The Characteristics of Tropical Cyclones and Associated High Winds in Shanghai from 2005 to 2020

Abstract

The characteristics of 21 tropical cyclones (TCs) and associated high winds in Shanghai from 2005 to 2020 are analyzed based on the TC best track data, surface observations, and ERA5 reanalysis dataset. Results indicate that 95% of the TCs causing high winds in Shanghai were generated during July to September, and located to the south of Shanghai for ~80% of the impact time. Divided into four categories according to tracks, the TCs show significant discrepancies in genesis month and location, intensity, size, and duration. The findings of high winds in Shanghai highlight the combination effect of TC intensity, track, and the distance between TCs and Shanghai normalized by TC size (ND_SH-TC). The maximum 10-min sustained (V_max) and extreme 3-s wind speeds (V_ex) in Shanghai only increase slightly with TC intensity, due to the fact that the stronger TCs are usually farther away from Shanghai. The V_max and V_ex differ in response to different TC tracks and reach the maximum value under TCs moving northward across the land west of Shanghai and the sea east of Shanghai, respectively. The urban and coastal wind speeds are the smallest and the largest respectively, indicating that the largest underlying surface roughness is associated with the most rapid urban construction. The wind speeds under the influence of TCs in four quadrants of Shanghai also present different variations with the ND_SH-TC. Additionally, the wind direction in Shanghai is regularly related to the relative position between the TC center and Shanghai. A case study combined with TC In-Fa (2106) verified the validity of the conclusions, which can provide a reference for TC forecasters to make more accurate wind field predictions.

This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction

Tropical cyclones (TCs) are one of the most severe weather systems in the world and led an average of 43 deaths and 780 million dollar losses every day of the past 50 years [1]. The TC winds and rains can not only directly or indirectly cause flood disasters but also seriously destroy buildings and service facilities and obstruct traffic [2, 3, 4], resulting in heavy casualties and economic losses to coastal cities [5, 6, 7, 8, 9, 10].

To reduce the threat of TCs, many spaceborne, airborne, and ground-based observation systems have been dedicated to monitoring TC location, intensity, and structure in real time [4, 11, 12, 13]. Meanwhile, the development of observations and global numerical weather prediction models have steadily improved the accuracy of TC track prediction in the past several decades [14, 15], with a mean error of less than 200 km at a lead time of 72 hr [16]. However, due to the complexity of TC physical structure, the interaction between TCs and atmospheric environment, sea–air energy exchange, and the impact of surface underlying after TC landed [17], TC intensity is still unable to be predicted perfectly [18, 19]. To alleviate the shortcomings in TC track and intensity forecasting, some TC warning centers, such as the National Hurricane Center (NHC), the Central Pacific Hurricane Center (CPHC), and the Joint Typhoon Warning Center (JTWC), provide estimates of wind radii (e.g., 34 kt, 50 kt, and 64 kt; 1 kt = 0.51444 m·s⁻¹), which can assist with making informed decisions about potential TC wind impacts at a given location [13, 20]. Nevertheless, the analysis of TC high wind (defined here as the winds induced by a TC that exceeds 10.8 m·s⁻¹) characteristics based on observations is still of great significance for forecasting operations, since the TC high wind forecasting for a coastal city is a combination of the TC intensity, size, track, and the interaction of TCs with local terrain.

Recently, Bloemendaal et al. [21] concluded that the distribution of maximum 10-min sustained wind speed (V_max) of global coastal cities under TC influence was related to the sea surface temperature and came in a variety of the return periods. Kruk et al. [22] also found the differences in TC wind speeds and return periods in coastal areas of the United States. Based on the TC landfall data in the United States from 1985 to 2008, Kovach and Konrad [23] found that the high winds occurred mostly in the right front sector of a TC, and the influence degree was related to TC size, intensity, and forward speed of movement, as well as the position relative to the TC track. Some researchers have focused on the relationship between the peak near-surface wind speed and the central pressure in TCs to improve the quality of operational forecasting, risk assessment, and basic research and understanding of the TCs [24, 25, 26]. As one of the countries severely affected by TCs, China summarizes and analyzes the number, duration, intensity, and other characteristics of TCs in the western North Pacific and South China Sea every year based on the TC best track data, in order to provide references for future TC forecasting operations [9, 27]. Studies show that 89% of the landfalling TCs in China caused high winds, and the influence range significantly expanded after TC landed [28]. In recent decades, both the frequency and the intensity of TCs causing high winds have increased significantly [29, 30]. Ni et al. [31] demonstrated that the southeastern coastal area of China has the highest frequency of TC high winds. Based on the TC landfall wind data in China during 2004–2018, Liang et al. [32] found that the wind at the Beaufort scale six and above mainly occurred within 300 km in the right front of a TC with typhoon intensity and above. Lu et al. [33] revealed that there were significant differences in the influence of TCs on China from different tracks, with relatively more negative impacts and serious disasters related to the TCs moving westward and northwestward. The previous work focused on the occurrence frequency, return period, and temporal and spatial distribution of TC winds and demonstrated the unique sensitivity of different regions to TCs, due to the influence of the TC track, landfalling location, intensity, size, atmospheric circulation, and underlying surface, which brings great challenges to the actual TC wind forecasting operations.

Shanghai is located on the east coast of China and is vulnerable to an average TC influence of approximately two each year [34]. As the largest economic center and the most urbanized city in China, Shanghai has suffered tremendous economic losses and urban security threats under the background of the increasing probability of extreme TC weather caused by global warming [30]. For example, TC Haikui (1211) in 2012 disrupted transport, with rail, subway, and maglev systems suspended; more than 30,000 trees blown down; numerous large billboards and tens of thousands of vegetable and fruit greenhouses damaged; and agricultural production losses of nearly 500 million RMB [35]. Under the influence of TC In-Fa (2106) in 2021, the entire rail traffic in Shanghai was affected, and more than 4,700 trees were blown down. Notably, there is still a high demand for TC high wind forecasting operations for development and safety guarantees in Shanghai. On the premise that there is a certain gap between the numerical forecast and the actual TC high winds [16, 18, 19], it is of great significance to analyze the TC high winds in Shanghai based on the observations. Xu [36] pointed out that the TC track influenced the dominant wind direction and maximum wind speed in Shanghai, and the urbanization significantly reduced the surface wind speed during TC impacts. Wang et al. [37, 38] demonstrated the differences in the wind field characteristics of several TCs in the near-coastal area of Pudong, Shanghai. Li and Du [39] analyzed the extreme wind speed distribution in the coastal area of Shanghai based on four TCs that had severely affected Shanghai and pointed out that the south coast is greatly affected by the landfalling TCs near Shanghai, with the north coast by TCs moving northward across the sea east of Shanghai. Of note, previous work on TC wind observations in Shanghai has provided meaningful results but is not fine enough, or only a case study has been carried out, which is inadequate and incomplete to figure out the Shanghai TC high wind characteristics.

