Natural hazards fatalities in Brazil, 1979–2019

The impact of natural hazards on nations and societies is a global challenge and concern. Worldwide, studies have been conducted within and between countries, to examine the spatial distribution and temporal evolution of fatalities and their impact on societies. In Brazil, no studies have comprehensively identified the fatalities associated with all natural hazards and their specificities by decade, region, sex, age, and other victim characteristics. This study carries out an in-depth analysis of the Brazilian Mortality Data of the Brazilian Ministry of Health, from 1979 to 2019, identifying the natural hazards that kill the most people in Brazil and their particularities. Lightning is the deadliest natural hazard in Brazil during this period, with a gradual decrease in the number of fatalities. The number of hydro-meteorological fatalities increases from 2000 onwards, with the highest number of fatalities occurring between 2010 and 2019. Although Brazil is a tropical country affected by severe droughts, extreme heat has the lowest number of fatalities compared to other natural hazards. The period from December to March has a higher number of fatalities, and the southeast is the most populous region where most people die. The number of male victims is twice as high as the number of female victims, across all ages groups, and unmarried victims are the most likely to die. It is therefore essential to recognize and disseminate the knowledge about the impact of different natural hazards on communities and societies, namely on people and their livelihoods, in order to assess the challenges and identify opportunities for reducing the effects of natural hazards in Brazil.


Introduction
Natural hazards, a process, phenomenon or human activity that may cause loss of life, injury or other health impacts, property damage, social and economic disruption or environmental degradation (UNDRR 2020) are recurring phenomena (WMO 2021). They can be, for example, a violent storm over an uninhabited region with no human consequences and can be triggers to disasters, to serious disruption of the functioning of a community or a society (UNISDR 2017). The increase in frequency and intensity of climatic extremes worsen the severity of natural hazards (IPCC 2022;2021) and, consequently, the challenges Extended author information available on the last page of the article 1 3 that social systems must manage, especially people who experience vulnerabilities in their daily lives .
Death due to natural hazards is the ultimate consequence and not only is associated with the nature and extent to which a system is exposed to a natural event but also linked to socioeconomic and environmental contexts and conditions (IPCC 2022). In Brazil, the number of individuals affected by natural hazards is high. Between 2001 and 2021, approximately 53 million people were affected by landslides, storms, droughts, floods, and flash floods (EM-DAT 2021).
Despite the number of deaths due to weather-, climate-, and water-related disasters decreasing worldwide from 1970 to 2019 (WMO 2021), two essential questions require answers: What is the most dangerous category of natural hazards? Which individuals are the most fatally affected by natural hazards? Thus, this study aims to provide, for the first time in Brazil, insights into temporal evolution, spatial distribution, and induced fatalities characteristics differentiated by sex, age, and marital status in the Brazilian population from 1979 to 2019.

Studies on natural hazards fatalities
Within the broader field of disaster and environmental change research, studies have been conducted on natural hazards impacts, measuring and qualifying fatal victims and their death circumstances (Froud and Petley 2018;Mahapatra et al. 2018;Salvati et al. 2018;Diakakis and Deligiannakis 2017;Badoux et al. 2016;Zagheni et al. 2015;Neumayer and Plümper 2007). Many empirical studies on mortality due to natural hazards have been conducted from perspectives based on different scales of analysis, differentiated by the category of natural hazards and/or focused on a single event.
• Global scale studies Global studies have estimated human losses, identifying the impacts of all categories of natural hazards on the human population and its subgroups, finding global trends. In the 1990s, Sapir (1993) provided insights into patterns and trends in disaster mortality and morbidity generated by natural and man-made hazards, affirming that they occur neither equally nor at random. Neumayer and Plümper (2007) indicate economic, cultural, and socially constructed gender-specific vulnerabilities as responsible for different strands of global male and female mortality.
On the global scale, some studies have focused on a specific category of natural hazard, evaluating its possible mortality patterns, its reasons, and the most affected subgroups. Zagheni et al. (2015) correlated the level of economic development of nations with the age and sex of fatal victims of hydro-meteorological disasters. Froud and Petley's (2018) analysis of global fatal landslides indicates a strong seasonal rainfall pattern throughout the annual cycle, and the most prevalent fatalities occurred in densely populated urban centers in the poorest countries. Likewise, Doocy et al. (2013) suggested, based on a historical review of floods from 1980 to 2009, a global increase in the frequency of flood events, and mortality varying annually and often concentrated around large-scale events. Holle (2016) analyzed how global lightning fatalities differ between developed and lesser-developed countries and suggested higher education and awareness as measures to reduce the number of deaths. Regarding the impact of extreme temperatures, a comparative multi-country analysis was conducted in Latin American cities to assess whether the risk of heat-related mortality is associated with age, sex, and disease status, amplified by social disparities (Bell et al. 2008). The results of the global analysis of natural hazards fatalities have demonstrated what is already known: the poorest nations and the most disadvantaged populations are those that suffer most severely (Hamza 2015) and have the largest number of deaths .
• National scale, multi-hazards Research within a given country allows comparison of mortality data among different categories of natural hazards (i.e., floods, landslides, lightning, storms, hurricanes, extreme temperatures as cold and heat waves), and the identification of the category that kills the most people and the regions that are most affected. Moreover, sociodemographic characteristics of the victims can be assessed.
An analysis by Thacker et al. (2008) highlights that extreme temperatures in the USA cause more deaths than other natural hazards. Despite this information, the amount of media attention and disaster relief resources is much higher for hurricane, storm, flood, and tornado events than for extreme temperatures. A study of Mexico (Jauregui-Dias et al. 2019) identified that from 2000 to 2015, the mortality from extreme weather events in Mexico remained constant. The differences between sex and age of the fatalities vary by the type of disaster and because of the sex division of labor. Mahapatra et al. (2018) analyzed Indian data on mortality by category of natural hazard from 2001 to 2014 for the states and union territories of India and found no consistent pattern in death rates. Extreme weather events vary in magnitude and region annually, and males are at greater risk of death than females due to their participation in outdoor work. Myung and Jang (2011) analyzed for South Korean the causes of fatalities due to meteorological disasters, the demographical characteristics of the victims, and regional distribution and found that the majority of fatalities are caused by floods due to drowning and that the place of the fatal accidents is distinct for male and female victims. Furthermore, a study on natural hazards fatalities in Switzerland, from 1946 to 2015 (Badoux et al. 2016), recognized that most fatalities occur due to avalanches, affecting male and female victims differently. This study also recognizes that improved forecasting, process detection, and warning systems are mainly responsible for the reduction of the annual natural hazards fatality number.
• National scale, hazard specific Other authors have investigated natural hazards mortality through multi-hazards lenses, and a group of studies has focused on mortality due to specific categories of natural hazards at the national scale, with a long-term dataset, or has investigated specific events. Coates (1999) was the first to publish a study on flood fatalities in Australia from 1788 to 1996 and found a reduction in fatalities due to improvements in warning systems and rescue services. Ashley and Ashley (2008), based on US mortality data from 1979 to 2004, attributed the high number of flood fatalities to "people walking purposely" through water. A study of Greece demonstrated the circumstances under which most of the flood fatalities, from 1970 to 2009, occurred: during nighttime, in outdoor areas, mostly vehicle-related, and from drowning . A study of landslide and flood fatalities in Italy from 1965 to 2014 correlated the age and sex of victims to the circumstances of their deaths (Salvati et al. 2018).

