Introduction

Tropical cyclones (TCs) are among the most severe natural disasters in terms of human casualties and direct economic losses1,2,3,4. The global average annual direct economic loss from TCs is about US$79 billion for the period 2010–20195; and the average annual death toll is about 6600 people between 1989 and 20196. TCs worldwide show a significant upward trend in direct economic losses, closely linked to their increasing intensity, growing economic exposure, and changing vulnerability7.

Limited by the resolution of TC-induced direct economic loss records (e.g., provincial scale or TC event scale records)8,9, it is difficult to conduct high-precision spatial and temporal analysis of TC risk in China. Understanding the economic exposure to TC is a vital factor to accurately assess TC-induced risk10,11. It is undeniable that the economic exposure to TCs is steadily increasing as people move to exposed regions along the coasts and the major urban centers and as the number and value of assets built in TC-exposed areas increase12. Meanwhile, in recent decades, systematic changes in TC activities are observed globally under changing climate, including the poleward shifts of TC maximum intensity to higher latitudes13 and the slower decay of landfalling TCs14, etc. Those systematic changes will pose a greater challenge to higher latitude and inland areas, which are usually less prepared for TCs and therefore more vulnerable to TC damage. But unexpected extreme events will likely generate surprising catastrophes, particularly in those unprepared regions. For example, a wake-up call from a TC-induced flooding event in Henan, a northern inland province of China, affected more than 14 million people, killed 398 people, and inundated 16 million hectares of crops, causing direct economic losses of US$20.69 billion and even higher magnitude of indirect losses in 202115. Therefore, understanding spatial and temporal patterns of present and future TCs’ economic exposure will substantially help us to determine which locations should be targeted for the mitigation and adaptation activities to alleviate their impact8,16.

China is one of the countries that are most severely affected by TCs, with an average of seven landfall TCs and about US$9 billion of direct economic losses annually. Considering China’s rapid urbanization and population shift to coastal cities in recent decades, it is one of the countries mostly impacted by TC hazards. Therefore, this article will investigate changes in the spatial and temporal patterns of economic exposure to TCs in China and factors contributing to those changes. We will also discuss the possible reasons for the differences in the degree of response of TC economic exposure to TC-induced direct economic losses in different regions.

Results

Northward shift of TC economic exposure in China

TC economic exposure is defined as the GDP in the TC impact area (details see Methods and Supplementary information), and the annual TC economic exposure is the sum of the economic exposure of all TCs each year. Here we first identified a rapid upward trend in the annual total TC economic exposure of China, with the average value of US$6.46 trillion, US$10.38 trillion, and US$20.45 trillion in the three sub-periods 2006–2010, 2011–2015, and 2016–2020, respectively (at the 2015 price level; Fig. 1a). Meanwhile, we also discovered that the TC economic exposure for the entire country is moving to higher latitudes (Fig. 1b, c), with a northward shifting ratio of 19.71 km per year for the exposure centroid from 2006 to 2020 (R2 = 0.29, p-value < 0.05). The economic exposure of TC in the sub-period 2006-2010 (Fig. 1d) is mostly lower than the 2006-2020 whole period average, while the exposure in the sub-period 2016–2020 (Fig. 1f) is mostly higher than the whole period average. In the sub-period 2011–2015, the area of TC economic exposure below and above the whole period average is roughly equivalent (Fig. 1e). Spatially, the total area of high value (e.g., above US$5.0 billion) of TC economic exposure shows a tremendous expansion from 2006 to 2020 (Supplementary Fig. 1). In the sub-period 2006-2010, the TC economic exposure of more than US$5 billion is concentrated in the eastern and southern coastal regions of China, while there is no area shown in northern and inland China (Supplementary Fig. 1b). In the sub-period 2011-2015, the region of the TC economic exposure of more than US$5 billion expands to some inland areas (e.g., Hunan, Jiangxi and Hubei, etc., details can be found in Supplementary Fig. 2) and northern China (e.g., Shandong and southern Liaoning, etc.) (Supplementary Fig. 1c), where it is particularly noteworthy that the relative change of TC economic exposure to the average level of more than 50% in most regions of Jilin and Liaoning (Fig. 1e and Supplementary Fig. 2). In the sub-period 2016-2020, there is clearly more expanded regions with TC economic exposure of more than US$5 billion (Supplementary Fig. 1d), especially in the northern China, such as Shandong, Henan, Beijing, Tianjin, Hebei, and Liaoning, etc., where the relative change of the TC economic exposure to the average level is more than 75% (Fig. 1f).

