1. Introduction
With increasing global mean temperature, more intensified and frequently extreme precipitation events will occur along the southeastern coastal region of China in the future (Guo et al. 2018). With socioeconomic development, the number of people and the value of assets in the urban coastal regions have increased. Widespread torrential floods caused by intense extreme precipitation events contribute significantly to casualties and economic costs. Therefore, analyzing the characteristics and variability in floods is essential for designing strategies to adapt to potential changes in coastal flood risk.
Many previous studies have assessed variability in extreme precipitation in coastal areas around the globe, such as in the United States (Knight and Davis 2009; Kunkel et al. 2010; Barlow 2011), Australia (Lavender and Abbs 2013; Villarini and Denniston 2016), the Korean Peninsula (Kim et al. 2006), and China (Ren et al. 2006, 2007; Chang et al. 2012; Su et al. 2015). However, few researchers have focused on changes in coastal flood behavior (Villarini et al. 2014a; Aryal et al. 2018; Barth et al. 2018). Precipitation is a major driver of flooding and hence of the associated losses (Pielke and Downton 2000; Hundecha and Merz 2012); however, observation evidence provided by some studies (Small et al. 2006; Ivancic and Shaw 2015) reveals that the upward trend in extreme rainfall amounts or intensity is not always associated with increasing flood magnitudes. The possible mechanisms that contribute to decoupling between the trends in flood magnitude and precipitation may be decreases in antecedent soil moisture, earlier snowmelt, increases in canopy storage capacity induced by higher temperatures and changing rainstorm characteristics (such as decreasing storm extent and shorter storm duration) (Sharma et al. 2018). Moreover, artificial facilities, such as constructed reservoirs, can also contribute to inconsistency in changes in flood magnitude and extreme precipitation (Batalla et al. 2004).
In the southeastern coastal region of China, precipitation is composed of TC rainfall caused by landfalling tropical cyclones in the western North Pacific Ocean and monsoon (non-TC) rainfall associated with the East Asian monsoon (Chou et al. 2009). In our study, TC and non-TC rainfall is defined as the situation in which rainfall in a catchment is dominantly caused by TC and non-TC, respectively. When rainfall in a catchment is approximately equally caused by TC and non-TC, this situation is called “mixed rainfall,” and the flooding caused by the mixed rainfall is called the “mixed flooding.” However, our objective is to compare the TC and non-TC flooding. Hence, the mixed flooding is not analyzed in our study. In southeastern coastal China, TC rainfall accounts for 20%–40% of yearly total precipitation (Ren et al. 2006) and non-TC rainfall constitutes at least 50% of yearly total precipitation(Day et al. 2018). TC extreme rainfall accounts for 50%–70% of the total extreme precipitation volume (i.e., larger than the rainfall value of the 95th percentile) on an interannual scale (Wu et al. 2007; Chang et al. 2012). Although the TC extreme precipitation has been analyzed in many previous studies (Chang et al. 2012; Su et al. 2015; Gu et al. 2017; Zhang et al. 2018), only Chang et al. (2012) and Su et al. (2015) calculated the trends in the TC and non-TC extreme precipitation amount and revealed that the annual TC extreme precipitation amount showed a slight decline, but the annual non-TC extreme precipitation amount showed a large increase. However, the hydrographic differences between TC and non-TC flooding and variability in TC and non-TC flooding remain unexplored.
When assessing the variability in TC and non-TC flooding, TC and non-TC flood events must be identified. The identification method of TC floods was first proposed by Waylen (1991), who considered flood events to be caused by a TC if the discharge station was within a 3-km-radius circle centered on the TC and the flood peak occurred within 7 days after the occurrence of the TC. Later, many studies modified the circle radius parameter in Waylen’s identification criteria (Villarini and Smith 2010; Villarini et al. 2014a; Zhu et al. 2015). The circle radius parameter reflects the spatial extent of rainfall associated with TCs (Villarini and Smith 2010, 2013). This parameter is usually set to 500 km in studies in the United States (Villarini and Smith 2010, 2013; Villarini et al. 2014a; Aryal et al. 2018; Zhu et al. 2015), and the rationality of the value was demonstrated by Kimball and Mulekar (2004), Dare et al. (2012), and Villarini et al. (2014b). However, the distance threshold of 500 km is limited and some severe TC-induced flood events cannot be detected. For example, predecessor rain events (PREs), which are typically located approximately 1000 km away from TCs were demonstrated to be triggered by TCs (Galarneau et al. 2010). Since the distances between the TC and location of flooding were larger than 500 km, some severe flooding events caused by PREs over the midwestern United States (Rowe and Villarini 2013) cannot be detected by Villarini’s method (Villarini and Smith 2010), thus underscoring the fact that TC identification based solely on the distance between the gauge location and the TC center is not appropriate. Besides, the identification method of non-TC floods does not exist.
Due to these limitations, we improved the identification procedure of TC flood events by linking flooding to rainfall. The improved procedure can also extract non-TC flood events. The main objective of this study is to compare the characteristics of TC and non-TC flood events and examine changes in TC and non-TC floods in the southeastern coastal region of China. This study is organized as follows. Section 2 presents the study area and data. Section 3 details the methods used to compare TC and non-TC flooding. Section 4 describes the percentage of TC and non-TC extreme precipitation and floods, the hydrographic differences between TC and non-TC rainfall–flood events and the trends in TC and non-TC extreme precipitation and floods. Finally, section 5 compares the results with studies in the southeastern coastal United States.
2. Study area and data
This study selected 19 catchments in areas where the ratio of the TC-induced rainfall to the total rainfall was greater than 10% in the coastal region of southeastern China, including the coastal area of the Chinese mainland and Hainan Island (Fig. 1) (defined as the region within 106°–120°E and 18°–32°N), according to the analysis of spatiotemporal variation in typhoon rainfall over South China (Lee et al. 2010). All 19 catchments are located in the humid subtropical monsoon climate zone. The mean annual rainfall ranges from 1200 to 2800 mm and heavy rainstorms are the major cause of floods in this region. The rainfall in these catchments is produced by two distinct systems: the East Asian monsoon and TCs in the North Pacific Ocean. The ratio of the TC rainfall to the total rainfall ranges from 10% to 50% in this region. The ratio on Hainan Island is higher than that in the coastal area of the Chinese mainland (Chang et al. 2012).
