Introduction

Tropical cyclones (TCs) are one of the most catastrophic atmospheric events, characterized by strong winds, heavy rainfall and storm surges. Intense TCs cause devastating losses to human life and property, particularly in the coastal regions of the most active TC basins, such as the northwestern Pacific and North Atlantic1,2,3. The question of how TC activity will respond to future climate warming is increasingly drawing the attention of science community, although significant uncertainty still remains (e.g., refs. 4,5,6,7). Observational records (e.g., refs. 8,9,10) and model simulations (e.g.6,11,12) indicate that climate warming will increase the intensity of global TCs, potentially suggesting a ‘temperature-TC intensity’ paradigm of anthropogenic climate change. Balanced against this, an alternative view proposes that the increased cyclone intensity is related to natural oscillations in the atmospheric-oceanic system13,14,15. Nevertheless, the limited length of observation records and the relatively low level of confidence in future frequency projections16,17,18 pose significant challenges to testing these hypotheses.

Coastal sedimentary archives offer a means to quantify historical fluctuations in TC activity and to uncover the main climate drivers, thus providing an alternative way to test these hypotheses. In the past few decades, various records of Holocene TC activity from the Atlantic, Pacific, and Indian oceans have been reported (e.g., refs. 19,20,21,22,23,24,25). These studies have primarily focused on detecting long-term frequency patterns of TC activity and associated climate forcings26,27,28,29,30. However, few pre-modern reconstructions have been used to understand the magnitude of past variations in TC intensity (e.g., refs. 31,32,33,34), resulting in an incomplete picture of the long-term behavior of TC activity.

In coastal and marine settings, TCs are able to deposit coarse-grained sediments in areas typically experiencing finer sediment deposition as a result of storm surge. These TC-related sediment layers, referred to as TC-event-beds, can be used to infer the energy of historical TC events considering specific assumptions on past sea levels, sediment availability, and local geomorphology35,36,37. To enhance the precision of TC intensity reconstructions for specific areas, it is important to quantify the sedimentological characteristics of modern TC-event-beds for which the atmospheric and marine conditions (e.g., storm surge height and wind speed, etc.) are known. Quantification of TC-event-bed intensity begins with a comparison of modern TC intensity with the thickness, distribution, and grain size of the event beds19,37,38. Numerical models have also been developed to help constrain the intensity of TC-event-beds31,33,39. For example, an inverse modeling approach (i.e., a simple advection-settling model) tested only by a limited number of modern TC-induced overwash layers has been proposed to quantify the magnitude of past TC-event-beds based on the height of storm surge31, and has been successfully applied at several landfall locations containing TC deposits (e.g.32,40,41,42). However, many of these conventional intensity reconstructions lack precise calibration against high-resolution instrumental records, as they do not span the entire period of the event bed records. This makes it challenging to make a direct quantitative comparison between the modern instrumental record and the long-term TC record. So far, the variability and driving mechanisms of past TC intensity remain highly uncertain, creating an urgent need for reliable proxy reconstructions of TC intensity on centennial to millennial scales. In contrast to the North Atlantic31,33 and Australia20,39, no-millennial-scale record is currently available to directly quantify TC intensity in the northwestern Pacific region.

In this study, we introduce a 2000-year-long record of TC intensity from two sites in eastern China, namely, the Jiangsu tidal flats and the Zhejiang-Fujian mud belt (Fig. 1). The deposition of muddy sediments at these sites has been both rapid and continuous over the past millennia43, enabling the reconstruction of historical TC intensity using an instrumental-calibrated technique. Specifically, we used TC intensity indices based on flow depth and wind speed (TCI_fd and TCI_ws) to quantify TC intensity at the Jiangsu tidal flats and the Zhejiang-Fujian mud belt, respectively. By integrating instrumental records and existing paleoclimate records, we evaluate the magnitude of past variations in TC intensities and associated climate forcings over the last 2000 years.

