1. Introduction
Floods are one of the most common and destructive natural hazards and have caused more than 317 000 deaths and more than 784 billion U.S. dollars in economic losses across the globe from 1970 to 2018 (Guha-Sapir et al. 2018). Floods (defined as the accumulation of water over areas that are not normally submerged; IPCC 2012a) can be further classified into different types based on the triggers: fluvial floods due to high upstream river discharge, pluvial floods resulting from localized rainstorms and inadequate drainage, and coastal floods due to extreme sea levels (Svensson and Jones 2004; Huntingford et al. 2014; Moftakhari et al. 2017). Traditional flood hazard assessment usually only considers one single type of flood at a time (e.g., pluvial or fluvial floods only) (Milly et al. 2002; Garner et al. 2017). However, different types of floods can occur in certain weather systems, such as the simultaneous occurrence of pluvial and coastal floods in Houston due to heavy rainstorms and storm surges driven by Hurricane Harvey in 2017 (Emanuel 2017; Zhang et al. 2018a). In fact, combinations of any two or more types of floods can be considered as compound floods: 1) pluvial and coastal, 2) pluvial and fluvial, and 3) fluvial and coastal. Such compound floods exacerbate the adverse impacts of single noncompound floods and, hence, pose greater threats to society and ecosystems. Coastal areas are usually the most densely populated and economically developed areas of a country because of their proximity to the sea. Additionally, many coastal areas are often threatened by compound floods from heavy precipitation and extreme storm surges, especially in areas affected by storms. For instance, during Supertyphoon Hato in 2017, the coastal floods caused by extreme storm surges were accompanied by severe waterlogging (i.e., pluvial flooding) in Macau (Hong Kong Observatory 2017; Wang et al. 2019). The co-occurrence of coastal and pluvial floods in Macau brought about much higher stress for discharging inland water into the sea and eventually caused 12 deaths and at least 200 injuries (HKO 2017). Given the substantial damage associated with compound floods from precipitation and storm surges around the world, the risks and mechanisms of compound floods are receiving increasing attention in recent years.
Different types of floods can be physically and statistically interdependent because compound floods are usually associated with certain types of weather systems. Zscheischler et al. (2018) studied a risk framework that links multiple climatic drivers with societal risk and indicated that the interaction of multiple hazards (i.e., drivers) is one of the primary causes of significant societal and economic impacts. In compound events, individual contributing variables may not be extreme themselves; however, their jointly dependent occurrence can cause severe impacts that are greater than those of the individual variables (Bevacqua et al. 2017; Zscheischler et al. 2018). Therefore, flood prevention management must thoroughly consider the statistical interdependence and joint probability distribution of compound floods. Dependence between different flood drivers (such as precipitation, river discharge, and storm surges) has been reported in previous studies, although the focus of these studies has not necessarily been on compound floods (Svensson and Jones 2002, 2004; Zheng et al. 2013; Paprotny et al. 2018; Xu et al. 2018; van den Hurk et al. 2015). Furthermore, constructing joint probability distributions of the dependent flood drivers is key to assessing the joint occurrence of compound floods. The methods for joint distribution have become more feasible and reliable with the recent development of joint statistical theories (Bevacqua et al. 2019; Li et al. 2015). Copula theory is one of the most widely used joint probability distribution models for the construction of multivariable distributions to assess joint statistical behaviors in the field of hydrology (Bevacqua et al. 2017, 2019, 2020a,b; Couasnon et al. 2020; Lian et al. 2013; Wahl et al. 2015; Ward et al. 2018; Xu et al. 2019). Previous regional studies based on joint distribution theories have evaluated the probability of compound floods by considering the dependence of individual variables, such as those in the United States, Australia, and Europe (Bevacqua et al. 2019; Wahl et al. 2015; Zheng et al. 2013). For example, Wahl et al. (2015) found an increasing risk of compound floods from storm surges and precipitation in the U.S.-based observations. Bevacqua et al. (2019) projected that the joint probability of compound floods would increase in Europe under future climate change scenarios based on model simulations. These studies assessed compound flood hazards at a regional scale, which is important in regional flood risk management. On a global scale, more recent studies have explored the application of simulated storm surge data on compound flood analysis, which provided useful information regarding global compound flood risks, especially for areas with scarce observations (Bevacqua et al. 2020a,b; Couasnon et al. 2020; Ikeuchi et al. 2017). However, global-scale studies based on long-term observations are rare because of limited available data.
Compound floods associated with heavy precipitation and storm surges can be triggered by different mechanisms. Tropical cyclones (TCs; i.e., intense circular storms that affect coastal areas in tropical and subtropical areas) are usually accompanied by strong winds, low pressure, and abundant moisture transportation and may simultaneously trigger heavy precipitation and storm surges (Irish et al. 2008; Khouakhi et al. 2017; Lin et al. 2013; Walsh et al. 2016). Additional studies have reported an increase in TC intensity and slowing of TC movement in recent decades (Emanuel 2005; Knutson et al. 2015; Kossin 2018; Lai et al. 2020; Zhang et al. 2020). Such changes in TCs lead to more extreme TC-induced precipitation and potentially result in a higher probability of compound floods associated with TCs (Kunkel et al. 2010; Zhang et al. 2018b). Moreover, extratropical cyclones (ETCs; another type of cyclones that mostly affect mid- and high-latitude areas) can also simultaneously generate damaging storm surges and heavy precipitation. Previous studies have identified the importance of ETCs in precipitation and storm surges individually over mid- and high-latitude areas (Booth et al. 2016; Catto and Pfahl 2013; Hawcroft et al. 2012, 2018). Booth et al. (2016) compared hurricane and extratropical storm surges on the northeast coast of the United States and found that hurricanes are more likely to create larger and heavier storm surge hazards, while ETC-induced storm surges are weaker but continue for a longer time.
Although TCs/ETCs may cause heavy precipitation and storm surges, not all TCs/ETCs trigger these two types of extremes at the same time. More specifically, TCs/ETCs may result in different combinations of precipitation and storm surges: 1) heavy precipitation but no extreme storm surge (i.e., noncompound floods); 2) extreme storm surge and no heavy precipitation (noncompound floods); 3) no heavy precipitation or extreme storm surge (nonextreme events); and 4) heavy precipitation and extreme storm surge (i.e., compound floods). Furthermore, not all compound floods are triggered by TCs/ETCs. For different regions across the globe, geographical locations (e.g., low-, mid-, or high-latitude areas) determine the possibility of occurrence of various weather systems, and geographical settings (e.g., topography and shape of coastlines) affect the statistics (e.g., occurrence and intensity) of precipitation and storm surges triggered by these weather systems (Chang et al. 2013; Hawcroft et al. 2012; Khouakhi et al. 2017). The extent to which different mechanisms (e.g., TCs and ETCs) contribute to compound floods associated with heavy precipitation and storm surges is not well understood. Given the complexity of the relationship between compound floods and TCs/ETCs, quantitative evidence of the degree of TC/ETC impacts on compound floods is important for enhancing the scientific understanding of compound floods and improving the forecasting, warning, and management of compound floods in different regions.
From the above discussions, it is of great scientific significance and practical importance to evaluate the global compound flood hazards and understand the physical mechanisms behind. This knowledge is key to formulating and implementing proactive measures in flood protection management to help prepare cities for future compound floods in a timely manner (Paprotny et al. 2018). Therefore, in this study, we aimed to assess the probability of compound floods from precipitation and storm surges, and quantified the contributions of TCs and ETCs to compound floods across the world. First, the dependence structures between precipitation and storm surge were examined and the return periods of compound floods were estimated. Then, the impacts of TCs and ETCs on compound floods were quantitatively evaluated. Finally, the weather conditions associated with compound floods, extreme precipitation only, and extreme storm surges were analyzed. This study provides a comprehensive global hazard assessment of compound floods and quantitative evidence for the contributions of TCs and ETCs to compound floods, which can serve as an important reference for flood risk management in coastal regions across the globe and a foundation for further studies.