• Natural hazard event specific
Additionally, case studies have assessed the singularities of the cyclone of 1991 in Bangladesh (Bern et al. 1993), the 2003 heat wave in France (Poumadere et al. 2005), the 2002 landslide in Chuuk, Micronesia (Sanchez et al. 2009), and the 1931 hurricane, and floods in Fiji (Yeo and Blong 2010), and each one has it single particularities.

Natural hazards in Brazil
Brazil is the largest country in South America and fifth in the world, with an area of 8,510,295,914 km2 and a population of 210,147,125 people (IBGE 2019). Located mostly in a tropical area with a hot climate, with a large variation in precipitation, from large accumulation to rainfall deficits, along the seasons and through the national territory (Debortoli et al. 2017). It is geopolitically divided into 26 states and one federal district, grouped into five regions ( Fig. 6): the North, characterized by an Equatorial climate and a population density of 4.72hab/km 2 ; the Northeast, with Equatorial and Tropical (humid and semiarid) climate with 36.26hab/km 2 ; the Central-West, Tropical humid climate and 10.01hab/ km 2 ; the Southeast, with highlands and humid climate, with 92.02hab/km 2 , and the South region, with subtropical climate and 52.09hab/km 2 (CEPED 2019). Its continental dimensions, the regional characteristics of biomes and socioeconomic inequality are conditioning factors for the impact of extreme events in the country (Perez et al. 2020).
Brazilian mortality due to natural hazards is characterized by storms, floods, flash floods, landslides, droughts, debris flows, heatwaves, extreme cold, tornados, erosion, forest fires (CEPED 2012), and lightning (Cardoso et al. 2014). In Brazil, extreme weather events and their medium and long-term impacts, materialized in diseases and health problems, are poorly evaluated and are misunderstood, and mortality associated with them is misregistered as a consequence of hazardous events (Xavier et al. 2014). Material damage to the health sector's service structures and functioning is also a consequence of natural hazards (Londe et al. 2018), which limits the capacity to account for an increased number of fatalities and fulfill demands (Freitas et al. 2020), with consequent underreporting of mortality data.
Crucial questions require answers. Which groups are fatally affected by natural hazards in Brazil? Are they distinct within the country's regions? Is the fatality data set influenced by the sex and age characteristics of the victims? Do victim's characteristics have patterns? Using a robust database of fatalities allows for an appropriate assessment of extreme events impacts (Petley 2012), an analysis of the spatial distribution and temporal evolution of fatalities (Pereira et al. 2016), and the sociodemographic characteristics of the victims and circumstances surrounding fatalities. The knowledge on income disparities, class, level of literacy, cultural references, health, disabilities, gender, and other characteristics are central to comprehending social vulnerabilities to manage the multiple threats (Fordham et al. 2013).
This study aims to assess natural hazards fatalities that occurred in Brazil by analyzing a mortality dataset, from 1979 to 2019, from the Brazilian Ministry of Health. By using this database, this study examined the fatalities distribution in the national territory, their periodicity, and the number and characteristics of the affected victims. It aims to recognize and make public the knowledge of different natural hazards' impacts on communities and societies, specially people and their livelihoods. Additionally, the study contributes to the understanding of the populations at risk in disasters in Brazil and supports official governmental structures to reduce human fatalities and impacts due to natural hazards.