Study catchments with the hydrological stations.
Citation: Journal of Hydrometeorology 21, 7; 10.1175/JHM-D-20-0002.1
Nineteen hydrological stations are available within the study catchments (Table 1). The daily river discharge data from 1960 to 2015 were collected from the Hydrological Yearbooks published by the Information Center of Ministry of Water Resources and collected by the Tsinghua University Library. The daily river discharge data in several years were missing, which may affect the trend detection. Hence, we applied the approach proposed by Slater and Villarini (2017) to analyze the influence of missing data. The applied method on the influence of missing data on trend detection is provided in the Supplementary material. Digital elevation data with a resolution of 90 m were downloaded from the Shuttle Radar Topography Mission (SRTM) database (Jarvis et al. 2008). Daily precipitation data from 1960 to 2015 were from the 0.25° × 0.25° China Gauge-Based Daily Precipitation (CGDPA) product (Shen et al. 2014). The catchment boundaries were extracted from the DEM, and the daily basin-average rainfall was calculated based on the catchment boundaries and CGDPA. To be consistent with the catchment resolution, the 0.25° × 0.25° precipitation data were resampled to a 1-km grid using the nearest neighbor assignment in ArcGIS Software. The basic information on tropical cyclones, including the central location, maximum sustained wind and minimum sea level pressure, were collected from the Shanghai Typhoon Institute (STI) of the China Meteorological Administration. The available period of tropical cyclone track data was from 1960 to 2015, and the temporal resolution was 6 h (Ying et al. 2011).
Basic characteristics of the study catchments and hydrological stations.
3. Methodology
a. Identification of TC and non-TC rainfall–runoff events
Figure 2 shows the three main steps in the identification procedure. First, based on the daily basin-average rainfall and runoff time series, we combined the POT approach (Smith 1984) with the HydRun package to extract rainfall–flood events (named “POT rainfall–flood events” hereafter) with flood peaks that exceeded a “threshold for flood events.” Subsequently, the objective synoptic analysis technique (OSAT) method, proposed by Ren et al. (2006) was applied to separate the gridded precipitation time series into TC rainfall and non-TC rainfall (section 2). Because of its convenience and high accuracy, this method has been widely used in China to distinguish TC precipitation (Wang et al. 2008; Chang et al. 2012; Yang et al. 2018; Wang et al. 2019). Finally, according to the ratio of TC rainfall to total rainfall from the beginning of the rainfall event to the flood peak time in one POT rainfall–flood event, the TC-induced and non-TC-induced rainfall–flood events were identified among the POT rainfall–flood events (section 3).
Schematic diagram of the identification method.
Citation: Journal of Hydrometeorology 21, 7; 10.1175/JHM-D-20-0002.1
1) Extraction of POT rainfall–flood event
The POT approach and MATLAB toolbox for rainfall–flood analysis named HydRun developed by Tang and Carey (2017) were combined to separate POT rainfall–flood events based on daily basin-averaged rainfall and runoff time series. This approach consists of four main steps: base flow separation, runoff event identification, rainfall event identification, and rainfall to runoff events attribution (Fig. 3).
Base flow separation separates streamflow time series into quick and base flows (Fig. 3a).
A local-minimum method and return threshold are used to identify runoff events based on quick flows (Fig. 3b). In this step, two parameters, namely, the peak threshold and return threshold, should be set in HydRun. We set the peak threshold to make sure that two flood events per year per average can be chosen. The return threshold is set to satisfy the constrain condition that two flood peaks of adjacent flood events are larger than the logarithmic value of the drainage area and follow the guidelines for determining flood flow frequency established by the U.S. Water Resources Council (1981). In this study, an appropriate return threshold of 0.1 was set by performing tests with values between 0.05 and 0.3 according to Tang and Carey (2017), and it satisfied the constraint condition.
The rainfall time series is separated into rainfall events by a rainless period whose duration exceeds the threshold of a rainless period (Fig. 3c), which is set as 1 day.
Rainfall events are attributed to the identified runoff events by finding the rainfall events that overlap on the search window of one runoff event (Fig. 3d). The search window parameter (n) was set to 2 days, which ensures that all identified runoff events have attributed rainfall events. POT rainfall–flood events that occur two times per year on average were selected.
Overview of the rainfall–runoff event extraction procedure.
Citation: Journal of Hydrometeorology 21, 7; 10.1175/JHM-D-20-0002.1
2) Separation of TC and non-TC rainfall
Based on the TC dataset and daily precipitation data with a resolution of 1 km, the OSAT method was applied to classify the gridded precipitation on each day as either TC or non-TC rainfall. This method is used because it considers two TC rainfall generation mechanisms (TC circulation rainfall and remote rainfall) whereas the widely used Villarini’s method (Villarini and Smith 2010) of TC precipitation separation only considers the TC circulation rainfall. In fact, TC rainfall is normally classified into two types: TC circulation rainfall and remote rainfall resulting from the interaction between the TC circulation and other weather systems outside of the TC region (Chen et al. 2010; Xing et al. 2016; Ren et al. 2006). TC circulation rainfall represents the main rain shield of TCs, and remote rain was defined as the precipitation occurring outside of the main body of the TC but having some physical relationship with the TC (Xing et al. 2016).
In the OSAT method, the TC circulation rainfall was identified by determining whether the grid is within a circle with a variable radius of D0 from the TC center. TC remote rainfall was identified by checking whether the grid is within a circle with a variable radius of D1 and belongs to the TC rain belts. TC rain belts were obtained by determining independent rain belts by grouping precipitation based on the spatial structure of the gridded precipitation and then identifying TC rain belts from independent rain belts according to the distance between the TC center and the weighted-precipitation center of the rain belts. The variable parameters D0 and D1 were set as piecewise functions of the TC maximum wind speed rather than a fixed value (Table 2). The piecewise functions of the two variable parameters were picked out from 12 alternative function schemes designed in Ren et al. (2007).