The remainder of this paper is structured as follows. The data and methods used in this paper are described in sections 2 and 3, respectively. Results about the return periods of compound floods from heavy rainfall and storm surge, and the impact of cyclones are provided in section 4. In section 5, we discussed the mechanisms and weather patterns associated with compound floods. The main conclusions are summarized in section 6.
2. Data
a. Storm surge and precipitation data
To estimate pluvial floods caused by precipitation extremes, daily precipitation data from more than 4900 stations with the longest time series of 126 years were collected from the Global Historical Climatology Network (GHCN–Daily; https://www.ncdc.noaa.gov/ghcn-daily-description; Menne et al. 2012). Hourly sea level data were derived from the Global Extreme Sea Level Analysis version 2 (GESLA-2; https://www.gesla.org/; Woodworth et al. 2016). Following Wahl et al. (2015) and Ward et al. (2018), for each tide gauge, we extracted the daily maximum storm surge and the mean precipitation of all stations within 25 km of the tide gauge. When there was no precipitation station within 25 km of the tide gauge, the search radius was expanded to 50 km. Only tide gauges with at least 75% of overlapping data completeness per year and at least 18 years of records during 1979–2014 were selected. The minimum record threshold of 18 years was selected with consideration of the number of available tide gauges and the quality of the analysis (Fig. S1 in the online supplemental information). This time period was used to estimate the percentiles of precipitation and storm surges to define extreme events. Eventually, 314 tide gauges were identified, which are mostly located along the coasts of North America, Europe, Australia, East Asia, and Southeast Asia (Fig. 1). To extract storm surges from sea level observations, the sea level was divided into three major components namely mean sea level, astronomical tide, and residual tide (which includes storm surge and wind waves); this was based on the year-by-year tidal harmonic analysis using the T-tide Toolbox in Matlab (Pawlowicz et al. 2002). Although high tides and earthquakes are also potential triggers of coastal floods, we only focused on those associated with storm surges related to weather extremes (i.e., TCs and ETCs) in this study (Vousdoukas et al. 2018).
Locations of tide gauges and the length of overlapping years. (a) Locations of tide gauges and the length of overlapping years, (b) number of tide gauges with a particular length of overlapping years, and (c) number of tide gauges with data available in a particular year. In (a), the colors denote the length of overlapping years. Only tide gauges with ≥18 years of data are shown.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
b. Cyclone track data and meteorological data
The TC track data (i.e., longitude and latitude of TC centers) between 1950 and 2014 was collected from the International Best Track Archive for Climate Stewardship (IBTrACS; https://www.ncdc.noaa.gov/ibtracs/; Knapp et al. 2010). This dataset, combining the observed track estimates from multiple sources, has been widely used in studies related with TCs (e.g., Lai et al. 2020; Wahl et al. 2015). However, there is no accepted single ETC track dataset owing to the complexity of ETCs (i.e., various shapes and structures, changeable motion, and wide size ranges; Neu et al. 2013). Previous studies on ETCs usually identified ETCs using objective feature tracking techniques based on meteorological variables such as mean sea level pressure (Wang et al. 2006) and relative vorticity at 850 hPa (Hoskins and Hodges 2002; Wang et al. 2006). By using different detection and tracking methods, the cyclone (i.e., including TC and ETC) tracks showed marked differences in characteristics such as track density, track agreement with other datasets, wind speed, and intensity (Neu et al. 2013). In this study, the ETC track data between 1979 and 2012 produced by the “M09” method [see Neu et al. (2013) for details] proposed in the Intercomparison of Mid-Latitude Storm Diagnostics project (IMILAST; www.proclim.ch/imilast/index.html; Neu et al. 2013) was used because this method has the most balanced performance in various aspects mentioned above. Cyclones including ETCs and TCs were identified using the automatic tracking methods mentioned above. Identical tracks in both the ETC track data and IBTrACS data were removed from the ETC catalog. When a TC/ETC center was located within 500 km of a tide gauge, the storm surges of that day and ±1 day were considered as TC/ETC-induced storm surges. The same procedure was applied to the definitions of TC/ETC-induced precipitation (Booth et al. 2016; Hawcroft et al. 2012; Khouakhi et al. 2017; Lai et al. 2020; Zhang et al. 2018b). To avoid the duplication of TC/ETC-associated events, if one event was found to be related to both TCs and ETCs, this event is identified as a TC-associated event and removed from the ETC-associated event catalog.
Sea level pressure, wind field, and precipitable water content data from 1948 to 2014 were collected from NCEP–NCAR reanalysis dataset (https://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.html; Kalnay et al. 1996) to analyze the weather patterns associated with compound floods.
3. Methodology
a. Compound flood return periods
Compound flood events were sampled based on the peak over threshold method (Gu et al. 2017). A compound flood event was detected if both daily precipitation and daily maximum storm surges exceeded their individual high percentiles. Compound flood hazards were characterized by the joint return periods of extreme precipitation and extreme storm surges (i.e., the average waiting times between compound flood events). The return period of compound floods is inversely proportional to the probability of extreme precipitation and extreme storm surges: a longer return period corresponds to a lower probability of compound floods, and vice versa (Bevacqua et al. 2019). In this study, the 98.5th percentile was chosen as the threshold to define compound floods. With setting 18 years as the minimum threshold of record length, the threshold used to identify compound floods should not be too high to ensure at least one compound flood can be sampled in most tide gauges. At the same time, the threshold should be as high as possible to ensure the compound floods we analyzed were rare enough to be considered as “extremes.” With consideration of these factors, the 98.5th percentile was chosen as the threshold to define the univariate extremes and compound floods. In this case, if extreme precipitation and extreme storm surges are independent univariate events (i.e., the occurrence of one does not affect the other), the return period of compound floods with precipitation and storm surges exceeding the 98.5th percentiles is ~12 years, which is shorter than the minimum record length (i.e., 18 years) so that at least one compound flood event can be sampled in most of the tide gauges. In addition, to test the sensitivity of our results to the choice of threshold and also provide more information about more extreme compound floods, the joint return periods of compound floods defined by the 99.5th percentile (i.e., expected return period of ~109 years) were analyzed and are shown in the online supplemental material. Under climate change, the statistics of hydrometeorological variables are nonstationary (i.e., the statistics change over time; Gu et al. 2017; She et al. 2015). Therefore, the percentiles of precipitation/storm surges at all stations were estimated based on the daily time series from 1979 to 2014 to mitigate the impacts of nonstationarity.
Copula-based return periods were computed based on the joint distribution of daily precipitation and daily maximum storm surges. For each station, the dependence between precipitation and storm surge was measured using Kendall’s rank correlation coefficient τ (Kendall 1938). To avoid the risk of biasing the representation of extreme tails by the bulk of the joint distribution (Bevacqua et al. 2019), we selected the pairs (Rsel, Ssel) with both precipitation and storm surge exceeding the 95th percentiles to fit the copula models. If the number of selected pairs was <20 in a tide gauge, the threshold was lowered to ensure that the number of selected pairs was ≥20, while the threshold could not be below the 90th percentile. The marginal distributions of the selected precipitation series Rsel and selected storm surge series Ssel were selected from five distributions: lognormal, normal, exponential, Weibull, and generalized extreme value distributions. Based on these selected pairs (Rsel, Ssel), five types of copulas (including normal, t, Clayton, Gumbel, and Frank copulas) were used to identify the dependence structure between precipitation and storm surges (Nelsen 2006). The copulas’ parameters were estimated using the maximum likelihood estimator (Nelsen 2006). The Akaike information criterion (AIC; Akaike 1974) was employed to select the best-fitting copula. Goodness of fit was assessed using the Cramer–von Mises test (Genest et al. 2009).
