Different types of weather events, including floods, lightning, tornadoes, hurricanes, and heat waves, can harm society and result in morbidities and mortalities. Flooding is one of the main weather-related disasters that cause a vast number of deaths every year across the globe [1
]. Based on the Emergency Events Database, 6.62 million people were killed by floods worldwide in the 20th century [5
]. Jonkman [6
] investigated fatalities related to different types of flooding and different regions around the world from 1975 to 2002. The study found floods to be the most significant natural hazard, and flash floods caused the highest fatalities on average per event than other flood types. More than 100,000 people died in flood events across the globe in the last 10 years of the 20th century according to Jonkman [6
Smith and Ward [7
] provided detailed information on the causes and consequences of river and coastal floods and the different approaches adopted by practitioners to respond to different types of flood hazards. French and Holt [8
] classified floods into three main categories: Flash floods, river floods, and coastal floods. The same classification method was adopted by other researchers who compiled data from the database maintained by the National Climatic Data Center (e.g., [9
]). Flash floods are defined as floods that rise and fall rapidly within six hours or even within three hours; river floods are caused by river water overflowing its banks; and coastal floods are caused by extratropical and tropical cyclone surges [8
]. Floods are caused by a variety of natural and man-made processes and conditions [9
]. Specifically, several factors can contribute to flooding, including rainfall amounts and intensity, regional topography, regional land use, soil types, and antecedent moisture conditions. For example, the absence of vegetation can contribute to flash flooding; steep slopes can cause fast surface runoff; rapid snowmelt in mountain areas can lead to the sharp rise of downstream rivers; and the breach of a dam or levee can lead to flash floods [8
]. These reasons have led to several researchers calling for the adoption of an integrated approach for studying devastating flooding events that includes observations and modeling e.g., [11
In a study that examined the detailed circumstances of more than 1000 flash flood fatalities across the US, Terti et al. [14
] observed that the fatality circumstances exhibited certain characteristics related to season, time of day, duration of flood, location, and age and gender groups. Hamilton et al. [15
] performed psycho-cognitive analysis of the beliefs of people who willingly drive into flood water in Australia. They identified key attitudinal, social expectations, and efficacy beliefs that guide willingness to drive through flooded waterways, including fear of being stranded, pressure from others, and seeing other drivers doing it, among others. Diakakis [16
] analyzed 60-year flood fatality data from Greece and found a strong association between the risk-taking behavior during floods and the demographics of the victims, the type of the surrounding environment, and vehicle use, and used this information to develop a statistical model to predict the behavior of a flood victim based on the characteristics of the individual and the environment. Vinet et al. [17
] examined fatalities resulting from two flood events in France—a total of 67 fatalities. They found that the individual vulnerability is the product of internal factors including personal knowledge, age, and health, and awareness of the risk, and external factors such as the availability of shelter and building type. They stressed the need to address all of these specific vulnerabilities in prevention and warning messages. Diakakis et al. [18
] studied flood fatalities in Greece and identified factors and behaviors leading to increased vulnerability and found them to be different between urban and nonurban environments.
