1. Introduction
As one of the costliest and deadliest natural disasters, tropical cyclones (TCs) pose a high risk for global coastal regions (Woodruff et al. 2013; Adger et al. 2005). On average, nine tropical cyclones make landfall in China each year (Yin et al. 2010), bringing strong winds, storm surge, and heavy rainfall leading to severe property damages and loss of lives. Supertyphoon Haiyan, which made landfall in the Philippines, Vietnam, and Southeast China in November 2013, is one of the strongest tropical cyclones ever recorded in the region, causing 6000 fatalities and estimated damages of USD 4.55 billion (in 2013 values) resulting from a combination of rainfall, coastal flooding, and wind damages (Long et al. 2016). From 1985 to 2002, an average of 483 deaths were caused by typhoons each year, and a total of 490 000 houses collapsed resulting in direct economic losses of CNY 34.7 billion (in 2000 values) (Shi 2016). These damages have been increasing in recent years over China; Chen et al. (2018) found a statistically significant increasing trend in the inflation-adjusted losses associated with tropical cyclones from 1983 to 2015.
Tropical cyclone precipitation (TCP) plays an important role in TC-related damages. TCP can cause flooding and landslides and could lead to compounding impacts on coastal communities already affected by storm surge and strong winds (Shi et al. 2014; Raymond et al. 2020), and extending flood-related damages to inland regions. Both the intensity and spatial extent of TCP are important factors in tropical cyclone regional impacts. Several studies have examined TCP climatology and trends (Klotzbach 2006; Prat and Nelson 2016; Kunkel et al. 2010; Guzman and Jiang 2021a), TCP spatial extent (Matyas 2010; Zhou et al. 2018; Touma et al. 2019), the relationship between TCP and TC strength (Zhou and Matyas 2017; Konrad and Perry 2010; Guzman and Jiang 2021b), and different meteorological factors contributing to TCP characteristics (W. Zhang et al. 2018; Wang et al. 2018; Guzman and Jiang 2021b). In the western North Pacific, especially for those TCs making landfall in China, many studies showed that TCP intensity and spatial extent are influenced by TC intensity, structure, and location. Feng and Shu (2018) found that higher TC intensity could lead to higher TCP intensity and larger spatial coverage of heavy rain before TC landfall. Yu et al. (2017) found that axisymmetric rainfall is closely related to TC intensity using satellite-derived rainfall data. TCs with higher intensity have higher averaged rain rates, higher averaged peak axisymmetric rain rates, more average total rain, and larger averaged rain areas. Xu et al. (2014) showed that vertical wind shear dominates the rainfall asymmetries, and the magnitude of the rainfall asymmetry increases with shear magnitude but decreases with increasing TC intensity. Although these studies cover many characteristics of TCP, there are few systematic studies that relate TC strength to TCP characteristics along TC tracks (before, during, and after landfall) and for all types of TCs. Increasing our understanding of TCP intensity and spatial extent could help inform strategies for mitigating associated risks.
Some studies have examined the historical variability in intensity and spatial extent of TCP for western North Pacific tropical cyclones. For example, Ren et al. (2007) employed the Objective Synoptic Analysis Technique (OSAT) to partition TCP according to TC intensity at specific locations using station observations. The OSAT system groups all precipitation observations into different rain belts and identifies whether the station precipitation is influenced by a TC by comparing distances between a station and the TC center, the specified maximum TC size, and specified minimum TC size. Based on the case study of two TCs that made landfall in China in 2003, Koni and Imbudo, Ren et al. (2007) found that, as compared with using a fixed TCP radius that is used in most studies, OSAT is able to distinguish precipitation rain belts and estimate the size of TCP rain belts associated with different TCs. Kubota and Wang (2009) used data from 22 rain gauge stations in the western North Pacific islands, the Philippines, Taiwan, and the southwest islands of Japan to investigate the effects of TCs on seasonal and interannual variability of the total rainfall. They found that, on average, TCP decreased at a relatively faster rate up to 1000 km away from the TC center, and beyond that, TC rainfall weakens and does not vary with distance. These studies show that not only does TCP spatial extent and TCP intensity vary along TC tracks, but that TCP intensity also varies with distance from the track center, with the highest intensities occurring within a 1000 km radius. These characteristics of TCP need to be considered when quantifying TCP spatial extent and intensity and their changes.
Given the significant contribution of TCP to the total annual rainfall in southern, southeastern, and eastern China (Ren et al. 2006), many studies have also assessed the historical changes of typhoons’ TCP characteristics using observational records. For example, Qiu et al. (2019) found that the intensity and frequency of tropical cyclone extreme precipitation have increased in southeastern China in the period spanning 1958–2016, based on the daily precipitation observations of 157 meteorological stations. Similarly, Zhang et al. (2017) showed an increasing trend of TCP in southeastern China and a strong connection between flooding and extreme precipitation caused by TCs. Additionally, Ying et al. (2011) found increasing trends in the number of rain days per tropical cyclone, total precipitation per storm, and maximum 1-h precipitation over mainland China since the 1970s.
When assessing variations in TCP intensity, previous studies mostly used a fixed radius (typically 500 or 1000 km) from the center of the TC track for all TC categories, whether strong or weak. However, TCP intensity and spatial extent, as well as trends in TCP characteristics, may vary depending on the intensity of the tropical cyclones. Therefore, to better understand the role of TC intensity in TCP characteristics and trends, it is important to determine TCP intensity and spatial extent for different TC categories (e.g., Touma et al. 2019). Building on previous work, our study aims to identify the climatological TCP intensity and spatial extent under different TC categories and time scales in China. Based on daily precipitation data and TC best-track data from 1951 to 2019, we quantify the TCP intensity and spatial extent in China by employing a geostatistical framework developed by Touma et al. (2018, 2019). In this study, we address the following scientific questions:
-
How does TCP intensity and spatial extent differ spatially and for different TC intensity categories in China during 1951–2019?