Workflow of estimating return periods of compound floods and evaluating impacts of TCs and ETCs.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
b. Contributions of TCs and ETCs to compound floods
The contribution of TCs or ETCs to compound floods was estimated as the ratio of the number of TC-induced or ETC-induced compound floods to the total number of compound floods. It should be noted that this was limited by the record length of ETC data, and that the fractional contributions of TCs and ETCs were calculated based on data from 1979 to 2012.
To further estimate the impact of TCs/ETCs on the return periods of compound floods, we examined the changes in return periods by removing TC-/ETC-induced cases. In this section, the analysis of the impacts of TCs was based on the time series between 1950 and 2014, while the time series between 1979 and 2012 was used to analyze the impact of ETCs.
TCs can affect the return periods of compound floods by altering 1) dependence between precipitation and storm surge, 2) marginal distribution of precipitation, and 3) marginal distribution of storm surges. To attribute the impact of TCs on the return periods of compound floods to these three factors, we calculated the changes in return periods of compound floods by changing only one of the three factors at a time while keeping the other two factors unchanged. Thus, the extent of the impact of a given factor can be estimated by the relative changes in the return periods due to the change in this factor. The details of the experiments are as follows.
To evaluate the impact of TC-induced dependence between precipitation and storm surge on compound floods, we estimated the changes in return periods by altering the dependence structure to that of non-TC series while keeping the marginal distributions unchanged. We first created the non-TC precipitation series Rnon-TC and non-TC storm surge series Snon-TC, and their empirical CDFs
and , respectively. Based on the original selected series Rsel and Ssel, and their corresponding empirical CDFs and , we defined time series and . In this case, R1 and S1 had the same marginal distributions as Rsel and Ssel, respectively. The return periods of compound floods, without considering TC-associated dependence , can be calculated based on (R1, S1). Since (R1, S1) has the same marginal distributions as (Rsel, Ssel), and will be the same as and , respectively; however, the joint distribution will be the same as that of ( , ). Therefore, indicates the return periods of compound floods without the impact of TC-related dependence between precipitation and storm surge. To estimate the impact of TC-induced precipitation on compound floods, we estimated the changes in return periods by changing the marginal distribution of precipitation to that of a non-TC series while keeping the marginal distribution of storm surge and the dependence structure unchanged. We defined time series
and the corresponding CDF . In this case, R2 had the same marginal distribution as Rnon-TC. The return periods of compounds without considering TC-associated precipitation can be calculated based on (R2, Ssel). Since R2 had the same marginal distributions as Rnon-TC, was equivalent to , while the marginal distributions of storm surge and dependence structure were unchanged. Therefore, indicates the return periods of compound floods without the impact of TC-induced precipitation. To estimate the impact of TC-induced storm surge on compound floods, we estimated the changes in return periods by changing the marginal distribution of storm surge to that of a non-TC series while keeping the marginal distribution of precipitation and the dependence structure unchanged. We defined time series
. In this case, S3 had the same marginal distribution as Snon-TC. The return periods of compound floods, without considering TC-associated storm surge , can be calculated based on Rsel and S3. Since S3 had the same marginal distributions as Snon-TC, was equivalent to , while the marginal distribution of precipitation and dependence structure were unchanged. Therefore, indicates the return periods of compound floods without the impact of a TC-induced storm surge.
c. Weather patterns associated with compound floods
Sea level pressure, wind field, and precipitable water content for three locations (Brest in France, Charleston in the southeastern United States, and Takamatsu in Japan) under three different extreme event types (including compound floods, extreme precipitation and nonextreme storm surge, and extreme storm surge and nonextreme precipitation) were compared to examine the weather conditions that favored compound floods. These three tide gauges were selected because 1) they distribute in different parts of the world and 2) they have relative long data record lengths during 1948–2014 (the record lengths are 61, 67, and 30 years in Brest, Charleston, and Takamatsu, respectively). For each location, the anomalies of sea level pressure, wind, and precipitable water content of all compound flood days (i.e., precipitation > 98.5th percentile and storm surge > 98.5th percentile), extreme precipitation without extreme storm surge days (i.e., precipitation > 98.5th percentile and storm surge < 95th percentile), and extreme storm surge without extreme precipitation days (i.e., storm surge > 98.5th percentile and precipitation < 95th percentile) during 1948–2014 were calculated. The significance levels of the anomalies were examined using the two-tailed Student’s t test to detect the characteristics of weather patterns objectively (Shen et al. 2020a,b; Xu et al. 2020).
4. Results
a. Joint return periods of compound floods from precipitation and storm surge
Before evaluating the joint probability of compound floods from precipitation and storm surge, we first estimated the magnitudes of the marginal variables (i.e., extreme precipitation and extreme storm surge) at the global scale. Figure 3 shows the univariate magnitudes of extreme precipitation and storm surge at the 98.5th percentile, which is equivalent to a 67-day return period of the univariate extreme events. At this percentile level, if extreme precipitation and storm surges are independent of each other, the return period of the co-occurrence of both extreme events is expected to be approximately 12 years. The results show that extreme precipitation was the highest (>100 mm) in Japan and southern China, followed by the coast of the Gulf of Mexico and northern Australia, where extreme precipitation exceeded 60 mm (Fig. 3a). In contrast, the highest extreme storm surge was found in northwestern Europe, where the 98.5th percentile exceeded 1 m. The extreme storm surge was 0.4–0.8 m on the U.S. coast, in southern Europe, on the southeast coast of China, and at some tide gauges in Australia (Fig. 3b). Therefore, in Southeast Asia and the Gulf of Mexico, the magnitude of pluvial floods from heavy precipitation was higher than other regions, while Europe experienced higher extreme storm surges than other regions. The fractional contributions of TCs and ETCs to extreme precipitation and extreme storm surges are shown in Figs. S2 and S3, respectively. More than 40% of extreme precipitation in East Asia and northwestern Australia was associated with TCs, while the contribution of TCs to extreme precipitation was 10%–30% on the east coast of the United States. The contribution of TCs to extreme storm surges exceeded their contribution to extreme precipitation. In East Asia, TCs contributed to >50% of extreme storm surges, while this value ranged from 30% to 50% on the coast of the Gulf of Mexico (Fig. S2). In mid- and high-latitude areas, >60% of extreme precipitation and extreme storm surges were associated with ETCs (Fig. S3).
Magnitude of extreme precipitation (mm) and extreme storm surge (m) at the 98.5th percentile. (a) Magnitude of extreme precipitation at the 98.5th percentile and (b) magnitude of extreme storm surge at the 98.5th percentile. The colors in (a) and (b) denote the values of precipitation and storm surge, respectively. The 98.5th percentiles of daily accumulated precipitation and daily maximum storm surge were estimated based on empirical cumulated distribution.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
The correlation and dependence structure are key to constructing copula models to estimate the joint return periods of compound floods (Li et al. 2015). Kendall’s correlation coefficient τ was estimated to characterize the dependence between daily precipitation and the corresponding daily maximum storm surge for each tide gauge (Fig. 4a). The correlations between the daily values have been examined and used as evidence of the high correlations between precipitation and storm surges in different regions, such as in Europe (Bevacqua et al. 2019). We found a significant positive dependence between daily precipitation and daily maximum storm surge in most locations (i.e., 306 out of 314 locations) around the world. From a regional perspective, northwestern North America and western Europe showed higher dependence between precipitation and storm surge, while Kendall’s τ was smaller on the east coast of North America, East Asia, and Australia. Regarding extreme events, the dependence between extreme precipitation and extreme storm surge exceeding the 95th (instead of 98.5th) percentiles was examined (Fig. 4b). The thresholds of the 95th percentiles were employed here because the copula-based joint distributions between precipitation and storm surge were modeled based on the selected pairs exceeding the 95th percentiles (Bevacqua et al. 2019; also see section 3a herein). In this case, the dependences between extreme precipitation and extreme storm surge exceeding the 95th percentiles were closely related to the copula-based joint distributions, and thus, the copula-based joint return periods of compound floods. As shown in Fig. 4b, the dependence between extreme precipitation and extreme storm surge was significant; however, the number of stations with significant dependence was not as high as that of daily values. Significant positive correlations between extreme precipitation and storm surges were identified in Japan, the U.S. coast, and several stations in Australia and Europe. This spatial pattern largely coincides with the spatial distribution of TC contributions to extreme precipitation and extreme storm surges (Fig. S2). Considering that TCs are one of the most common weather systems that simultaneously cause heavy precipitation and storm surges, they play a key role in determining the dependence between extreme precipitation and extreme storm surges. At the same time, other factors such as topography and seasonality of extreme precipitation and extreme storm surges should not be ignored (Ward et al. 2018). For example, in Europe, the dependence between extreme precipitation and extreme storm surges at most stations was insignificant, while their daily values were significantly interdependent. This can be explained by the fact that the main extreme precipitation and extreme storm surges occur in different seasons over Europe (see section 5a). In addition to the correlation, the dependence structure of the two marginal values is important for the estimation of the copula-based joint return period of compound floods. As shown in Fig. S4, the best-fitting copulas of the pairs of extreme precipitation and extreme storm surges at most of the gauges passed the significance test at a level of 0.05, demonstrating that copulas are capable of constructing the dependence structure of extreme precipitation and extreme storm surges.