Values of the parameters D0 and D1 as a function of the TC maximum wind speed; Dmin is the minimum distance between the TC center and the mainland of China, and Dmin = 0 at landfall or on the land.
3) Separation of TC and non-TC rainfall–flood events
We first calculated the rainfall ratio for POT rainfall–flood events, which was defined as the ratio of TC rainfall in the period Tflood to total rainfall in the period Tflood (the definition of Tflood is given in Fig. 4). Then, we used k-means clustering (Visual Numerics 1997), an unsupervised clustering technique to separate the POT rainfall–flood events into three groups based on the rainfall ratio by assuming that the POT rainfall–flood events can be classified into three types generated by “non-TC rainfall,” “non-TC and TC rainfall (mixed),” and “TC rainfall” (Vormoor et al. 2016). The events in the groups with the largest and smallest mean rainfall ratio were classified as TC-induced and non-TC-induced rainfall–flood events, respectively. The events in the remaining group were considered mixed events and not analyzed in the trend analysis.
Illustration of the time characteristics for a POT rainfall–flood event; tpeak is the date of flood peak, tstart_rain is the beginning of rainfall, and Tflood is the time period from tstart_rain to tpeak.
Citation: Journal of Hydrometeorology 21, 7; 10.1175/JHM-D-20-0002.1
b. Identification of the basin-average TC and non-TC extreme precipitation
We calculated the daily basin-average rainfall, TC rainfall, and non-TC rainfall time series separately by averaging the daily rainfall, TC rainfall, and non-TC rainfall in all grids within the catchment, respectively. To extract daily basin-average extreme precipitation, we selected the 95th percentile of the nonzero daily basin-average rainfall as the threshold for extreme precipitation. Then the daily basin-average rainfall exceeding this threshold was selected as the daily basin-average extreme precipitation. Subsequently, the daily basin-average TC extreme precipitation and non-TC extreme precipitation on the days with basin-average extreme precipitation were extracted from the basin-average TC rainfall and non-TC rainfall time series, respectively. According to the daily basin-average extreme precipitation, TC extreme precipitation and non-TC extreme precipitation, the ratio of the daily basin-average TC extreme precipitation to the daily basin-average extreme precipitation was calculated. Based on this ratio, we used k-means clustering to categorize the daily basin-average extreme precipitation into three groups. The daily basin-average extreme precipitation values in the groups with the largest and smallest mean ratios were classified as TC-induced and non-TC-induced extreme precipitation, respectively. The remaining extreme precipitation was considered mixed rainfall and not analyzed in the trend analysis.
c. Indicators describing the variability of TC and non-TC flooding
We analyzed the variability of TC and non-TC flooding based on two aspects: differences in event characteristics between TC and non-TC rainfall–flood events and long-term changes in the magnitude and frequency of TC-induced and non-TC-induced flooding and extreme precipitation.
To describe the differences in the characteristics between TC and non-TC rainfall–flood events, nine indicators of an rainfall–flood event were calculated based on the identified TC and non-TC rainfall–flood events according to the study by Tarasova et al. (2018) (Table 3). These indicators were event peak time, event runoff coefficient, event time scale, event rise time, normalized event peak, rainfall volume and maximum rainfall intensity, 10-day antecedent rainfall, and base flow at the beginning of the event:
The event peak time is the occurrence date of the flood peak in a rainfall–flood event and it represents the seasonality of flooding. To visualize the seasonality of TC and non-TC flooding, we calculated the percentage of floods per month and plotted the seasonality curves.
The event runoff coefficient is the ratio between the quick runoff volume (mm) and rainfall volume (mm) in a rainfall–flood event and it characterizes the portion of rainfall released from water storage as runoff.
The event time scale is the ratio between the quick runoff volume (mm) and the magnitude of flood peak (mm day−1) and it characterizes the shape of the hydrograph and the duration of an event, and is an indicator of catchment flashiness and the importance of fast runoff generation processes. A short event time scale shows that the fast runoff generation process prevails.
The event rise time characterizes the duration (day) from the beginning of the rainfall event to the day of the flood peak, and it is normalized by the total duration (day) of the event and shows how fast the peak is reached. A short event rise time represents a high contribution of fast runoff to the runoff generation process.
The normalized event peak is the maximum total discharge (mm day−1) of the identified runoff event normalized by the long-term average flow (mm day−1) derived from the observed discharge time series and it represents the magnitude of the runoff event.
The rainfall volume is the sum of the rainfall volume (mm) in the period prior to the date of the flood peak in a rainfall–flood event.
The maximum rainfall intensity is the maximum rainfall intensity (mm day−1) in the period prior to the date of the flood peak in a rainfall–flood event. Both the rainfall volume and maximum rainfall intensity describe the rainfall characteristics in a rainfall–flood event.
The 10-day antecedent rainfall is the sum of the rainfall volume (mm) over the 10 days prior to a rainfall–flood event.
The base flow at the beginning of the event is the base flow (mm day−1) on the day preceding the runoff event. Both the 10-day antecedent rainfall and the base flow at the beginning of the event describe the antecedent wetness states.
The abovementioned metrics were calculated for both TC and non-TC rainfall–flood events. Then the Wilcoxon test was used to examine the statistical significance of the mean differences between the non-TC and TC metrics except for the event peak time at a significance level of 0.05.
Rainfall–runoff event characteristics studied for all catchments.