-
How do the TCP intensity and trends differ when using a fixed radius from when TC-varying spatial extents are considered?
2. Data and method
a. Observational data
Daily precipitation from the observational records of 699 meteorological stations in China, provided by the National Meteorological Information Center, China Meteorological Data Service Center (Feng et al. 2004), are used in this study. The station precipitation data were available from 1951 to 2019 at the time of the analysis, and the 699 stations are well distributed over Northeast China, East China, and South China (Fig. 1). We also built two new binary datasets, Z1 and Z50, to identify stations that have exceeded precipitation thresholds of 1 and 50 mm for that day, respectively. A station has a binary value of 1 if it exceeds the threshold, and a value of 0 if it is below the threshold.
An example of using the 700-km neighborhood to determine the binary station dataset. This example shows the TCP (≥1 mm day−1) length scale calculation for Supertyphoon Rammasun at 1200 UTC 19 Jul 2014. At this location and this time step, Supertyphoon Rammasun is classified as TSST (a supertyphoon reduced to tropical storm at the location). Dark-blue and dark-green stations are for TCP equal to or exceeding 1 mm day−1 [Z(x) = 1]. Light-blue and light-green stations are for TCP less than 1 mm day−1 [Z(x) = 0]. Green stations are inside the 700-km radius (red circle), and blue stations are outsideit.
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
The TC tracks are taken from the western North Pacific best-track dataset from the Joint Typhoon Warning Center (JTWC; Chu et al. 2002). We only considered western North Pacific TCs that reach at least tropical storm intensity during their lifetime, that is, we excluded tropical depressions from our analysis, a total of 1788 TCs for the period 1951 to 2019. Each track contains information on the TC center latitude and longitude locations and their maximum sustained wind speed at each specific date and time. While the TC tracks are available on 6- or 3-hourly (typically only near landfall) time intervals, the precipitation data are only available on a daily resolution. Thus, we selected only the times in each TC track at 1200 UTC to match with the station precipitation. In this study we consider track locations for each TC, whether the position is before, during, or after landfall in China, as long as the sample size of precipitation data is large enough to apply the semivariogram method described below.
b. Relaxed moving neighborhood method for TCP spatial extent
A geostatistical framework used to analyze the intensity and spatial extent of TCP was developed by Touma et al. (2019) and applied to United States landfalling North Atlantic hurricanes. Here, we adopt the same approach to assess TCP characteristics for western North Pacific TCs making landfall over China. For each TC, we establish the TC location and intensity at each daily time step. We use the maximum sustained wind speed at each time step to classify the TC based on the TC descriptions from JTWC. A tropical storm (TS) is defined as those with maximum sustained wind speed between 34 and 63 kt (1 kt ≈ 0.51 m s−1). When the maximum wind speed equals to or exceeds 64 kt, the storms are categorized into two subcategories: a typhoon (TY) if the maximum wind speed is between 64 and 129 kt, and a supertyphoon (ST) if the maximum wind speed is 130 kt or higher. Then, we identify the lifetime maximum intensity (LMI) category for each TC using the maximum intensity that the TC reached during its lifetime, which is a good representation of the overall intensity of the TC. Although TCs may change intensity multiple times before landfall in China, here we are interested in the whole trajectory and not a smaller segment of the TC evolution. Figure 2 shows the changes of TCs number from 1951 to 2019, with different LMI categories. In total there are, 370 STs, 858 TYs, and 620 TSs that make landfall in China during the 68 years of our analysis period. We define a “point intensity-LMI” category for each TC track snapshot, which corresponds to the intensity at each TC track snapshot and LMI. As an example, TSTY indicates that the storm intensity at that specific time is a TS and that storm lifetime intensity reaches TY intensity. There are six possible point-LMI combinations, namely, TSTS, TSTY, TSST, TYTY, TYST, and STST. The second step is to employ a moving neighborhood and semivariogram framework to analyze TCP spatial extent and intensity under the different point-LMI categories (Touma et al. 2018, 2019). For each binary precipitation dataset (Z1 or Z50), we use a 700-km-radius neighborhood around the TC track point to identify 1–1 station pairs (two stations that exceed the threshold) and 1–0 station pairs (one station that exceeds the threshold and one that does not). Every station within the 700 km radius for that track point is paired up with every other station in the dataset (Fig. 1). This method allows us to “relax” the neighborhood, that is, by allowing one station to fall outside of the boundary, we prevent excluding stations that may experience TCP beyond the 700 km radius and allowing us to quantify a more accurate measure for the extent of TCP.
The year-to-year variations of the number of TCs per year from 1951 to 2019 for the different LMI categories. The red, purple, and blue lines are yearly numbers of TCs with ST, TY, and TS lifetime maximum intensity, respectively.
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
An example of the semivariogram value as a function of the separation distance (see section 2 for details on calculation of the length scales). This example is the same as Fig. 1, but for the fitted semivariogram for TCP extent at the track point of Supertyphoon Rammasun. The colored bars show the number of station pairs that have a raw semivariogram value of 0 or 0.5. The points show the experimental semivariogram value, which is an average of the raw semivariogram values over each separation distance interval (25 km). The line plot shows the fitted spherical semivariogram, with the practical range of 292 km labeled for this specific example.