Dependence (i.e., Kendall’s τ) between precipitation and storm surge. (a) Dependence between daily precipitation and storm surge and (b) dependence between extreme precipitation and storm surge (i.e., >95th percentile). The colors indicate the values of Kendall’s τ. Open circles denote insignificant Kendall’s τ (α = 0.05).
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
The empirical and copula-based joint return periods of compound floods were calculated to estimate the occurrence probability of compound floods. Figure 5a shows the empirical return periods of the compound floods estimated using Eq. (1). The return periods of compound floods, considering the dependence between precipitation and storm surge, were much shorter than the expected 12-yr return period, if the two types of extreme events were independent. Specifically, the return periods of compound floods were <2 years along the U.S. coast, East Asia, and some stations in western Europe and Australia (Fig. 5a). On the west coast of Canada and southern Europe, the return periods of compound floods were 4–8 years. The return periods of the compound floods were >8 years in northwestern Europe. There were 13 gauges in which the estimation of the empirical return periods failed (i.e., open circles in Fig. 5a). In these gauges, the return periods may be longer than the number of years in the observations. Using copula functions to model dependence was not constrained by the shortcomings of the empirical method (Fig. 5b). Comparing Figs. 5a and 5b shows that in most locations, the copula-based return periods of compound floods matched well with the empirical return periods, which can be further validated by scatterplots of empirical return periods and copula-based return periods shown in Fig. S5. The Pearson’s correlation coefficient between empirical and copula-based return periods is 0.83 at a significance level of 0.001. For locations in which the estimation of empirical return periods failed, the copula-based return periods were longer than the record lengths. Therefore, because of the dependence between precipitation and storm surges, for most of the coastal regions of the world the return periods of compound floods from precipitation and storm surges were <8 years, especially in Asia, Australia, North America, and southern Europe. These were shorter than the expected 12-yr return period, assuming that the marginal variables were independent. When a higher threshold (e.g., 99.5th percentile) was used to define extreme precipitation and extreme storm surges, the return periods of more extreme compound floods showed similar patterns (Fig. S6). The empirical return periods of compound floods ranged from 2 to 32 years on the coasts of the United States, southern Europe, East Asia, and Australia, which was much shorter than the expected 109-yr return period in the independent case. The copula-based return periods of compound floods showed a similar pattern. In locations where the empirical return periods estimation failed (e.g., open circles in northern Europe, northeast coast of North America), the copula-based return periods were longer than 32 years. Compared with the results of compound floods defined by 98.5th (Fig. 5), the empirical return periods estimation failed in more tide gauges when using a higher threshold (Fig. S6a). The Pearson’s correlation coefficient between empirical and cupula-based return periods was 0.44 at a significance level of 0.001. The higher failing ratio and lower correlation coefficient when using the threshold of 99.5th percentile indicated that if the threshold was too high, the relatively short data record length and sparseness of samples might cause larger uncertainty for the estimation of compound flood probability. In summary, both spatial patterns of joint return periods of compound floods defined by the higher or lower threshold suggested that coastal regions experienced a high probability of compound floods due to the interdependence of precipitation and storm surges, and flood prevention management should take this factor into consideration.
Joint return periods of compound floods from extreme precipitation and extreme storm surge across the globe: (a) empirical return periods based on Eq. (1) and (b) copula-based return periods based on Eq. (5). A compound flood is the co-occurrence of an extreme storm surge and precipitation exceeding the 98.5th percentiles. Open circles in (a) denote no compound flood was observed. Colors denote the return periods.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
The hotspot regions of compound floods identified in our observation-based study include the coasts of North America, East Asia, southwestern Europe, and Australia, which are in general consistent with the spatial patterns detected in previous simulation-based studies (Bevacqua et al. 2020a,b; Couasnon et al. 2020). Because simulation-based compound flood identification may be affected by model uncertainties and biases, our results provide important references for the evaluation of these simulation-based studies (Couasnon et al. 2020).
b. Contribution of TCs and ETCs to compound floods
TCs can bring strong winds, and heavy rainfall, triggering compound floods. In higher-latitude areas such as Europe, northeastern North America, and South Australia, storm surges and heavy precipitation are often related to ETCs (Booth et al. 2016; Hawcroft et al. 2012). Therefore, we quantified the contributions of TCs and ETCs to compound floods on a global scale. TCs contributed to >80% of compound floods in East Asia and northern Australia, and >50% on the southeast coast of the United States (Fig. 6a). In areas less affected by TCs (such as western Europe, the northeast coast of the United States, and the east coast of Australia) the contributions of TCs were <30%. By contrast, ETCs’ contributions to compound floods were between 60% and 80% in the northeastern United States and western Europe, and >90% in eastern Europe (Fig. 6b). These results are consistent with the seasonality characteristics of concurrence of compound floods proposed by Bevacqua et al. (2020b), who found that the compound floods in tropics tend to occur during TC seasons, while at midlatitudes the peak seasons of compound floods are around the autumn and winter, when the ETC activity is highest. In areas affected by both TCs and ETCs (such as the coasts of the United States, Japan, and Australia) the contributions of ETCs were <40%. Comparing Figs. 6a and 6b, the spatial patterns of TC and ETC contributions to compound floods were compensated for by each other; in other words, in areas where TC fractional contributions were high, the contributions of ETCs were lower, and vice versa. Figure 6c shows that >80% of compound floods on the east coast of the United States and in Europe, East Asia, and Australia were associated with TCs and/or ETCs. On the west coast of the North America, the total contributions of TCs and ETCs were 20%–80%. This result demonstrates the importance of TCs and ETCs in triggering compound floods from precipitation and storm surges.
Fractional contribution of (a) TCs, (b) ETCs to compound floods from extreme precipitation and storm surge exceeding the 98.5th percentiles, and (c) total fractional contributions of TCs and ETCs to compound floods [i.e., sum of (a) and (b)]. Data from 1979 to 2012 were used to estimate the fractional contributions of TCs and ETCs. Open circles denote that the fractional contributions are 0.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
To evaluate how TCs (or ETCs) affect the return periods of compound floods, we compared the return periods of compound floods associated with non-TC (or non-ETC) events and those of the original time series including TC (or ETC) and non-TC (or non-ETC) events. After removing the impact of TCs, the return periods of compound floods increased dramatically and were expected to be 5 times longer than those of all events in East Asia (Fig. 7a). On the coast of the Gulf of Mexico and northern Australia, the return periods of non-TC compound floods were 150% longer than those in the original series. On the southeast coast of the United States, the return periods were shortened by 40%–150% due to the impact of TCs. The impacts of TCs on compound flood hazard were relatively insignificant in Europe and the northeast coast of the United States, where the changes in return periods of compound floods were <40% after the TC-induced events were removed. When the impacts of ETCs were removed, the greatest changes in return periods occurred in northeastern Europe (150%–500%; Fig. 7b). The return periods of compound floods increased by 40%–150% after removing the impact of ETCs in western Europe, the northeast coast of the United States, the east coast of Canada, and southern Australia. In areas including Japan, the west coast of the United States, and northwestern Australia, removing the ETC-associated data shortened the return periods of compound floods (Fig. 7b). This may be because most precipitation/storm surges associated with ETCs in these places are nonextreme events, and removing these relatively small values will increase the ratio of high values, which in turn increases the probability of extreme events.