Brazilian mortality information system
Data on mortality have been registered in Brazil since 1975 by the Ministry of Health. The Brazilian Mortality Information System (SIM) was created following the standards of the International Classification of Diseases (ICD) of the World Health Organization (WHO), and its objective is collecting, registering, and disseminating quantitative and qualitative data on mortality in Brazil. According to the WHO, the ICD is an essential tool that allows the world to report and compare mortality data in a standardized manner, supporting evidence-based decision-making.
This Brazilian system is filled out with microdata reported on death certificates where causes of death are certified by a physician or medical examiner, along with up to 20 contributing variables, e.g., age, sex, residence municipality of the deceased, date of death, and others. These data are classified according to a standard set of codes: deaths from 1979 through 1995 are classified using the Ninth Revision of ICD- ICD-9 (WHO 1977), and provide information on the probable cause of death of Accidents Due To Natural And Environmental Factors, and deaths for 1996 and beyond are classified using the Tenth Revision (ICD-10) and present data classified as Exposure to Forces of Nature (WHO 2010). Both International Classifications, ICD-9 and ICD-10, only consider primary data fatalities that are a direct result of a hazardous event.

Brazilian natural hazards data
According with the Brazilian SIM, and using the categories and codes of the ICD of the World Health Organization, Table 1 describes the hazards categories that are contained in the analysis and phenomena included and excluded in each category. They are Lightning, Excessive Heat and Excessive Cold, Storms, Floods and Landslides. Deaths due to volcano activity are also presented in the database (0.02%) but are ignored in this analysis knowing it as a problem of data registration.
Data from 1979 to 2019 contain seven variables that allow recognizing the victims characteristics, place of death, day and cause of fatality that are: natural hazard category, death municipality, date of fatality, sex, age, locus of occurrence, and marital status ( Fig. 1). As a result, all categories can either have "Other", which represents "none of the options", and "Ignored", which are "not known information". The "NA," the nonapplicable variable, represents "no data value stored" and/or "improperly or mistakes information" and was excluded from the analysis.

Methods
Based on the preceding literature and to accomplish the desired analysis, this study used fatality data published on the Brazilian Mortality Information System by the Ministry of Health. The analysis of fatalities related to natural hazards in Brazil, from 1979 to 2019, was conducted in the following steps: Step 1. Separate by category of natural hazard: (i) lightning, (ii) excessive heat, (iii) excessive cold, (iv) storms and flood events, and (v) landslides. These five categories were defined based on original data grouping, allowing to compare the number of fatalities by category of natural hazards, and its increment through the years.
Step 2. The five categories were merged into three major categories of natural hazards: (a) Lightning, kept as a single category; (b) Extreme Temperatures, by grouping excessive natural heat and cold, and (c) Hydro-meteorological hazards, by grouping storms and floods together with landslides, as shown in Fig. 1. Merging storms, floods, and landslides into a single major category "Hydro-meteorological hazards" was done following the expert's premise that in Brazil, many of the landslides, mainly wet mass, are a direct result of frequent and intense precipitation (IBGE 2019; Lacerda et al. 2017;OPAS 2015) and that most fatal landslides are rainfall-triggered (Froud and Petley 2018). Analyses of the date of occurrence of fatalities and their spatial distribution were based on these three major categories defined in Step 2.
Step 3. The analysis of multidimensional identities of fatal victims was conducted for the two major categories, Extreme Temperature events and Hydro-meteorological hazards only, and not for the major category Lightning. This decision was based on the fact that lightning studies and demographic analysis have already been conducted by Cardoso et al. 2014   The central category of analysis is the cause of fatality. Causes, e.g., exposure to sunlight and other unspecified forces of nature were not considered in this analysis, as neither sequelae nor other and unspecified effects of external causes of fatalities. Fatality data associated to Dam collapses were also excluded from the study recognizing that it is not caused by a natural phenomenon.
The study was conducted based on the principles of Exploratory Data Analysis (Wickham and Grolemund 2016) and used the R language and environment for statistical computing and graphics (R Core Team 2022) program for Data Science Analysis, enabling the conversion of mortality data into knowledge.