In this paper, the characteristics of TCs and associated high winds affecting Shanghai from 2005 to 2020 are comprehensively investigated, based on the TC best track data, the surface observation, and the reanalysis data of ERA5, aiming to improve the ability of TC mitigation and disaster reduction in Shanghai and provide a more valuable method for TC wind forecasting in coastal cities of China. This paper is organized as follows: Section 2 introduces the required data including surface observations, best track data, and ERA5 reanalysis dataset; Section 3 introduces the data statistical methods; Section 4 focuses on the TC characteristics from 2005 to 2020, including the TC genesis month and location, intensity, size, and translation speed; Section 5 analyzes the distribution and variation characteristics of the 10-m wind field in Shanghai during the affecting period of the TCs; Section 6 carries out a case study to check the results of Sections 4 and 5; and a summary and discussion are given in Section 7.

2. Data

The data used in the paper include the surface observations, the TC best track data, and the reanalysis data.

The influence of TCs on Shanghai is determined based on the observed surface wind field from 36 automatic weather stations (AWSs) in Shanghai (Figure 1). Observations were made with Vaisala WA 15 wind speed sensors at 10 m above ground level and sampled at 1 s. All the AWSs are national meteorological observation stations that have collected meteorological data in Shanghai since 2005, and the data source and quality are ensured. TC high wind is defined when the V_max at an AWS exceeds 10.8 m·s⁻¹. Whether a TC has an impact on Shanghai is determined based on two steps. First, the satellite cloud pictures are used to confirm that Shanghai is covered by the cloud system of a TC, that is, controlled by a TC system. Second, this TC is considered to have an impact on Shanghai when two or more AWSs recorded TC high winds during its impact [36, 40]. Based on the observations of 36 AWSs from 2005 to 2020, there were 21 historical TCs that caused high winds in Shanghai (Figure 2). According to the track location with respect to Shanghai, they are divided into four groups: landing in Shanghai (four TCs), moving northward across the sea east of Shanghai (four TCs), moving northward across the land west of Shanghai (10 TCs), and moving westward across the land west of Shanghai (three TCs; hereafter LD, NAE, NAW, and WAW) [40]. Table 1 shows the TCs and the time span affecting Shanghai. The distance between a TC center and the People’s Square (121.47°E, 31.23°N; Figure 1) is defined as the distance between the TC and Shanghai (D_SH-TC) [40].

[figure(s) omitted; refer to PDF]

Table 1

Tropical cyclones (TCs) that affected Shanghai during 2005–2020.

Year	TC code	Name	Time span affecting Shanghai (LST)
2020	2004	Hagupit	4 Aug, 7:00–5 Aug, 5:00
2019	1909	Lekima	9 Aug, 12:00–11 Aug, 06:00
2019	1917	Tapah	21 Sep, 11:00–22 Sep, 14:00
2019	1918	Mitag	1 Oct, 11:00–2 Oct, 12:00
2018	1810	Ampil	21 Jul, 23:00–23 Jul, 05:00
2018	1812	Jongdari	3 Aug, 02:00–20:00
2018	1814	Yagi	12 Aug, 11:00–13 Aug, 20:00
2018	1818	Rumbia	16 Aug, 09:00–18 Aug, 16:00
2016	1614	Meranti	15 Sep, 16:00–16 Sep 20:00
2015	1509	Chan-hom	10 Jul, 22:00–12 Jul, 15:00
2015	1521	Dujuan	28 Sep, 20:00–29 Sep, 20:00
2014	1416	Fung-wong	22 Sep, 08:00–23 Sep, 20:00
2013	1323	Fitow	6 Oct, 10:00–7 Oct, 6:00
2012	1211	Haikui	6 Aug, 10:00–9 Aug, 23:00
2011	1109	Muifa	6 Aug, 20:00–7 Aug, 17:00
2009	0908	Morakot	9 Aug, 20:00–11 Aug, 19:00
2008	0807	Kalmaegi	19 Jul, 08:00–20 Jul, 05:00
2008	0808	Fung-wong	29 Jul, 16:00–31 Jul, 12:00
2007	0713	Wipha	19 Sep, 01:00–20 Sep, 11:00
2005	0509	Matsa	5 Aug, 11:00–8 Aug, 00:00
2005	0515	Khanun	11 Sep, 10:00–12 Sep, 07:00

These cyclones were identified based on the maximum 10-min average wind speed at two or more AWSs, which is greater than or equal to 10.8 m·s⁻¹.

The TC best track data of 6-hr intervals from the Shanghai Typhoon Institute of China Meteorological Administration (CMA) [41] and JTWC are used to analyze the characteristics of TC activities affecting Shanghai. The CMA dataset provides TC track (center latitude and longitude), maximum 2-min sustained wind speed at 10 m (VMAX_10 m), and minimum mean sea level pressure (MSLP_min). The TC intensity is stratified into five categories according to the VMAX_10 m near the TC center: tropical storms (TS, 17.2–24.4 m·s⁻¹), strong tropical storms (STS, 24.5–32.6 m·s⁻¹), typhoons (TY, 32.7–41.4 m·s⁻¹), strong typhoons (STY, 41.5–50.9 m·s⁻¹), and super typhoons (SuperTY, ≥51.0 m·s⁻¹) [40]. The JTWC dataset provides a wind radius of 34 kt (R₃₄) to indicate TC size [20].

The fifth-generation reanalysis dataset ERA5 [42], launched by the European Centre for Medium-Range Weather Forecasts (ECMWF), is used to analyze the 500-hPa synoptic situation at the moment of TC genesis identified by the TC best track data. The reanalysis dataset integrates the development achievements of ECMWF in the fields of physical model, core dynamics, and data assimilation in the past 10 years and can provide hourly 0.25° resolution global geophysical parameters with 137 vertical levels from the ground to 0.01 hPa.

3. Methodology

To combine the research results more effectively with the actual wind forecasting operations, the analysis of Shanghai wind field mainly focuses on the hourly V_max, extreme wind speed (V_ex), and the corresponding wind direction (WD_max and WD_ex, respectively). The definition and classification method are introduced below.

V_max is the maximum value of the 10-min sustained wind speed of the previous hour, and V_ex is the maximum value of the instantaneous wind speed (3 s) in the same time span. WD_max and WD_ex are the corresponding wind directions. During TC impacts, the surface wind speed in Shanghai is not only related to TC intensity but also affected by the complex underlying surface of Shanghai. In recent years, the large amount of construction in Shanghai has given rise to a significant distinction in the underlying surface, which leads to various wind speed distributions [36]. In Shanghai, there are ~29,000 buildings of 8–15 stories, ~21,000 buildings of 16–29 stories, and over 1,820 skyscrapers ( > 100 m), with more buildings in the center area [43]. To distinguish the effects of different underlying surfaces on the wind field, the AWSs are divided into urban and suburban ones according to the 2010 modified IGBP MODIS LULC data (with resolution of 15 s) with updated three urban types based on a 100-m spatial resolution land use dataset. This urban land use dataset is generated using the urban and rural settlements data from the Global Urban Footprint (GUF) project (https://www.dlr.de/eoc/en/desktopdefault.aspx/tabid-11725/20508_read-47944/) and the population density data [44]. Then, the coastal stations located within 7 km from the southeast boundary line of Shanghai are categorized as coastal stations. Thus, the 36 AWSs are divided into three categories: urban stations (17), suburban stations (10), and coastal stations (9) (Figure 1). Hourly mean values of V_max, V_ex, WD_max, and WD_ex during TC impacts are calculated separately for urban, suburban, and coastal stations.

