Climatology of Tropical Cyclone Extreme Rainfall over China from 1960 to 2019

The tropical cyclone (TC) is one of the most powerful rainfall-generating systems on Earth and many extreme rainfall events in China and elsewhere are related to TC activity (Tao, 1980). For example, the world record for 24-h rainfall of 1825 mm on Reunion Island in the western South Indian Ocean was triggered by tropical storm Denise in 1966 (Holland, 1993). The maximum 24-h rainfall in China of 1748.5 mm at Ali Mountain, Taiwan was triggered by the super typhoon Herb (1996) (Chen, 2012). The six highest records of 24-h accumulated rainfall in China were all caused by TCs (Chen et al., 2010). Therefore, TC rainstorms always have been a focus of both meteorological forecasts and scientific research (Gao et al., 2009). TC rainstorms are associated with multi-scale atmospheric interactions and complex physical processes and our knowledge of the mechanisms of their formation is still limited (Chen et al., 2010). It is important to accurately forecast extreme precipitation and its distribution when a TC makes landfall. However, Yu et al. (2020) showed that the equitable threat score for the forecast of the 24-h precipitation caused by TCs that make landfall is <0.1 when they reach the intensity of a torrential rainstorm (≥250 mm). It is clearly currently very difficult to forecast tropical cyclone extreme rainfall (TCER).

The contribution of TC precipitation to total precipitation is significant in the China and therefore the spatiotemporal distribution of TC precipitation has attracted much interest. Ren et al. (2002) showed that the southeastern coastal regions of China are most frequently affected by TCs, with >500 mm each year, accounting for 20%−40% of the total annual precipitation (Ren et al., 2006). The total TC precipitation in the China shows a significant decreasing trend, mainly because the frequency of TCs affecting China has decreased (Ren et al., 2006). The area influenced by TC precipitation is mainly determined by strong TCs (Yang et al., 2019). TC rainfall disasters could have long impact time and affect a wide area. Extreme rainfall disasters caused by TCs that make landfall are generally the most serious events. For example, the super typhoon Nina (1975) induced an unprecedented torrential rainstorm in Henan Province after landfall, which resulted in at least 26000 casualties, destroyed 102 km of the Beijing−Guangzhou railway and interrupted train operations for 18 days (Ding et al., 1978; Yang et al., 2017). The super typhoon Morakot (2009) caused serious destruction to many provinces and cities in the China. In Taiwan, there were 673 casualties, 26 people reported missing and the direct economic loss was 19.5 billion New Taiwan Dollars (Chen, 2012). Recently, the super typhoon Lekima (2019) made landfall, had a long duration over land, caused 70 casualties (or missing people), and direct losses reached 51.53 billion RMB (Li et al., 2020).

With the rapid development of detection instruments and the recent increase in the adverse effects of extreme weather events, there has been much research on extreme rainfall events (Wu and Luo, 2019). Given the current background of global warming, especially since the 1990s, the intensity and frequency of extreme rainfall events and flooding have been increasing in the China (Xia et al., 2019). The increases in the frequency and intensity of extreme rainfall events are more significant in monsoon regions than in the non-monsoon region (Han et al., 2019). Furthermore, extreme rainfall from TC has increased with recent global warming (Wu et al., 2020). Khouakhi et al. (2017) showed that 35%−50% of extreme rainfall in global regions worldwide is a result of TCs. Chang et al. (2012) indicated that the trend of TCER are clearly different in Chinese mainland and Taiwan, noting that TCs caused 75%−77% of the extreme hourly precipitation in Taiwan (Wu et al., 2017). Similarly, TC rainstorms are important extreme rainfall events on the southeastern coast and in some inland areas of China (Wang et al., 2008). Studies further showed that the average amount of precipitation per TC (Zhang et al., 2013) and the ratio of TCER frequency to the total frequency of extreme rainfall in southeastern coastal areas of China increased significantly with rainfall intensity, and that the frequency of TCER>300 mm increased from 1960 to 2012 (Su et al., 2015). Wu et al. (2007) showed that the number of TCER days and the total amount of extreme rainfall in Hainan Island increased from 1962 to 2005. In terms of the tracks, TCs causing extreme rainfall in the China occurred with two dominant tracks: (1) the TC makes landfall on the southeastern coast of Chinese mainland via Taiwan or the ocean north of Taiwan; and (2) the TC makes landfall on the coast of South China or hovers in the offshore area (Jiang and Qi, 2016). The frequency of TCs with extreme daily precipitation >250 mm in the Yangtze and Pearl River deltas has been increasing slowly since the 1980s (Chao and Chao, 2014). Extreme rainfall and the induced flood disasters have a great impact on both human society and economic development, which has become the forefront and hotspot of international academic research at present (Gu and Zhang, 2020).