To detect long-term changes in the magnitude and frequency of TC-induced and non-TC-induced flooding and extreme precipitation, four extreme precipitation and flood indicators were calculated based on the identified TC and non-TC rainfall–flood events and the identified basin-average TC and non-TC extreme precipitation (Table 4). The Mann–Kendall (MK) test (Mann 1945; Kendall 1975) was used to detect long-term trends in the magnitudes of flood peaks and extreme precipitation. Because the time series of the frequency of flood and extreme precipitation were discontinuous count values, we used a Poisson regression to evaluate the trends in the frequency, following the studies of Villarini and Smith (2013). The detected trends in four extreme precipitation and flood indicators were reported at the 5% significance level.
Extreme precipitation and flood indicators studied for all catchments.
4. Results
a. Percentages of TC and non-TC extreme precipitation and flooding
Figure 5 shows the ratio threshold for TC and non-TC extreme precipitation and flooding. In all catchments, the ratio threshold for TC flooding was larger than 0.56 and that for non-TC flooding was smaller than 0.3, and the ratio threshold for TC extreme precipitation was larger than 0.76 and that for non-TC extreme precipitation was smaller than 0.3. Figure 6 shows the percentages of identified TC, non-TC and mixed extreme precipitation to total extreme precipitation. The percentages of TC extreme precipitation ranged from 12% to 50%, while that of non-TC extreme precipitation ranged from 49% to 88%. The lowest percentage of TC extreme precipitation occurred in the coastal area of the Chinese mainland, while the highest value occurred in Hainan Island. The average percentages of TC and non-TC to extreme precipitation were 24% and 76%, respectively, and TC extreme precipitation accounted for a smaller proportion of the total extreme precipitation than that of non-TC extreme precipitation in all catchments.
Ratio threshold for non-TC and TC flood/extreme precipitation.
Citation: Journal of Hydrometeorology 21, 7; 10.1175/JHM-D-20-0002.1
Percentage of identified (a) TC, (b) non-TC, and (c) mixed extreme precipitation and POT flood events.
Citation: Journal of Hydrometeorology 21, 7; 10.1175/JHM-D-20-0002.1
With respect to flooding, the percentage of TC floods ranged from 9% to 47%, while that of non-TC floods ranged from 18% to 83% (Fig. 6). On average, the percentages of TC and non-TC floods were 28% and 53%, respectively. TC floods occupied larger proportions in Hainan Island, with an average value of 45% of the total floods, while non-TC floods dominated in the southeastern coast of the Chinese mainland, with an average value of 25% of the total floods.
The percentages of TC and non-TC floods were closely related to the percentages of TC and non-TC extreme precipitation among the catchments (Spearman’s rank correlation coefficient between the contribution of TC rainfall and flood events was equal to 0.91, while that of non-TC rainfall and floods was equal to 0.9 at the 5% significance level). The mean percentage of TC floods was significantly larger than that of TC extreme precipitation, whereas that of non-TC floods was significantly smaller than that of non-TC extreme precipitation. In addition, the percentage of mixed events was nearly zero for extreme precipitation but was much higher for floods than extreme precipitation. This finding suggests that daily extreme precipitation in the whole catchment belonged to a single type of extreme precipitation, but many flood events that lasted for several days belonged to the compound effects of both rainfall types.
b. Hydrographic characteristics of TC and non-TC rainfall–flood events
The catchments on Hainan Island had a different pattern from those in the coastal area of the Chinese mainland. The percentages of TC and non-TC flood events per month in the coastal area of the Chinese mainland and Hainan Island are shown in Fig. 7. The highest percentages of TC floods appeared in August in the coastal area of the Chinese mainland. On Hainan Island, however, the highest values appeared in October. The percentages of TC floods were larger than 10% from June to October in both the coastal area of the Chinese mainland and Hainan Island. The highest percentages of non-TC floods occurred in July in the coastal area of the Chinese mainland and in October on Hainan Island. Non-TC floods accounted for more than 10% of the events from May–August in the coastal area of the Chinese mainland, and from August–November on Hainan Island.
Percentage of numbers of TC and non-TC floods per month in the coastal area of the Chinese mainland and Hainan Island.
Citation: Journal of Hydrometeorology 21, 7; 10.1175/JHM-D-20-0002.1
Figure 8 shows comparisons of the remaining eight indicators between TC and non-TC rainfall–flood events over the southeastern coastal China. Our results showed obvious differences between TC and non-TC rainfall–flood events. The runoff in TC rainfall–flood events exhibited a significantly smaller event time scale [12 (63%) of the stations; 12 means the number of stations showing significantly smaller values and 63% represents the percentage of these stations], larger normalized peak discharge [10 (52%) of the stations], and shorter event rise time [8 (42%) of the stations] than that in non-TC rainfall–flood events. Furthermore, the rainfall in TC rainfall–flood events was characterized by a larger maximum rainfall intensity [15 (79%) of the stations] but a smaller rainfall event volume [8 (42%) of the stations] than that in non-TC rainfall–flood events. However, the event runoff coefficient and antecedent wetness condition indicators, including the 10-day antecedent rainfall and base flow at the beginning of the event, exhibited no obvious differences between TC and non-TC rainfall–flood events.
Boxplots with whiskers from minimum to maximum showing the non-TC and TC rainfall–flood event characteristics for all catchments. The boxes show the interquartile ranges of the values, and the short bold lines show the medians of the values. The asterisks indicate distributions with a statistically different median (Wilcoxon test, p value < 0.05). Outliers are not presented.
Citation: Journal of Hydrometeorology 21, 7; 10.1175/JHM-D-20-0002.1
c. Long-term trends in magnitude and frequency of TC and non-TC extreme precipitation and floods
The trends in the TC, non-TC, total extreme precipitation amounts and frequencies are given in Table 5. With respect to the trends of extreme precipitation amount, 4 stations (21%) exhibited significant decreasing trends in TC extreme precipitation amount, and no station showed a significant increasing trend. In addition, 10 (53%) and 7 (37%) of the stations reflected statistically significant increasing trends in the non-TC extreme precipitation amount and total precipitation amount, respectively, and no station showed a significant decreasing trend. The TC extreme precipitation amount significantly decreased by an average of −0.15 mm yr−1. Non-TC extreme precipitation showed significant positive trends, with an amount of 1.98 mm on average and 1.79 mm in total per year.