Fig. 1: Site location.
figure 1

Maps showing (a) the study area and core ZM01, and (b) the locations of cores YC01 and SA, and the GPS–RTK sections (P1 and P2). KC: Kuroshio Current; YSWC: Yellow Sea Warm Current. The white lines (a), and green and yellow areas (b) represent the major rivers discharging into the sea, modern tidal flats and exposed sand ridges, respectively.

Full size image

Results

TC intensity indices

Cores SA and YC01 were retrieved from the Jiangsu tidal flats, while core ZM01 was obtained from the Zhejiang-Fujian mud belt. As reported by Yang et al.30 core SA recorded 11 event beds spanning from 1941 to 2014 CE, corresponding to the 11 recent TCs that have impacted the Jiangsu coast (Supplementary Fig. 1 and Supplementary Table. 1). While not all of these 11 TCs directly hit the Jiangsu coast, they all caused storm surges along the coast due to their large maximum wind radius. These storm surges were recorded by a tide gauge station (Supplementary Table. 1) and ultimately contributed to the formation of the TC event layer on the tidal flats. The significant agreement between the event beds in core SA and recent TCs (Supplementary Fig. 1) strengthens our confidence in interpreting the event beds as the deposits left by TCs. TC-event-beds in core SA demonstrate a gradual-coarsening succession of sediment grain size from the upper part to the lower part of the tidal flats (Supplementary Table. 1). This suggests that the difference in grain size of TC-event-beds in different parts of the tidal flats could potentially be attributed to the influence of water depth on the transport distance. To verify this inference, we investigated the variation of grain size with seaward distance using high-accuracy positional data from two sections of the Jiangsu modern tidal flats (Fig. 1b) and the corresponding sediment grain size of surficial sediments. The results show that the mean grain size (Mz) and D90 of the P1 and P2 sections also exhibit a trend of coarsening toward the sea (Fig. 2c). Moreover, a strong logarithmic relationship was observed between D90 and relative distance (starting from the 2008 seawall). This finding, depicted in Fig. 2c, sheds light on how D90 and transport distance are connected for the TC-event-beds in different parts of the modern tidal flats.

Fig. 2: Establishment of TC intensity indices.
figure 2

Comparison (a) and correlation (b) of the sand content of core ZM01 with the instrumental TC wind speed (TCI_ws) affecting the Zhejiang coast (1984–2018 CE). The empirical relationships (c) between D90 and relative distance for P1 and P2 sections, and (d) between D90 and transport distance for TC-event-beds in core SA, both on the Jiangsu tidal flats.

Full size image

To enhance the advection-settling model (A-S model), we initially calculated the settling velocities and depth-averaged flow velocities for the grain sizes characterizing the 11 TC-event-beds in core SA. Using Eq. 2 (see Methods), we then integrated the instrumental TCI_fd to compute the transport distances of these 11 TC-event-beds. As these event beds correspond to sediment on different parts of the tidal flats (Supplementary Table. 1), we were able to determine the transport distances of the TC-event-beds in each area. The transport distances for the upper, middle, and lower parts of the tidal flats were approximately 5715 ± 428 m, 3250 ± 461 m, and 2663 ± 272 m (Supplementary Fig. 2), respectively, which exhibited an expected logarithmic relationship with the corresponding D90 values (Fig. 2d). For core YC01 in the Jiangsu coast, we selected appropriate transport distances to reconstruct TCI_fd for the 36 TC-event-beds from different parts of the tidal flats30. We also adjusted the reconstructed TCI_fd for the sea level change off the Jiangsu coast44. Furthermore, we utilized the developed wind speed-based TC intensity index (TCI_ws; Fig. 2a, b) to reconstruct TC intensity for core ZM01 located at the Zhejiang-Fujian mud belt (see Eq. 3 in “Methods” section).