There has been much research on the synoptic features of TCER (Ding et al., 1978; Gao et al., 2009; Chen and Xu, 2017), but less on its climatological characteristics. Previous research (Jiang et al., 2018; Qiu et al., 2019) on the climatological characteristics of TCER has mainly focused on specific areas (e.g., the southeastern and southern coasts of China) based on a definition with a fixed threshold for the maximum daily precipitation (e.g., 1000 mm or 50 mm) in the TC process. Previous studies considered TCs that made landfall, but not sideswiping TCs, and some of them only used data from 700 weather stations in the China. It is therefore still difficult to comprehensively understand and quantitatively compare the basic climatology of TCER in different areas of China.

In the present study, we used TC precipitation data from national-level daily rain gage data for China for the last 60 years (1960–2019) to define TCER using a percentile method. We then analyzed the spatiotemporal distribution of TCER in China (including Taiwan, Hong Kong and Macao) and the track and intensity of tropical cyclones resulting in TCER.

In this study, the TC best track datasets (1960−2019) developed by the Shanghai Typhoon Institute of China Meteorological Administration were used (Ying et al., 2014; Lu et al., 2021), along with datasets of rainfall induced by TCs from Ren et al. (2011), which were obtained using the objective synoptic analysis technique (OSAT). The basic idea of the OSAT is to divide stations with rainfall into different rain belts and then to identify which rain belts are associated with TCs. This dataset has been widely used in research on TC precipitation (Ren et al., 2006, 2007; Luo et al., 2016; Jiang et al., 2018; Feng et al., 2020).

The OSAT precipitation dataset covers 1888 stations in the China affected by TC precipitation, which includes stations in Taiwan (21 stations), Hong Kong and Macao besides stations in the Chinese mainland. It is necessary to select as many samples as possible for research into extreme events, so we used the period of 1960−2019. Among 1888 stations in the OSAT precipitation dataset, only 1742 stations built before January 1960 (including January 1960) are reserved. To ensure the effectiveness of precipitation at each station, the stations were sorted in descending order based on the total frequency of daily TC precipitation. The bottom 10% of stations were excluded, specifically, 170 stations with TC precipitation less than 23 days from 1960 to 2019 were removed. Finally, this gave a total 1572 (1742−170) stations used in this study. The daily TC precipitation data from 1200 UTC to 1200 UTC were used to analyze extreme precipitation events.

As a result of its vast territory and complex and diverse topography, precipitation in China shows a wide range of spatial distributions and therefore a relative threshold is more reasonable for defining TCER. The TCER threshold at a single station used here refers to the percentile method (Jiang and Qi, 2016), which is commonly used in studies of the climatic characteristics of extreme events. The 99-th percentile of TC rainfall was used to define TCER and the nonparametric scheme of Bonsal et al. (2001) was used to determine the threshold. The daily TC precipitation was composed of N samples sorted in ascending order (P₁, P₂, P₃, …, P_m, …, P_n). The probability that a certain value is less than or equal to P_m is:

where F is the percentile, M is the serial number and N is the number of samples. The precipitation threshold is determined as F = 99%. The threshold criterion used here is more stringent than in previous studies (Jiang and Qi, 2016; Jiang et al., 2018; Qiu et al., 2019). A TCER event was defined as when TCER occurred at a certain station on a certain day. TCER events refer to several episodes of TCER that occurred at more than one station or on more than one day.