Trends in the magnitude and frequency of TC, non-TC and total extreme precipitation. The numbers in bold indicate that the trend of values is significant at the 95% confidence level.
Furthermore, 2 (11%) of the stations exhibited a significant decrease in TC extreme precipitation frequency, and no station presented a significant increase. Moreover, 10 (53%) and 5 (26%) of the stations presented significant increases in the non-TC extreme precipitation frequency and total precipitation frequency, respectively, and no station showed a significant decreasing trend. The TC extreme precipitation frequency decreased significantly, with an average significant trend of approximately −0.003 days yr−1. The non-TC and total extreme precipitation frequencies increased significantly, with average significant trends of 0.005 and 0.003 days yr−1, respectively.
The impact of missing years on the trend analysis of flood magnitude and frequency is shown in Table S1 in the online supplemental material. Table S1 shows that the trend values of flood magnitude in 15 (79%) of the stations and flood frequency in 13 (68%) of the stations were reliable. Table 6 shows the trends in the magnitude and frequency of the TC, non-TC, and total floods. The unreliable trend values due to the missing years are marked in Table 6. The results show that 3 (16%) of the stations exhibited reliably significant decreasing trends in the magnitude of TC floods, only 2 (11%) of the stations showed reliably significant decreasing trends in the magnitude of non-TC floods, and 1 (5%) of the stations showed reliably significant decreasing trends in the magnitude of total floods. No stations showed a significant increasing trend in the magnitude of the TC floods.
Trends in the magnitude and frequency of TC, non-TC and total floods. The numbers in bold indicate that the trend of values is significant at the 95% confidence level. Asterisks indicate that the trend value may be not reliable due to the missing years according to the criteria (Slater and Villarini 2017).
Additionally, none of the stations showed significant trends in the frequency of the TC floods, 2 (11%) stations reflected reliably significant decreasing trends in the frequency of the non-TC floods, while 4 (21%) of the stations reflected reliably significant decreasing trends in the frequency of the total floods.
5. Discussion
This study found that the percentage of TC floods was significantly larger than that of TC to extreme precipitation in southeastern coastal China. This finding is different from the pattern observed in the southeastern United States, where the percentage of TC floods was lower than that of extreme precipitation (Aryal et al. 2018). In southeastern coastal China, non-TC extreme precipitation frequently occurred from April to July prior to TC precipitation season (Su et al. 2015), and it may have increased the soil water storage. Therefore, in addition to flooding events caused by TC extreme precipitation from July to September, additional TC flooding also occurred because of moderate TC precipitation on partially saturated areas due to the impact of high antecedent soil water (Su et al. 2015). In contrast, in the eastern United States, extreme precipitation mainly occurs in the summer season (from June to September) in association with frequent TCs (Nogueira and Keim 2011); however, most of the flooding (as much as 80%) occurs in the cold season (from October to March) and is associated with winter–spring extratropical cyclone systems (Villarini and Smith 2010; Villarini 2016) rather than TCs.
Our study also revealed the consistency of the decrease in TC floods (flood magnitude) and TC extreme precipitation (precipitation amount and frequency) in southeastern coastal China, which is in inconsistent with the pattern in the southeastern United States, where increasing trends in TC extreme precipitation amount and frequency occurred but obvious changes in TC flood magnitude and frequency did not occur (Knight and Davis 2009; Kunkel et al. 2010; Aryal et al. 2018). The pattern difference may be due to that the relatively wet soil moisture condition during the TC flood season in southeastern coastal China (Suon et al. 2019) but relatively dry conditions in the southeastern United States (Gao et al. 2006). Catchments with a wet soil moisture condition have a limited capacity to regulate the runoff response to rainfall. The decreases in extreme precipitation directly lead to decreases in flooding in southeastern coastal China. Conversely, increases in extreme precipitation may not necessarily result in increasing flooding in the southeastern United States during the TC flood season (Bennett et al. 2018).
Moreover, our study showed inconsistency in the trend directions between the flooding (decreasing flood magnitude and frequency) and extreme precipitation (increasing precipitation amount and frequency) events for the non-TC types in southeastern coastal China. This inconsistency may be attributed to regulation by the high-density reservoirs in southeastern China. Before the non-TC rainy season, the large storage capacity per area enables reservoirs to reduce the flood magnitude when floods occur (Yang and Lu 2014; Gao et al. 2012). Unfortunately, previous studies have not revealed the trends in the magnitude and frequency of non-TC flood events in the southeastern United States, although an increasing trend in the non-TC extreme precipitation (precipitation amount and frequency) has been revealed (Bishop et al. 2019).
In this study, we found significant different characteristics between the TC and non-TC flood events. In the future, the linkage between event characteristics and climate or landscape features can be further studied, thus to provide insights into the different physical drivers of TC and non-TC flood response (Blöschl et al. 2013). Such work can aid in the interpretations of the heterogeneity of flood generation processes and facilitate a better understanding of how they can be accurately captured by hydrological models (Fenicia et al. 2008; Rogger et al. 2013).
This study separated TC and non-TC floods and analyzed the trends in TC and non-TC floods. Since floods associated with different generation mechanisms may have different magnitudes and thus different frequency distributions, future studies can focus on deriving compound distributions that combine TC and non-TC floods, which can reduce uncertainty when designing floods for flood frequency analyses (Fischer et al. 2016; Yan et al.2019). However, after dividing floods into various types, the length of the flood time series for each type is limited, which leads to uncertainty in the flood frequency analysis. Also, future studies can attribute trends in floods, which has been a topic of high interest (Merz et al. 2012; Slater et al. 2015). The attribution of flood trends requires assumptions about possible drivers. Separation of TC and non-TC floods can provide a better physical basis on which to select the possible drivers that are responsible for floods.