To assess the accuracy of the two intensity indices, we directly compared the instrumental TC intensity with these two indices (Supplementary Fig. 3). We first reconstructed the TCI_fd of 11 TC-event-beds in core SA and compared them with corresponding instrumental TC flow depths affecting the Jiangsu coast. We, therefore, reconstructed the TCI_ws of core ZM01 for the period of 1984 to 2018 CE and compared them with the instrumental annual maximum TC wind speed affecting the Zhejiang coast. The results indicated a small relative error, on average of 6.1% and 6.2% for the two indexes, respectively, demonstrating that they have significant potential to quantify TC intensity for the Jiangsu tidal flats and Zhejiang-Fujian mud belt, respectively.

Study period and TC intensity

The top 1–10 m of core YC01 on the Jiangsu tidal flats recorded a total of 36 TC-event-beds spanning from 645 to 1950 yr BP, with a high sedimentation rate of 6.5 mm/yr on average and a temporal resolution of grain-size parameters at 4 cm spacing of about 6 years. Meanwhile, the upper 192 cm of offshore core ZM01, dated between approximately 1910 yr BP and −68 yr BP, has a temporal resolution of grain-size analyses at 1 cm spacing of about 10 years. The study period of 645 to 1910 yr BP was therefore selected to cover the overlap of the two records. Moreover, the stable sedimentary environments in the Jiangsu tidal flats and Zhejiang-Fujian mud belt, including tidal regime over the past 2 kyr30,45 permit the integration of longer TC intensity records into the instrumental record using the TCI_fd and TCI_ws indices.

The estimated TCI_fd and TCI_ws values, as well as the reconstructed TC intensity from34, exhibited similar trends despite chronological uncertainty (Supplementary Fig. 4), with average values of 5.0 m and 29.8 m/s, respectively. However, there was significant variability in these values during the study period, ranging from 2.7 to 9.8 m and 20.1 to 40.1 m/s, respectively. Interestingly, both records provide compelling evidence of a regime shift in TC intensity occurring at approximately 1440 and 1530 yr BP, respectively (Figs. 3 and S4), as detected by the popular iterated cumulative sums of squares algorithm46. Before this shift, the TCI_fd and TCI_ws varied from higher levels (6.4 m and 35.2 m/s on average) with a decreasing trend with time, but then abruptly decreased to lower levels (4.0 m and 27.5 m/s on average) with a more uniform distribution continuing to 645 yr BP. As a result of this shift, the average values of TCI_fd and TCI_ws have decreased by approximately 38% and 22%, respectively. It is worth noting that their maximum values never returned to the pre-shift average. Based on these two independent records, we conclude that the shift in TC intensity in eastern China occurred during an interval of approximately 1440–1530 yr BP, centered on approximately 1485 ± 45 yr BP. Following the shift, TC intensity in eastern China decreased ~30 ± 8% compared to that of pre-shift. The TCs in the northwestern Pacific exhibit large spatial inhomogeneity, with four clusters of TCs identified following different tracks47. The core sites along the Zhejiang and Jiangsu coasts are primarily influenced by TCs belonging to cluster 1, which account for ~34% of the northwestern Pacific TCs. These TCs generally travel in a northwest-to-northwards direction and predominantly make landfall over East Asia. As a result, we infer that the variations in TC intensity near our sites reflect the changes in the intensity of TCs affecting the east coast of China north of Zhejiang and, to some extent, the average intensity variation of cluster 1 TCs in the northwestern Pacific.

Fig. 3: Late Holocene TC intensity records in eastern China.
figure 3

Comparison of TC intensity between (a) Jiangsu coast and (b) Zhejiang coast. The gray line and gray bar show the trend and regime shift in TC intensity, respectively. RWP Roman Warm Period (~2200–1500 yr BP), DACP Dark Ages Cold Period (~1500–1150 yr BP), MWP Medieval Warm Period (~1150–700 yr BP).