4. Characteristics of the TCs Affecting Shanghai

What are the characteristics of TCs affecting Shanghai? In this section, the TC genesis month and location, intensity, size, translation speed, and the possible relationship with TC track will be analyzed.

4.1. TC Genesis Month

The TCs generated in the western North Pacific affect Shanghai with an average interannual frequency of 1.3 from 2005 to 2020, which has a maximum of four in 2018 and none in 2006, 2010, and 2017 (Figure 3(a)). The genesis month covers from June to September (Figure 3(b)). The number of TCs generated in July, August, and September is seven, five, and eight, respectively, with a 95% probability, while there is only one TC generated in June. From the perspective of TC tracks (Figure 3(b)), there are NAE TCs in June; LD, NAE, and NAW TCs in July; and four groups in August and September.

[figure(s) omitted; refer to PDF]

The above analysis shows that there is a clear genesis month range from June to September and an obvious relationship between TC genesis month and track [33]. TC forecasting should focus on LD, NAE, NAW, and WAW TCs from July to September, yet the TCs generated in June cannot be ignored, especially the NAE TCs. The genesis month variation with track could be due to the subtropical high position, which influences the movement of TCs approaching China, so that landfall is more likely in southern China in June. The subtropical high is typically over northern China during July to September, which makes TC impacts in Shanghai more likely.

4.2. TC Genesis Location

The genesis location is a sensitive factor for TC track development [45]. Overall, the genesis development domain for all the 21 TCs is 125–161°E and 8–23°N. Twenty TCs are generated in the range of 125–150°E and 8–23°N, while only one TC is generated to the east of 150° E and belongs to the NAE group (Figure 2). Categorized according to TC tracks, the genesis location of LD is typically further north, the NAE is usually further east, and the NAW is normally further west. In contrast, the typical WAW genesis location is closer to the center of TC development domain.

To explore the possible track development of four TC groups, the synthesis analysis of the 500-hPa synoptic situation is performed at the genesis moment. The center of each TC from the same TC track type is taken as the synthesis center, and the mean synoptic situation field at 500 hPa is calculated using EAR5 data [46] (Figure 4). The strong subtropical high corresponding to the genesis moment of LD TCs is located further northwest and completely covers Shanghai, which causes TCs to subsequently be steered by the southwesterly flow and make landfall in Shanghai. For NAE TCs, the subtropical high at the genesis moment has a further southeast location over the ocean, and the land is controlled by the westerly trough (i.e., the low trough in the midlatitude westerlies). Under the control of both large-scale systems, the NAE TCs keep moving over the ocean. At the genesis moment of the NAW TCs, the weak subtropical high is located over the ocean, and the westerly trough over the land is further north, which is conducive to these TCs landing in southern Shanghai and subsequently moving northward to affect Shanghai. The WAW TCs are associated with a stronger, further southwest subtropical high with its west ridge (i.e., a ridge where the isobars protrude toward the lower pressure) further west than the NAW TCs. Thus, the WAW TCs cannot move northward but westward after landing to the south of Shanghai.

[figure(s) omitted; refer to PDF]

In general, the genesis location of LD, NAE, and NAW TCs has typical characteristics, with further north, further east, and further west, respectively. And the 500-hPa synoptic situation at the TC genesis moment indicated that the location, strength, and ridge position of the subtropical high could directly affect the TC track. Thus, based on the genesis location and the synoptic situation at genesis moment, the possible TC track could be predicted, what’s more, combined with the genesis month variations with track and the TC track prediction, whether a TC will affect Shanghai could be judged in advance by TC forecasters.

4.3. TC Intensity, Size, and Translation Speed

The TC destructive potential is mainly due to the TC intensity, size, and duration [47]. To provide more detailed TC activity characteristics, a further analysis on TC VMAX_10 m, MSLP_min, R₃₄, the time span (T_span) affecting Shanghai, and the averaged translation speed for the entire TC life (V_{ts, avg}) is carried out in this section.

VMAX_10 m and MSLP_min are both crucial indices for TC intensity. The lower MSLP_min is, the greater the pressure gradient force within the TC is and the more significant the air flow converging toward the TC center at the lower layer is, which could form a higher wind speed and thus a stronger TC intensity. That is, VMAX_10 m and MSLP_min are negatively correlated variables, which is fully confirmed by the mean VMAX_10 m and MSLP_min trends of LD, NAE, and NAW TCs (Figure 5(a)). The NAE and WAW TCs have comparable larger mean VMAX_10 m and lower mean MSLP_min, which are 28.7 m·s⁻¹ and 978.9 hPa for the NAE TCs and 27.8 m·s⁻¹ and 979.4 hPa for the WAW TCs. The values for NAW TCs are 25.5 m·s⁻¹ and 983.1 hPa. The LD TCs have the smallest VMAX_10 m of 20.0 m·s⁻¹ and the highest MSLP_min of 990.7 hPa.

[figure(s) omitted; refer to PDF]

The TC size parameter R₃₄ is a widely used measure of the radial TC wind field structure [20]. The mean R₃₄ of LD, NAE, NAW, and WAW TCs is 87, 171, 104, and 110 km, respectively (Figure 5(a)). It is clear that the R₃₄ shows an increase trend with the TC intensity, that is, the stronger TC track group has relatively larger TC size. The strongest NAE TCs has the largest TC size; however, the WAW TCs has comparable TC intensity with the NAE TCs but obviously smaller R₃₄.

The value of T_span is another factor to assess the potential TC impacts on Shanghai in TC forecasting operation. The T_span is defined as the range of the start and end times when the V_max at two or more AWSs exceeds 10.8 m·s⁻¹ under the influence of a TC, without excluding the times when the V_max does not exceed 10.8 m·s⁻¹ within this range. The mean T_span of LD, NAE, NAW, and WAW TCs is 1.33, 1.32, 1.45, and 1.80 days, respectively (Figure 5(b)). The value of the first two TC groups is comparable, while WAW TCs have the longest T_span. The longer mean T_span may be contributed by the slower mean V_{ts, avg} [48]. The WAW TCs move the slowest with a mean V_{ts, avg} of 15.1 km·hr⁻¹, followed by NAW with 20.1 km·hr⁻¹, NAE with 21.2 km·hr⁻¹, and LD TCs with 23.6 km·hr⁻¹. Additionally, the TC intensity may also affect the T_span, because stronger TCs may survive longer over land. The mean intensity of WAW TCs is indeed stronger in the four TC groups (Figure 5(a)).