By category of natural hazards
An analysis was conducted of mortality data due to natural hazards reports on 8716 fatalities in Brazil from 1979 to 2019. Lightning is the natural hazard that kills the most people in Brazil (57.8%), followed by landslides (20.7%), and storms and floods (15.7%). For Brazilian fatalities, extreme temperatures represent 5.8% of the total deaths, caused by excessive cold (4.1%) and excessive heat (1.7%).
The high percentage of fatalities caused by lightning in the total number of fatalities follows the global pattern in which lightning accidents in lesser-developed countries are associated with working conditions in non-safe lightning structures (Holle 2016;Cardoso et al. 2014). Despite the increase in the frequency and density (flashes km −2 year −1 ) of lightning events in Brazil (Santos 2017;Pinto Jr. et al. 2007), a slow reduction in the number of lightning fatalities from 1979 to 2019 (Table 2; Fig. 2) occurs similarly to other nations (Holle 2016). This is possibly associated with urbanization and the increase in people spending their time inside lightning-safe structures (Badoux et al. 2016). The reduction of fatalities can also be associated with the improvement in medical care, and emergency communication and transportation (Ashley and Ashley 2008).
Events associated with landslides occur irregularly and have an increasing pattern. The increasing pattern of mortality starts in 1995, when the number of fatalities per year increases (Fig. 2) reaching it highest number of fatalities in the last decade of the study period (Table 2). Storm fatalities, associated with floods events, abruptly increase from 2000 to 2019 (Table 2; Fig. 2). An increment occurs due to a single outlier event: the 2011 Serrana Region tragedy, a single catastrophic event characterized by 24 h of torrential rain on January 11 and 12, 2011, that culminated in more than 905 deaths, over 300 missing persons, and affecting more than 300 thousand people, with tens of thousands of displaced and homeless people (Alvalá and Barbieri 2017 (2021) diagnoses flash floods, riverine floods, and general landslides as major responsible for natural hazards fatalities in Latin America. Extreme rainfall is recognized as the leading trigger of landslides (Froud and Petley 2018) and floods (Doocy et al. 2013), and the rising mortality trend is a result of high intensity storms associated with rainfall seasons (Petley 2012). The gradual and continuous rising trend observed in the data analyses reveals the role that storms and landslides play in increasing overall fatalities in Brazil from 1979 until 2019 (Fig. 3).
The analysis of the Cumulative Number of Fatalities due to different categories of natural hazards (Fig. 3) reveals that years that have peaks of fatalities (2008,2010,2011) are associated with landslides events, and the passage of high intensity rain in specific regions. They are listed in Table 3 (for more info see Sect. 4.5): Landslides in Brazil are characterized by few occurrences but are responsible for extensive impacts. They are responsible for a high number of fatalities, and are the main responsible for property damages, affecting mainly infrastructure, housing, education, and health structures and facilities (CEPED 2019). Despite the lower number of fatalities due to extreme heat and cold, these categories are the ones that most affect the country, over time and throughout the national territory, and produce greater numbers of damages and losses that affect agriculture, livestock and industry, and public services such as water supply and sanitation, health system, and energy supply (CEPED 2019).The frost phenomenon in Brazil is well-known and has been studied by different Brazilian scientific institutions since the 1980' (Algarve 1996). Cold fronts are recognized for causing heavy rainfall (Seluchi et al. 2017) and the steeply temperature decline, causing strong freezes (Satyamurty et al. 2002) over different regions in the Brazilian territory, affecting agriculture and production of maize, beans, coffee, sugar cane, fruit, and also vegetables (INMET 2022).
Studies recognize the positive association between temperature variability (TV) and hospitalization (Moraes et al. 2022;Zhao et al. 2018) and also mortality (Bitencourt et al. 2020;Guo et al. 2018) which both cold and hot temperatures are associated with increased risk of health outcomes and mortality in all season in multiple countries/regions  due to respiratory and circulatory, cerebrovascular and cardiovascular causes (Perez et al. 2020), but not as direct impact of extreme cold.
Despite the increase in the average annual temperature from 1961 to 2018 in Brazil (Ramos et al. 2018) and heat waves that affected Brazil as a whole since 2005 (Nobre et al. 2019), the analysis indicates that only 1.7% of the total fatalities are due to exposure to excessive natural heat. Excessive natural heat, classified as a heatwave, is as a silent mortal natural hazard and a worldwide leading killer among natural hazards (IFRC 2020;Poumadere et al. 2005). Tens of thousands of deaths due to heatwaves, over the last three decades, have been reported worldwide, but the reported deaths can be difficult to establish. Also, it is hard to estimate how heat exposure exacerbates other preexisting health conditions (WMO and WHO 2015;Hajat and Kosaky 2010).
Brazilian studies have demonstrated the direct association between temperature rise and the intensification of heat stress conditions (Bitencourt et al. 2021); mortality due to the increase in vector-borne, water-borne, and food-borne diseases; and cardiovascular and respiratory diseases (Hacon et al. 2018). All these mortality data are not qualified as a direct impact of extreme heat and represent a plausible explanation for why the number of fatalities due to exposure to excessive natural heat is so low. There is either the hypothesis that populations in warm climates are adapted to high temperature (Son et al. 2016) and there is a possible adaptive response to a hot environment ("acclimatization") in which an individual "learns" to better tolerate exposure to excessive heat (WMO and WHO 2015). For Brazil, there is no scientific evidence suggesting the human adaptation to temperature variation exposure and adverse health outcomes to temperature change under climate change scenarios (Zhao et al. 2018;Guo et al. 2018).