According to weather-related fatality and injury statistics from the National Weather Service (NWS) for the 10-year average of 2009–2018 and for the 30-year average of 1989–2018, floods caused the second-highest number of weather-related fatalities in the US, surpassed only by heat waves [19
]. However, Borden and Cutter [20
] listed flooding as the fourth deadliest weather-related disaster behind heat/drought, severe weather, and winter weather. Kunkel et al. [21
] observed and discussed a generally increasing trend of flood-related damages and fatalities in the last 25 years of the 20th century in the US. Freshwater flooding from 1970 to 1999 caused more than one-half of 600 water-related fatalities in the contiguous US [22
Several researches have examined flood-related fatalities in the United States e.g., [1
]. French et al. [23
] reported that flash floods contributed to the most flood fatalities, identifying 1185 fatalities caused by 32 flash flood events from 1977 to 1981. According to the study, forty-two percent (42%) of reported drowning deaths were vehicle-related. In four flood events involving dam breaks, warnings for heavy rain, and flash flooding were issued, but none for dam failure. Dittmann [1
] estimated a total of 3934 flood fatalities from 1959 to 1991 in the United States with an annual average of 119 fatalities, while Ashley and Ashley [10
] reported a total of 4586 fatalities related to flooding in the contiguous US from 1959 to 2005, with an annual average of 97.6 fatalities (excluding the data of Hurricane Katrina, which occurred in 2005). Ashley and Ashley [10
] suggested that heavy rain, snowmelt, structural failure, and a combination of these factors all contributed to flooding. They also found that flash floods accounted for the majority of flood-related fatalities and identified high-fatality regions, such as the Ohio River valley, the northeast Interstate-95 corridor, and near the Balcones Escarpment in south-central Texas. Sharif et al. [9
] conducted research on Texas from 1959 to 2008 and observed that the edge of Balcones Escarpment is a region of remarkably high fatalities. According to the study, a total of 840 flood-related fatalities occurred in Texas and flash floods caused a majority of those fatalities.
This study is motivated by several major floods that occurred in the US in recent years and resulted in a significant number of fatalities. The analysis investigates the demographic aspect (gender and age), spatial aspect, and temporal aspect (time of the day and month of the year), flood type, and circumstances leading to flood fatalities in the contiguous US between 1959 and 2019. The purpose of this study is to improve understanding of the situational conditions, demographics, and spatial and temporal characteristics of flood fatalities and describe how the results can be used to improve flood fatality prevention measures and warning messages taking into account compounding situations, such as the recent COVID-19 pandemic.
2. Study Area and Data Source
The study area includes the contiguous (conterminous) United States (CONUS), which consists of the 48 adjoining US states and the District of Columbia (DC). The study area excludes the non-contiguous states of Alaska and Hawaii. Together, the 48 contiguous states and the District of Columbia occupy an area of 8,080,464 km2
. The climate of CONUS is controlled to changes in latitude and a range of geographic features, including mountains and deserts. Generally, the climate of CONUS becomes warmer in the north-south direction and drier in the east-west direction except for very wet areas along the West Coast. West of 100° W, the climate is cold and semi-arid in the interior upper western states, whereas the climate is warm, hot desert, and semi-arid in southwestern US. East of 100° W, the climate becomes humid continental in northern areas, transitioning into a humid temperate climate from the Southern Plains and lower Midwest east to the Middle Atlantic states. In addition to local climate and type of precipitation, flooding potential is controlled by local topography and other land surface characteristics [24
]. For example, the steep mountainous terrain and major urban centers exposed to tropical storms and convective rainfall systems are more prone to flooding [25
]. Another example is the encounter of warm and humid air from the south, the Gulf, and the cold air from the north that occurs right above the Balcones escarpment curve in Texas causing heavy rains and floods (Flash Flood Alley of Texas). At the same time, tropical storms from the Gulf are another contributor to the heavy rains and floods in the Alley [9
The flood fatality data used in this study were compiled from Storm Data
]. Storm Data
is maintained by the National Climatic Data Center (NCDC) as monthly reports including detailed information on severe weather events across the US and their impacts, especially injuries and fatalities. The NWS started producing the “Climatological Data” publication, which only reported information on tornadoes and their impacts in 1950 and later added information on thunderstorms, wind, and hail in 1955. In 1959, the “Climatological Data” was officially renamed “Storm Data
” and started to include data on all storms and other severe weather phenomena. NWS-designated officials compile the information included in the Storm Data
publication from numerous sources. These sources include, but are not limited to, emergency management officials in different administration levels, law enforcement, media, insurance companies, and the public [26
]. Storm Data
provides some details about flood fatalities and victims. The description of the flood type in Strom Data is confusing to a large degree and include categories that seem to overlap including “flood”, “flash flood”, “flash flood and flood”, “flash flooding and river flooding”, “rain”, “heavy rain and flooding”, “flooding due to hurricane/tropical storm”, and “storm/tornadoes/cyclones”. The causal activity/setting/location of the fatality includes “in water”, “vehicle”, “permanent home”, “mobile home”, “outside”, “camping”, “boat”, “business”, and “equipment”. The gender and age of the victims are provided for some of the victims. The time of the fatality occurrence is often provided in broad terms. Detailed information is missing for many fatalities. For example, the time of occurrence was not mentioned for 62% of the fatalities.