Impacts of (a) TCs and (b) ETCs on return periods of compound floods. The impacts of TCs/ETCs were estimated in terms of relative changes (%) in return periods of compound floods after TC-/ETC-associated events are removed. Open circles denote locations that were not affected by TCs/ETCs.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
c. Attribution of changes in return periods to changes in marginal and joint distributions caused by TCs and ETCs
TCs and ETCs can affect the compound floods by changing the marginal distributions of extreme precipitation and extreme storm surges and altering the joint distributions between them. To attribute the changes in return periods of compound floods to these three factors, we estimated the relative changes in return periods in three cases when considering only (i) changes in the dependence between precipitation and storm surge by removing TC (or ETC) events, (ii) changes in the marginal distribution of precipitation by removing TC (or ETC) events, or (iii) changes in the marginal distribution of storm surges by removing TC (or ETC) events. For example, to estimate the impact of TC-caused changes in dependence on return periods of compound floods, we replaced the copula model of the original selected series with that of the non-TC series, and kept the marginal distributions of precipitation and storm surge unchanged. The impact of TC-caused changes in dependence on compound floods is the relative difference between the return periods of the non-TC series and the original selected series.
Figure 8a shows the changes in the return periods caused by changes in dependence due to the removal of TC-associated data. The changes in dependence by removing TC-associated events could increase the return periods of compound floods by >80% in Japan and South China; in contrast, this percentage was <40% on the east coast of the United States, Europe, and Australia, indicating the limited impact of TC-induced interactions on compound floods in these areas. When TC-induced precipitation was not considered, the return periods of compound floods increased by 150%–300% in East Asia and northern Australia (Fig. 8b). On the southeast coast of the United States, the probability of compound floods increased by 40%–150% due to TC-induced precipitation, whereas this percentage was <40% in the northeastern United States. For the changes caused by the removal of TC storm surges (Fig. 8c), the return periods of compound floods were at least 3 times longer in Japan. The return periods of compound floods increased by 80%–300% after TC-induced storm surge events were removed in the Gulf of Mexico, northern Australia, and Southeast Asia, whereas this value was <40% in the east coast of the United States, eastern Australia, and Europe. Comparatively speaking, TC-induced storm surges amplified the probability of compound floods more than TC-induced precipitation, which can also be expected due to the fact that the fractional contribution of TCs to extreme storm surges was higher than that of extreme precipitation (Fig. S2). Therefore, in East Asia (where TCs are frequently experienced), any changes in the three mechanisms associated with TCs can substantially affect the compound floods. TC-induced changes in storm surges amplify the probability of compound floods most substantially. For the southeast coast of the United States, the contributions of individual mechanisms to changes in the return periods of compound floods were <80%. The accumulated effect of the three mechanisms results in a double-compound flood hazard.
Attribution of changes in return periods of compound floods to the changes in (a) dependence, (b) precipitation, and (c) storm surge by removing events associated with TCs. Open circles denote the locations that were not affected by TCs.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
The results of the attribution analysis on ETCs’ impact on the return periods of compound floods are shown in Fig. 9. When removing ETC-associated dependence, the return periods of compound floods were shorter on the west coast of the United States, Europe, East Asia, and Australia (Fig. 9a), which implies that the dependence between extreme precipitation and extreme storm surge was stronger after removing ETC-associated events. The return periods increased (<60%) on the south and east coast of the United States after removing the impact of ETCs on dependence (Fig. 9a). When ignoring ETC-associated precipitation, the return periods were doubled in Europe and southeastern Australia, but were slightly shortened on the west coast of the United States, Japan, and northwestern Australia (Fig. 9b). In contrast, ETC-induced storm surges greatly increased the occurrence probability of compound floods in most ETC-affected areas except Japan (Fig. 9c), showing the significant impact of ETC-induced storm surges on compound floods in mid- and high-latitude areas. These results are consistent with the spatial patterns of the fractional contributions of ETCs to extreme precipitation and extreme storm surges (Fig. S3). Comparing the attribution results of the impacts of TCs and ETCs on compound floods, it is evident that the impacts of ETCs were more diverse among different factors (i.e., dependence and marginal distributions) and regions. Our analysis of ETCs was conducted based on the M09 track data from IMILAST (Neu et al. 2013). A caveat shall be made whereby the results may be changed if other track data are used, given the inconsistency among different ETC track datasets (Neu et al. 2013).
Attribution of changes in return periods of compound floods to the changes in (a) dependence, (b) precipitation, and (c) storm surge by removing events associated with ETCs. Open circles denote the locations that were not affected by ETCs.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
5. Analysis of mechanisms and weather patterns of compound floods
a. Different peak seasons of extreme precipitation and extreme storm surge in Europe
Our results show that the dependence between daily precipitation and daily maximum storm surge was significant in Europe, but the dependence of extreme precipitation and storm surge was much weaker (Fig. 4b). Due to the low dependence between the extremes of precipitation and storm surges, in some gauges in Europe, the return periods of compound floods were much longer than those in other regions (e.g., North America, East Asia, and Australia; Fig. 5). The low dependence of extreme values over Europe can be explained by the different seasons of extreme precipitation and extreme storm surges (Fig. 10); extreme precipitation tended to occur in August–October, while extreme storm surges mostly occurred in November–January, suggesting different peak seasons of extreme precipitation and extreme storm surges in Europe. The seasonal difference in the occurrences of extreme precipitation and storm surges largely reduced their interdependence, resulting in a lower occurrence of compound floods in Europe.
Probabilities of occurrence of extreme precipitation and extreme storm surge on day of the year over Europe. The contours and shades denote the density of data pairs (i.e., precipitation and storm surge exceeding the 95th percentile). The empirical cumulative probabilities of the stations in Europe (30°–60°E, 30°–70°N) were used to compute the density.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
b. Weather patterns associated with compound floods, extreme precipitation, and extreme storm surges
To reveal the weather patterns associated with compound flood events, extreme precipitation without extreme storm surge, and extreme storm surge without extreme precipitation, the anomalies of sea level pressure, wind, and precipitable water content of three tide gauges located in western Europe (Brest), the southeastern coast of the United States (Charleston), and East Asia (Takamatsu) were analyzed (Figs. 11 and 12). The weather patterns associated with compound floods were characterized by significant negative anomalous sea level pressure, stronger landward wind, and higher precipitable water content (Figs. 11a–c and 12a–c). During extreme precipitation events, the anomalies of sea level pressure and wind speed were smaller, and precipitable water content anomalies were higher, compared to that of compound floods and extreme storm surge (Figs. 11d–f and 12d–f). The occurrences of extreme storm surge events were associated with significant negative sea level pressure anomalies and strong wind, while there were no obvious changes in precipitable water content (Figs. 11g–i and 12g–i). The composited maps of weather patterns associated with compound floods, extreme precipitation without extreme storm surge, and extreme storm surge without extreme precipitation were shown in Figs. S7 and S8. The weather patterns associated with compound floods were characterized by deep low pressure, cyclonic wind, and high precipitable water content simultaneously occurring around the target locations (Figs. S7a–c and S8a–c). Abundant precipitable water was found in weather patterns during extreme precipitation, but there was no obvious low pressure or cyclonic wind (Figs. S7d–f and S8d–f). The occurrences of extreme storm surge events were associated with obvious cyclonic wind and deep low pressure systems (Figs. S7g–i and S8g–i). The low pressure systems associated with storm surges in Europe were larger and brought about stronger winds than those of the other sites, which, to some extent, explains the larger magnitudes of extreme storm surges in Europe (Fig. 3b). In the Takamatsu tide gauge in Japan, a low pressure system and cyclonic wind were identified in all the three types of extreme events, and the strengths of the systems varied. In this region, extreme events were highly related to TC activity. More than 90% of compound floods were associated with TC activities (Fig. 6a), and approximately 50% of extreme precipitation and extreme storm surges were induced by TCs (Fig. S2). Therefore, the weather mechanisms during the three types of extreme events were similar (i.e., TCs) in Takamatsu.