6. Conclusions
This study evaluated the different hydrographic characteristics between non-TC and TC rainfall–flood events, and analyzed the long-term trends in the frequency and magnitude of non-TC and TC flood events using an identification method for TC and non-TC floods that links flooding to rainfall. The major findings of this study are summarized as follows.
In southeastern coastal China, the average percentage of TC floods (28%) was significantly larger than that of TC extreme precipitation (24%), whereas that of non-TC floods (53%) was significantly smaller than that of non-TC extreme precipitation (76%).
The months for which the percentages of TC floods were larger than 10% were June–October in both the coastal area of the Chinese mainland and Hainan Island. The months for which the percentages of non-TC floods were larger than 10% were May–August in the coastal area of the Chinese mainland and August–November on Hainan Island. Compared to non-TC rainfall–flood events, TC rainfall–flood events had larger normalized peak discharges and maximum rainfall intensities but smaller event rise times, event time scales and rainfall event volumes. In addition, the event runoff coefficient and antecedent wetness condition indicators showed no obvious differences between the TC and non-TC rainfall–flood events. These indicators suggested that the TC rainfall–flood events were in a fast runoff generation process associated with large flood peaks, high rainfall intensity and small rainfall volume.
A significant decreasing trend in the magnitude of TC floods was observed for 3 (16%) of the stations, a reliably significant decreasing trend in the magnitude of non-TC floods was observed for 2 (11%) of the stations, and a reliably significant decreasing trend in the magnitude of total floods was observed for 1 (5%) of the stations. No station showed a significant increasing trend in the magnitude of the TC floods.
Additionally, none of the stations showed significant trends in the frequency of TC floods. Two (11%) of the stations reflected reliably significant decreasing trends in the frequency of non-TC floods, and 4 (21%) of the stations reflected reliably significant decreasing trends in the frequency of total floods. The magnitude of TC floods exhibited a decreasing trend similar to the trend for the amount and frequency of TC extreme precipitation. However, the magnitude and frequency of non-TC floods showed a decreasing trend while that of non-TC extreme precipitation showed an increasing trend.
Acknowledgments
This work is financially supported by the National Natural Science Foundation of China (Projects 41661144031, 51922063). The authors thank Mr. Weigang Tang from McMaster University for providing the MATLAB toolbox HydRun and Mr. Fuming Ren from the China Meteorological Administration for providing the code of the OSAT.
REFERENCES
Aryal, Y. N., G. Villarini, W. Zhang, and G. A. Vecchi, 2018: Long term changes in flooding and heavy rainfall associated with North Atlantic tropical cyclones: Roles of the North Atlantic Oscillation and El Niño-Southern Oscillation. J. Hydrol., 559, 698–710, https://doi.org/10.1016/j.jhydrol.2018.02.072.
Barlow, M., 2011: Influence of hurricane-related activity on North American extreme precipitation. Geophys. Res. Lett., 38, L04705, https://doi.org/10.1029/2010GL046258.
Barth, N. A., G. Villarini, and K. White, 2018: Contribution of eastern North Pacific tropical cyclones and their remnants on flooding in the western United States. Int. J. Climatol., 38, 5441–5446, https://doi.org/10.1002/joc.5735.
Batalla, R. J., C. M. Gomez, and G. M. Kondolf, 2004: Reservoir-induced hydrological changes in the Ebro River basin (NE Spain). J. Hydrol., 290, 117–136, https://doi.org/10.1016/j.jhydrol.2003.12.002.
Bennett, B., M. Leonard, Y. Deng, and S. Westra, 2018: An empirical investigate-on into the effect of antecedent precipitation on flood volume. J. Hydrol., 567, 435–445, https://doi.org/10.1016/j.jhydrol.2018.10.025.
Bishop, D. A., A. P. Williams, and R. Seager, 2019: Increased fall precipitation in the southeastern United States driven by higher-intensity, frontal precipitation. Geophys. Res. Lett., 46, 8300–8309, https://doi.org/10.1029/2019GL083177.
Blöschl, G., M. Sivapalan, T. Wagener, A. Viglione, and H. Savenije, 2013: Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales. Cambridge University Press, 465 pp., https://doi.org/10.1017/CBO9781139235761.
Chang, C. P., Y. H. Lei, C. H. Sui, X. H. Lin, and F. M. Ren, 2012: Tropical cyclone and extreme rainfall trends in East Asian summer monsoon since mid-20th century. Geophys. Res. Lett., 39, L18702, https://doi.org/10.1029/2012GL052945.
Chen, L., Y. Li, and Z. Cheng, 2010: An overview of research and forecasting on rainfall associated with landfalling tropical cyclones. Adv. Atmos. Sci., 27, 967–976, https://doi.org/10.1007/s00376-010-8171-y.
Chou, C., L. F. Huang, L. S. Tseng, J. Y. Tu, and P. H. Tan, 2009: Annual cycle of rainfall in the western North Pacific and East Asian sector. J. Climate, 22, 2073–2094, https://doi.org/10.1175/2008JCLI2538.1.
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.
Day, J. A., I. Fung, and W. Liu, 2018: Changing character of rainfall in eastern China, 1951-2007. Proc. Natl. Acad. Sci. USA, 115, 2016–2021, https://doi.org/10.1073/pnas.1715386115.
Fenicia, F., J. J. McDonnell, and H. H. G. Savenije, 2008: Learning from model improvement: On the contribution of complementary data to process understanding. Water Resour. Res., 44, W06419, https://doi.org/10.1029/2006WR005563.
Fischer, S., A. Schumann, and M. Schulte, 2016: Characterization of seasonal flood types according to timescales in mixed probability distributions. J. Hydrol., 539, 38–56, https://doi.org/10.1016/j.jhydrol.2016.05.005.
Galarneau, T. J., L. F. Bosart, and R. S. Schumacher, 2010: Predecessor rain events ahead of tropical cyclones. Mon. Wea. Rev., 138, 3272–3297, https://doi.org/10.1175/2010MWR3243.1.