By date of occurrence
From now on, the analysis run considering the three major categories of natural hazards defined at Step 2: i. Lightning, ii. Extreme Temperatures, and iii. Hydro-meteorological hazards. The monthly distribution of fatalities registered in Brazil from 1979 to 2019 indicates January as the month with a disproportionately high number of fatalities due to natural hazards (Fig. 4). Hydro-meteorological hazards (60.2%) and lightning (38.7%) are responsible events for the high number of deaths. The month with lowest number of fatalities is August, in which extreme temperatures increase, namely, 16.2% of total deaths. During July, extreme temperatures are the most responsible for fatalities (43.5%), followed by hydro-meteorological hazards (30.3%). Lightning events reach the smallest number of fatalities in July (Fig. 4) because this month represents the peak of the dry season, and is characterized by fewest lightning strikes. The seasonality analysis of fatalities displayed in Fig. 5 also presents the period from late December (21st) to March (20th) (Southern Hemisphere (SH) Summer) with the highest number of fatalities. For that period, lightning hazards are responsible for the majority of fatalities (54%), followed by hydro-meteorological causes (45.2%). The period from late March (21st) to June, 21st (SH autumn) hydro-meteorological events are responsible for higher number of fatalities (47%) than lightning (40%).
In Brazil, the different seasonal distribution of precipitation varies along the region and period of the year, varying from wet to dry conditions. The SH summer is the season with Fig. 4 Monthly distribution of natural hazards fatalities (cumulative data over the studied period) the most significant rainfall in the Southeast region. It is characterized by isolated heavy rains (high intensity with a short duration; CEMADEN 2019), and the main characteristic is the occurrence of rains for several days (CPTEC 2021). The SH autumn is the rainiest period in the Northeast region and, despite being mainly characterized by a dry climate , is when extreme precipitation occurs and when the highest number of fatalities due to extreme precipitation is concentrated. Each region has particularities in fatalities, and they are presented in Table 4.

Spatial distribution
The analysis of fatalities due to the natural hazards of the different categories in the Brazilian territory shows a predominance of fatalities in the Southeast (47.7%). The South concentrates 20.2% of fatalities; the Central-West, 13.2%; the Northeast, 10.2%; and the North region has the lowest percentage (8.7%) (Fig. 6).
Lightning is the deadliest hazard in almost all Brazilian regions, with the exception of the Southeast, where hydro-meteorological hazards are responsible for more than half of fatalities (Table 4). In the Central-West, the number of fatalities associated with lightning events is disproportionally high with an annual death rates the average probability of a person being struck by lightning and dying (Cardoso et al. 2014).
Fatalities associated with hydro-meteorological events occur predominantly in the Southeast: ranked first is the state of Rio de Janeiro (RJ), followed by Minas Gerais (MG), and São Paulo (SP) states. According to Froud and Petley (2018) the number of  fatalities is strongly related to seasonal rainfall regimes, but other studies demonstrate that the spatial distribution of fatalities is not exclusively linked to rain. In Portugal, the principal mortality hotspot is related to high population numbers and natural conditions (Pereira et al. 2016). In the USA, the spatial distribution of flood fatalities is also linked with heavily populated urban centers and steep topographical (Ashley and Ashley 2008). They are also associated with individuals' behavior: fatalities are aggravated by deliberate decisions regarding risk-taking, which may vary by sex, age, and circumstances (Zagheny et al. 2015;Doocy et al. 2013).
In Brazil, the most affected regions are often those with the greatest population density, and the poor quality of urban infrastructure services, combined with intense rainfall regimes (Perez et al. 2020). The Southeast, in which most of Brazil 's population is concentrated, is where the hydro-meteorological fatalities are concentrated. The states of Santa Catarina (SC) in the South region and Pernambuco in the Northeast, occupy the fourth and fifth positions, respectively. The state of Pará (PA), in the North region, is in the sixth position, with a high number of hydro-meteorological fatalities due to landslides (Fig. 7). The sixth position of Pará State is due to landslides in the Itaituba municipality, recorded since 2006. A very brief data analysis of Itaituba' fatalities indicate that 63% of the fatal victims are gold miners and all of them males.
Lightning fatalities are concentrated as follows: in the Southeast region, SP and MG; in the South region, Paraná (PR) and Rio Grande do Sul (RS); and in the Central-West region, Goiás and Mato Grosso do Sul (MS). Those six states are responsible for 66.5% of all fatalities associated with lightning in Brazil from 1979 to 2019, and are associated with open space labor activities and is related to a high percentage of the population being involved in rural activities (Cardoso et al. 2014). Brazil is also affected by extreme droughts related to extremely hot temperatures and the distribution of rainfall. The Amazon region presents a natural climate variability, producing droughts and floods (Marengo and Espinoza 2016); however, recently, it is presenting changes in its rainfall conditions (Perez et al. 2020), experiencing two extreme dry spells (2005 and 2010 droughts) and three floods, in 2009floods, in , 2012floods, in and, 2021floods, in (Espinoza et al. 2022Satyamurty et al. 2013;Marengo et al. 2011). The Northeast have been the most affected by high intense droughts from, at least, 2000 (Cunha et al. 2019;Cuartas et al. 2022); but no fatalities were associated with these high intense droughts in the database.

