and 2). The notable reduction in TC tracks over the Korean peninsula and south of Japan may primarily be associated with the substantial decrease in TC genesis south of 25°N. The surge in TC genesis east of 160°E results in a considerable increase in TC tracks east of this longitude (Fig. 1e, f and Fig. 3e, f).

In the following, we analyze the future change of the steering flows during JASO as depicted in Fig. 4a-f. The steering flows involve vertically integrating of the flows from 850 hPa to 300 hPa48,49,50. Typically, TC tracks are influenced by both the beta drift and steering flows18,47. The beta drift stems from the interaction between the gradient of planetary vorticity and the primary circulation of TCs, typically causing a westward and poleward movement in the Northern Hemisphere (Murakami and Wang, 2010). At first, we assess the capacity of model simulation to recreate the steering flows observed in historical simulation with the ERA5 data (Fig. 4a-b). The steering flows in HST run are slightly weakened and shrunk compared with reanalysis data. With the global warming intensifying, the steering flows become further weaker (Fig. 4b-d).

Fig. 4: The climatological mean horizontal winds in observation and models along with the change in wind and temperature in future warming scenarios.
figure 4

The horizontal winds (vector, m s−1) averaged from 850 hPa to 300 hPa during JASO in the (a) ERA5 data, (b) HST run, (c) W2K run, and (d) W4K run along the horizontal winds (shading, m s−1) at 300 hPa. The contrast of the (e, f) horizontal winds and (g, h) air temperature averaged from 850 hPa to 300 hPa between (e, g) W2K and HST and (f, h) W4K and HST during JASO over the WNP based on the d4PDF ensemble for last 32 years. The red dots in (e, f) denote that the difference in the horizontal winds is significant at the 95% confidence level. All the regions in (g, h) are significant at the 95% confidence level.

Full size image

A comparison of steering flows between future warming run and historical run reveals a prominent anomalous easterly north of 30°N, which is conducive to slowdown of the steering flows and thus more TC track density there (Fig. 4e-f). Meantime, a significant anomalous cyclonic circulation is located to east of 130°E (Fig. 4f), which is conducive to occurrence of more TCs in this region. Consequently, the increase in TC tracks over northeastern China and Hokkaido of Japan can be attributed to the slowdown of steering flows. On the other hand, anomalous easterlies are also detected east of the Philippines and over the South China Sea (Fig. 4e-f), particularly under future 4 K warming, which is conducive to steering TCs westward and making landfall. Nevertheless, a considerable decline in TC genesis frequency south of 20°N has markedly reduced the number of the TCs approaching southeast China, Indo-China Peninsula and the Philippines. This suggests that the reduction in TC tracks in these regions is primarily attributable to the decline in the frequency of TC genesis, rather than changes in the steering flows.

Figure 5a-d further illustrates the horizontal wind contrast at 850 hPa and 300 hPa between W2K and HST and between W4K and HST. It highlights that anomalous easterly north of 30°N is predominantly influenced by the upper level winds (Fig. 5c-d), whereas anomalous cyclone and the prominent anomalous easterly east of the Philippines and over the South China Sea are contributed by the lower level winds (Fig. 5a-b). The lower-level anomalous cyclone east of 130°E is associated with the upper-level jet stream change. Based on quasi-geostrophic theory, jet streaks experience significant ageostrophic flows in the entrance and exit regions51,52. When the upper-level westerly jet weakens, it results in the decreased ageostrophic northerly (southerly) wind at upper level in the entrance (exit) region of the jet, accompanied by the reduced upper-level divergence (convergence), lower-level convergence (divergence) and mid-level upward (downward) motion south of the jet streak. As such, the reduced lower-level convergence and divergence are accompanied by lower-level anomalous anticyclonic and cyclonic circulations south of the jet flows as shown in Fig. 5a-b.

Fig. 5: The change in lower-level and upper-level horizontal winds and SST in future warming scenarios.
figure 5

The contrast of the horizontal winds at (a, b) 850 hPa and (c, d) 300 hPa and (e, f) SST between (a, c, e) W2K and HST and (b, d, f) W4K and HST during JASO based on the d4PDF ensemble for last 32 years. The black dots in (a–d) and all the regions in (e, f) denote that the difference is significant at the 95% confidence level. The red line indicates the 25 m s−1 contour for 300-hPa winds in HST run, blue line for W2K run, green line for W4K run.

Full size image

The above analysis has highlighted the emergence of anomalous easterly north of 30°N. The question arises as to what cause the anomalous easterly in the extratropical troposphere over the WNP. The total steering flows in the troposphere north of 30°N are predominantly contributed by changes in the upper-level zonal wind (Fig. 5c-d). Across the whole WNP, the tropospheric temperature displays a warming tendency, with a larger increase north of 35°N (Fig. 4g-h). This substantial warming north of 35°N leads to a weakening in the meridional temperature gradient. The reduced meridional temperature gradient results in negative vertical zonal wind shear based on the thermal wind balance, ultimately inducing easterly wind anomalies in the upper level (Fig. 5c-d).

Figure 4 has elucidated a physical linkage between anomalous easterly flow trend and the pronounced warming trend observed in the mid-latitude troposphere. What is cause of this greater warming tendency in the mid-latitude troposphere? Fig. 5e-f reveals the future change of SST spatial pattern during JASO in W2K and W4K relative to HST. A pronounced warming tendency occurs across the entire WNP, with the most significant increase concentrated in the mid-latitude region. The magnitude of SST change is more pronounced in W4K than in W2K (Fig. 5f and 5e). Consequently, this SST warming pattern in the mid-latitude region leads to the overlying tropospheric warming through the surface air-sea heat flux and moist convection.

