Detecting the impact of climate change on tropical cyclones in Southern China

It is well known that tropical cyclones (TCs) making landfall in Southern China (SC) account for more than half of all TCs making landfall in China. Therefore, it is important to have an in-depth understanding of the activities of TCs in SC under climate warming. Our results show that there have been no significant changes in the frequency and duration of these TCs, but their intensities have unexpectedly decreased by ∼20% since 1980, which is inconsistent with the previous understanding that climate change increases TC intensity. The results consistently show a significant decrease in the different TC intensity percentiles, which is related to the intensity distribution that shows a significant decrease in the proportion of tropical storms and a significant increase in the proportion of tropical depressions, as well as a slight decrease in the proportion of category 1–2. Because of the locations of those TCs activity show a clear shoreward migration tendency, indicating that land friction can suppress TC intensification, so that TC intensity has weakened. In addition, results also suggest that TC development is strongly suppressed and is mainly related to the enhancement of atmospheric stability, vertical wind shear and subtropical high under global warming. These results are quite different from the previous understanding of the changes in TC intensity under global warming. Such knowledge can help us better understand the relationship between climate change and the impact of TC activity at the regional scale.


Introduction
Disasters caused by tropical cyclones (TCs) have a major impact on human activities in coastal areas and countries (Mendelsohn et al 2012).In 2023, several TCs, including TCs Talim (2304), Doksuri (2305), Saola (2309), andHaikui (2311), hit most coastal cities in China.These TCs caused incalculable economic losses to metropolises such as Beijing, Shanghai, Hong Kong, Shenzhen, Guangzhou, and their surrounding areas by bringing recordbreaking heavy rainfall disasters.Therefore, it is very important to have an in-depth understanding of their activities for disaster mitigation purposes (Wang et al 2019).
According to statistics, global warming caused by human activities is 1 • C higher than before the industrial revolution, and the temperature increase occurred mainly after the middle of the last century (Masson-Delmotte et al 2022).How climate change affects TC activity has always been a research topic of great interest (Emanuel 1987, Knutson et al 2010, 2019, 2020, Walsh et al 2015, 2016).At present, there is sufficient evidence to suggest that global warming may have influenced changes in TC activity at global and regional scales (Emanuel 2005, Chu and Murakami 2022, Shan et al 2023).Observational and numerical model studies suggest that the frequency of TC genesis will gradually decrease with increasing global warming, while the frequency of strong TCs will increase significantly (Elsner et al 2008, Wang and Zhou 2008, Mendelsohn et al 2012, Kossin et al 2013).Therefore, average TC intensity and TC destructive potential are likely to increase over the western North Pacific (WNP) basin (Knutson and Tuleya 2004, Emanual 2005, Knutson et al 2010, Park et al 2013, Knutson et al 2019, 2020).
Furthermore, with global warming, the location of the lifetime maximum intensity (LMI) of TCs has exhibited a significant trend of poleward migration (Kossin et al 2014), which may be related to the tropical expansion caused by global warming (Seidel et al 2008, Rajaud andNoblet-Ducoudré 2017).Such poleward migration has also been observed in the vicinity of Taiwan and the East China Sea since 2000 (Tu et al 2009).In addition, the location of the LMI of global TCs has also shown an apparent coastward migration (Wang and Toumi 2021), with an additional two TCs entering the offshore region every decade.Thus, there are more TCs in the offshore region, posing more potential threats to coastal regions.
Located on the western coast of the WNP, China is affected by many TC activities originating from the WNP and the South China Sea (SCS).Statistics show that the average frequency of TCs making landfall in China is about 7-8 per year (Wang et al 2019, Shan and Yu 2021).They are mainly concentrated in southern and eastern China (SC and EC), and relatively few occur in northern China.Furthermore, observational studies have revealed that the destructive potential of TCs making landfall has increased significantly in China in recent decades (Liu et al 2020).Li et al (2017) pointed out that the increase in the destructive potential of TCs mainly occurred in the EC due to the increase in the frequency and intensity of TCs that make landfall in the EC, but this phenomenon is not obvious in the SC.
Statistically, the SC is probably one of the most affected regions by TC activities in China and even in the world, because more than half of the TCs landfalling in China each year may landfall in the SC region (figure 1).However, it is still not completely clear how climate change affects TC changes in the SC.Therefore, the study of TC activities in the SC is important for understanding the changes in TC activities in China and for disaster prevention and mitigation in the SC region.In this study, for statistical feasibility, we considered TCs making landfall in the SC region (including Guangdong, Guangxi, Hainan, Hong Kong, and Macao), and examined the response of these TCs to large-scale climate change and its possible causes.Section 2 presents the data and methods used in this study.The main results and findings are presented in section 3. Section 4 presents the conclusions of this study and a brief discussion.