Figure 1a shows the spatial distribution of the TCER threshold in China. The regions affected by TCs include most areas in central and eastern China, although they extend north to central and eastern Inner Mongolia and northeast China and west to the east of southwestern China, including Sichuan, Yunnan and Guizhou. The threshold of TCER decreases from the southeastern coast to the northwest inland, with regional heterogeneity. The threshold of TCER in Taiwan reaches the magnitude of a torrential rainstorm (250 mm d⁻¹) and in coastal regions reaches the intensity of a heavy rainstorm (100 mm d⁻¹). At latitudes 30°−40°N, the threshold for exceeding the magnitude of heavy rainstorms extends westward from the coastal areas to the provinces of Shandong, Henan, southeastern Hebei, Jiangsu and southern Anhui. This is closely related to the extratropical transition process under the interaction of a mid-latitude trough and the northward movement of TCs that make landfall, as has been demonstrated previously in many case studies (Ding et al., 1978; Chen et al., 2012). The extreme threshold in some areas of northeast China, which is less affected by TCs, also reaches the magnitude of a heavy rainstorm (100 mm). The threshold of TCER in most of the remaining regions reaches the magnitude of a rainstorm (50 mm) and only 10.8% of the stations have TCER thresholds <50 mm. The maximum TCER threshold in the study area is 578 mm at Ali Mountain (23.51°N, 120.81°E) in Taiwan, 35 times higher than the minimum of 16.1 mm at Wuzhai (38.92°N, 111.82°E) in Shanxi.

Figure 1.  Spatial distribution of the (a) threshold intensity of the 99-th percentile (mm d⁻¹), (b) frequency (d yr⁻¹), (c) average intensity (mm d⁻¹) and (d) historical maximum (mm d⁻¹) of TCER affecting China from 1960 to 2019.

Figure 1b shows the average annual frequency of TCER, which has a spatial distribution consistent with that of the annual average number of TC precipitation days. The average annual frequency of TCER is >13 d yr⁻¹ in the southeastern coastal region and decreases inland. In the area south of the Yangtze River, the frequency of TCER is generally between 5 and 13 d yr⁻¹, whereas it is generally <3 d yr⁻¹ to the north. The maximum annual frequency of TCER in Taiwan is 16 days in Hengchun (22.01°N, 120.74°E), Dawu (22.36°N, 120.90°E), Lanyu (22.04°N, 121.55°E) and Ali Mountain (23.51°N, 120.81°E) in the south and Yushan (23.49°N, 120.95°E) in the north. The maximum annual frequency in the Chinese mainland is 15 days at Jiuxian Mountain (25.72°N, 118.10°E) in Fujian Province.

The extreme rainfall intensity at a single station was defined as the average rainfall rate of TCER for that station and is the ratio of the total extreme daily rainfall to the number of extreme rainfall days. Figure 1c shows the distribution of the extreme rainfall intensity, which is similar to the distribution of the 99-th percentile threshold. In addition to the strong intensity centers in Taiwan and the coastal areas of southern Chinese mainland, the intensity in the area south of the Yangtze River and eastern Sichuan is also >100 mm. The intensity is concentrated locally in inland regions, especially in the north, with scattered high-value centers. This may be due to the extreme values of daily precipitation in a TC and the low frequency of TC precipitation in this area. For example, there were only two TCER events in Zhumadian (33.00°N, 114.02°E) in Henan during the period 1960–2019, with daily precipitation values of 420.4 mm (typhoon Cecil in 1982) and 258.1 mm (typhoon Nina in 1975). The average intensity was 339.2 mm, which exceeded the intensity at some coastal stations.

According to the distribution of the historical maximum of TC precipitation (Fig. 1d), the daily precipitation is >250 mm at numerous coastal stations in southern China. The daily precipitation is >500 mm at some stations in Taiwan, Guangdong and Hainan. There are also some stations with TC daily precipitation >500 mm in the northern inland areas, such as Henan and Shandong. The maximum TC daily precipitation occurrs at Ali Mountain (23.51°N, 120.81°E) in Taiwan; influenced by the typhoon Morakot (2009), the daily precipitation reached 1165.5 mm on 9 August 2009 and 1161.5 mm on 8 August 2009, ranking the first and second in history.