Full size image

According to the TCI_fd reconstruction, two TC-event-beds along the Jiangsu coast occurring around approximately 1745 and 1840 yr BP (i.e., TCI_fd of 8.5 and 9.8 m) seem to be linked to anomalously large TC events (Fig. 3a). The TC intensity associated with these events exceeds that of recent TCs observed since 1949 CE, as the instrumental maximum TC flow depth recorded along the Jiangsu coast is only 7.4 m48. However, there is no evidence of an anomalously large TC event with TCI_ws greater than the instrumental maximum TC wind speed affecting the Zhejiang coast (Fig. 2a). This suggests that the present TC wind speed impacting the Zhejiang coast, with a maximum of 42.5 m/s, is considerably higher than the maximum reconstructed TCI_ws of only 40.1 m/s over the past 2,000 years (Figs. 2a and 3b). Nonetheless, the onset of the highest TCI_ws values of 40.1 m/s and 39.6 m/s in Zhejiang coast aligns with the timing of these two anomalously large TCI_fd values in Jiangsu coast, which took place approximately 1745 and 1780 yr BP, respectively. Consequently, relying solely on instrumental or sedimentary evidence to assess the threat of TCs to coastal communities may significantly reduce the accuracy of such assessments. The inconsistency between the two reconstructions also highlights the importance of understanding TC activity on an inter-regional scale, which is critical to ensuring the safety of life and property in coastal communities.

Discussion

Theoretically, Emanuel (1987) proposed that the maximum potential intensity of TCs would increase in a greenhouse gas-warming climate49. Furthermore, observations and modeling support the ‘temperature-TC intensity’ paradigm, indicating that TC intensity should increase with Anthropogenic climate warming due to the rise in sea surface temperature (SST)8,11,34,50. SSTs in the low-latitude region are believed to be significant factors that influence TC intensity in the northwestern Pacific, as they determine the rate and duration of TC intensification10,47. Our record of TC intensity covers a sufficiently wide range of time and temperature to test this paradigm. Our two TC intensity indices, along with the tropical SST reconstructions from the Indo-Pacific warm pool, reveal that the highest reconstructed TCI_fd and TCI_ws both occur during the middle phase of the RWP (Fig. 3). Additionally, the observed 1485 ± 45 yr BP shift in TC intensity corresponds to the transition from the RWP to the DACP. However, this shift does not occur during the transition from the DACP to the MWP (Fig. 4a). Notably, there was no significant change in tropical SST magnitude before and after the 1485 ± 45 yr BP shift. For instance, while SST returned to and exceeded its pre-shift level during the late MWP (i.e., 700–950 yr BP), the TC intensity remained low. This makes it difficult to attribute the intensity shift to the temperature paradigm, as tropical SST did not undergo a shift before and after 1485 ± 45 yr BP. This is in line with observational studies, which suggest that the thermodynamic effect of the tropical SST alone cannot account for the documented changes in TC intensity (i.e., the annual accumulated power dissipation index) over the past 30 years51. Instead, this finding raises questions about whether other key large‐scale climatic factors altered the state of TC intensity at around 1485 ± 45 yr BP. Studies relying on sedimentary archives indicate that long-term variations in the frequency, intensity and track of TC activity are tied with some key large-scale oceanic and atmospheric factors, such as El Niño–Southern Oscillation (ENSO), the Atlantic Meridional Overturning Circulation (AMOC), and the Sahara and West African Monsoon (WAM) (e.g., refs. 21,22,45,52,53,54). Reconstructions of past climate show some shifts in the magnitude of these factors during the past 2000 years (e.g., refs. 55,56,57,58,59,60,61), which may have caused a shift in TC intensity along the east coast of China. To test this hypothesis, we compared the reconstructions of TC intensity with proxy records of these factors.