Fig. 1: Time series and changes in spatial patterns of economic exposure to TCs.
figure 1

a Time series of annual TC economic exposure of China. b Time series of the distance between equator and the annual TC economic exposure centroid of China. c Footprint shifting in the centroid of annual economic exposure to TCs in China in three sub-periods 2006–2010, 2011–2015, and 2016–2020. The relative changes of TC economic exposure in three sub-periods (d, 2006-2010. e, 2011-2015. f, 2016-2020) to the average values in the period 2006–2020. In panel a, we also show the average values of TC economic exposure in the three sub-periods (red lines). In panel b, we also show the linear regression line (solid black line), and the 95% confidence intervals (gray shade).

Full size image

China’s regional TC economic exposure also shows significant increases in 2006–2020 (Fig. 2). The temperate zone (between 40°N and 66.5°N; Fig. 2a) has average exposure of US$0.07 trillion, US$0.38 trillion, and US$0.93 trillion in three sub-periods of 2006–2010, 2011–2015, and 2016–2020, respectively. The average exposure for the subtropical zone (between 23.5°N and 40°N; Fig. 2c) are US$4.04 trillion, US$7.14 trillion, and US$14.14 trillion in the same three sub-periods. And the tropical zone (between 0° and 23.5°N; Fig. 2e) observe US$2.36 trillion, US$2.86 trillion, and US$6.33 trillion in average exposure in three sub-periods. Therefore, the increase of TC economic exposure is a nation-wide phenomenon over China. The distance to the equator of the annual TC economic exposure centroid in the subtropical zone in China shows a significant northward shift trend, with 13.31 km per year (Fig. 2f; R2 = 0.30, p-value < 0.05). Those results imply that the northward shift in TC economic exposure is particularly evident in the subtropical zone, where the degree of influence of TCs has intensified from 2006 to 2020. The distance to the equator of the annual TC economic exposure centroid in the temperate zone in China shows no significant trend due to the generally low TC frequency (Fig. 2a) and short period (15 years) of observation time. However, the recent dramatic increase in TC economic exposure in this region and the global northward shift of TC intensity13 suggests a potential northward spreading of risk to this region. The distance to the equator for the centroid of annual TC economic exposure in tropical zone shows no change (Fig. 2f), possibly can be attributed by this region’s proximity to equator. The consistently positive contribution from the temperate and subtropical zones to the poleward shift of TC economic risk suggests that the northward expansion of TC economic risk is a China-wide phenomenon, although the regional differences are evident.

Fig. 2: Time series of economic exposure to TCs in three different zones of China.
figure 2

a,c,e are annual total TC economic exposure and b, d, f are distances between the equator and the annual TC economic exposure centroids. a, b Temperate zone in China. c, d Subtropical zone in China. e, f Tropical zone in China. In panels a, c, and e, red lines are the average values of TC economic exposure in the three sub-periods (red lines). In panels b, d, and f, solid black lines are linear regression lines, and gray shades are the 95% confidence intervals .