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
In this function, α is the practical range, b is the partial still, and c is the nugget. The partial still is the maximum value of the semivariogram beyond the practical range and the nugget reflects microscale variability, that is, errors due to the choice of the 25-km interval used to calculate the experimental semivariogram. In our example of Supertyphoon Rammasun in Fig. 1, we focus on the practical range (marked by a vertical arrow in Fig. 3), which we use to define the length scale of Z1 or Z50 TCP for a given track point (Touma et al. 2018, 2019).
c. Tropical cyclone precipitation intensity
We calculate the distribution of TCP for the 1951–2019 period in two ways. First, we use a 500-km fixed neighborhood around each TC location to identify the stations that are located near the TC track. A 500-km radius is believed to capture TC-induced rainfall around a TC track (Q. Zhang et al. 2018; Lau et al. 2008; Jiang and Zipser 2010; Dare et al. 2012; Yokoyama and Takayabu 2008; Lonfat et al. 2004). Second, we use the length scale calculated in the geostatistical framework as the radius of the neighborhood and calculate the distribution of TCP using this TC-varying neighborhood. This allows us to quantify the discrepancies in TCP characteristics when using a fixed neighborhood in comparison with the TC-varying length scale of TCP.
d. Significance testing and trend calculation
To understand the differences in TCP characteristics for different categories, we employ the Mann–Whitney–Wilcoxon (MWW) test to identify statistically significant nonzero differences in the medians and full distributions of TCP intensity and spatial extent for all categories (Bauer 1972). The Wilcoxon test is used to solve the two-sample problem by constructing a confidence set for a shifting parameter with Wilcoxon two-sample statistics and the p value result shows the difference between the two samples. In this study, when the p value of the Wilcoxon test is smaller than 0.001, the distributions of the two samples are statistically significantly different. For the trend calculation at each station, we use the linear regression model (l m) function in R to calculate the regression coefficient (Searle 1997), and we also use the Mann–Kendall (MK) trend test to check whether the trend is statistically significant. The basic principle of MK tests for the trend is to examine the sign of all pairwise differences of observed values (Libiseller and Grimvall 2002). In this study, when the p value of the MK tests is smaller than 0.05, the trend is considered to be statistically significant.
3. Results
a. Climatology of tropical cyclone precipitation spatial extent
We calculate the spatial extent of the 1 mm day−1 (Z1) and 50 mm day−1 (Z50) rainfall over China for our analysis period. Similar to Touma et al. (2019), who assessed the spatial extent of TCP in the eastern United States, the spatial extent of Z1 precipitation calculated in our study reach up to ∼1500 km, whereas those of Z50 reach up to ∼800 km. We find that TCs that are centered on the mainland or near the coastline of China have a larger TCP spatial extent (bigger than 500 km) than those centered over the sea with 1 mm day−1 rainfall (Z1) (Fig. 4). In other words, for most TCs, as a tropical cyclone moves from the ocean toward land, the rainfall area expands in size and as its storm intensity weakens at the same time. Additionally, we find that weaker LMI categories (TY and TS) with the same point intensity, have a smaller TCP extent than the ST LMI category, for both Z1 and Z50 (i.e., TYTY < TYST, TSTS < TSTY < TSST, etc.) (Figs. 5a,b). Meanwhile, for the same LMI categories, the TC snapshot with the lower intensity tends to have the largest extent in Z1 (i.e., TSST > TYST > STST, TSTY > TYTY, etc.) (Fig. 5a). The median extent of the 1 mm day−1 rainfall intensity ranges from 250 km to close to 500 km, with a large spread within each storm category. The median of the Z1 TCP extent of TSST is 165 km larger than the median of TYST and 311 km larger than STST, and the average of the Z1 TCP extent of TSST is 103 km larger than the average of TYST and 201 km larger than STST. Similarly, the median of the Z1 TCP extent of TSTY is 68 km larger than the median of TYTY, and the average of the Z1 TCP extent of TSTY is 37 km larger than the average of TYTY. While only the TSST category is significantly distinct from other point-LMI categories (p < 0.001 for the MWW test; Fig. 5c), these results show that the TCs that reach their maximum intensity of ST or TY and then weaken to TS strength have a larger TCP spatial extent than other point-LMI categories. For the more intense rainfall category (50 mm day−1), the median spatial extent is much smaller than the size of the 1 mm day−1 one and ranges from 110 to 210 km. The smaller spatial extent for the intense rainfall region is expected, as the strongest rainfall is associated with the eyewall of a TC. In this case, the median and average of the Z50 TCP extent for the TSST storms is not the most extensive spatially, while the median is 16 km larger than that of TYST, and the average is 53 km larger than that of TYST. Meanwhile, the median is 54 km smaller than STST, and the average is 68 km smaller than STST. However, these differences are not statistically significant, and relatively similar to the separation distance interval used in the study, and therefore may be subject to microscale variability. Therefore, the spatial extent of only the TSTS category is statistically significantly distinct from other point-LMI categories (p < 0.001; Fig. 5d).
The distribution of all TCP spatial extent values at all TC track points. Red, orange, and yellow colors are for TCP spatial extent smaller than 500 km, and cyan, blue, and purple colors are for TCP spatial extent larger than 500 km.
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
(a) The distribution of Z1 (1 mm day−1) and (b) Z50 (50 mm day−1) TCP spatial extent for each point-LMI category; the line in the box shows the median value, and the crosses in the box show the average value. Numbers shown above each boxplot are the total number of different TCP extent in each “point-LMI” category. Also shown are the difference of the median of (c) Z1 and (d) Z50 TCP spatial extent among the different point-LMI classifications. The number and shading of each grid represent the difference of the median TCP intensity of the row category minus the column category. The p value of the MWW test when comparing the distribution among the categories is shown in italics.