Fig. 4: Comparisons of TC intensity reconstructions with other paleoclimatic records.
figure 4

a TC intensity from the Jiangsu and Zhejiang coasts. Z-score was realized by averaging the Z-cores of these two records. b SST reconstructions from the Indo-Pacific warm pool55,89 (c) Reconstructed ENSO from the equatorial Pacific55,56 and the TRACE simulation90. d AMOC reconstructions from the Bermuda Rise57 and northern Iceland Basin58. e Reconstructed Saharan dust and vegetation from West and East Africa59,60. f Reconstruction of the WAM from equatorial Cameroon61. The gray bar shows the shift in TC intensity in eastern China. The shifts in the magnitude of ESNO, AMOC, and Saharan dust were detected using the popular iterated cumulative sums of squares algorithm46.

Full size image

ENSO is widely recognized as one of the primary drivers of global TC variability, significantly impacting TC activity in the northwestern Pacific, Australia, and the North Atlantic27. During El Niño years, when SSTs in the central and eastern equatorial Pacific are higher than normal, the mean genesis location for TCs in the northwestern Pacific generally shifts to the southeast, resulting in longer lifetime and greater intensity of TCs; In La Niña years, there is a tendency toward more short-lived TCs, many of which do not reach high TC intensity62,63. Based on the best-track dataset for the northwestern Pacific region for the period 1950–2002 CE, Camargo and Sobel (2005) suggested that the lifetime, intensity, and number of TCs contribute significantly to the ENSO signal in TC intensity, although the lifetime effect appears to be the most important of the three63. It seems that we can speculate that the mean lifetime of TCs tends to increase during El Niño phase as more TCs form in the southeast quadrant of the northwestern Pacific region. As a result, TCs will experience a longer traveling time (westward and northward) before encountering the continent or colder mid-latitude water, giving them a greater chance to develop higher intensity. Favorable atmospheric conditions (e.g. greater vorticity and weaker vertical wind shear) also exist in the southeast quadrant of the northwestern Pacific during El Niño years, which favor TC intensification64. However, the intensity effect in particular could also be due, to some extent, to other influences of ENSO on the mean regional climate of the northwestern Pacific63. Recent research has shown that during El Niño years, upper ocean heat content decreases by as much as 20%–30% (as compared to normal) over the northwestern Pacific65. This offsets the favorable atmospheric conditions and longer tracks, but the end result is still a slightly higher than normal TC intensity during El Niño years. Additionally, the AMOC plays a crucial role in redistributing heat on Earth and has a major impact on climate. Observational and modeling studies suggest that the weakening of AMOC leads to strong ENSO variability through the background warming over the central equatorial Pacific, resulting in an El Niño-like pattern in the tropical Pacific66,67,68. This pattern also contributes to the enhancement of TC intensity in the northwestern Pacific region. Our reconstructions demonstrate a consistent relationship between higher TC intensity and the strengthening of ENSO, as well as the weakening of AMOC (Fig. 4c, d), as evidenced by the observational data63,68. Specially, prior to the 1485 ± 45 yr BP shift, a gradual decrease in TC intensity is linked with a gradual decrease and increase in ENSO and AMOC, respectively. Conversely, lower and stable TC intensity after 1485 ± 45 yr BP coincides with weaker ENSO and stronger AMOC. Additionally, the slight increase in TC intensity after the 1485 ± 45 yr BP shift is coeval with a slight increase and decrease in ENSO and AMOC, respectively. It is noteworthy that the magnitude of ENSO and AMOC did not return to their pre-shift levels after 1485 ± 45 yr BP, suggesting that they are potentially key factors controlling the shift in TC intensity along the east coast of China.