Full size image

Factors contributing to northward shift of TC economic exposure

We identified a northward shift in the TC economic exposure centroid for China from the previous section. Since the TC economic exposure combines information from both the TC impact area and GDP distribution12, here we constructed regression models to understand how shifts in TC frequency and GDP contribute to the TC exposure shift, both individually and collectively. Individual models show that he northward shift of TC economic exposure centroid is more than twice as sensitive to that of the GDP distribution as to that of the annual TC impact distribution. The TC economic exposure centroid distance to equator increases at a ratio of 0.43 (R2 = 0.68, p-value < 0.05) with the annual TC frequency centroid distance to the equator, while the same ratio to the GDP centroid distance to equator is at a 0.89 (R2 = 0.74, p-value < 0.05), as shown as two fitted lines in Fig. 3. If we considering both factors together in a regression model (Supplementary Table 1), the TC frequency has a slope of 0.24 and the GDP has a slope of 0.58. The northward shift of the centroid has a speed of 46.53 km per year (R2 = 0.44, p-value < 0.05) for the annual TC frequency distribution but 13.35 km per year (R2 = 0.14, p-value = 0.16) for the GDP distribution from 2006 to 2020 (Supplementary Fig. 3 and Supplementary Table 2). Based on our estimation, the TC impact contributed a northward shift of 11.22 km per year, the GDP distribution contributed a northward shift of 7.75 km per year, and their joint change contributed a northward shift of 0.74 km per year to the northward shift of TC economic exposure in China during the period 2006-2020 (19.71 km per year; details see Methods). The result shows that the shift of TC impact distribution is the main cause for the northward shift of TC economic exposure in China, followed by the shift of GDP distribution.

Fig. 3: Effect of TC impact distribution and GDP distribution on TC economic exposure.
figure 3

Scatters DISTCESS versus DISTCGDP (red) and DISTCESS versus DISTCIF (blue). We also show the linear regression line (solid line), and the 95% confidence intervals (shaded). DISTCEE: Distance between the equator and the centroid of annual TC economic exposure (unit: km); DISTCIF: Distance between the equator and the centroid of annual TC impact distribution (unit: km), DISTCGDP: Distance between the equator and the centroid of GDP distribution within the spatial range where the number of TC impact was at least 1 in given year (unit: km).

Full size image

Conclusion and Discussion

The northward shift of TC economic exposure can be partially explained by the systematic change in the global TCs behaviors, such as the recently discovered poleward shifting of TC maximum intensity of TCs13, and partially explained by the recent northward movement of China’s economy hot spots17. In the recent 15 years, the area of TC impact is rapidly expanding towards northern China, with its centroid moving northwards at a rate of 46.53 km per year. The frequency of TCs causing direct economic losses to the northern provinces of China (Shandong, Henan and further northern provinces) shows an increasing trend, with a total frequency of 2, 14, and 37 in the three sub-periods (2006–2010, 2011–2015, and 2016–2020), respectively (CMA various years, e.g., CMA 201518). For similar TC intensities, northern regions tend to suffer higher direct economic losses than southern regions, even though the TC economic exposure in northern regions is lower than that in southern regions. For example, Typhoon Maysak (No. 2009) in 2020 had 10-minute maximum sustained winds of 45.91 knots and maximum daily precipitation of 124.87 mm in Jilin, and Typhoon Wipha (No. 1907) in 2019 had 10-minute maximum sustained winds of 45.99 knots and maximum daily precipitation of 127.29 mm in Guangdong, which were about the similar magnitude of TC hazards, but the direct economic loss in Jilin (US$568.33 million) was much higher than the direct economic loss in Guangdong (US$19.17 million), although the economic exposure in Jilin (US$105.94 billion) was much lower than that in Guangdong (US$1,333.76 billion). Thus, the concerns of some recent global scale studies13,14 is becoming realized regionally in China: higher latitudes are less prepared for TCs and therefore more susceptible to TC economic loss than lower latitudes, exacerbated by the increased risk from TC northward shift.