Data and methodology
In this study, we mainly used the best track data in the International Best Track Archive for Climate Stewardship (IBTrACS) (Knapp et al 2010) from 1950 to 2021, including the records of China Meteorological Administration (CMA), Hong Kong Observatory (HKO), Joint Typhoon Warning Center (JTWC), and Japan Meteorological Administration (JMA) for the WNP basin, to investigate the changes in TCs landfalling the SC region.Since the TCs recorded by JMA start as tropical storms (TSs: >35 kt), the frequency of TCs counted by JMA will be smaller than the other three records.Because of the larger number of observations available, the CMA database is likely to be more accurate and complete over the offshore and land areas of China than over the open ocean (Ying et al 2014), therefore, we mainly analyzed and studied the results from the CMA records, and the HKO, JTWC and JMA records were used to verify the results of the CMA.All TCs (usually ⩾ 24 kt) that reached the tropical depression (TD) level and above were also included.
To characterize the activities of these TCs that make landfall in the SC region, this study used the power dissipation index (PDI) defined by Emanuel (2005), which is the cubic integral of the TC intensity (the wind speed).For statistical convenience, we calculated the changes in the total PDI on the annual scale, which can reflect the combined contribution of the annual TC frequency, average duration, and average intensity (Emanuel 2005, Lin and Chan 2015, Tu et al 2018, 2020).
In addition, to explore the change in the TC intensity, we considered the LMIs of TCs and the proportion distributions of the TC intensity at different intensity levels.Among them, for the TC LMI and its corresponding positions, we selected the position where the TC reaches its LMI for the first time, and the stage before the LMI was defined as the TC development stage.Regarding the TC intensity category, we utilized the Saffir-Simpson hurricane wind scale (Taylor et al 2010) and utilized the statistics to divide all the TC intensities into four classifications: TD, tropical storm (TS), categories 1-2 (CAT 12), and categories 3-5 (CAT 35).Then, we calculated their corresponding proportion distributions.In order to further explore the changes of TC activity with current climate warming, taking the phase change of the Pacific Interdecadal Oscillation (PDO) occurring in 1998 (Meehl et al 2011, England et al 2014, Tu et al 2018), we divide the whole study period into two subperiods (i.e. P1: 1980-1997 and P2: 1998-2021) for comparison.
To explore the internal reasons for how climate warming affects the changes in TC activity in the SC region, we mainly used the monthly average of the fifth-generation European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalysis (ERA5) data (Hersbach et al 2020) from 1980 to 2021, including sea surface temperature (SST), water vapor, air temperature, geopotential, convective available potential energy and horizontal wind.It is well known that atmospheric vertical wind shear and atmospheric stability are two important factors influencing the occurrence and development of TCs (Gray 1968, Tuleya and Kurihara 1981, Wong and Chan 2004, Balling and Cerveny 2006).According to conventional definitions, we defined the horizontal wind difference between 200 hPa and 850 hPa as vertical wind shear (Velden and Sears 2014), and we defined the atmospheric potential temperature difference between 300 and 900 hPa as atmospheric stability (Sharmila and Walsh 2018, Tu et al 2021).Since we mainly examined climate change affecting TC activity, the differences in each influencing factor in June-October between P1 and P2 subperiods were also calculated.