In summary, the NAE and WAW TCs have comparable larger mean VMAX_10 m and lower mean MSLP_min. The largest R₃₄ is with the NAE TCs. The mean T_span of four TC groups is within 2 days, with the longest value from the WAW TCs, which may be contributed by the relatively slower V_{ts, avg} and stronger intensity. Obviously, the WAW TCs have both relatively stronger intensity and longer time span affecting Shanghai in four TC groups and should be paid more attention to by the forecasters in Shanghai. However, this does not mean that the other TC groups can be taken lightly. In addition to the TC intensity and duration, the distance between the TC center and Shanghai also has a nonnegligible impact, which will be discussed in Section 5.

5. Characteristics of the Shanghai Wind Field during TC Impacts

The previous section mainly focused on the characteristics of TCs that caused high winds in Shanghai. This section will analyze the characteristics of the wind field in Shanghai during the impact of the TCs. The wind speed spatial distribution characteristics and variation with TC intensity, track, size, and the distance between the TC center and Shanghai; the wind direction variation with TC track and the relative position between the TC center and Shanghai are the major concerns.

5.1. Spatial Distributions of the Wind Speed

Based on the high wind definition (Section 2), a total of 727 observation times corresponding to 21 TCs can be taken as sample data for the analysis of the surface wind field in Shanghai. Selecting Shanghai as the center, the horizontal space is divided into four quadrants: right rear, left rear, left front, and right front (Figure 2). Figure 6(a) shows that the TC locations corresponding to the observation times are 36.7%, 42.5%, 13.6%, and 7.2% in each quadrant, respectively, with a total of ~80% in the first two quadrants, that is, to the south of Shanghai. The TC intensities affecting Shanghai include TS, STS, TY, and STY, with a ratio of 42.4%, 25.4%, 18.6%, and 12.6%, respectively. The stronger TCs are mainly in the right rear quadrant, and the left front quadrant only contains TS.

[figure(s) omitted; refer to PDF]

Under the influence of TCs, the AWSs do not always record TC high wind. Figure 6(b) shows the ratio of observation times of TC high winds to all the 727 observation times for each of the urban, suburban, and coastal AWSs. Only two AWSs reach a ratio of 0.4 and both located in the coastal region of Shanghai, while the ratios for the other AWSs are ~0.1. Correspondingly, the mean V_max of the two AWSs is both above 10 m·s⁻¹, and the spatial distribution of the mean V_max shows a clear decrease from the coastal region to the urban region, with a maximum of 13.1 m·s⁻¹ and a minimum of 1.6 m·s⁻¹ (Figure 6(c)). Comparatively, the mean V_ex is obviously larger than the mean V_max, and 78% of the AWSs get a mean V_ex of above 10 m·s⁻¹ (Figure 6(d)).

The above analysis shows that TCs are located to the south of Shanghai for ~80% of the impact time. The TC intensity affecting Shanghai includes TS, STS, TY, and STY, and the impact time of the first two accounts for ~70%. Under the influence of TCs, the coastal AWSs have relatively higher frequencies to record TC high winds and larger mean V_max.

5.2. Variation of the Wind Speed

In this section, the variation of the 10-min wind speed in Shanghai during TC impacts is analyzed in detail. For all TC intensities (i.e., TS, STS, TY, and STY), the mean V_max variation ranges of urban, suburban, and coastal stations are 4.6–5.7 m·s⁻¹, 5.0–6.3 m·s⁻¹, and 7.6–9.2 m·s⁻¹, respectively, with the corresponding mean V_ex of 9.9–12.0 m·s⁻¹, 9.9–12.5 m·s⁻¹, and 12.9–15.4 m·s⁻¹, respectively (Figure 7(a)). The mean V_max and V_ex at the urban, suburban, and coastal stations show a slightly increasing trend with TC intensity, which is consistent with the forecasting experience. In addition to the TC intensity, the D_SH-TC can also affect the V_max and V_ex in Shanghai, and the closer the TC is to Shanghai, the larger V_max and V_ex will be. Figure 7(a) shows that the stronger TCs are usually farther away from Shanghai, so the impact on the V_max and V_ex could be weakened. The combination of the stronger TCs with larger D_SH-TC and weaker TCs with smaller D_SH-TC results in the insignificant increasing relationship between surface winds and TC intensity in Shanghai. From a geographical perspective, the urban region has the smallest mean V_max and V_ex, followed by suburban and coastal regions, indicating that the largest underlying surface roughness is associated with the most rapid urban construction. Xu [36] found that the increased urbanization of Shanghai led to an increase in the underlying surface roughness, which induced a decrease in coastal V_max from 14.8 m·s⁻¹ to 10.6 m·s⁻¹ during 1961–2002. Therefore, the maximum mean coastal V_max of 9.2 m·s⁻¹ in this paper indicates that the urbanization of Shanghai may have further reduced the wind speed in recent years. However, the maximum mean coastal V_ex of 15.4 m/s still reaches Beaufort scale seven, indicating that the TC winds still need special attention.

[figure(s) omitted; refer to PDF]

In addition to TC intensity, the impact of TC winds on Shanghai is also closely related to its track. Figure 7(b) shows that the mean urban, suburban, and coastal V_max have a similar relationship with the TC track. That is, the mean V_max of the three AWS categories reaches a minimum value under LD TC influence and a maximum value under NAW TC influence, with variation ranges of 4.1–5.3 m·s⁻¹, 4.4–5.7 m·s⁻¹, and 7.3–8.3 m·s⁻¹, respectively. Similar to V_max, the mean V_ex of the three AWS categories also reaches a minimum value under LD TC influence. The difference is that the mean V_ex corresponding to NAE TCs is the maximum value (Figure 7(b)). The variation ranges for the mean urban, suburban, and coastal V_ex are 9.3–11.2 m·s⁻¹, 9.2–11.2 m·s⁻¹, and 12.8–14.8 m·s⁻¹, respectively. In closing, the Shanghai surface wind speed differs in response to different TC tracks; the wind speed under LD TC influence is the smallest, with the maximum mean V_max under NAW TC influence and the maximum mean V_ex under NAE TC influence. This distinction could be due to the TC intensity and D_SH-TC of the four TC tack groups. The WAW TCs have relatively stronger intensity (Figure 5(a)) but significantly larger D_SH-TC (Figure 7(b)), which directly reduced its impact on the surface wind in Shanghai. There is an interesting phenomenon that the urban V_max is obviously lower than the suburban V_max due to the blocking effect of the underlying, but the difference of the urban and suburban V_ex is relatively smaller. This may reveal the comparable perturbation effect of the underlying terrain on urban and suburban wind field to rise the extreme wind.