Gao, B., D. Yang, T. Zhao, and H. Yang, 2012: Changes in the eco-flow metrics of the Upper Yangtze River from 1961 to 2008. J. Hydrol., 448–449, 30–38, https://doi.org/10.1016/j.jhydrol.2012.03.045.
Gao, H., E. F. Wood, T. J. Jackson, M. Drusch, and R. Bindlish, 2006: Using TRMM/TMI to retrieve surface soil moisture over the southern United States from 1998 to 2002. J. Hydrometeor., 7, 23–38, https://doi.org/10.1175/JHM473.1.
Gu, X., Q. Zhang, and V. P. Singh, 2017: Spatiotemporal patterns of annual and seasonal precipitation extreme distributions across China and potential impact of tropical cyclones. Int. J. Climatol., 37, 3949–3962, https://doi.org/10.1002/joc.4969.
Guo, J., G. H. Huang, and X. Wang, 2018: Future changes in precipitation extremes over China projected by a regional climate model ensemble. Atmos. Environ., 188, 142–156, https://doi.org/10.1016/j.atmosenv.2018.06.026.
Hundecha, Y., and B. Merz, 2012: Exploring the relationship between changes in climate and floods using a model-based analysis. Water Resour. Res., 48, W04512, https://doi.org/10.1029/2011WR010527.
Ivancic, T. J., and S. B. Shaw, 2015: Examining why trends in very heavy precipitation should not be mistaken for trends in very high river discharge. Climatic Change, 133, 681–693, https://doi.org/10.1007/s10584-015-1476-1.
Jarvis, A., H. Reuter, A. Nelson, and E. Guevara, 2008: Hole-filled SRTM for the globe version 4. CGIAR-CSI SRTM 90 m Database, CGIAR-CSI, accessed 18 March 2014, https://cgiarcsi.community/data/srtm-90m-digital-elevation-database-v4-1/.
Kendall, M. G., 1975. Rank Correlation Methods. 4th ed. Charles Griffin, 202 pp.
Kim, J. H., C. H. Ho, M. H. Lee, J. H. Jeong, and D. L. Chen, 2006: Large increase in heavy rainfall associated with tropical cyclone landfalls in Korea after the late 1970s. Geophys. Res. Lett., 33, L18706, https://doi.org/10.1029/2006GL027430.
Kimball, S. K., and M. S. Mulekar, 2004: A 15-year climatology of North Atlantic tropical cyclones. Part I: Size parameters. J. Climate, 17, 3555–3575, https://doi.org/10.1175/1520-0442(2004)017<3555:AYCONA>2.0.CO;2.
Knight, D. B., and R. E. Davis, 2009: Contribution of tropical cyclones to extreme rainfall events in the southeastern United States. J. Geophys. Res., 114, D23102, https://doi.org/10.1029/2009JD012511.
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 associa-ted with tropical cyclones. Geophys. Res. Lett., 37, L24706, https://doi.org/10.1029/2010GL045164.
Lavender, S. L., and D. J. Abbs, 2013: Trends in Australian rainfall: Contribution of tropical cyclones and closed lows. Climate Dyn., 40, 317–326, https://doi.org/10.1007/s00382-012-1566-y.
Lee, M. H., C. H. Ho, and J. H. Kim, 2010: Influence of tropical cyclone landfalls on spatiotemporal variations in typhoon season rainfall over South China. Adv. Atmos. Sci., 27, 443–454, https://doi.org/10.1007/s00376-009-9106-3.
Mann, H. B., 1945: Nonparametric tests again trend. Econometrica, 13, 245–259, https://doi.org/10.2307/1907187.
Merz, B., S. Vorogushyn, S. Uhlemann, J. Delgado, and Y. Hundecha, 2012: More efforts and scientific rigour are needed to attribute trends in flood time series. Hydrol. Earth Syst. Sci., 16, 1379–1387, https://doi.org/10.5194/hess-16-1379-2012.
Nogueira, R. C., and B. D. Keim, 2011: Contributions of Atlantic tropical cyclones to monthly and seasonal rainfall in the eastern United States 1960–2007. Theor. Appl. Climatol., 103, 213–227, https://doi.org/10.1007/s00704-010-0292-9.
Pielke, R. A., Jr., and M. W. Downton, 2000: Precipitation and damaging floods: Trends in the United States, 1932–97. J. Climate, 13, 3625–3637, https://doi.org/10.1175/1520-0442(2000)013<3625:PADFTI>2.0.CO;2.
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. M. Wang, and X. L. Wang, 2007: Estimating tropical cyclone precipitation from station observations. Adv. Atmos. Sci., 24, 700–711, https://doi.org/10.1007/s00376-007-0700-y.
Rogger, M., A. Viglione, J. Derx, and G. Blöschl, 2013: Quantifying effects of catchments storage thresholds on step changes in the flood frequency curve. Water Resour. Res., 49, 6946–6958, https://doi.org/10.1002/wrcr.20553.
Rowe, S. T., and G. Villarini, 2013: Flooding associated with predecessor rain events over the Midwest United States. Environ. Res. Lett., 8, 024007, https://doi.org/10.1088/1748-9326/8/2/024007.
Sharma, A., C. Wasko, and D. P. Lettenmaier, 2018: If precipitation extremes are increasing, why aren’t floods? Water Resour. Res., 54, 8545–8551, https://doi.org/10.1029/2018WR023749.
Shen, Y., P. Zhao, Y. Pan, and J. Yu, 2014: A high spatiotemporal gauge-satellite merged precipitation analysis over China. J. Geophys. Res. Atmos., 119, 3063–3075, https://doi.org/10.1002/2013JD020686.
Slater, L., and G. Villarini, 2017: On the impact of gaps on trend detection in extreme streamflow time series. Int. J. Climatol., 37, 3976–3983, https://doi.org/10.1002/joc.4954.
Slater, L., M. B. Singer, and J. W. Kirchner, 2015: Hydrologic versus geomorphic drivers of trends in flood hazard. Geophys. Res. Lett., 42, 370–376, https://doi.org/10.1002/2014GL062482.