With respect to the future change in TC genesis ___location, Cao et al. 43 have examined the latitudinal shift in TC genesis ___location around the globe under climate change utilizing the d4PDF simulations including the WNP basin. Their finding revealed a consistent poleward migration in TC genesis ___location in the future warming scenarios, attributed to the weakening of the Hadley circulation. Therefore, the primary driver of future change in TC genesis ___location within the WNP is the modification of large-scale vertical atmospheric motion.

Discussion

Our research elucidates the future projection of TC tracks within the WNP based on the large ensemble high-resolution simulation, as an extension of our previous study43 which mainly focused on the future projection of TC genesis ___location around the globe. The schematic diagram shown in Fig. 6 displays the future projection in TC tracks and associated physical mechanism. Our results reveal that under future warming, the change in TC track density is not spatially homogeneous with an increase north of 40°N and east of 160°E and a decrease south of 40°N and west of 160°E. Correspondingly, there is a larger possibility that more TCs will affect northeastern China and Hokkaido of Japan and a significant reduction in Korean Peninsula, East and South China, the Philippines and Indo-China Peninsula.

Fig. 6: The schematic diagram.
figure 6

The schematic diagram showing the future projection of TC track changes and the related atmospheric flow patterns and SST anomalies under future warming. The red curly line indicates the jet flows at upper level in historical simulation.

Full size image

Two primary factors could explain the predicted changes in TC track frequency: TC genesis ___location changes and the large-scale steering flow changes. By comparing future warming runs with historical run, we deduce that the increase in TC tracks north of 40°N is primarily associated with change in tropospheric steering flows. Under future warming, anomalous easterly is present north of 40°N, which is conducive to slowdown of climatological westerly steering flows, so that TCs have more time to stay in the middle latitude, leading to increase the TC tracks here. The increase in TC tracks east of 160°E is predominantly associated with changes in TC genesis. An increased number of TC genesis events is projected in the subtropical region (north of 25°N) as well as east of 160°E, with a decrease in the tropical region (south of 25°N). Expectedly, the rise in TC genesis east of 160°E results in a substantial increase in TC tracks east of 160°E. Additionally, the decrease in TC tracks south of 40°N including TC landfall along the southeastern Asian coast is predominantly associated with the decrease in TC genesis south of 25°N rather than the steering flows.

Further analysis reveals that anomalous easterly in the mid-latitude region is linked to the spatial pattern of future SST warming, particularly the substantial warming in the middle latitude. The larger mid-latitude SST warming favors the weakening of meridional temperature gradient, resulting in anomalous easterly in the upper level according to thermal wind balance.

In conclusion, our projection results indicate that the spatial pattern of the TC track density poised to undergo significant changes under future warming. However, we acknowledge the inherent uncertainties associated with these projections, particularly stemming from model configuration. To enhance the reliability of our future projections, a thorough examination and refinement of the modeling configuration is imperative. In the future, we will conduct a series of experiments with various SST anomalies and model configurations to mitigate the associated uncertainty in the projection of TC tracks. Meantime, our projection is based on the atmospheric general circulation model. In the future, we will need to employ a fully-coupled atmosphere-ocean model to make the projection efforts.

Methods

Observational TC data

The observational data of TC geneses over the WNP originate from the IBTrACS53, which encompasses the latitude and longitude coordinates of TC tracks as well as the maximum sustained wind speed recorded at six-hour interval. Our primary focus is on the TCs that surpass the intensity threshold of 34 knots. For the purpose of this present study, we employ the TC best track data spanning from 1979 to 2010 to ensure compatibility with the simulation dataset. Our analysis is centered on TC genesis over the WNP from July to October (JASO for brevity)43,54,55,56.

Simulation data

The d4PDF dataset was produced by the MRI-AGCM3.2H on a grid with 60-km spacing36. The historical climate simulation, known as HST, spans from 1951 to 2010 and is forced by observed sea surface temperature (SST) and sea ice concentration data along with observed greenhouse gas concentrations. The simulation encompasses a 100-member ensemble generated through perturbations in both the initial atmosphere and the SST.

The future warming simulations, designated as W2K and W4K, are carried out under constant warming conditions corresponding to the 2040 and 2090 values, respectively, of the RCP8.5 scenario adopted in CMIP5. In both W2K and W4K, the SST forcing is derived from the detrended SST observations plus six climatological SST warming patterns from six CMIP5 models: MRI-CGCM3, MIROC5, GFDL-CM3, MPI-ESM-MR, NCAR-CCSM4, and HadGEM2-AO. Each warming pattern is scaled to represent a 2-K and 4-K increase, respectively, relative to the preindustrial climate. The W2K and W4K runs are composed of 54 and 90 ensemble members in total, respectively. Mizuta et al. 36 provided comprehensive details on the model configuration and experimental design. The analyses are concentrated on the final 32-year climatology of the historical run and the two warming scenarios. The warming-induced responses are determined by subtracting the historical run from the 2-K and 4-K warming runs, respectively.

TC tracking method in the model

The TC tracking method along with the threshold values utilized in the present study follows the approaches outlined in the previous research19,41. The tracking algorithm incorporates multiple detection criteria. To qualify as a TC, a candidate system must satisfy specified thresholds for maximum relative vorticity and wind speed at lower level, possess a warm core at upper level and exhibit a particular vertical wind speed profile. If a candidate consistently meets all these criteria over a minimum of 36 hours, it is classified as a TC by the algorithm. The threshold values selected in the tracking process are tailored to the model to ensure that the global count of detected TC formations in HST run aligns closely with the observations. The TC formation and track data are then interpolated into a 5° × 5° grid box to facilitate the calculation of the climatological mean and contrast patterns.

The significance test

We evaluate the statistical significance in composite analysis by employing a two-tailed Student’s t-test at a 95% confidence level through all samples from the 100-member historical run and the 54- or 90-member warming runs.