Currently, Storm Data
is the primary and most comprehensive source for flood fatality data, despite it containing several problems. For example, quality control is an important concern for all large databases such as Storm Data
. The flood fatality data in Storm Data
is generally conservative and underreported [9
]. Some of the flood fatality records lack certain details, such as the age and gender of the victims, time of the fatality occurrence, and the specific location. Moreover, the description of the events is not always consistent, and the amount of details may vary among states. Nevertheless, the fatality data listed in Storm Data
are well documented in general and the dataset is the best available source of this kind of flood fatality data [7
]. Lastly, a review of event descriptions in Storm Data
reveals that they have become more thorough over the years. Population data was obtained for the US Census Bureau [28
Flood fatality data used in this study were compiled from Storm Data.
The fatality data is listed in five broad categories that define the main flood types. These categories are coastal flood, flash flood, flood, heavy rain, and tropical storm. Moreover, fatality data from 1959 to 1995 were only available by manually reviewing the PDF files. We downloaded and reviewed these files to manually extract the needed data from each file. Data for years after 1995, were directly downloaded from the website in a tabular format. To compare the flood fatality risk for each state in the contiguous United States, the number of flood-related fatalities of each state is standardized by the corresponding population estimate for the state in the corresponding year (annual fatalities/population of the year). The description provided in Storm Data
was used to categorize fatality data, e.g., by flood type, activity/setting/location, time, and age/gender although these descriptions were missing from a significant number of fatalities as will be explained in the analysis. The population data of each state in each year was based on estimates of the US Census Bureau [28
]. The percentages of population in different age groups were obtained from the average of population percentages for different age groups of 1960, 1970, 1980, 1990, 2000, and 2010 following the approach described by Dittmann [1
]. To verify the age-related vulnerability to flood hazards, the fatality numbers were standardized by multiplying fatality numbers in each age group and the corresponding percentage of population in each age group. Flood fatality data for each state were downloaded for each year and then combined when necessary. In-house R scripts were developed to extract the data needed for the analysis. Excel was used to perform analysis of variance (ANOVA) to compare variables, such as gender. One-sample chi-square test was used to confirm the seasonality of fatalities.
5. Summary and Conclusions
This paper examined flood-related fatality data in the contiguous US from 1959 through 2019. The study extends the period of analysis of US flood fatalities reported in previous studies, computes the fatality rate for each state, and provides more details on the situational conditions, demographics, and spatial and temporal characteristics of flood fatalities. A total of 6478 flood fatalities occurred during the study period. The last two decades witnessed major flood events that changed the ranking of the top states compared to previous studies except for Texas (1069) with significantly higher flood-related fatalities than any other state. Rankings of counties within some states changed as well. The analysis included the temporal patterns of flood fatalities, the circumstances leading to the fatalities, the demographics of the victims, and the types of fatal flooding. The spatial analysis identified clusters of flood fatalities related to local conditions (e.g., climate or terrain) such as the Flash Flood Alley in Texas and the Gulf Coast region.
The analysis indicated that most flood fatalities could be prevented as most fatalities were related to voluntary contact with floodwaters. This conclusion is supported by other research. For example, Diakakis [16
] found that more than 74% of flood fatalities in Greece resulted from victims moving into or approaching floodwaters from an initial position of safety and Coates [30
] reported that only about 31% of flood victims in Australia died while waiting in their homes or camps, some unaware of the flood. Similar to almost all flood studies conducted in the US, this comprehensive analysis showed that the majority of fatalities were vehicle-related events and males are much more likely to be killed in a flood than females. Alderman et al. [50
] cited several studies that demonstrated the overrepresentation of males and older individuals among flood victims, especially in medium- and high-income countries.