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
We note here, that for five of the point-LMI categories of TCs, the median spatial extent is smaller than 500 km and only the median spatial extent of the “TSST” category is 501 km, confirming that a 500-km fixed radius around a storm is sufficient to capture the full distribution of the intensity of TCP for the majority of TCs. However, there are large spreads in the sizes of Z1 for the different categories of TCs (Fig. 5a). In fact, for approximately 34% of TCs, 500 km is smaller than the calculated spatial extent of the storm. Additionally, the spatial extent varies (usually increases) as the storms weaken, pointing to the need to expand or narrow the size considered when assessing the associated TCP. Therefore, we investigate the importance of varying the radius around the TC track in capturing the climatology and trends of TCP intensity.
b. Climatology of tropical cyclone precipitation
We first calculate the total TCP associated with all TCs impacting the China coast each year from 1951 to 2019 using a fixed 500-km radius neighborhood and then recalculate the total TCP using a varying neighborhood equal to the length scale calculated using the geostatistical method previously described (Figs. 6a,b). In both cases, the largest amount of TCP is found at stations located in coastal regions including Guangdong, Fujian, Zhejiang Province, and Hainan Island, reaching more than 80 mm of rainfall per year, contributing to about 10%–20% of the total yearly rainfall in these regions (Figs. 7a,b). The TCP amount decreases rapidly and quickly reduces to 20 mm or less per year 500 km inland. Relative to the fixed-radius method, the total TCP amount is greater at most stations, but reduced at some inland stations, when using a TC-varying radius (Fig. 6c), and this difference leads to the different TCP contribution ratio in each station (Fig. 7c). The fixed 500-km-radius method underestimates the TCP amount where the spatial extent is larger than 500 km and overestimates the TCP amount where the spatial extent might be smaller than the 500 km. Therefore, our findings show that using a TC-varying extent yields a more accurate estimate of the TCP amount.
The average annual TCP amount in each station from 1951 to 2019 based on TCP spatial extent of (a) fixed 500-km radius and (b) fitted semivariogram values. (c) The difference in total TCP amount from 1951 to 2019 between fitted semivariogram and fixed 500-km-radius extent [(b) minus (a)].
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
The distribution of the ratio of TCP to total precipitation during 1951–2019 based on the two types of TCP spatial extent estimates: (a) fixed 500-km radius and (b) fitted semivariogram radius. (c) The difference in distribution between fitted semivariogram and fixed 500-km-radius extent [(b) minus (a)].
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
We also examine the relationship between TC intensity categories and the associated precipitation intensity, and the effect of using a TC-varying radius. Figures 8a and 8c illustrate the TCP intensity distribution for each of the six point-LMI categories for all track locations over land. Surprisingly, the maximum rainfall intensity does not coincide with the strongest storm intensity at that specific track point. In terms of the median TCP intensity, the TSST category has the most intense rainfall among all categories, which means that the most intense rainfall over land occurs when the storm is characterized as a tropical storm for a lifetime maximum intensity of a supertyphoon track. In contrast, for all the TCs with LMI reaching supertyphoon intensity, the median rainfall intensity is much weaker for locations where the storm has a supertyphoon intensity (STST) relative to those weakened to tropical storm intensity. The median TCP intensity of TSST exceeds the median of TYST by 4.6 mm day−1 with a fixed 500-km radius and 5.3 mm day−1 with a TC-varying radius. The median TSST TCP also exceeds the STST TCP by 3.9 mm day−1 with a fixed 500-km radius and 6.25 mm day−1 with a TC-varying radius. This result implies that for a supertyphoon, the most intense rainfall over land occurs when the storm weakens to a tropical storm strength, rather than when the storm has supertyphoon intensity. Similarly, in comparing TSTY and TYTY categories, it is seen that the largest TCP intensity occurs when the typhoon weakens to a tropical storm strength. On the other hand, for storms with TS intensity, a stronger LMI leads to more intense rainfall amounts (TSST) than in storms with a weaker LMI (TSTS). The comparison between the categories TYST and TYTY also supports this conclusion. Furthermore, in the upper percentiles of the distributions TSST has the highest TCP values, namely, 86 mm day−1 (fixed 500-km radius) and 79 mm day−1 (varying radius) for the 90th percentile, while TSTS only reach 53 mm day−1 (fixed 500-km radius) and 45 mm day−1 (varying radius). Based on the p value of the MK test, using a fixed 500-km radius, the TCP intensity distribution of the STST category is not statistically significantly different from the others, but in the case of the TYST category, the distribution is statistically significantly distinct from the other categories, except STST. Considering a varying radius, the TCP intensity distributions of the TSST, TSTY, and TSTS categories are statistically significantly different from the others (p < 0.001; Figs. 8b,d).
The distribution of Z1 TCP intensity for each point-LMI category based on two types of TCP spatial extent: (a) fixed 500-km radius and (c) varying (fitted semivariogram) radius. The 90th, 95th, and 99th percentiles of the outliers are indicated by circles, triangles and squares, respectively, the line in the box shows the median value, and the crosses in the box show the average value. The number above each boxplot is the total number of all observed precipitation in all stations located in TCP spatial extent in each “point-LMI” category. Also shown is the difference in Z1 TCP intensity among the different point-LMI categories based on two types of TCP spatial extent: (b) fixed 500-km radius and (d) varying (fitted semivariogram) radius. The number and shading in each grid represent the difference of the median TCP intensity of the row category minus the column category. The p values of the MWW test of the differences are shown in italics.
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
c. Tropical cyclone precipitation trends
Given the fact that TCP can be more accurately estimated when using a varying size for the storms, it is relevant to ask if the temporal trend of TCP may be affected by the storm size as well. We show linear trends for TCP (Z1) and heavy TCP (Z50) for the period from 1951 to 2019 in Table 1, Figs. 9 and 10, respectively, for both the fixed radius (Figs. 9a and 10a) and the varying estimates (Figs. 9b and 10b). In both the fixed-radius and varied-radius estimates, there is an increasing trend of daily TCP over most stations in South, Southeast, and East China, and a decreasing trend in central and Northeast China. The distribution of heavy TCP (Z50) trends estimated using both fixed and varied radius, are very different from those of all TCP (Z1). In this case, increasing trends are found for both coastal stations and inland stations. However, when compared with other locations, heavy TCP increases at a much higher rate inland, especially in the north of China, such as Henan Province, Shanxi Province, and Inner Mongolia. Based on the MK test results, the increasing trends in Z1 TCP trends are significant for stations located along the South, Southeast, and East China coastal regions, and most stations in Guizhou Province. A comparison shows that the TC-varying-radius estimate and the fixed-radius estimate have generally similar trends of both Z1 TCP and heavy TCP (Table 1). There is a slight tendency for varying radius estimate to have more stations with significant decreasing trend of both TCP and heavy TCP; for example, there are 23 stations with significant decreasing trend of TCP for TC-varying-radius estimate, but only 6 stations in the fixed-radius estimate. Therefore, we show that using TC-varying radius may result in more accurate estimate of the TCP trend.