Results
Based on the results shown in figure 1, a total of 318 TCs have made landfall in China since 1950, with an average of about 7.57 TCs per year (annual averages of 7.19, 7.21 and 6.14 for the JTWC, HKO and JMA data, respectively).However, it should be noted that during the study period, 189 TCs made landfall in the SC, accounting for about 60% of all TCs that made landfall in China.This indicates that more than half of the TCs that made landfall in China affected the SC.
Similarly, the JTWC, HKO and JMA datasets indicate that the proportion of TCs affecting the SC was 58%, 60%, and 56%, respectively.Therefore, it is crucial to study the changes in TCs affecting the SC, despite the relatively small size of the area.Unfortunately, this issue has not received enough attention in previous research.
On average, the frequency of TCs that made landfall in SC was 4.5 per year (4.31, 4.17, and 3.43 per year in the HKO, JTWC and JMA datasets, respectively).Figures 1(c) and (d) show the changes in the frequency of TCs that made landfall in China and affected the SC region during 1980-2021.We found that both datasets showed obvious interannual changes.Among them, the number of TCs that made landfall in China reached 13 in 1994 and was the lowest (only 3 TCs) in 1987; while the frequency of TCs that made landfall in SC reached 8 in 1980/1993, and 1 TC made landfall in this region in 1987.Overall, the change in the frequency remained constant, and there was no obvious increasing or decreasing trend.
The frequency of TCs that made landfall in SC did not change significantly.We investigated whether the total destructive potential of these TCs changed.We used the power dissipation index (PDI) proposed by Emanuel (2005) to investigate this change.The total PDI time series of TCs landfalling the SC from 1950 to According to the linear trend, the average TC intensity reaches 51 kt in the early period and weakened to about 41 kt at the end of the study period, with a decrease rate of about 20%.Considering the uncertainty of the single dataset, we further examined the other three TC best-track datasets and found that the decrease in the intensity of these TCs was also reflected in the JTWC (22%, P = 0.03), JMA (15%, P = 0.03) and HKO (10%, P = 0.21) datasets.This shows that the conclusion that there was a significant decrease in the intensity of TCs that made landfall in SC is robust and independent on a single dataset.Many studies (Sriver and Huber 2007, Pasquero and Emanuel 2008, Cheng et al 2019) have suggested that global warming has significantly increased the heat stored in the upper ocean over the past few decades, which promotes the development and rapid intensification of TCs, and increases in TC intensity.This result differs from previous studies, indicating that the impacts of large-scale climate change are not completely uniformly reflected at the regional scale.
Changes in the LMIs of TCs may also reflect the impact of climate change on their intensities.We analyzed the changes in LMI of these TCs (figure 2(d)).The results show that the annual average LMI of TCs that made landfall in SC also showed a decreasing trend (P = 0.06), which also shows that the LMI that TCs can reach is gradually weakening.
In addition to the LMI, what about the TC intensity distributions?As shown in figure 3, we found that the intensity of TCs landfalling in the SC region was significantly decreased at different percentiles, i.e. 90%, 75%, 50%, and 25%.Furthermore, we counted the probability of TC occurrence for the different TC intensity classifications.Based on the linear regressions, these percentiles decreased by 21%, 21%, 23%, and 18%, respectively, in this study period.In all the records, we divided the TCs into TD, TS, CAT 12, and CAT 35 based on the Saffir-Simpson Hurricane Wind Scale and counted the frequency changes between the two subperiods.According to the statistical results, the frequencies of TD, TS, CAT 12 and CAT35 changed by 144%, −8%, −28%, and −13%, respectively.
Furthermore, the proportions of these classifications in all records in each year were used to conduct statistical analysis (figure 4).First, we analyzed the proportion of TD occurrence (figure 4(a)) for all  the TCs with intensities less than 35 kt.This proportion only 20%-30% in the early stage and gradually increases to 50%-60% during this study period, with a rate of 8.35% per decade (P < 0.01).In some years, such as 1998 and 2019, greater than 70% of the TC recorded did not exceed 35 kt.Correspondingly, decreased rapidly from 50%-60% in the early years to about 30%.Furthermore, we found that the proportion of CAT 12 occurrence decreased relatively little (P = 0.16; figure 4(c)) during 1980-2021, but the occurrence proportion of CAT 35 have no significant increase or decrease trend (P = 0.91; figure 4(d)).
In addition, we investigated the differences in TC intensity of these TCs that made landfall in SC between P1 and P2 subperiods.As shown in figures 5(a) and (e), the results show that both the LMI and the average TC intensity decreased in the P2 period (P = 0.10 and P = 0.00, respectively).Similarly, significant differences in TC intensity between P1, P2 periods can be found in the time series of TC intensity analyzed previously (figures 2(c), 3 and 4(a), (b)).So, what caused the decrease of TC intensity?And why the proportion of TD increased and the proportion of TS and CAT 12 decreased?
One of the reasons for the changes in TC intensity is that the locations of TC activity have changed.We found that the average LMI positions of the TCs in P2 exhibited significant northward and westward migration compared with those in P1 (P ⩽ 0.05, figures 5(b) and (c)), which is consistent with the poleward and coastward migration of the LMI position suggested by Kossin et al (2014) and Wang and Toumi (2021).The LMI positions of the TCs shifted closer to the land (figure 5(d)), indicating that the TCs were more influenced by the land friction during the TC development stage.In addition, the average TC locations also show a significant westward and northward migration (figures 5(f) and (g)), indicating that the location of TC activity has obviously migrated toward the coast (figure 5(h)).This phenomenon can be clearly found from the distribution of TC tracks, according to a comparison of the distribution of the tracks of the TCs landfalling in the SC during the two subperiods, the locations of these landfalling TCs are getting closer to the land in P2 period than that in P1 (figures 5(i) and (j)).Therefore, the TC activity became increasingly influenced by land friction, which in turn caused the LMIs of the TCs to also show a weakening trend (figure 2(d)).We then considered why the TCs in the later subperiod reached their LMI so close to land rather than in the open ocean.
There is no doubt that global warming will significantly increase the SST, significantly increase the heat content of the upper ocean, and rapidly increase the moisture content of the atmosphere (Herweijer  C −1 ), the increase in temperature will lead to more moisture entering the atmosphere, so the WNP basin also exhibited a significant increase in atmospheric moisture content during the second subperiod (figure 6(b)).So why do these factors seem to be very conducive to the development and rapid intensification of TCs, but do not lead to a higher intensity of TCs landfalling in the SC region?
We assume that the internal cause comes from the changes in the other conditions instead of the increase in SST and moisture factors.The results show that the vertical wind shear in the middle of the SCS showed a significant increase during P2 than that in P1 (figure 6(c)).It is well known that the formation and development of TCs are inhibited under a stronger wind shear environment.Under the background of the current climate warming, the increase in vertical wind shear in the middle of the SCS makes it difficult for TC to form and develop, resulting in a slow increase in the intensity of the TC development S Tu et al stage.In addition, as shown in figure 6(d), atmospheric stability in the WNP and SCS continues to increase under global warming, which also inhibits TC development and intensification (Sharmila and Walsh 2018, Tu et al 2021).Meanwhile, we found that the convective available potential energy is weakened during the P2 period (figure 6(e)), and the geopotential at 500 hPa (i.e.subtropical high) is obviously increased (figure 6(f)), and these conditions can also inhibit TC development.Thus, during the P2 period, although the necessary SST and moisture conditions for TC activity are sufficient, the vertical wind shear, atmospheric stability, convective available potential energy and subtropical high play an unfavorable role, resulting in a decrease in TC intensity.