Multidimensional identities of fatal victims and circumstances of death in Brazil
Studies of the demography of lightning fatalities in Brazil have been developed by the ELAT group-the Brazilian Atmospheric Electricity Group of the National Space Research Institute (INPE)-a global reference research group specialized in lighting studies and providing scientific knowledge on lightning and casualty demographics in Brazil. Cardoso et al. (2014) is the principal reference, and their studies are updated each year and published on their platform ELAT (2019). Their scientific knowledge indicates that on average, 78 million cloud-to-ground lightning strikes occur in Brazil per year. The majority of fatalities due to lighting occurs in rural areas (26%), followed by those inside a home (21%), in which the victim is on the phone, next to electric equipment, and/or close to windows and doors (Cardoso et al. 2014). More than threequarters of fatal victims due to lightning are males (82%), and the most affected group is aged from 20 to 29 years. The fatal victims aged less than 10 years and older than 70 years are few, representing 4% and 2%, respectively, in Brazil (ELAT 2019).

Sex and age analysis
Of the total number of fatalities due to natural hazards in Brazil from 1979 to 2019, male victims account for 6.418 (representing nearly 74%). When disaggregated by extreme temperatures and hydro-meteorological hazards, males account for 63.5% of fatal victims. Comparatively, male victims are the majority within every age group, except for the 85 + age group (Fig. 8). In the female group, the most fatally affected are under 20 years old, representing 36.2% of the total female fatal victims of natural hazards events in Brazil from 1979 to 2019. The female newborns, infants, and toddlers (under 5 years old) have the highest number of fatalities (Fig. 8). The group comprising females aged up to 75 years accounts for 4.8% of the total female fatal victims. For the male fatalities, those aged between 30 and 50 years represents 34.1% of the total male fatal victims; those among 40 and 44 had the highest number of fatalities. The male group aged more than 75 years represents 3.2% of the total victims. The number of female fatalities associated with extreme temperature events differs very much from the male victims. For every 100 fatalities due to extreme temperatures, 83 are male victims. This large difference between male and female fatalities is the highest for adults, especially for the age range 40-60 years (Figs. 8,9a). When associated with hydrometeorological events male fatalities represent 61% of the total (Fig. 8).
Are these characteristics of the Brazilian fatal victims of natural hazards consistent with the literature's findings? Zhao et al. (2018) study analyses the association between temperature variability (TV) and hospitalization in Brazil (2000Brazil ( -2015 and their findings indicate that in Brazil men 10-49 year old are more affected by exposure to TV than women. They affirm that this may be explained by male greater exposure to TV (e.g., higher level of outdoor activities) than female. Some authors also support the hypothesis that for certain types of disasters, females and males have a different propensity toward risk-taking. Many studies have associated the level of country income with the physical and behavioral Fig. 8 Age pyramid of number of fatalities by sex (female on the left and male on the right) and age group disaggregated by category of natural hazard; extreme temperature in green and hydro-meteorological in purple characteristics of the victims and their exposures to natural hazards. For high-income countries like Italy (Salvati et al. 2018), Greece , and Australia (Coates 1999), the higher number of males fatalities is justified by their propensity toward risk-taking, due to their risky behaviors. Alternatively, from another perspective, the higher risk perception of Swiss women is what leads them to exhibit more cautious and less adventurous behavior than men (Badoux et al. 2016). Another perspective in the literature has attempted to understand the social processes of gender inequalities and indicates that social norms and role behaviors, potentialized by socioeconomic status, are responsible for differences in the number of male and female fatalities due to natural hazards (Neumayer and Plümper 2007).

Locus of fatality
The majority of fatal victims of natural hazards in Brazil die at home (42.2%), and 16% die in a hospital. When desegregated by sex, locus of fatality, age, and category of natural hazards, the majority of female fatal victims affected by extreme temperatures (52.3%) die in a hospital, and female newborns, infants, and toddlers (under 5 years old) die the most. Adult males are likely to perish in a public place, and the adolescents and the elderly groups, in a hospital and at home (Fig. 9a). The male higher mortality associated to extreme temperatures is clearly apparent at Fig. 9a, and the major difference between male and female Fig. 9 a Number of fatalities disaggregated by sex, age group, locus of fatalities due to extreme temperature in Brazil, from 1979 to 2019 (please note that y-axis scale is different from b). b Number of fatalities disaggregated by sex, age group, locus of fatalities due to hydro-meteorological hazards in Brazil, from 1979 to 2019 (please note that y-axis scale is different from a) fatalities happens in the group between 40 to 49 years old, as suggested by Zhao et al. 2018.
When affected by a hydro-meteorological hazard, the majority of female victims die at home (58.5%), independently of their age. 38.1% of male fatal victims also die at home, and 28.8% of adults male's locus of fatality are qualified as "other," which indicates that they do not die at home, a public place, or at the hospital (Fig. 9b).
In Mexico, the risk of dying due to extreme weather events is associated with genderspecific division of labor, with men's work regularly occurring in open spaces (Jauregui-Dias et al. 2019). In India, a higher proportion of males than females participate in the labor force, and most work is performed outdoors (Mahapatra et al. 2018). Fiji´s worst natural disaster, the 1931 hurricane and flood, suggests cultural norms imposed differently on men and women are the main element that determined the distinctive patterns of fatalities (Yeo and Blong 2010). In South Korea, female death occurs mostly in residential areas, whereas for males, it occurs mainly outdoors (Myung and Jung 2011).
The age and locus of fatalities have also been a focus of research and are generally attributed based on the overrepresentation of the ages and their capacities. The high mortality among elderly individuals and young children is associated with their lack of physical strength and capacities (Bern et al. 1993;Coates 1999;Poumadere et al. 2005;Thacker et al. 2008;Yeo and Blong 2010;Doocy et al. 2013) or awareness (Sanchez et al. 2009).