The distribution of normal TCP (Z1) trend at each station based on two types of TCP spatial extent: (a) fixed 500-km radius and (b) fitted semivariogram values. The different color in each station shows the regression coefficient value, and different shape in each station shows the p value of the MK test result: circle means p value is smaller than 0.05 and the trend at this station is significant, whereas crosses mean p value is bigger than 0.05 and the trend at this station is not significant.
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
The distribution of heavy TCP (Z50) trend at each station based on two types of TCP spatial extent: (a) fixed 500-km radius and (b) fitted semivariogram values. The different color in each station shows the regression coefficient value, and different shape in each station shows the p value of the MK test result: circle means p value is smaller than 0.05 and the trend in this station is significant, whereas crosses mean p value is bigger than 0.05 and the trend in this station is not significant.
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
The Z1 (1 mm day−1) TCP intensity of each “point-LMI” category. Each category has two groups, one is pre-LMI (blue), which contains those TCP stations at which TCP occur before TC reaching the maximum wind speed along the track; another is post-LMI (red), which contains those TCP stations at which TCP occur during or after TC reaching the maximum wind speed along the track.
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
Summary of the total number of stations with increasing and decreasing TCP trends for the two different TCP categories and the fixed-radius and varying-radius estimates. The numbers inside the parentheses indicate the total number of stations having a significant trend (p < 0.05 in MK test results).
4. Discussion and conclusions
In this study, the tropical cyclone precipitation intensity and spatial extent for tropical cyclones in the western North Pacific are analyzed with daily precipitation records and tropical cyclones’ best-track data. The highlights of our research are summarized as follows:
-
For all precipitation caused by tropical cyclones (Z1 TCP), the strongest TCP intensities and largest TCP spatial extents occur when tropical cyclones reach the supertyphoon intensity and then weaken to tropical storm strength after making landfall in China, which is like the TCP in the North Atlantic (Touma et al. 2019).
-
For heavy precipitation caused by tropical cyclones (Z50 TCP), the strongest TCP intensities also occur when tropical cyclones reach supertyphoon strength and then weaken to tropical storms after making landfall in China. However, the largest TCP spatial extents occur when tropical cyclones reach supertyphoon intensities based on a very small sample of heavy precipitation in STST category.
-
These results indicate that the intensity of TC rainfall depends on the lifetime maximum of the storm, which is usually achieved over the oceans. This may be related to the fact that these storms are able to accumulate more moisture while over the oceans. Upon landfall, the TC weakens, becomes more asymmetrical, and expands in diameter, releasing heavier precipitation and expanding the spatial extent of associated TCP. This phenomenon has been documented for many TCs in the region (Chen et al. 2006; Dong et al. 2010; Yu and Cheng 2013).
-
When compared with inland areas, most coastal stations in China have increasing TCP trends, and stations in southern and southeastern China have higher positive trends than stations in northeastern China. However, for some inland stations, heavy TCP is increasing at high rates.
-
In comparison with using a calculation method with a fixed 500-km radius, using a TC-varying radius based on different TC intensities and TC lifetime maximum intensities could estimate TCP amount more accurately. Our results show that the most intense and widespread TCP occurs when TCs that have the strongest LMI weaken to tropical storm intensity, similar to the results for the North Atlantic TCs (Touma et al. 2019). Overall, the storms that are of tropical storm strength near landfall but achieved a lifetime maximum intensity of a supertyphoon or typhoon strengths consistently produce the heaviest rainfall. For TC tracks with the same LMI, TC points have a smaller TCP spatial extent and TCP intensity before reaching the LMI (Figs. 4 and 11; Table 2), than after they reach LMI. However, the lowest median TCP intensity is found to be for supertyphoons when they are at their strongest (STST). This result differs from North Atlantic TCs, where the weakest TCP occurs for the weakest tropical cyclone category (i.e., tropical storms; Touma et al. 2019). The discrepancy between the two basins could potentially be due to sampling errors, or to different definitions of supertyphoons and major hurricanes, as the intensity threshold of supertyphoons is 130 kt, whereas the threshold of Atlantic major hurricanes is 96 kt. Thus, the sample size of the supertyphoons is much smaller than Atlantic major hurricanes, which could magnify uncertainty and errors. Additionally, by using a varying TCP extent, our study was able to estimate TCP intensity more accurately, especially for extreme TCs with stronger intensity and small probability.
Table 2Summary of the total number of stations with different point-LMI categories under two different scenarios: one is pre-LMI, which contains those TCP stations for which TCP occur before TC reaching the maximum wind speed along the track; another is post-LMI, which contains those TCP stations for which TCP occur during or after TC reaching the maximum wind speed along the track. The ratio here is the value of pre-LMI divided by the value of post LMI.
To test the sensitivity of our results to the station data coverage, we choose supertyphoon Rammasun in 2014 as an example (Fig. 12) to illustrate the sensitivity of the semivariogram method (Figs. 1 and 3). To validate the estimates of TCP spatial extent based on rain gauge station data, we use the same best-track data during Supertyphoon Rammasun (Fig. 12b), and the daily precipitation data from CPC Global Unified Gauge-Based Analysis of Daily Precipitation (https://psl.noaa.gov/data/gridded/data.cpc.globalprecip.html), which is quality controlled against independent information from concurrent radar/satellite observations, as well as numerical model forecasts. We use the target area from 70° to 135°E, and from 0° to 55°N, with a total of 14 541 gridpoint “stations,” located at the center of each grid. The result shows that both calculations with different datasets can capture the changing trend of TCP spatial extent with different TC intensity. However, with the increasing station density and decreasing distance between the stations, the TCP spatial extent calculated from the CPC dataset is smaller than that calculated from observational stations, with the comparison showing that the TCP extent between different datasets in TC location 36–38 (Fig. 12a) with the CPC result is 120–140 km smaller than observation.