Conclusions and discussion
It is well known that SC is one of the most affected areas by TCs in China and even in the world.Our results show that the frequency of TCs landfalling in this region accounts for about three-fifths of the total TCs landfalling in China.In this study, we focused on how TCs landfalling in the SC region change with the current climate warming.
We found that the intensity of TCs landfalling in the SC has become weaker over the past four decades, while the duration and frequency have not changed significantly (figures 1-3).Among them, the decrease in TC intensity is mainly contributed by the increase in the proportion of TD and the decrease in the proportion of TS and CAT 12 (figure 4).Although the proportion of CAT 35 remains unchanged, we found that the LMI has also decreased, as well as the decrease in the frequency of CAT 35 during P2.We noticed that the location of TC activity has moved closer to the coast (figure 5).The increase in land friction has prevented the development of TCs and then partly led to the decrease in TC intensity (Wong and Chan 2006).
In addition to the effect of increased land friction caused by the shoreward migration of TC activity locations, changes in environmental conditions between these two periods also played a crucial role in reducing the intensity of TCs that made landfall in SC.In P2, vertical wind shear, atmospheric stability, subtropical high are enhanced, and the convective available potential energy is weakened (figures 6(c)-(f)), which is not conducive to TC intensification.The increased SST and water vapor conditions (figures 6(a) and (b)), which did not lead to an increase in TC intensity, were probably offset by the combined effects of these factors (figures 6(c)-(f)) and land friction.
In fact, previous studies have also found that deep convection, vertical wind shear, and midtropospheric humidity may create a more unfavorable environment for TC formation under the warming climate (Held & Zhao 2011, Sugi et al 2012), which is consistent with the results of our study.
Although the average TC intensity in SC has decreased, we found that there was no significant change in TC landfall intensity (figure not shown).Moreover, under the current climate warming conditions, the atmospheric moisture content is sufficient, so TC-induced heavy precipitation is likely to remain extreme (Guzman and Jiang 2021, Maxwell et al 2021).Sometimes, weaker TCs can also generate stronger heavy precipitation and even reach recordbreaking levels.Many of these events are residual vortex precipitation events that occur after the TC records, such as TC Haikui (2311) in 2023.Therefore, more research is needed to explore the TC activity in the SC, and it is still necessary to maintain sufficient vigilance regarding the impacts of TCs and their residual vortices.