Observational and numerical simulations have shown that vegetation cover and dust emission over the Sahara can impact TC intensity by triggering changes in the atmospheric circulation that affect the entire tropical region52,54,69. Their results suggest that TC intensity is positively correlated with Sahara vegetation and negatively correlated with Saharan dust concentration, both of which provide more favorable conditions for TC development52. Studies have demonstrated that Saharan dust, acting as cloud condensation nuclei, can lead to a significant reduction in the intensity of an idealized TC; This reduction is caused by the redistribution of precipitation and latent heating to more vigorous convection in the storm periphery that cools the low levels and disrupts the inflow of energy to the eyewall, hence making the eye larger and the maximum winds weaker69,70. Previous studies have also indicated that approximately one quarter of dust cases in East Asia between 2007 and 2020 CE originating from the Sahara Desert71, primarily distributed in the upper troposphere, influence TC intensity through aerosol-cloud-radiation interactions70. So that, prior to the shift in our TC intensity reconstruction, lower Saharan dust emissions were associated with higher TC intensity (Fig. 4a, e), which is consistent with mid-Holocene TC modeling studies52. The relatively stable and high Saharan dust emissions after 1485 ± 45 yr BP may have contributed to TC intensity remaining at a low level, also in line with our intensity reconstructions. Furthermore, the magnitude of Saharan dust after the shift did not return to pre-shift levels, suggesting its potential role in controlling this intensity shift.

A gradual decrease and subsequent increase in Sahara greening before and after the 1485 ± 45 yr BP shift would result in corresponding changes in TC intensity, which can roughly explain the trends in our TC intensity records. However, the magnitude of Sahara greening did not produce a significant change before and after the shift, which would not account for the significant shift between higher and lower TC intensity (Fig. 4a, e). Notably, the greening of the Sahara reached its peak around 900 yr BP, but TC intensity did not reach its highest level during that time. This contrasts with the findings of Pausata et al.52, which suggest that the impact of Sahara greening is more significant for TC activity than the role of reduced dust loadings. Pausata et al.52 examined the effect of Sahara greening and dust on TC activity during a warm climate state (mid-Holocene, 6000 yr BP) characterized by increased boreal summer insolation, while our TC intensity records focus on the past two millennia marked by decreased boreal summer insolation. The different climate state may cause the role of Sahara greening and dust on TC activity to vary across different time scales. For instance, the influence of Saharan dust on TC activity in the North Atlantic varies at different time scales54, suggesting that substantial increases in Saharan dust over the last 200 years and in the early Holocene are coeval with higher Atlantic TC activity, but are inversely correlated at other times over the last 3000 years.

Previous studies have also shown a correlation between the strength of WAM and TC intensity, suggesting that an intensified WAM increases TC intensity by enhancing cyclonic vorticity21,52,72. This correlation is consistent with our reconstructions, linking higher TC intensity before the 1485 ± 45 yr BP shift with a stronger WAM (Fig. 4a, f). However, the gradual weakening of the WAM after the shift contradicts our TC intensity records, as it would lead to a corresponding decrease in TC intensity. For example, the WAM reaches a maximum around 1300 yr BP within the study interval, but the corresponding TC intensity does not follow this trend. In addition, recent observational diagnostics and numerical simulations by Xu et al.73 show that the increasing intensity of TCs in the northwestern Pacific over the past decades is partly driven by the warming of the central-eastern Tibetan Plateau by reducing vertical wind shear within the monsoon trough area of the tropical northwestern Pacific. However, after comparing our reconstructed TC intensity with the temperature record from the central-eastern Tibetan Plateau over the past two millennia (ref. 74; Supplementary Fig. 5), we have observed that increased TC intensity in eastern China corresponds to the warming of the Tibetan Plateau, particularly since 1600 yr BP. However, it is not the primary factor driving the shift in TC intensity around 1485 ± 45 yr BP, as its temperature did not shift at the same time. The aforementioned lines of evidence suggest that Sahara greening, the WAM, and the temperature of the Tibetan Plateau are unlikely to be the key factors controlling the shift in TC intensity in eastern China.