As compared with most previous studies8,9,10,13,14,19,20 independently focused on changes in physical TC risk or in TC economic impact/exposure, our analysis holistically quantified both the changes in the physical TC impact and changes in economic exposure to TCs over China. We find that the northward shift of TC economic exposure centroid is more sensitive to the GDP distribution shift than to the annual TC impact shift. At the same TC intensity, TC economic exposure at high latitudes is more likely to turn into actual direct economic losses due to a poor preparedness for TC disaster reduction at high latitudes. Therefore, the increase in GDP will result in more severe direct economic losses when higher latitudes are affected by TCs. Our result demonstrates a high confidence that the total TC economic exposure in the temperate zone in China has increased during 2006–2020 (Fig. 1d–f, Fig. 2a, and Supplementary Fig. 1), with strongest increasing pattern in the last 5 years. Both increases in TC impact and TC economic exposure may lead to the negative impacts on the economic development of the higher latitude zone in China. Therefore, we need to pay particular attention to disaster mitigation and adaptation for possible extreme weather events triggered by TCs in northern China. For instance, there were no disaster event triggered by TCs until 2010 from the record from the perspective of Northeast China (Heilongjiang, Jilin, and Liaoning), but there were three severe TC-induced disaster events (e.g., with direct economic losses exceeding US$100 million) between 2011 and 2015. And the number of severe TC-induced disaster events went up to seven in 2016-2020.

In the future, the adverse impact of TCs on China’s economic development will likely to be further intensified21. TC locations and intensity will likely expand further northward in the western North Pacific22,23. Worse still, there is a high degree of confidence that TC intensity will continue to be stronger24. Meanwhile, China’s economy will continue to grow25, including the northern China. Those predictions suggest that the potential risk of TC-induced economic losses will likely continue to expand further northward in China. Since there is a lack of capacity and willingness to implement TC risk reduction policies because of very few previous experiences of TC disasters in northern China12, disaster mitigation and adaptation could be challenging with the changing risk into future. Our analysis provides a quantitative assessment of the changing TC risk in the recent 15 years and provides a reference for national disaster management. It is imperative that local governments in China can benefit from taking measures and preparing to mitigate the effects of TC-induced disasters based our estimate of changing pattern in TC economic risk. In addition, mitigating global warming by seeking CO2 reduction pathways will reduce intensity and frequencies of extreme TCs in the future TCs4, which in turn alleviate their economic impact to the society.

Wind field model is a very important component in our analysis for the TC exposure. Therefore, we finished a separate sensitivity analysis to test the stability and reliability of our results by using a different wind field model (Chavas model: CLE15)26,27. We observed the same rapid upward trend in the annual total TC economic exposure of China, with the average value of US$6.34 trillion, US$10.21 trillion, and US$19.71 trillion in the three sub-periods 2006–2010, 2011–2015, and 2016–2020, respectively (Supplementary Fig. 4a). Meanwhile, we also discovered that the TC economic exposure for the entire country is moving to higher latitudes (Supplementary Fig. 4b), with a northward shifting ratio of 19.44 km per year for the exposure centroid from 2006 to 2020 (R2 = 0.29, p-value < 0.05). In addition, we also examined the annual TC-mean economic exposure to fully understand how annual TC frequency influences our interpretation of the results. Here we find that there is also a rapid upward trend in the annual TC-mean economic exposure of China, with the average value of US$0.54 trillion, US$0.74 trillion, and US$1.18 trillion in the three sub-periods 2006–2010, 2011–2015, and 201–2020, respectively (Supplementary Fig. 5a). Meanwhile, the trend of the distance between equator and the annual TC-mean economic exposure centroid of China (Supplementary Fig. 5b) is very similar to the trend of the distance between equator and the annual TC economic exposure centroid of China (Fig. 1b). Therefore, our findings are consistent with different wind field models and the annual TC frequency has minor effect on the trends of the TC exposure in our study period. In the future research, we plan to integrate the storm surge damage28,29,30 into our current method to estimate the fully integrated TC multi-hazards exposure, as well as use a similar framework to estimate how TC exposure over China will likely to change under warming climate.