Group occurrence is an important indicator of extreme events. The total frequency of precipitation exceeding the extreme rainfall threshold in a single TC was calculated to indicate the group occurrence instead of the area of the extreme event. A total of 1623 TCs were generated in the South China Sea and the northwest Pacific from 1960 to 2019, with 1056 TCs producing precipitation in China. In addition, 581 TCs led to extreme rainfall, accounting for 55.0% of the TCs with precipitation and 35.8% of all TCs. On average, 12 stations broke the extreme rainfall threshold in each TC. Figure 2 shows there were only 99 TCs (17.0%) in which only one station reached the extreme threshold, while the remaining 482 TCs (83.0%) caused more than one TCER event, with prominent group occurrence. Among these, 223 (38.4%) and 22 (3.8%) TCs led to more than 10 and 50 TCER events, respectively. Three TCs caused more than 100 extreme rainfall events. Specifically, the typhoon Herb (1996) resulted in 117 extreme rainfall events. Table 1 shows the top ten TCs causing the most extreme rainfall events from 1960 to 2019 in detail.

Figure 2.  Frequency and percentage of stations exceeding the extreme threshold in a single tropical cyclone.

Extreme rainfall is unlikely to persist at the same station because of the release of energy from the TC, the weakening of water vapor support and/or the movement of TC system. However, this situation does occur. There were 180 continuous extreme rainfall events from 1960 to 2019, but extreme rainfall for three consecutive days only occurred at two stations—namely, Qionghai (19.23°N, 110.47°E) in Hainan (tropical depression in 2010) and Yushan (23.49°N, 120.95°E) in Taiwan (the typhoon Morakot in 2009). This was mainly due to the slower translational speed of the TCs. Specifically, the average translational speed of consecutive TCER (TCER) was 14.2 (19.8) km h⁻¹, which is statistically significant at the 99% confidence level based on the Student’s t-test. Multi-station continuous extreme rainfall was also observed in a single TC. For example, the typhoon Sepat (2007) caused two-day consecutive extreme rainfall events at 11 stations.

The frequency of TCER can be an important indicator of its impact area. There were on average 116 TCER events per year in China from 1960 to 2019. The maximum annual frequency of TCER events was 293 in 1994, followed by 274 in 2018 and 256 in 1975 (Fig. 3). In these years, several TCs generated multiple extreme rainfall events; for example, the typhoon Tim (1994), the typhoon Russ (1994), the typhoon Nina (1975), the typhoon Ora (1975), the typhoon Rumbia (2018), and the typhoon Ewiniar (2018) induced 90, 80, 83, 78, 102 and 52 TCER events, respectively.

Figure 3.  Interannual variation of the annual average frequency of TCER (green bars), its five-year moving average series (blue solid line) and linear trend (red solid line) in China from 1960 to 2019.

The correlation coefficients of the annual TCER frequency with the number of TCs generated and the number of TCs making landfall were 0.40 and 0.22, which passed the 99% and 90% confidence levels, respectively. There were 23 tropical cyclones in 1975, leading to 256 TCER events, with 11.1 events in each TC on average, the maximum number recorded. The TCER frequency showed an increasing trend but did not pass the 95% confidence level. The TCER frequency fluctuated during the period 1960−2019, with high values from the 1960s to the mid-1970s and low values from the late 1970s to the early 1990s. It reached a high value again in the mid-1990s, then decreased from the late 1990s to the first five years of the 21st century. The TCER frequency rose again after 2006.

The average intensity of TCER in China during the period 1960−2019 was 147.1 mm d⁻¹ and the strongest intensity was 227.7 mm d⁻¹ in 1998 (Fig. 4). Although there were only 33 TCER events in 1998, these stations were mainly located in Taiwan and the southeastern coastal areas. The TCER intensities in these areas were generally strong, resulting in the historical maximum TCER intensity in this year. The TCER intensity was 102.8 mm d⁻¹ in 1982, the historical minimum. Although there were 92 TCER events in this year, these stations were mostly concentrated in northern China with a weak TCER intensity.

Figure 4.  Interannual variation of the average intensity of TCER (green bars), its five-year moving average series (blue solid line) and linear trend (red solid line) in China from 1960 to 2019.

The intensity of TCER showed a weak increasing trend not passing the 95% confidence level. It was relatively weak before the mid-1990s and then strengthened from the mid-1990s to the 2010s, before weakening again. Ren et al. (2006, 2007) previously showed that the total amount of precipitation caused by TCs in China has shown a significant decreasing trend in recent decades, including TCER and non-extreme TC precipitation. This shows the uniqueness of TCER and it is worth comprehensive further study. Zhang et al. (2018) and Wang et al. (2020) concluded that the frequency and magnitude of TCER are modulated by the neutral phase of the El Niño-Southern Oscillation.