Anomalies of sea level pressure and wind associated with extreme events. Composite maps of sea level pressure anomalies (shading; hPa) and wind anomalies (black arrows) based on days where extreme events (i.e., >98.5th percentile) occurred in (left) Brest, (center) Charleston, and (right) Takamatsu between 1948 and 2014. Extreme events are (a)–(c) compound floods, (d)–(f) extreme precipitation without extreme storm surge (i.e., <95th percentile), and (g)–(i) extreme storm surge without extreme precipitation (i.e., <95th percentile). Shadings and arrows indicate that the anomalies are significant at the 95% confidence level based on a two-tailed Student’s t test. Green dots indicate the locations of tide gauges.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
Anomalies of precipitable water content associated with extreme events. Composite maps of precipitable water content (shading; kg m−2) based on days where extreme events (i.e., >98.5th percentile) occurred in (left) Brest, (center) Charleston, and (right) Takamatsu between 1948 and 2014. Extreme events are (a)–(c) compound floods, (d)–(f) extreme precipitation without extreme storm surge (i.e., <95th percentile), and (g)–(i) extreme storm surge without extreme precipitation (i.e., <95th percentile). Stippled areas represent the areas with anomalies being significant at the 95% confidence level based on a two-tailed Student’s t test. Green dots indicate the locations of tide gauges.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
Both TCs and ETCs bring about heavy rainfall and storm surges and, consequently, compound floods. The weather patterns are also similar for TCs and ETCs, which are characterized by low pressure systems. However, there were differences between the extreme events associated with TCs and ETCs in different locations. Figure 13 shows the accumulative probability of precipitation and storm surges for Brest, Charleston, and Takamatsu, and the magnitudes of precipitation and storm surges are shown in Fig. S9. In Brest, most compound floods and almost all extreme storm surge events were associated with ETCs (Fig. 13a). Almost all compound floods were caused by TCs in Takamatsu, and the most extreme storm surges and precipitation were associated with TCs (Fig. 13c). For the tide gauge Charleston in the southeastern United States (where the contribution of TCs and ETCs to compound floods were comparable, i.e., ~47%), TCs contributed to the most extreme compound floods, and ETCs caused a number of compound floods that were second only to the most severe flood (Fig. 13b). This result is consistent with Booth et al. (2016), who compared hurricane and ETC-induced storm surges on the northeast coast of the United States and found that TCs are more likely to generate the most extreme storm surges, while ETCs relate to slightly weaker events.
Scatterplots of accumulative probability (%) of precipitation and storm surge for tide gauges in (a) Brest in France, (b) Charleston in the United States, and (c) Takamatsu in Japan. The colors of points (boxes) denote different types of extreme events (zones). Extreme events associated with TCs and ETCs were denoted by rhombuses and triangles, respectively. The smaller open circles denote other events.
Citation: Journal of Climate 34, 20; 10.1175/JCLI-D-21-0050.1
6. Conclusions
In this study, the return periods of compound floods from heavy precipitation and storm surges were assessed at the global scale based on observed storm surges and precipitation with the longest overlapping record of >120 years. We estimated the empirical and copula-based return periods of compound floods and found that the dependence between precipitation and storm surge is a critical factor that affects the occurrence of compound floods. The dependence between daily precipitation and daily maximum storm surge was significant in most of the studied gauges, while significant dependence between extreme precipitation and extreme storm surge was identified on the U.S. coast and Japan. Because of the dependence between precipitation and storm surges, the return periods of compound floods from extreme precipitation and storm surges exceeding the 98.5th percentiles were substantially shorter than the expected 12-yr return period when the marginal variables were considered independent. Specifically, East Asia, southwestern Europe, the coast of North America, and the coast of Australia experienced a higher probability of compound floods with joint return periods < 2 years due to the strong dependence in these regions. Over Europe, the return periods of compound floods were longer than in other regions because of the low dependence of extreme precipitation and storm surges, which tended to occur in different seasons of the year. Our findings on the dependence between precipitation and storm surge and the return periods of compound floods in general agree with those in previous studies based on observations and simulations (Bevacqua et al. 2020a,b; Couasnon et al. 2020; Ward et al. 2018). These results present the global perspective of compound flood hazards from heavy precipitation and storm surges, and highlight the necessity of considering cases of compound floods when designing flood management strategies.
Our analyses of the contributions of TCs and ETCs to compound floods show that cyclones (including TCs and ETCs) play important roles in the occurrence of compound floods in many coastal regions. In East Asia, >80% of compound floods are associated with TC activity. In mid- and high-latitude areas that are less affected by TCs, ETCs play a key role in the occurrence of compound floods. TCs and ETCs can impact the compound floods in two ways: on one hand, TCs and ETCs raise the marginal probabilities of extreme precipitation and storm surges, and on the other hand TCs and ETCs can simultaneously cause heavy precipitation and extreme storm surges, showing the dependence of the two extremes. Our results show that the likelihoods of compound floods increase by 40%–500% or more in different areas of the world due to TCs. Meanwhile, ETC-associated precipitation and storm surge can amplify the probability of compound floods over Europe by 100%–300%. These findings provide preliminary insights into the interactions between storm activities and compound floods.
The weather patterns associated with compound floods, extreme precipitation, and extreme storm surges in three tide gauges located in Europe, the United States, and East Asia were compared to further investigate the physical mechanisms associated with compound floods. The results showed that the occurrence of compound floods was closely related to deep low pressure, cyclonic wind, and abundant precipitable water. During extreme precipitation events, weather patterns were characterized by abundant precipitable water but no low pressure or cyclonic winds. Extreme storm surge events were triggered by obvious cyclonic winds and deep low pressure systems. The weather conditions associated with compound floods, extreme precipitation, and extreme storm surges in East Asia were highly similar because TCs contributed to most extreme events in this region. In Europe, most compound floods or univariate extreme events (i.e., extreme precipitation and extreme storm surge) were associated with ETCs, while on the east coast of the United States and Japan TCs were likely to create most extreme hazards. Extreme precipitation and extreme storm surges in Europe tended to occur in different seasons, which explains the low interdependence of extreme precipitation and extreme storm surges in this region.
Thus, this study depicts the spatial pattern of compound flood return periods based on observational precipitation and storm surge data, and quantifies the impact of TCs and ETCs on compound floods across the globe. The weather conditions of the compound flood events were analyzed. Our results provide an important reference for global and regional compound flood risk management, which is a major challenge of climate change adaptation, as indicated in the IPCC report (IPCC 2012b).
Acknowledgments
The work described in this paper was supported by grants from the Research Grants Council of the Hong Kong Special Administrative Region, China (HKBU12303517 and HKBU12302518), the National Key R&D Program of China (Project No. 2019YFC1510400), Guangdong-Hong Kong Joint Laboratory for Water Security (2020B1212030005), and the National Natural Science Foundation of China (U1911205 and 41901041). The daily precipitation data from Global Historical Climatology Network (GHCN–Daily) are available at https://www.ncdc.noaa.gov/ghcn-daily-description. The hourly sea level from the Global Extreme Sea Level Analysis version 2 (GESLA-2) can be obtained from https://www.gesla.org/. The International Best Track Archive for Climate Stewardship is available at https://www.ncdc.noaa.gov/ibtracs/. The extratropical cyclone track data from the intercomparison project Intercomparison of Mid-Latitude Storm Diagnostics (IMILAST) is available at https://proclim.scnat.ch/en/activities/project_imilast. The NCEP–NCAR reanalysis dataset can be obtained from https://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.html.