Methods

Data

The best track dataset for TCs from 2006 to 2020 was acquired from the China Meteorological Administration31. The precipitation dataset of the Multi-Source Weighted-Ensemble Precipitation (MSWEP) were acquired from the GloH2O platform32,33. GDP data in China with 1 km×1 km grids in 2005, 2010, 2015, and 2019, were acquired from the Resource and Environment Data Cloud Platform of China34. Due to a lack of reported asset distributions for other years (2006–2009, 2011–2014, and 2016–2018), we assume that the GDP ratio is constant within each 5 years, therefore the GDP data for the missing years were linearly interpolated10. The recorded TC affected counties data used in MTDC was obtained from 3 sources: (1) the research articles8,35; (2) the annual Climate Bulletin of Guangxi Province and news reports, which contain information on the affected counties by each TC in Guangxi Province over the period 2012–2019; (3) the Atlas of Major Natural Disasters in China in 2010, which was published by the Surveying and Mapping Press and China Map Press, and contains information on the affected counties by Super Typhoon Fanapi in 201036. Consumer Price Index (CPI) from 2006 to 2020 was acquired from China Statistical Yearbooks Database. All data have been aligned to the same time span between 2006 and 2020.

Wind field model

The TC track data from the CMA does not have enough temporal resolution to cover the TC affected area. Therefore, we interpolate 35 points with equal distance between every two consecutive TC recorded points by a linear spline method8, providing a temporal resolution of 10 minutes. The gradient balance velocity \({V}_{g}\) for a stationary cyclone is calculated by using the Holland wind field model37 as follows:

$${V}_{g}=\sqrt{\frac{B({P}_{a}-{P}_{c})}{\rho }\times {\left(\frac{{RMW}}{r}\right)}^{B}\times {e}^{-{\left(\frac{{RMW}}{r}\right)}^{B}}+{\left(\frac{r\times f}{2}\right)}^{2}}-\frac{\left|r\right|\times f}{2}$$
(1)

where \(B\) is the parameter of air pressure profile; \({P}_{a}\) and \({P}_{c}\) are the pressures at the periphery and the cyclone center, respectively; \(\rho\) is the air density; \({RMW}\) is the radius to maximum wind speed; \(r\) is the distance between the cyclone center and target point; and \(f\) is the Coriolis parameter.

We then calculate the wind speed near the surface. According to Harper et al. 38, the wind speed at 10-m surface (\({V}_{z}\), 10-minute maximum sustained winds) is calculated according to Eq. (2):

$${V}_{z}=\left\{\begin{array}{c}\begin{array}{cc}0.81\times {V}_{g}, & {V}_{g}\, < \,6.00m/s\\ \left[0.81-2.69\times {10}^{-3}\times ({V}_{g}-6.00)\right]\times {V}_{g}, & 6.00m/s\,\le \,{V}_{g}\, < \,19.50m/s\end{array}\\ \begin{array}{cc}\left[0.77-4.31\times {10}^{-3}\times ({V}_{g}-19.50)\right]\times {V}_{g}, & 19.50m/s\le {V}_{g} \,< \,45.00m/s\\ 0.66\times {V}_{g}, & {V}_{g}\,\ge\, 45.00m/s\end{array}\end{array}\right.$$
(2)

Other wind field model for sensitivity tests

In this study, we also model and run the analysis using CLE15 model26,27 to estimate TC economic exposure. CLE15 model is given as:

$${\left(\frac{{M}_{{inner}}}{{M}_{m}}\right)}^{2-\left(\frac{{c}_{k}}{{C}_{d}}\right)}=\frac{2\times {\left(\frac{r}{{R}_{m}}\right)}^{2}}{2-\frac{{c}_{k}}{{C}_{d}}+\frac{{c}_{k}}{{C}_{d}}\times {\left(\frac{r}{{R}_{m}}\right)}^{2}}$$
(3)
$$\frac{\partial {M}_{{outer}}}{\partial r}=\chi \frac{{\left(r\times V\right)}^{2}}{{r}_{0}^{2}-{r}^{2}}$$
(4)

where \({M}_{{inner}}\) and \({M}_{{outer}}\) are the angular moment in the inner and outer wind regimes, \({M}_{m}\) is the angular momentum at \({R}_{m}\), \(\frac{{c}_{k}}{{C}_{d}}\) is the ratio of the exchange coefficients of enthalpy and momentum, \(\chi\) is calculate by the surface drag coefficient and the magnitude of the radiative-subsidence rate in the free troposphere, and \({r}_{0}\) is the radius of vanishing wind. The inner and outer wind profiles are connected at a merge point where the angular momentum and its radial derivative are mathematically continuous.