Like the seasonal variation of TC precipitation, the seasonal variation of TCER frequency also presents a unimodal structure, with the peak and sub-peak in August and July, respectively, followed by September (Fig. 5). The seasonal distribution of TCER is uneven, with 81.8% of events occurring during the peak period of TC activity (July–September). A total of 41.0% of TCER events occurred in August, which is higher than the proportion of TC precipitation (33.8%) in August, but less than the proportion of TC precipitation in all other months during the whole TC season. This is probably related to the climatology of the strong East Asian summer monsoon and the westward and northward extension of the western Pacific subtropical high.

Figure 5.  Seasonal variations in the TCER frequency (histogram) and all the tropical cyclone precipitation samples (green line) in China during the time period 1960–2019. The abscissa represents the month and the ordinate represents the percentage of samples in each month in the total number of annual samples (units: %).

Precipitation associated with TCs can last for several days over land. Are there any distinct characteristics of TC precipitation processes producing extreme precipitation? Figure 6 shows that TC precipitation processes with extreme rainfall generally lasted for five days over land, accounting for 23.2% of the total TC precipitation processes over land, followed by the proportion of six-day (19.6%) and four-day (19.0%) precipitation processes. This is different from the duration of total TC precipitation (including extreme and non-extreme precipitation) processes over land, which generally lasted for two to five days (not shown). The proportion of three-day processes was largest (20.4%), followed by four-day (20.0%) and five-day (15.7%) processes. Statistically, TCs with extreme precipitation may therefore have a longer duration over land than TCs associated with non-extreme precipitation.

Figure 6.  Frequencies of precipitation duration (blue line; units: %) and the occurrence day of extreme rainfall (red line; units: %) in a single tropical cyclone. The abscissa represents the number of days.

We investigated on which day of the whole TC precipitation process over land TCER usually occurred. Figure 6 shows that TCER mainly occurred on the fourth day, accounting for 30.4% of the total TC precipitation process over land. TCER occurred on the third and fifth days in 23.6% and 19.0% of the processes over land, respectively. This indicates that TCER generally occurred in the second half of the TC precipitation process over land. The heat map of TCER frequency (Fig. 7) shows that although the duration of TC precipitation processes with TCER varied from 1 to 19 days, TCER mostly occurred in the second half of the TC precipitation process over land, except for 10-day and 17-day precipitation processes. In five-day TC precipitation processes with the most TCER, there were 1618 extreme rainfall events, with 698 occurring on the fourth day of the process, accounting for 43.1% of the total frequency.

Figure 7.  Heat map of TCER frequency. The ordinate is the duration of precipitation in a single tropical cyclone, the abscissa is the occurrence day of extreme rainfall and the values in the figure are frequencies. The shading represents the percentage of different occurrence days of extreme rainfall in tropical cyclones of the same duration (units: %).

Figure 8 shows the superposition of the track densities of all the TCs that affected and brought rainfall to China and the TCs with extreme rainfall affecting China. There were two main centers of activity of the former: (1) the eastern side of Hainan Island (18°N, 114°E), which corresponds to TCs with a westward track; and (2) the eastern parts of the Philippines (15°N, 125°E), which corresponds to TCs with northward and northwestward tracks. TCs with extreme rainfall showed a similar track distribution and dominant track. However, the high-value area of the contour was more westward and northward—that is, the activity center of TC with extreme rainfall was closer to Chinese mainland.

Figure 8.  Track densities of all the tropical cyclones that affected and brought rainfall to China (black contours) and tropical cyclones with extreme rainfall (red contours) affecting China from 1960 to 2019 (grid interpolation at a spatial resolution of 1°×1°; frequency).