How to judge the possible impact of TCs on Shanghai according to the orientation of TCs relative to Shanghai is one of the priorities of forecasters and the greatest concern of decision-makers. In the following, the detailed variation of V_max and V_ex with the distance between the TC center and Shanghai in the coastal regions, which are most affected by the TCs, is focused on the perspective of the orientation of TCs relative to Shanghai. Due to that the TCs exhibit large variations in size at a particular TC intensity, the impact of TC size on the Shanghai wind field should not be ignored [20]. For example, under the influence of TCs with similar intensity and distance from Shanghai, the influence range of TCs with smaller size is relatively smaller, and thus the wind speed in Shanghai is relatively lower, with the opposite for TCs with larger size. Therefore, the D_SH-TC normalized by R₃₄ (ND_SH-TC) is used to analyze the wind speed variation in the four quadrants (Figure 8). However, due to the missing R₃₄ dataset, especially near and over land, the hourly observation samples without matched R34 data have to be discarded, resulting in only 70% of the 727 samples being usable here. Under the influence of the TCs in the right rear quadrant, V_max and V_ex increase with TC intensity and decrease with ND_SH-TC, reaching a maximum of 13.7 m·s⁻¹ and 21.0 m·s⁻¹, respectively. Corresponding to the TCs in the left rear quadrant, the V_max and V_ex variation with ND_SH-TC is similar with those under the TCs in the right rear quadrant but with greater maximum V_max and V_ex of 16.4 m·s⁻¹ and 25.0 m·s⁻¹, respectively. Notably, V_max and V_ex are relatively smaller under the influence of STY in the left rear quadrant, which may be due to fewer samples from a single TC. For the left front quadrant, only TS affects Shanghai, and V_max and V_ex present inconspicuous variations with the ND_SH-TC. In the right front quadrant, V_max and V_ex decrease with ND_SH-TC under the influence of TS and STS. In addition, the V_max and V_ex under the influence of TY increased with ND_SH-TC, which may be related to the fewer samples and should be further verified by more samples in the future. In summary, the maximum V_max and V_ex in the right rear quadrant are greater than those in the left rear quadrant. Generally, V_max and V_ex increase with TC intensity and decrease with ND_SH-TC. Based on the above variation, V_max and V_ex are linearly fitted to provide effective support for TC forecasting. It should be pointed out that the fitting formula may be inaccurate due to the limitation of sample size and will be updated with more samples in the future.

[figure(s) omitted; refer to PDF]

5.3. Variation of the Wind Direction

Wind direction is another fatal parameter in actual TC wind forecasting. Theoretically, the wind direction is directly affected by the TC rotational motion. It is well known that if Shanghai is located at different positions relative to a TC center, the wind direction will be significantly different or even the opposite. Therefore, the wind direction in Shanghai is related to the relative position between the TC center and Shanghai. Thus, the TC wind direction frequency in Shanghai is firstly presented in terms of TC location in four quadrants of Shanghai (Figure 9). When TCs are in the right rear quadrant, Shanghai is located in the northwest of TCs. Consequently, the dominant WD_max and WD_ex are 45° (northeast wind), with proportions of 51.3% and 43.9% over urban regions, 52.3% and 48.0% over suburban regions, and 48.1% and 47.1% over coastal regions. For the left rear quadrant, Shanghai is located in the northeast of the TCs, and accordingly, the dominant wind direction is 90°–135° (east wind and southeast wind) for both WD_max and WD_ex, with proportions of 77.9% and 73.5% over urban regions, 80.6% and 74.6% over suburban regions, and 81.8% and 79.6% over coastal regions, respectively. Corresponding to the left front quadrant, Shanghai is in the southeast of TCs, and the wind direction of 135°–180° (southeast wind and south wind) prevails for both WD_max and WD_ex, accounting for 61.7% and 62.7% over urban regions, 68.3% and 66.4% over suburban regions, and 73.4% and 72.1% over coastal regions, respectively. Of note, a supposed prevailing southwest wind is absent in this quadrant, which may be related to the obstruction from the mountainous area in Zhejiang to the southwest of Shanghai. When TCs are in the right front quadrant, Shanghai is located in the southwest of the TCs, and the dominant WD_max and WD_ex are 270°–315° (west wind and northwest wind), with proportions of 69.4% and 66.5% over urban regions, 80.1% and 78.0% over suburban regions, and 77.3% and 75.3% over coastal regions. In summary, the TCs in the four quadrants can lead to different dominant wind directions in Shanghai, in which northeast winds prevail in the right rear quadrant, east winds and southeast winds prevail in the left rear quadrant, southeast winds and south winds prevail in the left front quadrant, and west winds and northwest winds prevail in the right front quadrant. The WD_max and WD_ex have similar dominant wind direction over urban, suburban, and coastal regions.

[figure(s) omitted; refer to PDF]

The previous study proved that there is a clear relationship between TC wind direction and the relative position of the TC center and Shanghai. Thus, the significant difference of the relative positions between the TC of four tracks and Shanghai must lead to distinctive TC wind direction. Statistics shows that the northwest winds prevail under LD TC influence, with dominant wind directions of 0°–45° (north wind and northeast wind), 90°–135° (east wind and southeast wind), and 45°–90° (northeast wind and east wind), corresponding to NAE, NAW, and WAW TCs, respectively (Figure 10). In summary, the TC track has a significant impact on the dominant wind direction in Shanghai, and the prevailing wind direction is northeast for LD TCs, north and northeast for NAE TCs, east and southeast for NAW TCs, and northeast and east for WAW TCs.

[figure(s) omitted; refer to PDF]

6. Case Study

For reasonable application, the above analysis is checked based on the best track data and observations in Shanghai during the influence of TC In-Fa (2106) in 2021. TC In-Fa (2106) was generated at 134.8° E, 18.8° N in the western North Pacific at 20:00 on July 16, 2021 (LST, the same below), and then landed in the coastal area of Zhoushan, Zhejiang Province at ~12:30 on July 25. At 9:50 on July 26, it again landed in the coastal area of Jiaxing, Zhejiang Province, and then moved northwest on the land west of Shanghai. At 12:13 on July 26, the TC was nearest to Shanghai with a distance of 72 km. There is no doubt that TC In-Fa (2106) belongs to NAW TCs (Figure 11).

[figure(s) omitted; refer to PDF]

At the genesis moment of TC In-Fa (2106), the subtropical high was weak and located over the sea, while the westerly trough was over north land (Figure 11). The 500-hPa synoptic situation at genesis moment of TC In-Fa (2106) also conforms to the characteristics of NAW TCs. The mean VMAX_10 m, MSLP_min, R₃₄, V_ts, _avg, and T_span of TC In-Fa (2106) are 25.2 m·s⁻¹, 979.7 hPa, 172 km, 12.1 km·hr⁻¹, and 3 days, respectively. TC In-Fa (2106) moved relatively slowly and had a longer impact on Shanghai than the statistical value of NAW TCs. Its size was relatively larger, but the intensity was close to the statistical value.

The wind speed in Shanghai under the influence of TC In-Fa (2106) and NAW TCs is also compared. Under the influence of TC In-Fa (2106), the mean urban, suburban, and coastal V_max is 5.10 m·s⁻¹, 5.84 m·s⁻¹, and 9.48 m·s⁻¹, respectively, with 12.20, 12.62, and 16.54 m·s⁻¹ for mean V_ex (Figure 12). The mean V_max is in good agreement with the statistical value of NAW TCs, with a mean bias of 0.37 m·s⁻¹. However, the mean V_ex is slightly larger than the statistical value, with a mean bias of 1.83 m·s⁻¹. The coastal wind speed is the largest, followed by suburban and urban wind speeds. Moreover, all show a commonly increasing trend with TC intensity, but the difference under STS and TY is insignificant.