TC precipitation

TC precipitation usually refers to precipitation within a certain range from the center of a TC, and most studies consider the range of 500 km radius39,40 because it covers most of precipitation related to TCs. This study also uses the radius of 500 km, namely all precipitation within 500 km of the TC center is defined as TC precipitation.

TC impact area extraction

TC economic exposure is usually defined as the GDP in the TC impact area8. Therefore, it is necessary to determine the TC impact area before extracting economic exposure of TC. In this study, we propose a Methodology for setting hazard intensity Thresholds by considering the Distance from Coastline (MTDC) to extract the TC impact area.

Methodology for setting hazard intensity Thresholds by considering the Distance from Coastline (MTDC)

Two types of methods based on hazard types are normally considered when determining the TC impact area, namely, wind-precipitation method and wind method. The wind-precipitation method defines the TC impact area as the area where the wind speed or precipitation is greater than a fixed intensity threshold; while for the wind method defines the TC impact area as the area with wind speed above a fixed intensity threshold8,10. Tropical storm level wind speed (30 knots, 10-minute maximum sustained winds) is usually used as the wind threshold, and heavy rain level rainfall (25 mm per day) is used as the precipitation threshold8,10. However, Ye et al.8 reported that the wind method with the threshold of 30 knots will underestimate the TC impact area while the wind-precipitation method can improve the extraction accuracy of TC hazard exposure (ratio of the number of affected counties in the extracted impact area of TC to the number of reported TC affected counties).

On important issue here is that the impact area extracted by the wind-precipitation method usually not cover all affected counties. For example, appropriately 20% of the affected counties in China are still not include within the impact area obtained using the wind-precipitation method with the thresholds of 30 knots and 25 mm per day, respectively8. Therefore, it is necessary to adjust the TC hazard intensity threshold to accurately extract the TC impact area by using other additional information. The frequency of disaster occurrence is one of the key indicators of disaster resilience41,42. In general, areas with a higher frequency of occurrence of a particular natural disaster trend to be more prepared to that natural disaster. So, a relatively higher hazard intensity is usually required to cause significant amount of damage or impact in those areas with better preparedness. Here, we show a negative correlation between the frequency of TC impact and the distance from the coastline during the period 2010–2020 in China (Supplementary Fig. 6), and the distance does not have a lot of change by itself with the selected time period. This means that the distance is a very important factor controlling the damages from TC hazards and should be considered when calculating the exposed areas from those hazards. Because the wind speed of 30 knots and the rainfall of 25 mm are commonly set as the thresholds in the previous study8, so we use them as a baseline to ensure the comparability to our new method, which are calculated as follows:

$$W=\log \left(\frac{{wind}}{30}\right)$$
(5)
$$R=\log \left(\frac{{rain}}{25}\right)$$
(6)
$$C1=\max (W,R)$$
(7)
$$C2=\log \left(\sqrt{{\left(\frac{{wind}}{30}\right)}^{2}+{\left(\frac{{rain}}{25}\right)}^{2}}\right)$$
(8)

where \({wind}\) and \({rain}\) are the TC winds and precipitation intensities, respectively. \(W\) and \(R\) are the adjusted TC wind and precipitation intensities, respectively. \(C1\) is the adjusted TC comprehensive intensity constructed by the wind-precipitation method. \(C2\) is the adjusted TC comprehensive intensity constructed by the modified quadrant method43. Note that the 0 value of \(W\) and \(C1\) represent the adjusted intensity reaching the fixed intensity thresholds of the wind method and the wind-precipitation method, respectively. Here, we assume that the value of 0 is the fixed threshold of \(C2\) proposed in this study.