Figure 9 shows the track and intensity of TCs causing extreme rainfall at >100, >70, >50 and >30 stations during the period 1960−2019. Based on the China national standard [GB/T 19201-2006 in General Administration of Quality Supervision, Inspection and Quarantine of the People’s Republic of China (2006)], TC intensity is divided into six grades: tropical depression; tropical storm; severe tropical storm; typhoon; strong typhoon; and super typhoon. Figure 9a shows that three TCs caused TCER at >100 stations in the last 60 years: the super typhoon Herb (1996); the severe tropical storm Bilis (2006); and the severe tropical storm Rumbia (2018). These TCs caused TCER at 117, 105 and 102 stations, respectively. The three TCs made landfall on the southeastern coast of China with a northwestward track, which persisted after entering inland China. Figure 9b shows that there were 10 TCs causing TCER at >70 stations in the last 60 years and their tracks were highly consistent. Apart from one TC that made landfall with a westward track and affected the provinces of Hainan, Guangdong and Guangxi, the remaining nine TCs made landfall on the southeastern coast of China with a northwestward track and then entered inland China and persisted. The tracks of TCs generating extreme rainfall at >50 stations (Fig. 9c) and >30 stations (Fig. 9d) were similar and could be divided into two categories. The first category was a northwestward track in which the TC passed through Taiwan and made landfall in the southeastern coastal areas of Fujian and Zhejiang before moving inland. The second category was a westward or northwestward track in which the TC entered the South China Sea and made landfall in Hainan, Guangdong and Guangxi. There were two dominant tracks for TCs causing large-scale extreme rainfall in China. The first was a northwestward track in which the TC affected (made landfall in) Taiwan, then made landfall on the southeastern coast of Chinese mainland before moving inland. The TC then moved northward or southward and lasted for a long period of time on the continent. The second was the westward or northwestward track in which the TC affected (made landfall in) Hainan Island, Guangdong and Guangxi.

Figure 9.  Tracks and intensities of tropical cyclones causing extreme rainfall at (a) >100 stations, (b) >70 stations, (c) >50 stations and (d) >30 stations during the period 1960–2019. TD, tropical depression; TS, tropical storm; STS, severe tropical storm; TY, typhoon; STY, strong typhoon; SupTY, super typhoon.

We analyzed the TC intensity grade at the beginning of TCER to investigate the intensity characteristics of TCs with TCER (Fig. 10a). The extreme rainfall processes were divided into two groups based on the intensity of the TC. In the first group, the average amounts of extreme rainfall were almost the same in tropical depressions (42.9% of the TCs), tropical storms (17.9% of the TCs) and severe tropical storms (16.4% of the TCs) when TCER occurred, with average daily precipitation of 118.2, 121.8 and 123.9 mm, respectively. In the second group, the average amount of precipitation increased as the TC strengthened. When TCER occurred, 14.8% of the TCs had the intensity of a typhoon, 5.4% a strong typhoon and 2.5% a super typhoon, with average daily precipitation of 148.8, 172.8 and 157.2 mm, respectively. This is consistent with previous studies on the relationship between the intensity of TC rainfall and the intensity of TCs (Jiang, 2012; Liu et al., 2019).

Figure 10.  (a) Box-plot of the intensity grade of tropical cyclones at the beginning of the extreme rainfall event. (b) Scatter diagram of the change in tropical cyclone intensity during the period of extreme rainfall. The abscissa in part (a) shows the intensity grade and corresponding sample proportion (%) of the tropical cyclones and the ordinate shows the daily extreme precipitation (mm). The abscissa in part (b) shows the 24-hourly change rate of the maximum wind speed near the center of the tropical cyclone (m s⁻¹) with positive (negative) values representing strengthening (weakening) of the intensity of the tropical cyclone and 0 representing the maintenance of the intensity of the tropical cyclone. The ordinate is the daily extreme precipitation (mm). TD, tropical depression; TS, tropical storm; STS, severe tropical storm; TY, typhoon; STY, strong typhoon; SupTY, super typhoon.

By contrast, the strongest extreme precipitation was not always caused by the strongest TC. For example, in the famous “75·8” rainstorm process in Henan Province, the TCER occurred at Shangcai station (33.28°N, 114.27°E) in Henan on 7 August 1975, with daily precipitation of 755.1 mm, but the TC was only classified as a tropical depression (12.0 m s⁻¹). Therefore, the relationship between the intensity of TCs and the intensity of extreme rainfall is uncertain. Strong TCs are not always accompanied by excessive precipitation and weak TCs often result in extreme rainfall events (Yu et al., 2017).