Marital status analysis
Per the Brazilian Institute of Geography and Statistics (in Portuguese: Instituto Brasileiro de Geografia e Estatística; IBGE), the marital status analysis is on individuals aged 15 years and above. From those whose marital status was registered (14.5% were not available), the majority of fatalities were single (48.7%). Married victims represent 35.2%of the total and separated victims 4.8% and widow represent 6.7% each one. The marital statuses of 4.6% of the total victims were ignored. The analysis desegregated by marital status and sex generates a notable and singular finding. In Brazil, despite the number of male victims to be bigger than the female group, widow number of fatalities of the female victims (69.8%) is higher than the male group (30.2%) (Fig. 10). The analysis desegregated by marital status, sex, and category of natural hazards reproduces the same pattern; the number of widow female victims is higher than the male group on both categories of natural hazards. Our study reveals that in Brazil, males are the majority of fatal victims due to natural hazards. The female youth group (aged under 20 years) and female babies and infants are the most affected within the female group. Elderly individuals of both sexes account for a small percentage of the total. Single males and females represent more than half of the victims, but female widows are highly affected. The majority of the fatal victims die at home, and "other" is an important locus category where male victims affected by hydrometeorological events die much more than their female counterparts.
In Brazil, the last 30 years have seen important changes in the position of women, with an increase in their educational level and their increased participation in the labor market. There is also a growing number of families headed by women (Cavenaghi and Alves 2018), however, gender inequalities are still a structural feature in Brazil. Social processes of gender inequalities (Neumayer and Plümper 2007) could justify a large number of female fatalities due to natural hazards, but in Brazil, it is mostly men who die as a result of natural hazards. The "risk-taking behavior" hypothesis (Salvati et al. 2018;Diakakis and Deligiannakis 2017;Ashley and Ashley 2008;Coates 1999) could explain the higher number of male fatalities, but the majority of the male victims due to hydro-meteorological events die at home. The "division of labor" hypothesis (Jauregui-Dias et al. 2019;Mahapatra et al. 2018) can justify the higher number of male fatalities from extreme temperatures than female, but do not explain why adults male perish in a public place and at "other" place than the hospital or at home. Natural hazards and extreme events can open spaces for understanding vulnerabilities, and also capabilities to deal with it (Moreno and Shaw 2018), but further investigation is needed to understand these findings.

Natural hazards and Brazilian disasters
Natural hazards are induced by natural phenomena and occur daily. Are multi-category and affect the population and territory differently over time causing losses and damage. They are recognized as leading triggers that increase and worsen the preexisting socioeconomic conditions of vulnerable communities and provoke disasters (Hummel et al. 2016). Disaster is conceptualized as "a serious disruption of the functioning of a community or a society at any scale due to hazardous events interacting with conditions of exposure, vulnerability, and capacity, leading to one or more of the following: human, material, economic and environmental losses, and impacts" (UNISDR 2017).
In Brazil, approximately 85% of disasters are related to too much or too little rainfall. Growth in the number of occurrences of disasters has occurred, either due to the intensification of natural hazards or the increase in vulnerable regions exposed to weather extremes (Alvalá and Barbieri 2017). Southern Brazil is the region most exposed and vulnerable to disasters triggered by extreme rainfall (Marengo et al. 2021), and the Northeast is most affected by too little rainfall (Cunha et al. 2019).
Droughts are widespread throughout the Brazilian territory, causing a conjunct of enormous impacts, affecting more people than any other natural hazard  and provoking the highest impacts on economic and social structures in Brazil (Cunha et al. 2019). Droughts generate, as the first stage, a reduction in water supply, affecting agriculture, industry, and energy generation (Nobre et al. 2016) and health care facilities, schools, and other essential services (Freitas et al. 2020). The impacts on human health are also multiple and extend to the medium and long term, worsening existing diseases and facilitating the development of new diseases (OPAS 2015).