(a) The comparison of TCP spatial extent during Supertyphoon Rammasun based on two different datasets. The blue line is the TCP spatial extent computed from 699 observation stations in China. The red line is the TCP spatial extent computed from CPC Global Unified Gauge-Based Analysis of Daily Precipitation dataset. (b) The best track of Supertyphoon Rammasun, and the location of TC centers used in (a) as indicated by the different color dots.
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
Meanwhile, in this case when we compare the TCP extent between different datasets in TC location 42 and 43, the CPC result is much smaller than observation, with a gap of 143 and 208 km, respectively. One possible reason for this may be the inclusion of precipitation data in Vietnam when calculating the TCP extent using the CPC dataset. In our research we could not identify the uncertainty of TCP extent calculation without the precipitation data in the ocean. However, we are able to have a more accurate TCP extent calculation through the addition of precipitation data observed over surrounding countries in South, Southeast, and East Asia, as well as over the islands located away from the continent.
Our results also show that the annual TCP volume and the percentage of TCP to total precipitation decreases from the coastal region to inland China, in both fixed and varying TCP spatial extent estimates. A comparison of the trends of annual precipitation (AP) during 1951–2019 (Fig. 13) shows that both TCP trend and AP trend are increasing rapidly in the coastal area of southern and southeastern China. However, for inland stations, the AP trend is increasing rapidly while the TCP trend is increasing at lower rates or even decreasing. Previous studies (Li and Zhou 2015; Wang et al. 2020; Ren et al. 2006; Zhang et al. 2013; Liu and Wang 2020) reached a similar conclusion when comparing coastal and inland China. Our results also agree with the geographical distribution of TCP trends found by Liu and Wang (2020) over the period 1980–2017, with increasing TCP trends for stations in coastal China. Similar to our findings, Liu and Wang (2020) also found that the TCP trends for strong TCs were increasing in Guangdong and Hainan Province.
The distribution of annual precipitation trend at each station. The different color in each station shows the regression coefficient value, and different shape in each station shows the p value of the MK test result: circle means p value is smaller than 0.05 and the trend in this station is significant; crosses mean p value is bigger than 0.05 and the trend in this station is not significant.
Citation: Journal of Applied Meteorology and Climatology 61, 5; 10.1175/JAMC-D-21-0166.1
Our results provide nuanced understanding of TCP in southern and Southeast China, adding to previous literature (Li et al. 2017; Liu et al. 2020; Yin et al. 2010; Sparks and Toumi 2020). Moreover, our study quantifies TCP characteristics and trends in northern China, which has recently suffered damages from heavy rain caused by strong landfalling TCs, such as Supertyphoon Lekima in 2019 (Jia et al. 2020), and the inland flooding in Henan in 2021, which the heavy precipitation maybe linked to TC (Guo et al. 2021). Our findings provide a key understanding of the characteristics of TCP that contribute to the devastating impacts of TCs and managing future TC risks given the recent trends in TCP.
Acknowledgments.
This research was funded by a research project of “Assessing the risk of compound extreme events and large-scale natural disasters in megacities of China” from the Ministry of Science and Technology of the People’s Republic of China (Grant 2017YFC1503001). We thank the National Meteorological Information Centre of China Meteorological Administration for providing the weather station data and Naval Meteorology and Oceanography Command for providing the best-track data from the Joint Typhoon Warning Center Tropical Cyclone best tracks.
Data availability statement.
All of the best-track data of tropical cyclones used during this study are openly available from Joint Typhoon Warning Center (https://www.metoc.navy.mil/jtwc/jtwc.html?western-pacific). Because of confidentiality agreements, supporting data can only be made available to real-name registered researchers subject to a nondisclosure agreement. Details of the data and how to request access are available online (https://data.cma.cn/en) at the National Meteorological Information Centre, China Meteorological Data Service Centre.
REFERENCES
Adger, W. N., T. P. Hughes, C. Folke, S. R. Carpenter, and J. Rockström, 2005: Social-ecological resilience to coastal disasters. Science, 309, 1036–1039, https://doi.org/10.1126/science.1112122.
Bauer, D. F., 1972: Constructing confidence sets using rank statistics. J. Amer. Stat. Assoc., 67, 687–690, https://doi.org/10.1080/01621459.1972.10481279.
Chen, S. S., J. A. Knaff, and F. D. Marks, 2006: Effects of vertical wind shear and storm motion on tropical cyclone rainfall asymmetries deduced from TRMM. Mon. Wea. Rev., 134, 3190–3208, https://doi.org/10.1175/MWR3245.1.
Chen, W., Y. Lu, S. Sun, Y. Duan, and G. C. Leckebusch, 2018: Hazard footprint-based normalization of economic losses from tropical cyclones in China during 1983–2015. Int. J. Disaster Risk Sci., 9, 195–206, https://doi.org/10.1007/s13753-018-0172-y.
Chu, J.-H., C. R. Sampson, A. S. Levine, and E. Fukada, 2002: The Joint Typhoon Warning Center tropical cyclone best-tracks, 1945–2000. Naval Research Laboratory Rep. NRL/MR/7540‐02, 16 pp.
Dare, R. A., N. E. Davidson, and J. L. McBride, 2012: Tropical cyclone contribution to rainfall over Australia. Mon. Wea. Rev., 140, 3606–3619, https://doi.org/10.1175/MWR-D-11-00340.1.