REFERENCES
Akaike, H., 1974: A new look at the statistical model identification. IEEE Trans. Autom. Control, 19, 716–723, https://doi.org/10.1109/TAC.1974.1100705.
Bevacqua, E., D. Maraun, I. Hobæk Haff, M. Widmann, and M. Vrac, 2017: Multivariate statistical modelling of compound events via pair-copula constructions: Analysis of floods in Ravenna (Italy). Hydrol. Earth Syst. Sci., 21, 2701–2723, https://doi.org/10.5194/hess-21-2701-2017.
Bevacqua, E., D. Maraun, M. I. Vousdoukas, E. Voukouvalas, M. Vrac, L. Mentaschi, and M. Widmann, 2019: Higher probability of compound flooding from precipitation and storm surge in Europe under anthropogenic climate change. Sci. Adv., 5, eaaw5531, https://doi.org/10.1126/sciadv.aaw5531.
Bevacqua, E., M. I. Vousdoukas, T. G. Shepherd, and M. Vrac, 2020a: Brief communication: The role of using precipitation or river discharge data when assessing global coastal compound flooding. Nat. Hazards Earth Syst. Sci., 20, 1765–1782, https://doi.org/10.5194/nhess-20-1765-2020.
Bevacqua, E., and Coauthors, 2020b: More meteorological events that drive compound coastal flooding are projected under climate change. Commun. Earth Environ., 1, 47, https://doi.org/10.1038/s43247-020-00044-z.
Booth, J. F., H. E. Rieder, and Y. Kushnir, 2016: Comparing hurricane and extratropical storm surge for the mid-Atlantic and northeast coast of the United States for 1979–2013. Environ. Res. Lett., 11, 094004, https://doi.org/10.1088/1748-9326/11/9/094004.
Catto, J. L., and S. Pfahl, 2013: The importance of fronts for extreme precipitation. J. Geophys. Res., 118, 10 791–10 801, https://doi.org/10.1002/jgrd.50852.
Chang, C. P., Y. T. Yang, and H. C. Kuo, 2013: Large increasing trend of tropical cyclone rainfall in Taiwan and the roles of terrain. J. Climate, 26, 4138–4147, https://doi.org/10.1175/JCLI-D-12-00463.1.
Couasnon, A., and Coauthors, 2020: Measuring compound flood potential from river discharge and storm surge extremes at the global scale. Nat. Hazards Earth Syst. Sci., 20, 489–504, https://doi.org/10.5194/nhess-20-489-2020.
Emanuel, K., 2005: Increasing destructiveness of tropical cyclones over the past 30 years. Nature, 436, 686–688, https://doi.org/10.1038/nature03906.
Emanuel, K., 2017: Assessing the present and future probability of Hurricane Harvey’s rainfall. Proc. Natl. Acad. Sci. USA, 114, 12 681–12 684, https://doi.org/10.1073/pnas.1716222114.
Garner, A. J., and Coauthors, 2017: Impact of climate change on New York City’s coastal flood hazard: Increasing flood heights from the preindustrial to 2300 CE. Proc. Natl. Acad. Sci. USA, 114, 11 861–11 866, https://doi.org/10.1073/pnas.1703568114.
Genest, C., B. Rémillard, and D. Beaudoin, 2009: Goodness-of-fit tests for copulas: A review and a power study. Insur. Math. Econ., 44, 199–213, https://doi.org/10.1016/j.insmatheco.2007.10.005.
Gu, X., Q. Zhang, V. P. Singh, and P. Shi, 2017: Nonstationarity in timing of extreme precipitation across China and impact of tropical cyclones. Global Planet. Change, 149, 153–165, https://doi.org/10.1016/j.gloplacha.2016.12.019.
Guha-Sapir, D., R. Below, and P. Hoyois, 2018: EM-DAT: The CRED/OFDA International Disaster Database. Université Catholique de Louvain, accessed 9 September 2019, https://public.emdat.be/.
Hawcroft, M. K., L. C. Shaffrey, K. I. Hodges, and H. F. Dacre, 2012: How much Northern Hemisphere precipitation is associated with extratropical cyclones? Geophys. Res. Lett., 39, 24809, https://doi.org/10.1029/2012GL053866.
Hawcroft, M. K., E. Walsh, K. Hodges, and G. Zappa, 2018: Significantly increased extreme precipitation expected in Europe and North America from extratropical cyclones. Environ. Res. Lett., 13, 124006, https://doi.org/10.1088/1748-9326/aaed59.
Hong Kong Observatory, 2017: Super Typhoon Hato (1713) 20 to 24 August 2017. Accessed 24 September 2019, https://www.weather.gov.hk/informtc/hato17/report.html.
Hoskins, B. J., and K. I. Hodges, 2002: New perspectives on the Northern Hemisphere winter storm tracks. J. Atmos. Sci., 59, 1041–1061, https://doi.org/10.1175/1520-0469(2002)059<1041:NPOTNH>2.0.CO;2.
Huntingford, C., and Coauthors, 2014: Potential influences on the United Kingdom’s floods of winter 2013/14. Nat. Climate Change, 4, 769–777, https://doi.org/10.1038/nclimate2314.
Ikeuchi, H., Y. Hirabayashi, D. Yamazaki, S. Muis, P. J. Ward, H. C. Winsemius, M. Verlaan, and S. Kanae, 2017: Compound simulation of fluvial floods and storm surges in a global coupled river-coast flood model: Model development and its application to 2007 Cyclone Sidr in Bangladesh. J. Adv. Model. Earth Syst., 9, 1847–1862, https://doi.org/10.1002/2017MS000943.
IPCC, 2012a: Changes in climate extremes and their impacts on the natural physical environment. Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation, C. B. Field et al., Eds., Cambridge University Press, 109–230.
IPCC, 2012b: Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation, C. B. Field et al., Eds., Cambridge University Press, 582 pp.
Irish, J. L., D. T. Resio, and J. J. Ratcliff, 2008: The influence of storm size on hurricane surge. J. Phys. Oceanogr., 38, 2003–2013, https://doi.org/10.1175/2008JPO3727.1.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, https://doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.
Kendall, M. G., 1938: A new measure of rank correlation. Biometrika, 30, 81–93, https://doi.org/10.2307/2332226.
Khouakhi, A., G. Villarini, and G. A. Vecchi, 2017: Contribution of tropical cyclones to rainfall at the global scale. J. Climate, 30, 359–372, https://doi.org/10.1175/JCLI-D-16-0298.1.
Knapp, K. R., M. C. Kruk, D. H. Levinson, H. J. Diamond, and C. J. Neumann, 2010: The International Best Track Archive for Climate Stewardship (IBTrACS) unifying tropical cyclone data. Bull. Amer. Meteor. Soc., 91, 363–376, https://doi.org/10.1175/2009BAMS2755.1.
Knutson, T. R., and Coauthors, 2015: Global projections of intense tropical cyclone activity for the late twenty-first century from dynamical downscaling of CMIP5/RCP4.5 scenarios. J. Climate, 28, 7203–7224, https://doi.org/10.1175/JCLI-D-15-0129.1.
Kossin, J. P., 2018: A global slowdown of tropical-cyclone translation speed. Nature, 558, 104–107, https://doi.org/10.1038/s41586-018-0158-3.
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.
Lai, Y., and Coauthors, 2020: Greater flood risks in response to slowdown of tropical cyclones over the coast of China. Proc. Natl. Acad. Sci. USA, 117, 14 751–14 755, https://doi.org/10.1073/pnas.1918987117.
Li, J., Q. Zhang, Y. D. Chen, and V. P. Singh, 2015: Future joint probability behaviors of precipitation extremes across China: Spatiotemporal patterns and implications for flood and drought hazards. Global Planet. Change, 124, 107–122, https://doi.org/10.1016/j.gloplacha.2014.11.012.