3
The herein-developed analysis of hazards fatalities in Brazil from 1979 to 2019 identifies the many fatalities associated with storms, floods, and landslides triggered by heavy rain. Table 5 presents a list of selected Brazilian large-scale disasters (CEPED 2019) that affected more than 2 K people each, had declared state of emergency and required national assistance, and their characteristics: the per disaster number of fatalities and the number of displaced and homeless victims. There are some other important Brazilian disasters, such as the 2010 floodings in Alagoas and Pernambuco (CEPED 2019), storms in the municipalities of Tubarão (SC) in 2016 and São Francisco de Paula (SC) in 2017, the 2022 Petrópolis landslides (CNM 2022), but they are not shown in Table 5 due to a lack of available data. Figure 11 presents a brief analysis of mortality data, with the identities of the victims and circumstances of deaths for all disasters in Table 5. The first column in the graphics named "Brazil" represents the profile of Brazilian victims due to hydro-meteorological hazards from 2000 to 2019.
The result presents no standards on victims' profiles. The analysis of disaster mortality data reveals that each disaster has its distinct patterns. Each disaster presents different fatal victims characteristics and can be associated, i.e., with the timing of occurrence, when individuals are usually at home sleeping, as the major influence over mortality patterns (Carmo and Anazawa 2014). Mortality patterns can also associated with the magnitude of the disaster, and all inhabitants are equally affected, beyond their social vulnerabilities (Cardozo 2018). The place where the disaster event occurs also affects outcomes, i.e., the Angra dos Reis disaster in which the victims were in a lodge celebrating New Year's.
A comparative analysis of each disaster and Brazilian hydro-meteorological hazards fatalities from 2000 to 2019 present different patterns for age category, locus of fatality, sex, and marital status (Fig. 11). The sex data analysis indicates that in most of the disaster situations here presented, the number of males and female fatalities is very similar. By contrast, the number of female mortalities in large-scale disasters worldwide is much higher than that of males (UN 2018). For Cyclone Nargis (Myanmar) in 2008, female fatalities were 61% of deaths; in the 2004 Indian Ocean Tsunami in Banda Aceh (Indonesia), 70%; and in Cyclone Gorky in Bangladesh in 1991 (CRED-UNISDR 2015), 91%. These analyses incite new questions and demonstrate the need for new research for Brazil. In addition to fatalities, there is the need to recognize that the severity of climate change is more than "the number of fatalities." Climate change is underway (IPCC 2012, IPCC 2021, IPCC 2022, and particularly for Brazil, extreme climatic events namely heavy rains, droughts, and floods will become intensely and frequently (Marengo 2014), and consequently, will amplify the impacts and challenges that social systems meet, especially for individuals who already experience vulnerabilities in their daily lives (Perez et al. 2020). These ongoing climatic changes can also exacerbate social inequalities and the consequences for public health, promoting the development of water-borne and vector-borne diseases, specially post-traumatic psychosocial problems and cardiovascular and respiratory diseases (Hacon et al. 2018).

Data bias
The use of secondary data requires careful consideration. The quality of the database and the quantity of registered data can vary. Qualitatively, the causes of deaths can be misjudged because of medical uncertainties; thus, what is declared on a death certificate might not be the cause of death. They can also be affected by coding errors because personnel are sometimes insufficiently qualified for this type of activity (Laurenti et al. 2008). Multiple causes of death can be associated with weather events, but they are rarely registered as a direct result of it (Xavier et al. 2014).
Quantitatively, differences in the two classification systems of the cause of death can also be responsible for data variation. The ICD 09, used to qualify mortality data from 1979 to 1995, and the ICD10, from 1996 up to now, use a distinct standard set of codes, and this difference might be responsible for the increment in the number of mortality data of ICD10 when compared with ICD09 data (Thacker et al. 2008).
The inconsistent differences in the amount of data within the same country may reflect what the literature calls geographical bias in the data sources (Pereira et al. 2017), that is, data fatality are reported and registered with different degree of accuracy, producing an unsatisfactory coverage in many municipalities in the country, and resulting in distortions in the proportionality of the reported causes (DATASUS/IDB 2000). This inconsistency can also be associated with the lack of skills in data collection (Petley 2012) in different regions of the country. There is also a need to recognize and consider that the passage of extreme natural hazards can damage local health establishments or disrupt the whole health service, affecting data fatalities' records and management (Xavier et al. 2014).

Conclusions
Although climate change is often framed as a global problem for all humanity, the heterogeneity of its manifestations, impacts, and response strategies must be carefully considered. The first challenge is to identify the individuals who die from natural hazards and the elements that make some residents within a community more vulnerable and/or better to anticipate, respond and recover from the impacts than others.
The use of a systematic approach and standardized protocols for collecting and registering mortality data due to natural hazards is fundamental to improve the assessment of the diverse aspects of multiple and different hazards processes. This will make it possible to anticipate the impacts and prevent potential consequences. However, international standards do not consider local realities, and constraints may not work equally for each country.
For Brazil, the classification of landslides as a geophysical event that is not associated with storms and floods may bias comparative studies on the global impact of water-related disasters. The detachment of extreme heat fatalities and the increment of long-term droughts also underestimates the impact that Brazilian droughts may have on people's lives.
This study demonstrates that national patterns of fatalities due to natural hazards in Brazil are associated with the region-across the national territory-and the period which they occur. Lightning is the deadliest natural hazard in Brazil, but the last decade of the study is marked by the sudden increase of hydro-meteorological fatalities as consequence of very few but large-scale disasters. The highest number of fatalities occurs in the Southeast region, and in the period from December to March, in areas of high population density and poor quality urban infrastructure services, combined with excessive rainfall. The number of male fatalities is higher for every category of natural hazards, and single men and women are the most likely to die.
Despite the uniqueness of each disaster situation in terms of context (e.g., the importance of the time, place and circumstances in which people live) and hazard typology, there is a need to recognize and correlate 'fragilities' with the 'capacities' of individuals, families, societies, and socioeconomic systems to cope with threats. This strengthens the potential resources, knowledge and leadership of individuals to manage hazards and build resilient places for living. According to United Nations women are agents and leaders of collaborative change in climatic extreme scenarios, developing strategies for and key roles in environmental management, disaster risk reduction, and climate change resilience. As long as the number of deaths caused by natural hazards continues to rise, there is room and opportunity for the responsible authorities to urgently implement new measures to deal with the threats posed by natural hazards.