Since a significant negative correlation exists between the adjusted TC hazard density and its distance from the coastline (Supplementary Fig. 7 and Supplementary Table 3), we propose a new Methodology for setting hazard intensity Thresholds by considering the Distance from Coastline (MTDC) to extract the TC impact area. The MTDC has two assumptions: (1) the TC intensity threshold for extracting the TC impact area decreases as the distance from the coastline increase, with the rate of decrease and the slope of the fit between the adjusted TC intensity and the distance from the coastline remaining constant; (2) the threshold of adjusted TC intensity is 0 for coastal region. Supplementary Fig. 7 shows the fixed threshold (dashed red lines) and the threshold of the MTDC (solid red lines). Extraction accuracy by the MTDC. The sample size of reported TC-affected counties is 1223 between 2010 and 2020. Supplementary Table 4 shows the extraction accuracy of reported TC-affected counties by the fixed threshold method and the MTDC. The extraction accuracy of the MTDC improves by 16.6%, 6.8%, and 5.7% compared with that of the fixed threshold method when the adjusted TC hazard is \(W\), \(C1\), and \(C2\), respectively. And the highest extraction accuracy of 98.0% when the adjusted TC hazard is \(C2\) (Supplementary Fig. 7 and Supplementary Table 4). Therefore, in this study, we choose the combination of \(C2\) and MTDC as the hazard intensity threshold to extract the TC impact area.

TC economic exposure extraction

In this study, we define the total GDP of the TC impact area as the economic exposure of a TC. The annual TC economic exposure is defined as the sum of the economic exposure of all TCs in a given year. If a place (or a raster in map) was hit by two or multiple TCs, the annual TC economic exposure in the place (or the raster in map) was calculated as GDP multiplied by the frequency of TC hits.

TC economic exposure and GDP at the 2015 price level

In this study, we adjusted the TC economic exposure and GDP to the price level of 2015 using the consumer price index (CPI) in China to make the inflation-adjusted data comparable.

$${{GDP}}_{2015}=\frac{{{CPI}}_{2015}\times {{GDP}}_{Y}}{{{CPI}}_{Y}}$$
(9)
$${{Expo}}_{2015}=\frac{{{CPI}}_{2015}\times {{Expo}}_{Y}}{{{CPI}}_{Y}}$$
(10)

where \({{GDP}}_{2015}\) and \({{Expo}}_{2015}\) are the GDP and TC economic exposure at the 2015 price level (unit: RMB at the 2015 price level), respectively. \({{GDP}}_{Y}\) and \({{Expo}}_{Y}\) are the GDP and TC economic exposure at the year \(Y\) price level (unit: RMB at the year \(Y\) price level), respectively. \({{CPI}}_{2015}\) and \({{CPI}}_{Y}\) are the CPI data at 2015 and year \(Y\) in China, respectively.

Then, we translated the adjusted TC economic exposure and GDP in RMB to US$ using the exchange rate in 2015 (US$1.0\(\approx\)6.2284 RMB).

TC economic exposure centroid

The location \(\left({C}_{{Lat}}{,C}_{{Lon}}\right)\) of the TC economic exposure centroid is calculated as follows:

$${C}_{{Lat}}=\frac{{\sum }_{i=1}^{N}{{Lat}}_{i}\times {{Expo}}_{i}}{{\sum }_{i=1}^{N}{{Expo}}_{i}}$$
(11)
$${C}_{{Lon}}=\frac{{\sum }_{i=1}^{N}{{Lon}}_{i}\times {{Expo}}_{i}}{{\sum }_{i=1}^{N}{{Expo}}_{i}}$$
(12)

where \({C}_{{Lat}}\) and \({C}_{{Lon}}\) are the latitude and longitude coordinates of the location of the TC economic exposure centroid, respectively. \(N\) is the total number of grids in the TC economic exposure. \({{Expo}}_{i}\) is the economic exposure of the i-th grid. \({{Lat}}_{i}\) and \({{Lon}}_{i}\) are the latitude and longitude coordinates of the economic exposure of the i-th grid, respectively.