In summary, this study presents a late Holocene record of TC intensity in eastern China using an instrumental-calibrated technique, which allows for the calibration of long-term records against high-resolution instrumental TC records. The record reveals a significant regime shift in TC intensity around 1485 ± 45 yr BP along the east coast of China, marked by a 30 ± 8% reduction in intensity. This reduction can be attributed to substantial changes in the magnitudes of ENSO, AMOC, and Saharan dust, which may compensate for the effects of the ‘temperature-TC intensity’ paradigm. Moreover, as we assess the impacts of climate change, there is an increasing focus on “tipping elements” –components of the Earth’s climate system where state shifts are likely to occur75. Interestingly, all three factors that we proposed to influence the shift of TC intensity belong to these tipping elements of the climate76,77,78. It is suggested that ENSO, AMOC, and Saharan dust may have crossed a similar tipping point around 1485 ± 45 yr BP, as their magnitudes have not returned to their pre-shift levels since then. This may have resulted in their influence on TC intensity exceeding that of temperature by triggering changes in the oceanic and atmospheric state within the tropical Pacific the region where TCs originate. These findings highlight the importance of understanding TC intensity variability over long time scales and under different climatic conditions. Further research is needed to evaluate TC intensity variability at different time scales in different basins and globally, as well as its connection with climate tipping elements. Additionally, the instrumental-calibrated technique offers the possibility of predicting future trends in TC activity under changing climate conditions. It enables the assessment of changes in TC activity from a long-term perspective, allowing for the discrimination between natural variability and anthropogenic changes in TC activity. We expect this technique to serve as a starting point for more accurate and quantitative analysis of paleotempestology and other paleoclimatology on a global scale.

Methods

Study site and field sampling

The study focuses on the Jiangsu tidal flats and Zhejiang-Fujian mud belt (Fig. 1), located on the eastern coast of China, as these areas frequently intersect with TC paths. Core ZM01 (28°41.4′ N, 122°24.6′ E) is 5.08-m-long core was collected from the Zhejiang-Fujian mud belt in 2018 CE (Fig. 1a). Core YC01 (33°23.2′N, 120°12.3′E) and core SA (33°34.7′N, 120°33.3′E) are 39.75-m-long and 1.93-m-long cores, respectively, recovered from the Jiangsu coastal plain and modern tidal flats in 2014 CE (Fig. 1b). Additionally, two sections of the Jiangsu modern tidal flats (Fig. 1b) were surveyed in 2008 CE (section P1) and 2009 CE (section P2) using a Magellan Z-MAX GPS-RTK (a differential GPS system with a dynamic accuracy of 10 mm ± 0.5 ppm), resulting in high-precision positional data. Thirty-seven surficial sediment samples were also collected along these two sections.

Laboratory analysis

Cores ZM01, YC01, and SA were sliced into intervals of 1 cm, 4 cm, and 1 cm, respectively. All subsamples from the three cores and thirty-seven surficial sediments were measured for grain size using a laser Malvern Mastersizer 2000 with a duplicate measurement error of less than 3%. Grain size parameters were then calculated from the distribution curves using moment statistics. The age model for cores ZM01, YC01, and SA, presented by Yang et al.30,45, were established using two isotopic dating methods. Twenty-one and thirteen samples from the top of core ZM01 and core SA were selected for 210Pb analysis to quantify the sedimentation rate. The centennial-to-millennial scale chronologies of cores ZM01 and YC01 were constrained using eight and seven 14C-Accelerator Mass Spectrometry (14C AMS) dates, respectively. The top 1 m of sediments in core YC01 consists of yellowish-brown sandy silt, belonging to the supratidal zone, which was insensitive to recording TC events and therefore was not included for analysis.

Determination of past TC intensity

Sedimentary systems in shelf-coastal environments, such as tidal flats and shelf mud belts, require different methods to delimit the intensity of TC-event-beds79. To address this issue and enable direct quantitative comparisons between instrumental and long-term TC records, we have developed two TC intensity indices for the Jiangsu tidal flats and Zhejiang-Fujian mud belt using a technique that combines instrumental and sedimentary records. This technique allows us to calibrate long-term records of past TC intensity against high-resolution instrumental TC records.