Small, D., S. Islam, and R. M. Vogel, 2006: Trends in precipitation and streamflow in the eastern U.S.: Paradox or perception. Geophys. Res. Lett., 33, L03403, https://doi.org/10.1029/2005GL024995.
Smith, R. L., 1984: Threshold methods for sample extremes. Statistical Extremes and Applications. J. T. de Oliveira, Ed., Springer, 621–638.
Su, Z., F. Ren, J. Wei, X. Lin, S. Shi, and X. Zhou, 2015: Changes in monsoon and tropical cyclone extreme precipitation in Southeast China from 1960 to 2012. Trop. Cyclone Res. Rev., 4, 12–17, https://doi.org/10.6057/2015TCRR01.02.
Suon, S., Y. Li, L. Porn, and T. Javed, 2019: Spatiotemporal analysis of soil moisture drought over China during 2008-2016. J. Water Resource Prot., 11, 700–712, https://doi.org/10.4236/jwarp.2019.116041.
Tang, W. G., and S. K. Carey, 2017: HydRun: A MATLAB toolbox for rainfall-runoff analysis. Hydrol. Processes, 31, 2670–2682, https://doi.org/10.1002/hyp.11185.
Tarasova, L., S. Basso, M. Zink, and R. Merz, 2018: Exploring controls on rainfall-runoff events: 1. Time series- based event separation and temporal dynamics of event runoff response in Germany. Water Resour. Res., 54, 7711–7732, https://doi.org/10.1029/2018WR022587.
U.S. Water Resources Council, 1981: Guidelines for determining flood flow frequency. Bulletin 17B, 194 pp.
Villarini, G., 2016: On the seasonality of flooding across the continental United States. Adv. Water Resour., 87, 80–91, https://doi.org/10.1016/j.advwatres.2015.11.009.
Villarini, G., and J. A. Smith, 2010: Flood peak distributions for the eastern United States. Water Resour. Res., 46, W06504, https://doi.org/10.1029/2009WR008395.
Villarini, G., and J. A. Smith, 2013: Flooding in Texas: Examination of temporal changes and impacts of tropical cyclones. J. Amer. Water Resour. Assoc., 49, 825–837, https://doi.org/10.1111/jawr.12042.
Villarini, G., and R. F. Denniston, 2016: Contribution of tropical cyclones to extreme rainfall in Australia. Int. J. Climatol., 36, 1019–1025, https://doi.org/10.1002/joc.4393.
Villarini, G., R. Goska, J. A. Smith, and G. A. Vecchi, 2014a: North Atlantic tropical cyclones and U.S. flooding. Bull. Amer. Meteor. Soc., 95, 1381–1388, https://doi.org/10.1175/BAMS-D-13-00060.1.
Villarini, G., D. A. Lavers, E. Scoccimarro, M. Zhao, M. F. Wehner, G. A. Vecchi, T. R. Knutson, and K. A. Reed, 2014b: Sensitivity of tropical cyclone rainfall to idealized global-scale forcings. J. Climate, 27, 4622–4641, https://doi.org/10.1175/JCLI-D-13-00780.1.
Visual Numerics, 1997: IMSL Computational Technology Toolkit. Vol. II, Visual Numerics, 823 pp.
Vormoor, K., D. Lawrence, L. Schlichting, D. Wilson, and W. K. Wong, 2016: Evidence for changes in the magnitude and frequency of observed rainfall vs. snowmelt driven floods in Norway. J. Hydrol., 538, 33–48, https://doi.org/10.1016/j.jhydrol.2016.03.066.
Wang, X., L. Wu, F. Ren, Y. Wang, and W. Li, 2008: Influences of tropical cyclones on China during 1965-2004. Adv. Atmos. Sci., 25, 417–426, https://doi.org/10.1007/s00376-008-0417-6.
Wang, X. L., S. C. Liu, R. Liu, C.-J. Shiu, C. He, and X. Zhong, 2019: Observed changes in precipitation extremes and effects of tropical cyclones in South China during 1955-2013. Int. J. Climatol., 39, 2677–2684, https://doi.org/10.1002/joc.5980.
Waylen, P. R., 1991: Modeling the effects of tropical cyclones on flooding in the Santa Fe river basin, Florida. GeoJournal, 23, 361–373, https://doi.org/10.1007/BF00193609.
Wu, Y. J., S. G. Wu, and P. M. Zhai, 2007: The impact of tropical cyclones on Hainan Island’s extreme and total precipitation. Int. J. Climatol., 27, 1059–1064, https://doi.org/10.1002/joc.1464.
Xing, R., Z. Ding, S. You, and H. Xu, 2016: Relationship of tropical-cyclone-induced remote precipitation with tropical cyclones and the subtropical high. Front Earth Sci., 10, 595–606, https://doi.org/10.1007/s11707-015-0530-7.
Yan, L., L. Xiong, G. Ruan, C. Xu, P. Yan, and P. Liu, 2019: Reducing uncertainty of design floods of two-component mixture distributions by utilizing flood timescale to classify flood types in seasonally snow covered region. J. Hydrol., 574, 588–608, https://doi.org/10.1016/j.jhydrol.2019.04.056.
Yang, X., and X. Lu, 2014: Drastic change in China’s lakes and reservoirs over the past decades. Sci. Rep., 4, 6041, https://doi.org/10.1038/srep06041.
Yang, X., L. J. Zhou, C. F. Zhao, and J. Yang, 2018: Impact of aerosols on tropical cyclone-induced precipitation over the mainland of China. Climatic Change, 148, 173–185, https://doi.org/10.1007/s10584-018-2175-5.
Ying, M., B. D. Chen, and G. X. 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.
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.
Zhu, L. Y., S. M. Quiring, I. Guneralp, and W. G. Peacock, 2015: Variations in tropical cyclone-related discharge in four watersheds near Houston, Texas. Climate Risk Manage., 7, 1–10, https://doi.org/10.1016/j.crm.2015.01.002.