Dong, M., L. Chen, Y. Li, and C. Lu, 2010: Rainfall reinforcement associated with landfalling tropical cyclones. J. Atmos. Sci., 67, 3541–3558, https://doi.org/10.1175/2010JAS3268.1.
Feng, S., Q. Hu, and W. Qian, 2004: Quality control of daily meteorological data in China, 1951–2000: A new dataset. Int. J. Climatol., 24, 853–870, https://doi.org/10.1002/joc.1047.
Feng, X., and S. Shu, 2018: How do weak tropical cyclones produce heavy rainfall when making landfall over China. J. Geophys. Res. Atmos., 123, 11 830–11 848, https://doi.org/10.1029/2018JD029228.
Guo, X., A. Zhu, Q. Li, and R. Chen, 2021: Improving the response to inland flooding. Science, 374, 831–832, https://doi.org/10.1126/science.abm7149.
Guzman, O., and H. Jiang, 2021a: Global increase in tropical cyclone rain rate. Nat. Commun., 12, 5344, https://doi.org/10.1038/s41467-021-25685-2.
Guzman, O., and H. Jiang, 2021b: Heavier inner-core rainfall of major hurricanes in the North Atlantic basin than other global basins. J. Climate, 34, 5707–5721, https://doi.org/10.1175/JCLI-D-20-0668.1.
Jia, L., Z. Jia, F. Ren, C. Ding, M. Wang, and T. Feng, 2020: Introducing TC intensity into the DSAEF_LTP model and simulating precipitation of super-typhoon Lekima (2019). Quart. J. Roy. Meteor. Soc., 146, 3965–3979, https://doi.org/10.1002/qj.3882.
Jiang, H., and E. J. Zipser, 2010: Contribution of tropical cyclones to the global precipitation from eight seasons of TRMM data: Regional, seasonal, and interannual variations. J. Climate, 23, 1526–1543, https://doi.org/10.1175/2009JCLI3303.1.
Klotzbach, P. J., 2006: Trends in global tropical cyclone activity over the past twenty years (1986–2005). Geophys. Res. Lett., 33, L10805, https://doi.org/10.1029/2006GL025881.
Konrad, C. E., II, and L. B. Perry, 2010: Relationships between tropical cyclones and heavy rainfall in the Carolina region of the USA. Int. J. Climatol., 30, 522–534, https://doi.org/10.1002/joc.1894.
Kubota, H., and B. Wang, 2009: How much do tropical cyclones affect seasonal and interannual rainfall variability over the western North Pacific? J. Climate, 22, 5495–5510, https://doi.org/10.1175/2009JCLI2646.1.
Kunkel, K. E., D. R. Easterling, D. A. R. Kristovich, B. Gleason, L. Stoecker, and R. Smith, 2010: Recent increases in U.S. heavy precipitation associated with tropical cyclones. Geophys. Res. Lett., 37, L24706, https://doi.org/10.1029/2010GL045164.
Lau, K.-M., Y. P. Zhou, and H.-T. Wu, 2008: Have tropical cyclones been feeding more extreme rainfall? J. Geophys. Res., 113, D23113, https://doi.org/10.1029/2008JD009963.
Li, R. C. Y., and W. Zhou, 2015: Interdecadal changes in summertime tropical cyclone precipitation over southeast China during 1960–2009. J. Climate, 28, 1494–1509, https://doi.org/10.1175/JCLI-D-14-00246.1.
Li, R. C. Y., W. Zhou, C. M. Shun, and T. C. Lee, 2017: Change in destructiveness of landfalling tropical cyclones over China in recent decades. J. Climate, 30, 3367–3379, https://doi.org/10.1175/JCLI-D-16-0258.1.
Libiseller, C., and A. Grimvall, 2002: Performance of partial Mann–Kendall tests for trend detection in the presence of covariates. Environmetrics, 13, 71–84, https://doi.org/10.1002/env.507.
Liu, L., and Y. Wang, 2020: Trends in landfalling tropical cyclone–induced precipitation over China. J. Climate, 33, 2223–2235, https://doi.org/10.1175/JCLI-D-19-0693.1.
Liu, L., Y. Wang, R. Zhan, J. Xu, and Y. Duan, 2020: Increasing destructive potential of landfalling tropical cyclones over China. J. Climate, 33, 3731–3743, https://doi.org/10.1175/JCLI-D-19-0451.1.
Lonfat, M., F. D. Marks Jr., and S. S. Chen, 2004: Precipitation distribution in tropical cyclones using the Tropical Rainfall Measuring Mission (TRMM) microwave imager: A global perspective. Mon. Wea. Rev., 132, 1645–1660, https://doi.org/10.1175/1520-0493(2004)132<1645:PDITCU>2.0.CO;2.
Long, J., C. Giri, J. Primavera, and M. Trivedi, 2016: Damage and recovery assessment of the Philippines’ mangroves following Super Typhoon Haiyan. Mar. Pollut. Bull., 109, 734–743, https://doi.org/10.1016/j.marpolbul.2016.06.080.
Matyas, C. J., 2010: Associations between the size of hurricane rain fields at landfall and their surrounding environments. Meteor. Atmos. Phys., 106, 135–148, https://doi.org/10.1007/s00703-009-0056-1.
Prat, O. P., and B. R. Nelson, 2016: On the link between tropical cyclones and daily rainfall extremes derived from global satellite observations. J. Climate, 29, 6127–6135, https://doi.org/10.1175/JCLI-D-16-0289.1.
Qiu, W., F. Ren, L. Wu, L. Chen, and C. Ding, 2019: Characteristics of tropical cyclone extreme precipitation and its preliminary causes in southeast China. Meteor. Atmos. Phys., 131, 613–626, https://doi.org/10.1007/s00703-018-0594-5.