Jiangsu tidal flats

For the Jiangsu tidal flats, Yang et al.30 presented a 2 kyr continuous activity record of TCs by identifying 36 coarse-grained event beds in core YC01. However, this work focused on revealing the frequency of TC-event-beds without quantifying the intensity of individual event beds. The magnitude of past TC-event-beds can be reproduced using a simple A-S model31,80,81. The A-S model is based on the balance between longitudinal sediment transport by the flow and gravity-driven sediment settling through the water column. It assumes that the distance that grains are advected longitudinally from the top of the flow to the bed depends on flow depth, flow velocity, and settling velocity82:

$$\frac{h}{{w}_{s}}=t=\frac{{x}_{L}}{U}$$
(1)

where h is flow depth during TC-induced flooding, ws is the still–water particle settling velocity, t is settling time, and xL is the advection length scale for particles of a given grain size. U is depth-averaged flow velocity and can be calculated using the equation in Moore et al.82. For a given grain size and shape, ws can be calculated using equation in Ferguson and Church83. In this analysis, we determined the settling velocity for the D90 size class (defined as the grain size for which 90% of sample has smaller grain sizes) as it best reflects the maximum grain size transported by flooding events associated with TCs. The availability of coarse sand on the offshore sand ridges84 presumably allows the assumption that D90 is controlled by flow.

The A-S model depends on the supercritical flow occurring along the backside of a barrier (i.e., Froude number Fr = U/(gh)0.5 = 1, g is acceleration due to gravity), but could theoretically apply to at any transition to supercritical flow, including tidal flat environments42,85. Woodruff et al.31 assumed a constant xL during storm surges to yield a unique solution for quantifying the flow depth over a barrier during flooding. This assumption is appropriate for environments where the topographic or bathymetric changes are insignificant over time, such as coastal lagoons and lakes32,40,42. However, in meso- to macro-tidal settings like Jiangsu tidal flats, water depths are variable, and sediment grain sizes differ in different parts of the tidal flats86. Therefore, a constant xL is not suitable for calculating the flow depth during flooding in different parts of the tidal flats. As a result, Eq. 1 requires an additional water depth constraint to estimate the flow depth during flooding for different parts of the tidal flats (Supplementary Fig. 2).

To enhance the A-S model, we hypothesized that transport distance is dependent on water depth, as different parts of the tidal flats have varying xL. By analyzing the TC-event-beds identified in core SA and the corresponding instrumented flow depths during TC-induced flooding87,88, we were able to determine the transport distances for different parts of the Jiangsu tidal flats. Taking into account sea level variations, we can reconstruct the flow depth-based TC intensity (TCI_fd, m) recorded in core YC01 using the improved A-S model that incorporates different transport distances. The model can be expressed as follows:

$${\rm{TCI}}\_fd={\left(\frac{{{x}_{L}}^{2}{{w}_{s}}^{2}}{g}\right)}^{1/3}$$
(2)

Zhejiang-Fujian mud belt

Yang et al.45 developed a simple yet effective method for core ZM01 from the Zhejiang-Fujian mud belt by correlating sediment grain size with instrumental records of TC-induced wind speed. By combining the instrumental and sedimentary records, they discovered a significant and positive correlation (R = 0.86, P < 0.001, n = 35) between the content of the sensitive coarse-grained fraction (i.e., >63 μm fraction; sand content) in core ZM01 and the annual maximum wind speed of TCs that impacted the Zhejiang coast (120–124°E, 26–30°N) from 1984 to 2018 CE (Fig. 2a). However, the reconstruction of Yang et al.45 based only on the sand content cannot directly quantify the intensity changes of TCs over the last 2000 years. To address this issue, we developed an index called the wind speed-based TC intensity index (TCI_ws, m/s), which is based on the relationship between the sand content and TC wind speed. The index can be expressed as follows:

$${\rm{TCI}}\_ws=6.3449\ast \,\mathrm{LN}(sand\,content)+31.52$$
(3)