The results will help identify the risk factors associated with different types of flooding and the vulnerability of exposed communities. However, more detailed analysis is needed to identify individual vulnerability factors of exposed communities as conditions may be very different for different regions. Nonetheless, the study helps identify states with more conditions that lead to flood fatalities and detailed studies at a higher level of detail are necessary. Understanding the influence of local conditions and flooding risk factors is the first step to reducing flood fatalities. Flooding risk in urban areas will increase as the US continues to urbanize, limiting natural runoff mitigation mechanisms. Scientific and engineering knowledge, modeling tools, and data can be used not only to better characterize the frequency of severe storms, accurately delineate flood plains, and predict flood impacts, but also to plan quick responses when a major storm is imminent. Hydrometeorological forecasting needs to be improved. Policy makers, city planners, and engineers need to take proactive measures to protect urban communities against these impacts.
The spatial analysis of flood fatalities at multiple scales is necessary because most mitigation efforts are better to implement at the community level through various strategies, such as flood-proofing buildings [51
] and acquiring open spaces, conserving wetlands, land use management, and continuous monitoring [52
]. NWS offices can coordinate with local agencies in regards to flood risk communication and response strategies and evacuation training [53
]. States can coordinate with the federal government regarding insurance programs and tax incentives for flood risk [54
]. Moreover, one of the most common problems with flash floods is the unexpected overbank flow, which leads to city and settlement flooding. In many cases, the overbank flow is the result of human intervention in streams’ hydraulic characteristics (reduction of stream cross section) due to intense and unplanned urban sprawl [11
Non-stationary methods of flood frequency analysis [57
] can help communities and policy makers understand the rising flood risk and the need for investment in robust flood protection measures. Identifying and prioritizing runoff mitigation projects can provide some solutions to flooding problems (e.g., [58
]). However, structural solutions are not enough. The information provided in this study can help tailor educational campaigns to specific vulnerable groups. In addition to education campaigns typically organized by local authorities and NWS offices, transportation agencies, insurance companies, and auto manufacturers can help educate the public about the risks of driving into flood water. Most vulnerable individuals, such as drivers and certain age groups, must be the main target of awareness campaigns. However, there remains the question: Will improvement of flood warning technology and education programs necessarily lead to a significant change in behavior?
A timely issue that should be addressed by researchers is the compounding of disasters. Compound flooding (pluvial, fluvial, and storm surge) can increase the human toll of flooding events. However, flooding can be combined with other life-threatening occurrences such as power loss and interruptions of health and emergency services. The COVID-19 pandemic is the latest example of compounding disasters due to its significant impact on medical and emergency services and the lockdown requirements that can greatly restrict mass evacuation and mass sheltering that might become necessary during a flooding event. Extreme weather warnings may need to consider these and other impacts of a concurrent pandemic and direct individuals on what to do during a compound extreme event. Emergency officials should also be aware of and prepared for compound disasters. Disaster kits should now also include supplies recommended by the US Centers for Disease Control and Prevention to help prevent the spread of COVID-19, such as facemasks, gloves, and hand sanitizer.
Finally, there is evidence that major tropical storms have been increasing in the last four decades (e.g., [59
]), which underscores the need for prone communities to adapt to the shifting climate and weather realities. Flood zones are being expanded, placing financial burden on local governments and causing some residents to leave the affected localities altogether, resulting in loss of taxes and/or loss of eligibility for state and federal assistance due to the decrease in population. Moreover, repeated disasters prompt some relatively wealthier people to migrate, leaving citizens with lower capacity to adapt behind (e.g., [60
]). Multidisciplinary research is needed to better understand this problem and develop appropriate policies and interventions.