Figure 1 .
Figure 1.Distributions in TC tracks and changes in annual TC frequency.(a) Tracks of TCs that made landfall in China; and (b) Tracks of TCs that made landfall in South China (SC).The black triangles represent the locations of the TCs when they first reach 24 kt (which is considered the TC genesis location in this study).The gray curves represent the TC trajectories, and the blue area in the map represents the SC region.(c) Annual TC frequency of TCs that made landfall in China; and (d) Annual frequency of TCs that made landfall in the SC.The black dashed lines are the averages during the period 1980-1997 and 1998-2021 respectively.The red line is the linear regression of the time series.The significance of the linear regression is noted in the up-right corner of each plot.

Figure 2 .
Figure 2. Changes in relevant parameters of TCs landfalling in SC.(a) TC destructiveness (PDI), (b) annual average duration, (c) annual average TC intensity, and (d) annual average lifetime maximum intensity.The black dashed lines are the averages for the period 1980-1997 and 1998-2021, respectively.The red line is the linear regression of the time series.The significance of the linear regression is noted in the up-right corner of each plot.

Figure 3 .
Figure 3. Changes in different percentiles of TC intensity.(a) 90th, (b) 75th, (c) 50th, and (d) 25th percentiles.The percentiles are obtained from each individual TC, and then the annual average is calculated.The black dashed lines are the averages for the period 1980-1997 and 1998-2021, respectively.The red line is the linear regression of the time series.The significance of the linear regression is noted in the up-right corner of each plot.

Figure 4 .
Figure 4. Changes in different TC intensities and proportions.(a) Tropical depression (TD), (b) tropical storm (TS), (c) categories 1-2 (CAT 12), and (d) categories 3-5 (CAT 35).The black dashed lines are the averages for the period 1980-1997 and 1998-2021, respectively.The red line is the linear regression of the time series.The significance of the linear regression is noted in the up-right corner of each plot.

Figure 5 .
Figure 5. Changes in the intensity and trajectories of TCs landfalling in the SC.Differences in (a) TC intensity, (b) latitude, (c) longitude, and (d) distance to land of LMI during the P1 and P2 subperiods.The significance of these differences is noted in the up-right corner of each plot.(e)-(h) are the same as (a)-(d), but for TC records over the entire lifetime.Distribution of TC trajectories in (i) P1 and (j) P2.The black dots represent the genesis locations of these TCs, and the red dots represent the locations where the TCs reached their LMIs.

Figure 6 .
Figure 6.Comparison of the thermal and dynamic environmental conditions during P1 and P2 subperiods.(a) SST, (b) total column water vapor, (c) vertical wind shear, (d) atmospheric stability, and (e) convective available potential energy.The shaded areas denote that the differences are significant at the 95% confidence level.(f) Geopotential height curves at 500 hPa.The blue and red curves are for subperiods P1 and P2, respectively.
figure 6(a), the SSTs in the WNP and SCS both exhibit an increase during P2.According to the Clausius-Clapeyron relation (∼7% • C −1 ), the increase in temperature will lead to more moisture entering the atmosphere, so the WNP basin also exhibited a significant increase in atmospheric moisture content during the second subperiod (figure6(b)).So why do these factors seem to be very conducive to the development and rapid intensification of TCs, but do not lead to a higher intensity of TCs landfalling in the SC region?We assume that the internal cause comes from the changes in the other conditions instead of the increase in SST and moisture factors.The results show that the vertical wind shear in the middle of the SCS showed a significant increase during P2 than that in P1 (figure 6(c)).It is well known that the formation and development of TCs are inhibited under a stronger wind shear environment.Under the background of the current climate warming, the increase in vertical wind shear in the middle of the SCS makes it difficult for TC to form and develop, resulting in a slow increase in the intensity of the TC development