Raymond, R., and Coauthors, 2020: Understanding and managing connected extreme events. Nat. Climate Change, 10, 611–621, https://doi.org/10.1038/s41558-020-0790-4.
Ren, F., G. Wu, W. Dong, X. Wang, Y. Wang, W. Ai, and W. Li, 2006: Changes in tropical cyclone precipitation over China. Geophys. Res. Lett., 33, L20702, https://doi.org/10.1029/2006GL027951.
Ren, F., Y. Wang, X. Wang, and W. Li, 2007: Estimating tropical cyclone precipitation from station observations. Adv. Atmos. Sci., 24, 700–711, https://doi.org/10.1007/s00376-007-0700-y.
Searle, S. R., 1997: Linear Models. John Wiley and Sons, 560 pp.
Shi, P., 2016: Natural Disasters in China. Springer, 288 pp.
Shi, P., L. Lu, M. Wang, J. A. Wang, and W. Chen, 2014: Disaster system: Disaster cluster, disaster chain and disaster compound. J. Nat. Disasters, 23, 1–12, https://doi.org/10.13577/j.jnd.2014.0601.
Sparks, N., and R. Toumi, 2020: Pacific subsurface ocean temperature as a long-range predictor of South China tropical cyclone landfall. Commun. Earth Environ., 1, 30, https://doi.org/10.1038/s43247-020-00033-2.
Touma, D., A. M. Michalak, D. L. Swain, and N. S. Diffenbaugh, 2018: Characterizing the spatial scales of extreme daily precipitation in the United States. J. Climate, 31, 8023–8037, https://doi.org/10.1175/JCLI-D-18-0019.1.
Touma, D., S. Stevenson, S. J. Camargo, D. E. Horton, and N. S. Diffenbaugh, 2019: Variations in the intensity and spatial extent of tropical cyclone precipitation. Geophys. Res. Lett., 46, 13 992–14 002, https://doi.org/10.1029/2019GL083452.
Wang, L., J. Chen, and L. Wu, 2020: Impact of winter SST anomaly in tropical Eastern Pacific on following summer tropical cyclone precipitation in China. Int. J. Climatol., 40, 739–749, https://doi.org/10.1002/joc.6234.
Wang, S. Y. S., L. Zhao, J.-H. Yoon, P. Klotzbach, and R. R. Gillies, 2018: Quantitative attribution of climate effects on Hurricane Harvey’s extreme rainfall in Texas. Environ. Res. Lett., 13, 054014, https://doi.org/10.1088/1748-9326/aabb85.
Woodruff, J. D., J. L. Irish, and S. J. Camargo, 2013: Coastal flooding by tropical cyclones and sea-level rise. Nature, 504, 44–52, https://doi.org/10.1038/nature12855.
Xu, W., H. Jiang, and X. Kang, 2014: Rainfall asymmetries of tropical cyclones prior to, during, and after making landfall in South China and southeast United States. Atmos. Res., 139, 18–26, https://doi.org/10.1016/j.atmosres.2013.12.015.
Yin, Y., M. Gemmer, Y. Luo, and Y. Wang, 2010: Tropical cyclones and heavy rainfall in Fujian Province, China. Quat. Int., 226, 122–128, https://doi.org/10.1016/j.quaint.2010.03.015.
Ying, M., B. Chen, and G. Wu, 2011: Climate trends in tropical cyclone‐induced wind and precipitation over mainland China. Geophys. Res. Lett., 38, L01702, https://doi.org/10.1029/2010GL045729.
Yokoyama, C., and Y. N. Takayabu, 2008: A statistical study on rain characteristics of tropical cyclones using TRMM satellite data. Mon. Wea. Rev., 136, 3848–3862, https://doi.org/10.1175/2008MWR2408.1.
Yu, C.-K., and L.-W. Cheng, 2013: Distribution and mechanisms of orographic precipitation associated with Typhoon Morakot (2009). J. Atmos. Sci., 70, 2894–2915, https://doi.org/10.1175/JAS-D-12-0340.1.
Yu, Z., Y. Wang, H. Xu, N. Davidson, Y. Chen, Y. Chen, and H. Yu, 2017: On the relationship between intensity and rainfall distribution in tropical cyclones making landfall over China. J. Appl. Meteor. Climatol., 56, 2883–2901, https://doi.org/10.1175/JAMC-D-16-0334.1.
Zhang, J., L. Wu, F. Ren, and X. Cui, 2013: Changes in tropical cyclone rainfall in China. J. Meteor. Soc. Japan, 91, 585–595, https://doi.org/10.2151/jmsj.2013-502.
Zhang, Q., X. Gu, P. Shi, and V. P. Singh, 2017: Impact of tropical cyclones on flood risk in southeastern China: Spatial patterns, causes and implications. Global Planet. Change, 150, 81–93, https://doi.org/10.1016/j.gloplacha.2017.02.004.
Zhang, Q., Y. Lai, X. Gu, P. Shi, and V. P. Singh, 2018: Tropical cyclonic rainfall in China: Changing properties, seasonality, and causes. J. Geophys. Res. Atmos., 123, 4476–4489, https://doi.org/10.1029/2017JD028119.
Zhang, W., G. Villarini, G. A. Vecchi, and J. A. Smith, 2018: Urbanization exacerbated the rainfall and flooding caused by Hurricane Harvey in Houston. Nature, 563, 384–388, https://doi.org/10.1038/s41586-018-0676-z.
Zhou, Y., and C. J. Matyas, 2017: Spatial characteristics of storm‐total rainfall swaths associated with tropical cyclones over the eastern United States. Int. J. Climatol., 37, 557–569, https://doi.org/10.1002/joc.5021.
Zhou, Y., C. J. Matyas, H. Li, and J. Tang, 2018: Conditions associated with rain field size for tropical cyclones landfalling over the eastern United States. Atmos. Res., 214, 375–385, https://doi.org/10.1016/j.atmosres.2018.08.019.