The centroid of annual TC impact

Same as the definition of TC economic exposure centroid, the location \(\left({C}_{{Lat}}{,C}_{{Lon}}\right)\) of the annual TC impact centroid is calculated as follows:

$${C}_{{Lat}}=\frac{{\sum }_{i=1}^{N}{{Lat}}_{i}\times {{TCI}}_{i}}{{\sum }_{i=1}^{N}{{TCI}}_{i}}$$
(13)
$${C}_{{Lon}}=\frac{{\sum }_{i=1}^{N}{{Lon}}_{i}\times {{TCI}}_{i}}{{\sum }_{i=1}^{N}{{TCI}}_{i}}$$
(14)

where \({C}_{{Lat}}\) and \({C}_{{Lon}}\) are the latitude and longitude coordinates of the location of the annual TC impact centroid, respectively. \(N\) is the total number of grids in the TC impacts. \({{TCI}}_{i}\) is the annual TC impacts of the ith grid. \({{Lat}}_{i}\) and \({{Lon}}_{i}\) are the latitude and longitude coordinates of the annual TC impacts of the ith grid, respectively.

The centroid of GDP distribution

Same as the definition of TC economic exposure centroid, the location \(\left({C}_{{Lat}}{,C}_{{Lon}}\right)\) of the GDP distribution centroid is calculated as follows:

$${C}_{{Lat}}=\frac{{\sum }_{i=1}^{N}{{Lat}}_{i}\times {{GDP}}_{i}}{{\sum }_{i=1}^{N}{{GDP}}_{i}}$$
(15)
$${C}_{{Lon}}=\frac{{\sum }_{i=1}^{N}{{Lon}}_{i}\times {{GDP}}_{i}}{{\sum }_{i=1}^{N}{{GDP}}_{i}}$$
(16)

where \({C}_{{Lat}}\) and \({C}_{{Lon}}\) are the latitude and longitude coordinates of the location of the GDP distribution centroid, respectively. \(N\) is the total number of grids in the GDP distribution. \({{GDP}}_{i}\) is the GDP of the i-th grid. \({{Lat}}_{i}\) and \({{Lon}}_{i}\) are the latitude and longitude coordinates of the GDP distribution of the ith grid, respectively.

Factors contribution

The TC economic exposure combines information from both the TC impact area and GDP distribution. According to Liu et al. 44. and Liao et al. 45, the influence on exposure change can be divided into three parts in this study, including the TC impact change effect, the GDP distribution change effect, and the joint change effect. The decomposition for exposure change is calculated according to Eq. 17.

$${D}_{E}={D}_{{TC}}+{D}_{{GC}}+{D}_{{JC}}$$
(17)

where \({D}_{E}\) is the northward shift speed in the annual TC economic exposure centroid for China, and \({D}_{E}=19.71\) (unit: km per year) in this study (Fig. 1b and Supplementary Table 2). \({D}_{{TC}}\) is the contribution of TC impacts shift speed. \({D}_{{GC}}\) is the contribution of GDP distribution shift speed. \({D}_{{JC}}\) is the contribution of joint change shift speed.

$${D}_{{TC}}={D}_{T}\times {R}_{T}$$
(18)
$${D}_{{GC}}={D}_{G}\times {R}_{G}$$
(19)
$${D}_{{JC}}={D}_{E}-{D}_{{TC}}-{D}_{{GC}}$$
(20)

where \({R}_{T}\) is the ratio of TC economic exposure centroid distance to equator increases with the annual TC frequency centroid distance to the equator, and \({R}_{T}=0.24\) in this study (Supplementary Table 1). \({R}_{G}\) is the ratio of TC economic exposure centroid distance to equator increases with the GDP distribution centroid distance to the equator, and \({R}_{G}=0.58\) in this study (Supplementary Table 1). \({D}_{T}\) is the northward shift speed in the annual TC impact distribution in China and \({D}_{T}=46.53\) (unit: km per year) in this study (Supplementary Table 2). \({D}_{G}\) is the northward shift speed in the GDP distribution in China and \({D}_{G}=13.35\) (unit: km per year) in this study (Supplementary Table 2).