We used the 24-h change rate of the maximum wind speed near the center of the TC to characterize the change in TC intensity during TCER. Figure 10b shows that 66.7%, 25.3% and 8.0% of the TCs weakened, maintained their original intensity or strengthened, respectively, during the period of extreme rainfall. More than half (53.9%) of the TCs weakened by 1.0−15.0 m s⁻¹. In extreme rainfall processes with intensities >250 mm, 65.6%, 25.3% and 9.1% of the TCs weakened, maintained their original intensity, or strengthened, respectively. In extreme rainfall processes with intensities >600 mm, 68.5%, 28.9% and 2.6% of the TCs weakened, maintained their original intensity or strengthened, respectively.

TCs affecting China often interact with the topography, the westerly trough or the mei-yu front, which influences the location and intensity of precipitation. Precipitation mainly occurs in the left-front quadrant of the tropical cyclone before it makes landfall and then in the right-front quadrant after it has made landfall (Meng et al., 2019). In terms of the location of the rainfall maximum, cyclonic rotation has been further identified from south China to east China (Yu et al., 2015). We analyzed the location of TCER relative to the center of the TC. The statistics showed that during the period of extreme rainfall, the TC centers were located on land or sea in 64.7% and 35.3% of the samples, respectively. Thus, TCs were more likely to cause extreme rainfall after making landfall. Figure 11 shows the locations of extreme rainfall corresponding to the centers of TCs over the sea (Fig. 11a) and land (Fig. 11b). It can be concluded that, irrespective of the location of the centers of the TC, the higher the magnitude of extreme rainfall, the closer it is to the center of the TC. In particular, the distance between the location of extreme rainfall (>250 mm) and the center of the TC was ≤3°.

Figure 11.  Locations of extreme rainfall relative to the tropical cyclone centers over the (a) sea and (b) land. The coordinate origin represents the average location of the tropical cyclone center during the extreme rainfall. The X direction stands for the longitude (positive represents east and negative represents west) and the Y direction stands for the latitude (positive represents north and negative represents south). The circles denote the distances of 3°, 6° and 12° to the tropical cyclone center.

When the center of the TC was located over the sea (Fig. 11a), the TCER affecting China was more likely to appear on the northern side of the center of the TC, accounting for 85.1% of the total, whereas only 14.9% of the TCER occurred on the southern side. Specifically, 51.1%, 34.0%, 10.7% and 4.2% of the extreme rainfall occurred to the northwest, northeast, southwest and southeast of the TC, respectively. Therefore, when the center of the TC was located over the sea, the extreme rainfall tended to occur to the northwest of the TC affecting China. This further verifies our previous conclusion that TCs causing extreme rainfall in China have a dominant northwestward track.

When the center of the TC was over land (Fig. 11b), the TCER affecting China was also more likely to appear on the northern side of the center of the TC, accounting for 57.7%. In detail, 24.2%, 33.5%, 23.6% and 18.7% of the extreme rainfall occurred in the northwest, northeast, southwest and southeast of the TC, respectively.

We conclude that when the center of the TC was located over land, extreme rainfall was more likely to occur to the northeast of the TC affecting China. The change in the structure of the TC before and after landfall and the extratropical transition process, which are influenced by the mid-latitude trough and the topography, might cause these differences in the location of TCER relative to the center of TCs between the ocean and land. Analysis of the distance between the location of extreme rainfall and the center of the TC indicates that when the center of the TC was over the sea (land), 7.4% (16.5%), 22.5% (46.8%), 32.5% (67.1%), 57.8% (91.4%) and 99.5% (99.9%) of the extreme rainfall occurred within 1°, 2°, 3°, 6° and 12° of the center of the TC, respectively. The distribution of extreme rainfall was more dispersed (concentrated) when the center of the TC was located over the sea (land). For example, in the northwest quadrant of TCs over the sea with the most TCERs, extreme rainfall occurred within 12° of the center of the TC. By contrast, in the northeast quadrant of TCs over the land with the most TCERs, extreme rainfall occurred within 6° of the center of the TC.

We defined the 99-th percentile of daily TC precipitation as the TCER threshold at each single station and analyzed the spatiotemporal distribution of TCER affecting China based on the TC best-track dataset from the China Meteorological Administration and TC precipitation national-level daily rain gauge data derived from OSAT in the last 60 years (1960–2019). Our results can be summarized as follows.