[figure(s) omitted; refer to PDF]

During the whole affecting process, TC In-Fa (2106) mainly passed through the right and left rear quadrants of Shanghai. Thus, the observations in Shanghai are compared with the statistical TC high wind speed for the two quadrants (Figure 13). In the right rear quadrant, the statistical V_max and V_ex corresponding to STS intensity from the fitting formula (Figures 8(a) and 8(b)) are 3.8 and 5.1 m·s⁻¹ lower than the observations, while the statistical V_max and V_ex corresponding to TY intensity are closer to the observations, with mean V_max and V_ex biases of −1.1 and −2.2 m·s⁻¹. In the left rear quadrant, the statistical V_max and V_ex corresponding to STS are greater than the observations, with V_max biases of 4.3 m·s⁻¹ and V_ex biases of 4.9 m·s⁻¹, respectively. The larger bias for STS may be due to an insufficient sample size for the fitting formula (Figure 8), and it is hoped that this will be gradually corrected in the future as sample size increases.

[figure(s) omitted; refer to PDF]

Furthermore, the WD_max and WD_ex of urban, suburban, and coastal stations were checked when TC In-Fa (2106) was in the right and left rear quadrants (Figure 14). For the right rear quadrant, the dominant WD_max and WD_ex are 45°, consistent with the statistical values. When TC In-Fa (2106) is in the left rear quadrant, the dominant WD_max and WD_ex are 45°–135°, which is slightly different from the statistical value of 90°–135°. This may be due to the fewer samples of TC In-Fa (2106) in the left rear quadrant.

[figure(s) omitted; refer to PDF]

Generally, the genesis month and location and track of TC In-Fa (2106) conform to the characteristics of NAW TCs, and the intensity is close to the statistical value of NAW TCs. Under its influence, the mean V_max and V_ex of urban, suburban, and coastal stations are in good agreement with the statistical values of NAW TCs. Notably, when TC In-Fa (2106) is in the right rear quadrant of Shanghai, the dominant wind direction is consistent with the statistical value, and the wind speed is greater than the statistical value. When TC In-Fa (2106) is in the left rear quadrant, the dominant wind direction deviates slightly from the corresponding NAW statistical value, and the wind speed is less than the corresponding statistical value. In addition, TC In-Fa (2106) moves relatively slowly, resulting in a large T_span bias with the NAW statistics.

7. Conclusions and Discussion

This paper first presents the genesis month and location, intensity, size, and translation speed of 21 TCs affecting Shanghai from 2005 to 2020, using TC best track data and ERA5 reanalysis dataset. The results show that 95% of the TCs causing high winds in Shanghai are generated in July, August, and September. There are LD, NAE, NAW, and WAW TCs in August and September and the first three groups in July. Moreover, the genesis location of LD is further north, with NAE of a further east genesis location and NAW of a further west genesis location. Otherwise, the WAW TC genesis typically occurs in the central part of the TC development domain. The location, strength, and ridge position of the subtropical high could directly affect the TC track. The NAE and WAW TCs have comparable larger mean VMAX_10 m and lower mean MSLP_min. The largest R₃₄ is with the NAE TCs. The mean T_span of four TC groups is within 2 days, with the longest value from the WAW TCs, which may be contributed by the relatively slower V_{ts, avg} and stronger intensity.

Second, the distribution and variation characteristics of the high winds in Shanghai during TC impacts are analyzed based on surface observations. Spatial distribution shows that TCs are located to the south of Shanghai for ~80% of the impact time. The TC intensity affecting Shanghai includes TS, STS, TY, and STY, and the impact time of the first two accounts for ~70%. V_max and V_ex increase slightly with TC intensity, which may be due to the fact that the stronger TCs are usually farther away from Shanghai, so their impact on the V_max and V_ex may be weakened. For all TC intensities (i.e., TS, STS, TY, and STY), the mean V_max variation ranges of urban, suburban, and coastal stations are 4.6–5.7 m·s⁻¹, 5.0–6.3 m·s⁻¹, and 7.6–9.2 m·s⁻¹, respectively, with mean V_ex variation ranges of 9.9–12.0 m·s⁻¹, 9.9–12.5 m·s⁻¹, and 12.9–15.4 m·s⁻¹, respectively. The coastal wind speed is the largest, and the urban wind speed is the smallest, indicating that the largest underlying surface roughness is associated with most rapid urban construction. In terms of the TC track impact on Shanghai wind speed variation, the wind speed under LD TC influence is the smallest and that under NAW and NAE TC influence is relatively larger. From the perspective of the orientation of TCs relative to Shanghai, V_max and V_ex induced by TCs in the right and left rear quadrant of Shanghai increase with TC intensity and decrease with ND_SH-TC. And the wind direction shows regular variation, that is, the northeast wind prevails in the right rear quadrant, east wind and southeast wind in the left rear quadrant, southeast wind and south wind in the left front quadrant, and west wind and northwest wind in the right front quadrant.

Finally, a case study, based on TC In-Fa (2106), is carried out to check the conclusions obtained in the previous two parts. The genesis month and location and the track of TC In-Fa (2106) conform to the characteristics of NAW TCs, and the intensity is close to the statistical value of NAW TCs. The mean V_max and V_ex of urban, suburban, and coastal stations are in good agreement with the statistical values of NAW TCs. Notably, when TC In-Fa (2106) is in the right rear quadrant of Shanghai, the dominant wind direction is consistent with the statistical value, and the wind speed is greater than the statistical value. When TC In-Fa (2106) is in the left rear quadrant, the dominant wind direction deviates slightly from the corresponding statistical value, and the wind speed is less than the corresponding statistical value. In addition, TC In-Fa (2106) moves relatively slowly, resulting in a large T_span bias with the NAW statistics.

This study comprehensively analyzes the characteristics of the TCs and associated high winds in Shanghai based on the latest observations. The statistical characteristics of the TCs affecting Shanghai show obvious discrepancies in genesis month and location, intensity, size, and translation speed. The findings of high winds in Shanghai highlight the combination effect of TC intensity, track, and normalized distance between the TC center and Shanghai, which should be closely noted by the TC wind forecasting in Shanghai. The case study of the characteristics of TCs and associated high winds in Shanghai shows that the TC genesis month and location, the track, and wind field characteristics in Shanghai, especially the wind speed fitting formula for different TC intensities, can provide a reference for TC forecasters to make more accurate wind field predictions.

However, the uncertainty of our results mainly comes from the insufficient TC sample size, especially in the fitting of the surface wind speed and the normalized distance between the TC center and Shanghai, which is aggravated by the lack of TC size data near and over land. Moreover, the case study is only for NAW TCs, and the effectiveness of the research results for other TC groups needs to be verified with more samples in the future.