Lian, J. J., K. Xu, and C. Ma, 2013: Joint impact of rainfall and tidal level on flood risk in a coastal city with a complex river network: A case study of Fuzhou City, China. Hydrol. Earth Syst. Sci., 17, 679–689, https://doi.org/10.5194/hess-17-679-2013.
Lin, I. I., G. J. Goni, J. A. Knaff, C. Forbes, and M. M. Ali, 2013: Ocean heat content for tropical cyclone intensity forecasting and its impact on storm surge. Nat. Hazards, 66, 1481–1500, https://doi.org/10.1007/s11069-012-0214-5.
Menne, M. J., I. Durre, R. S. Vose, B. E. Gleason, and T. G. Houston, 2012: An overview of the Global Historical Climatology Network–Daily database. J. Atmos. Oceanic Technol., 29, 897–910, https://doi.org/10.1175/JTECH-D-11-00103.1.
Milly, P. C. D., R. T. Wetherald, K. A. Dunne, and T. L. Delworth, 2002: Increasing risk of great floods in a changing climate. Nature, 415, 514–517, https://doi.org/10.1038/415514a.
Moftakhari, H. R., G. Salvadori, A. AghaKouchak, B. F. Sanders, and R. A. Matthew, 2017: Compounding effects of sea level rise and fluvial flooding. Proc. Natl. Acad. Sci. USA, 114, 9785–9790, https://doi.org/10.1073/pnas.1620325114.
Nelsen, R. B., 2006: An Introduction to Copulas. 2nd ed., Springer, 272 pp.
Neu, U., and Coauthors, 2013: IMILAST: A community effort to intercompare extratropical cyclone detection and tracking algorithms. Bull. Amer. Meteor. Soc., 94, 529–547, https://doi.org/10.1175/BAMS-D-11-00154.1.
Paprotny, D., O. Morales-Nápoles, and S. N. Jonkman, 2018: HANZE: A pan-European database of exposure to natural hazards and damaging historical floods since 1870. Earth Syst. Sci. Data, 10, 565–581, https://doi.org/10.5194/essd-10-565-2018.
Pawlowicz, R., B. Beardsley, and S. Lentz, 2002: Classical tidal harmonic analysis including error estimates in MATLAB using T_TIDE. Comput. Geosci., 28, 929–937, https://doi.org/10.1016/S0098-3004(02)00013-4.
Salvadori, G., and C. De Michele, 2004: Frequency analysis via copulas: Theoretical aspects and applications to hydrological events. Water Resour. Res., 40, W12511, https://doi.org/10.1029/2004WR003133.
She, D., and Coauthors, 2015: Investigating the variation and non-stationarity in precipitation extremes based on the concept of event-based extreme precipitation. J. Hydrol., 530, 785–798, https://doi.org/10.1016/j.jhydrol.2015.10.029.
Shen, X., L. Wang, and S. Osprey, 2020a: The Southern Hemisphere sudden stratospheric warming of September 2019. Sci. Bull., 65, 1800–1802, https://doi.org/10.1016/j.scib.2020.06.028.
Shen, X., L. Wang, and S. Osprey, 2020b: Tropospheric forcing of the 2019 Antarctic sudden stratospheric warming. Geophys. Res. Lett., 47, e2020GL089343, https://doi.org/10.1029/2020GL089343.
Svensson, C., and D. A. Jones, 2002: Dependence between extreme sea surge, river flow and precipitation in eastern Britain. Int. J. Climatol., 22, 1149–1168, https://doi.org/10.1002/joc.794.
Svensson, C., and D. A. Jones, 2004: Dependence between sea surge, river flow and precipitation in south and west Britain. Hydrol. Earth Syst. Sci., 8, 973–992, https://doi.org/10.5194/hess-8-973-2004.
van den Hurk, B., E. van Meijgaard, P. de Valk, K. J. van Heeringen, and J. Gooijer, 2015: Analysis of a compounding surge and precipitation event in the Netherlands. Environ. Res. Lett., 10, 035001, https://doi.org/10.1088/1748-9326/10/3/035001.
Vousdoukas, M. I., L. Mentaschi, E. Voukouvalas, M. Verlaan, S. Jevrejeva, L. P. Jackson, and L. Feyen, 2018: Global probabilistic projections of extreme sea levels show intensification of coastal flood hazard. Nat. Commun., 9, 2360, https://doi.org/10.1038/s41467-018-04692-w.
Wahl, T., S. Jain, J. Bender, S. D. Meyers, and M. E. Luther, 2015: Increasing risk of compound flooding from storm surge and rainfall for major US cities. Nat. Climate Change, 5, 1093–1097, https://doi.org/10.1038/nclimate2736.
Walsh, K. J. E., and Coauthors, 2016: Tropical cyclones and climate change. Wiley Interdiscip. Rev.: Climate Change, 7, 65–89, https://doi.org/10.1002/wcc.371.
Wang, Q., Y. Xu, N. Wei, S. Wang, and H. Hu, 2019: Forecast and service performance on rapidly intensification process of Typhoons Rammasun (2014) and Hato (2017). Trop. Cyclone Res. Rev., 8, 18–26, https://doi.org/10.1016/j.tcrr.2019.07.002.
Wang, X. L., V. R. Swail, and F. W. Zwiers, 2006: Climatology and changes of extratropical cyclone activity: Comparison of ERA-40 with NCEP–NCAR reanalysis for 1958–2001. J. Climate, 19, 3145–3166, https://doi.org/10.1175/JCLI3781.1.
Ward, P. J., and Coauthors, 2018: Dependence between high sea-level and high river discharge increases flood hazard in global deltas and estuaries. Environ. Res. Lett., 13, 084012, https://doi.org/10.1088/1748-9326/aad400.
Woodworth, P. L., J. R. Hunter, M. Marcos, P. Caldwell, M. Menéndez, and I. Haigh, 2016: Towards a global higher-frequency sea level dataset. Geosci. Data J., 3, 50–59, https://doi.org/10.1002/gdj3.42.
Xu, H., K. Xu, L. Bin, J. Lian, and C. Ma, 2018: Joint risk of rainfall and storm surges during typhoons in a coastal city of Haidian Island, China. Int. J. Environ. Res. Public Health, 15, 1377, https://doi.org/10.3390/ijerph15071377.
Xu, H., K. Xu, J. Lian, and C. Ma, 2019: Compound effects of rainfall and storm tides on coastal flooding risk. Stochastic Environ. Res. Risk Assess., 33, 1249–1261, https://doi.org/10.1007/s00477-019-01695-x.
Xu, P., L. Wang, W. Chen, G. Chen, and I. S. Kang, 2020: Intraseasonal variations of the British–Baikal corridor pattern. J. Climate, 33, 2183–2200, https://doi.org/10.1175/JCLI-D-19-0458.1.
Zhang, G., H. Murakami, T. R. Knutson, R. Mizuta, and K. Yoshida, 2020: Tropical cyclone motion in a changing climate. Sci. Adv., 6, eaaz7610, https://doi.org/10.1126/sciadv.aaz7610.
Zhang, Q., Y. Lai, X. Gu, P. Shi, and V. P. Singh, 2018b: Tropical cyclonic rainfall in China: Changing properties, seasonality, and causes. J. Geophys. Res., 123, 4476–4489, https://doi.org/10.1029/2017JD028119.
Zhang, W., G. Villarini, G. A. Vecchi, and J. A. Smith, 2018a: 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.
Zheng, F., S. Westra, and S. A. Sisson, 2013: Quantifying the dependence between extreme rainfall and storm surge in the coastal zone. J. Hydrol., 505, 172–187, https://doi.org/10.1016/j.jhydrol.2013.09.054.
Zscheischler, J., and Coauthors, 2018: Future climate risk from compound events. Nat. Climate Change, 8, 469–477, https://doi.org/10.1038/s41558-018-0156-3.