(1) The TCER threshold decreased from the southeastern coast to the northwest inland, with regional inhomogeneities. The TCER threshold in Taiwan reached the magnitude of a torrential rainstorm (250 mm), whereas it was less than the magnitude of a rainstorm (50 mm) in the northwest. The maximum threshold (578 mm, Ali Mountain in Taiwan) was about 35 times the minimum threshold (16.1 mm, Wuzhai in Shanxi Province). There were some scattered strong centers of TCER inland in addition to those in Taiwan and the coastal regions of southern Chinese mainland. Although the TCER frequency in northern China was not high, it had a high intensity and prominent locality. This is consistent with the conclusions of Zhang et al. (2018), who reported that “TCs are responsible for heavy extreme precipitation over large parts of China including both coastal and inland areas.”

(2) TCER showed a significant group occurrence. On average, there were 12 TCER events in each TC. Both the frequency and intensity of TCER showed slightly increasing trends that did not pass the 95% confidence level. The TCERs mainly occurred during the peak period of TC activity (July−September), accounting for 81.8%; 41.0% of them occurred in August. Most of the TC precipitation processes with extreme rainfall lasted for four to six days and the TCER mainly occurred in the second half of the TC precipitation process over land.

(3) The TCs causing extreme rainfall over a wide area of China had two dominant track patterns. One was a northwestward track, in which the TC affected (made landfall in) Taiwan and subsequently made secondary landfall over the southeastern coast of Chinese mainland, before moving inland and lasting for a long period of time on the continent. The other pattern was a westward or northwestward track, in which the TC affected (made landfall in) Hainan Island, Guangdong and Guangxi.

(4) No statistical linear relationship was found between the intensity of TC and the intensity of TCER. Strong TCs are not always accompanied by strong precipitation and weak TCs can result in extreme rainfall events. During the period of extreme precipitation, 66.7%, 25.3% and 8.0% of TCs weakened, maintained their original intensity, or strengthened, respectively.

(5) More TCER samples (64.7%) occurred when the TC was centered over land than over sea. When the center of the TC was located over the sea (land), the extreme rainfall over land was more likely to appear on its northwestern (northeastern) side with a dispersed (concentrated) distribution. Irrespective of where the center of the TC was located, extreme rainfall over land was unlikely to occur on its southeastern side. The greater the magnitude of extreme rainfall, the closer it was to the center of the TC. In particular, the distance between the location of extreme rainfall >250 mm and the center of the TC was ≤3°, which indicates that TCER mainly occurred in the inner core region of the TC.

We analyzed a long time series of TCER affecting China over the last 60 years to reveal its spatiotemporal distribution, which is helpful in further understanding the climatic background of TCER. The OSAT tropical cyclone precipitation data used here were daily precipitation data derived from the national-level rain gauge datasets, which present the accumulated precipitation from 1200 UTC to 1200 UTC the next day. This is different from the 24-h accumulated precipitation in any period (Chen and Xu, 2017) and the data from the regional meteorological (Cheng et al., 2014) and hydrological (Ding, 2015) stations. Therefore, the exact extreme values for the same TCER event may differ among the different datasets. However, this does not affect the climatology analysis given here. Our conclusions were limited by the rain gauge data datasets used, which means precipitation estimates were missing over the sea. This may have led to the uneven spatial distribution of TCER.

In addition, we only considered the basic spatiotemporal distribution of TCER. It is necessary to study the dominant meteorological factors. Because the distribution of precipitation in China is jointly regulated by monsoon activity and the local topography, with clear regional characteristics, the spatial distribution of extreme precipitation varies significantly. It is therefore essential to investigate further the regional differences in TCER.

Acknowledgements. This work was supported by the National Key Research and Development Program of China (GrantNo. 2019YFC1510205), National Natural Science Foundation of China (Grant Nos. 42175008, 41775048 and 41930972), National Basic Research Program of China (Grant No. 2015CB452804), the Open Grants of the State Key Laboratory of Severe Weather (Grant Nos. 2021LASW-A12 and 2020LASW-B06) and Huafeng Meteorological Media Group Essential Research Project (Grant No. CY-J2020002). The best-track data is from http://tcdata.typhoon.org.cn. The authors are grateful to the insightful editors and the two anonymous reviewers for their constructive comments that highly improved the quality of this paper.

Electronic supplementary material: Supplementary material is available in the online version of this article at https://doi.org/10.1007/s00376-021-1080-4.