Acknowledgments

This work was supported by Natural Science Foundation of Shanghai (grant nos. 22ZR1456100), National Natural Science Foundation of China (grant nos. 41875059 and 41875071), and Natural Science Foundation of Shanghai (grant nos. 21ZR1457700).

References

[1] WMO, "WMO atlas of mortality and economic losses from weather, climate and water extremes (1970–2019) (WMO-No. 1267)," 2021. https://wmo.int/publication-series/wmo-atlas-of-mortality-and-economic-losses-from-weather-climate-and-water-extremes-1970-2019

[2] Y. Zong, X. Chen, "Typhoon hazards in the Shanghai area," Disasters, vol. 23 no. 1, pp. 66-80, DOI: 10.1111/1467-7717.00105, 1999.

[3] L. S. Chen, Y. H. Duan, L. L. Song, Y. L. Xu, Typhoon Forecast and its Disaster, 2012.

[4] R. F. Rogers, "Recent advances in our understanding of tropical cyclone intensity change processes from airborne observations," Atmosphere, vol. 12 no. 5,DOI: 10.3390/atmos12050650, 2021.

[5] K. Emanuel, "Increasing destructiveness of tropical cyclones over the past 30 years," Nature, vol. 436 no. 7051, pp. 686-688, DOI: 10.1038/nature03906, 2005.

[6] L. Wu, B. Wang, "What has changed the proportion of intense hurricanes in the last 30 years?," Journal of Climate, vol. 21 no. 6, pp. 1432-1439, DOI: 10.1175/2007JCLI1715.1, 2008.

[7] P. J. Shi, X. Yang, W. Xu, J. Wang, "Mapping global mortality and affected population risks for multiple natural hazards," International Journal of Disaster Risk Science, vol. 7 no. 1, pp. 54-62, DOI: 10.1007/s13753-016-0079-4, 2016.

[8] R. Chen, W. Zhang, X. Wang, "Machine learning in tropical cyclone forecast modeling: a review," Atmosphere, vol. 11 no. 7,DOI: 10.3390/atmos11070676, 2020.

[9] X. Y. Lyu, Y. L. Xu, L. Dong, X. Z. Gao, "Analysis of characteristics and forecast difficulties of TCs over Northwestern Pacific in 2018," Meteorological Monographs, vol. 47, pp. 359-372, 2021.

[10] M. Xiao, "Change in the occurrence frequency of landfalling and non-landfalling tropical cyclones over the Northwest Pacific," Journal of Climate, vol. 34 no. 8, pp. 3145-3155, DOI: 10.1175/JCLI-D-20-0647.1, 2021.

[11] C. Velden, B. Harper, F. Wells, J. L. Beven, R. Zehr, T. Olander, M. Mayfield, C. C. H. I. P. Guard, M. Lander, R. Edson, L. Avila, A. Burton, M. Turk, A. Kikuchi, A. Christian, P. Caroff, P. McCrone, "The Dvorak tropical cyclone intensity estimation technique: a satellite-based method that has endured for over 30 years," Bulletin of the American Meteorological Society, vol. 87 no. 9, pp. 1195-1210, DOI: 10.1175/BAMS-87-9-1195, 2006.

[12] K. Emanuel, "100 years of progress in tropical cyclone research," Meteorological Monographs, vol. 59, pp. 15.1-15.68, DOI: 10.1175/AMSMONOGRAPHS-D-18-0016.1, 2018.

[13] J. A. Knaff, C. R. Sampson, M. E. Kucas, C. J. Slocum, M. J. Brennan, T. Meissner, L. Ricciardulli, A. Mouche, N. Reul, M. Morris, G. Chirokova, P. Caroff, "Estimating tropical cyclone surface winds: current status, emerging technologies, historical evolution, and a look to the future," Tropical Cyclone Research and Review, vol. 10 no. 3, pp. 125-150, DOI: 10.1016/j.tcrr.2021.09.002, 2021.

[14] L. Dong, F. Zhang, "OBEST: an observation-based ensemble subsetting technique for tropical cyclone track prediction," Weather and Forecasting, vol. 31 no. 1, pp. 57-70, DOI: 10.1175/WAF-D-15-0056.1, 2016.

[15] J. P. Cangialosi, E. Blake, M. DeMaria, A. Penny, A. Latto, E. Rappaport, V. Tallapragada, "Recent progress in tropical cyclone intensity forecasting at the National Hurricane Center," Weather and Forecasting, vol. 35 no. 5, pp. 1913-1922, DOI: 10.1175/WAF-D-20-0059.1, 2020.

[16] J. Moon, J. Park, D. H. Cha, "Does increasing model resolution improve the real-time forecasts of western north pacific tropical cyclones?," Atmosphere, vol. 12 no. 6,DOI: 10.3390/atmos12060776, 2021.

[17] I.-J. Moon, I. Ginis, T. Hara, B. Thomas, "A physics-based parameterization of air-sea momentum flux at high wind speeds and its impact on hurricane intensity predictions," Monthly Weather Review, vol. 135 no. 8, pp. 2869-2878, DOI: 10.1175/MWR3432.1, 2007.

[18] R. L. Elsberry, L. Chen, J. Davidson, R. F. Rogers, Y. Wang, L. Wu, "Advances in understanding and forecasting rapidly changing phenomena in tropical cyclones," Tropical Cyclone Research and Review, vol. 2, pp. 13-24, 2013.

[19] S. Gao, L. Dong, Y. Xu, Q. Qian, "Analysis of the characteristics and forecast difficulties of typhoons in western North Pacific in 2016," Meteorological Monographs, vol. 44, pp. 284-293, DOI: 10.7519/j.issn.1000-0526.2018.02.008, 2018.

[20] Y. S. Nekkali, K. K. Osuri, M. Mohapatra, "Physical understanding of the tropical cyclone intensity and size relations over the North Indian Ocean," Climate Dynamics,DOI: 10.1007/s00382-024-07118-0, 2024.

[21] N. Bloemendaal, H. de Moel, S. Muis, I. D. Haigh, J. C. J. H. Aerts, "Estimation of global tropical cyclone wind speed probabilities using the STORM dataset," Scientific Data, vol. 7 no. 1,DOI: 10.1038/s41597-020-00720-x, 2020.

[22] M. C. Kruk, E. J. Gibney, D. H. Levinson, M. Squires, "A climatology of inland winds from tropical cyclones for the eastern United States," Journal of Applied Meteorology and Climatology, vol. 49 no. 7, pp. 1538-1547, DOI: 10.1175/2010JAMC2389.1, 2010.

[23] M. M. Kovach, C. E. Konrad, "The spatial distribution of tornadoes and high wind impacts associated with inland-moving tropical cyclones in the southeastern United States," Physical Geography, vol. 35 no. 3, pp. 245-271, DOI: 10.1080/02723646.2014.909231, 2014.

[24] J. W. Choi, Y. Cha, H. Kim, R. Lu, "Relationship between the maximum wind speed and the minimum sea level pressure for tropical cyclones in the western North Pacific," Journal of Climatology & Weather Forecasting, vol. 4,DOI: 10.4172/2332-2594, 2016.

[25] D. R. Chavas, K. A. Reed, J. A. Knaff, "Physical understanding of the tropical cyclone wind-pressure relationship," Nature Communications, vol. 8 no. 1,DOI: 10.1038/s41467-017-01546-9, 2017.

[26] L. Bai, H. Yu, P. Black, Y. Xu, M. Ying, J. Tang, R. Guo, "Re-examination of the tropical cyclone wind-pressure relationship based on pre-1987 aircraft data in the western North Pacific," Weather and Forecasting, vol. 34 no. 6, pp. 1939-1954, DOI: 10.1175/WAF-D-18-0002.1, 2019.

[27] H. P. Wang, L. Dong, Y. L. Xu, G. Z. Nie, "Analysis on the characteristics of typhoon activity and forecasting difficulties in Western North Pacific in 2019," Meteorological Monographs, vol. 47, pp. 1009-1020, 2021.

[28] Y. H. Yang, X. T. Lei, "Statistics of strong wind distribution caused by landfall typhoon in China," Journal of Tropical Meteorology, vol. 20, pp. 633-642, 2004.

[29] Y. Lu, W. Zhu, F. Ren, X. Wang, "Changes of tropical cyclone high winds and extreme winds during 1980–2014 over China," Climate Change Research, vol. 12, pp. 413-421, 2016.

[30] C. Yao, Z. Xiao, S. Yang, X. Luo, "Increased severe landfall typhoons in China since 2004," International Journal of Climatology, vol. 41 no. S1,DOI: 10.1002/joc.6746, 2020.

[31] X. Ni, Q. Zhang, D. Ma, L. Wu, F. Ren, "Climatology and trends of tropical cyclone high wind in mainland China: 1959–2011," Journal of Geophysical Research: Atmospheres, vol. 120 no. 24, pp. 12378-12393, DOI: 10.1002/2015JD024031, 2015.

[32] L. Liang, X. Cui, C. Wang, L. Bai, "Characteristic distribution of surface winds associated with landfalling tropical cyclones in mainland China," Chinese Journal of Atmospheric Sciences, vol. 42, pp. 96-108, DOI: 10.3878/j.issn.1006-9895.1708.16224, 2018.

[33] Y. Lu, H. Zhao, D. Zhao, Q. Li, "Spatial-temporal characteristic of tropical cyclone disasters in China during 1984−2017," Haiyang Xuebao, vol. 43, pp. 45-61, 2021.

[34] C. Yue, Y. Tang, W. Gu, Z. Han, X. Wang, "Study of urban barrier effect on local typhoon precipitation," Meteorological Monographs, vol. 45, pp. 1611-1620, 2019.

[35] Z. Han, J. Tan, C. S. B. Grimmond, B. Ma, T. Yang, C. Weng, "An integrated wind risk warning model for urban rail transport in Shanghai, China," Atmosphere, vol. 11 no. 1,DOI: 10.3390/atmos11010053, 2020.

[36] J. Xu, "Distribution of wind speed and direction when typhoons influencing Shanghai," Meteorological Monographs, vol. 8, pp. 66-70, 2005.

[37] X. Wang, P. Huang, M. Gu, "Field measurements about integral scales of near-ground turbulence during typhoon ‘Mufia’," Journal of Tongji University (Natural Science), vol. 10, pp. 1491-1497, 2012.

[38] X. Wang, P. Huang, Y. Dai, M. Gu, "Comparison study on wind characteristics under typhoon and normal climate conditions in Shanghai area," Journal of Building Materials and Structures, vol. 36, pp. 93-98, 2015.

[39] L. Li, X. Du, "The distribution of peak wind speeds induced by typhoons along the coasts of the Changjiang Estuary and Hangzhou Bay," Journal of East China Normal University (Natural Science), vol. 2, 2021.

[40] Z. Han, C. Yue, C. Liu, W. Gu, Y. Tang, Y. Li, "Evaluation on the applicability of ERA5 reanalysis dataset to tropical cyclones affecting Shanghai," Frontiers of Earth Science, vol. 16 no. 4, pp. 1025-1039, DOI: 10.1007/s11707-022-0972-7, 2022.

[41] M. Ying, W. Zhang, H. Yu, X. Lu, J. Feng, Y. Fan, Y. Zhu, D. Chen, "An overview of the China meteorological administration tropical cyclone database," Journal of Atmospheric and Oceanic Technology, vol. 31 no. 2, pp. 287-301, DOI: 10.1175/JTECH-D-12-00119.1, 2014.

[42] H. Hersbach, B. Bell, P. Berrisford, S. Hirahara, A. Horányi, J. Muñoz-Sabater, J. Nicolas, C. Peubey, R. Radu, D. Schepers, A. Simmons, C. Soci, S. Abdalla, X. Abellan, G. Balsamo, P. Bechtold, G. Biavati, J. Bidlot, M. Bonavita, G. Chiara, P. Dahlgren, D. Dee, M. Diamantakis, R. Dragani, J. Flemming, R. Forbes, M. Fuentes, A. Geer, L. Haimberger, S. Healy, R. J. Hogan, E. Hólm, M. Janisková, S. Keeley, P. Laloyaux, P. Lopez, C. Lupu, G. Radnoti, P. Rosnay, I. Rozum, F. Vamborg, S. Villaume, J. N. Thépaut, "The ERA5 global reanalysis," Quarterly Journal of the Royal Meteorological Society, vol. 146 no. 730, pp. 1999-2049, DOI: 10.1002/qj.3803, 2020.

[43] Shanghai Statistics Bureau, Shanghai Statistical Yearbook, 2021.

[44] T. Ye, N. Zhao, X. Yang, Z. Ouyang, X. Liu, Q. Chen, K. Hu, W. Yue, J. Qi, Z. Li, P. Jia, "Improved population mapping for China using remotely sensed and points-of-interest data within a random forests model," Science of The Total Environment, vol. 658, pp. 936-946, DOI: 10.1016/j.scitotenv.2018.12.276, 2019.

[45] L. Chen, Y. Ding, Introduction to the Western Pacific Typhoon, 1979.

[46] Y. Li, L. Chen, J. Wang, "The diagnostic analysis on the characteristics of large-scale circulation corresponding to the sustaining and decaying of tropical cyclone after it’s landfall," Acta Meteorologica Sinica, vol. 62 no. 2, pp. 167-179, 2004.

[47] R. C. Li, W. Zhou, C. M. Shun, T. C. Lee, "Change in destructiveness of landfalling tropical cyclones over China in recent decades," Journal of Climate, vol. 30 no. 9, pp. 3367-3379, DOI: 10.1175/JCLI-D-16-0258.1, 2017.

[48] X. Chen, L. Wu, J. Zhang, "Increasing duration of tropical cyclones over China," Geophysical Research Letters, vol. 38 no. 2,DOI: 10.1029/2010GL046137, 2011.