This paper presents evidence of the geographical distribution of deaths due to lightning over Mexico for the period 1979–2011. Over 7300 deaths occurred during this period, an average of 230 per year, which translates into an average fatality rate per million inhabitants of 2.72 (1979–2011). A total of 60% of the fatal victims occur in only 7 out of the 32 states in Mexico, with the largest fraction occurring in the state of Estado de México (24%). The largest death toll is found in the young male population, in rural regions of the states of Estado de México, Michoacán, and Oaxaca, where the population density is low. The results have indicated a clear bias in the fatal victims toward boys and young males (under the age of 25), with more than 45% of the total deaths in that segment of the population. While female deaths constitute a small fraction of the total number, the under-25 age segment also has the largest number of fatal victims.
A county-level analysis of socioeconomic indicators clearly suggests that the geographical distribution of deaths is not correlated with population density nor with the maximum lightning density, but rather with vulnerability. The spatial distribution of deaths is better correlated with exposure to thunderstorms, agricultural activities, and low education levels. The large social vulnerability of those regions combined with the lack of recognition of the problem by society and the government are more likely responsible for the large death toll.
Lightning is one of the most spectacular weather phenomena and has fascinated people for centuries; several early societies associated the phenomenon with gods in their mythologies. But lightning constitutes a very real threat to society in modern times, leading to severe injuries and death and should be considered seriously by the population. Holle and López (2003) and Holle (2008) have estimated that up to 24 000 deaths occur globally due to lightning strikes, with an estimate of 240 000 injuries. These numbers were determined extrapolating from a death rate of 0.3 per million people in more developed countries, while other regions of the world were assumed an annual fatality rate of 6 deaths per million (Holle 2008). While some countries keep detailed records of deaths, others, particularly in the developing countries, have less comprehensive reporting, as indicated by Cooper and Kadir (2010).
Mexico is located in the tropics, with a marked rainy season where large cumulus clouds are prevalent. These clouds generate lightning affecting a large fraction of its territory (Christian et al. 2003; ,Kucienska et al. 2010).
The existence of a vulnerable population converts the natural lightning phenomenon into a serious risk (Wilches-Chaux 1993). It is necessary to explore the vulnerability of the society and to understand how it may become a determining factor in leading to casualties. Another paradigm of the concept of disaster develops around the lack of attention by government to small and moderate disasters, as opposed to huge disasters that may elicit international humanitarian response. Small or moderate disasters have a slow impact and occur at the local/county level. These are not perceived as disasters: they can accumulate over time and become significant, but the government does not appear prepared to take action to prevent these deaths. Measures such as the introduction of lightning injury prevention programs, as have been implemented in other countries, would seem logical to reduce the number of casualties. However, the study by Cooper and Kadir (2010) cautions that while it would be reasonable to suggest measures that have been successful in reducing the number of deaths in more developed countries to developing ones, often the socioeconomic and social factors that separate the countries in such categories are the ones that lead to the large number of casualties in developing countries and are, thus, nonapplicable.
The objectives of this study are (i) to analyze what appears to be a fairly unique dataset of lightning casualties in Mexico, collected by health care centers and hospitals throughout the country (not based on newspaper reports) for the period 1979–2011; (ii) to relate the spatial patterns of the natural threat to the spatial distribution of fatal victims; and (iii) to relate the fatal victims (by age, gender, ethnicity, and activities) to the patterns of social vulnerability in the country. In particular, we address whether the death toll is related to the threat of the natural phenomenon or to the vulnerability of the population and to possibly evaluate the role of government in reducing the number of injuries and fatalities due to lightning.
2. Datasets and analysis methodology
The data analyzed in this study are part of a government health database that contains the information on all deaths reported by hospitals and can be found online (SINAIS 2010). While many studies have relied on statistics based upon reports of lightning strokes in newspapers (Gomes et al. 2006; Cooper and Kadir 2010), this study uses the official Mexican government database, which includes information from 1979 to 2011. Two variables of the database were considered: (i) lightning stroke (code 907X) for the period 1979–97 and (ii) lightning stroke victim (code X33) for the period 1998–2011. It is important to mention that a change occurred in the reporting methodology after 1998 that included more categories for deaths and more information on the victim. The tables and graphs using this information were produced with the software DesInventar (La Red/CIESAS 2010), allowing the analysis from the national level, to the state and county levels. It is worth noting that other studies (Lopez et al. 1993) have pointed out that there is a severe underreporting of lightning fatalities. This underreporting may affect the statistical calculations and introduce a bias for which it is very hard to correct. The database used in this study is gathered from information generated at hospitals and health care centers located throughout the Mexican territory.
Additional information was obtained from official census data in Mexico, taken by the National Institute for Statistics and Geography (INEGI 2005, 2011a), for population and information on income, main activities, ethnicity, education levels, etc., that help support the assessment of the level of vulnerability of the population.
Finally, we also use a lightning database recorded by the World Wide Lightning Location Network (WWLLN) that detects the very low-frequency radiation (“sferics”) associated with lightning strokes. The network is currently formed by about 50 stations distributed around the world, located primarily at universities and research centers. WWLLN started operating in 2002 (Dowden et al. 2008) but only from 2005 onward did the detection efficiency reach an acceptable level (Jacobson et al. 2006; Abarca et al. 2010). Currently, the University of Washington coordinates the WWLLN network (http://webflash.ess.washington.edu/indexnew.html).
Kucienska et al. (2010) evaluated data derived from WWLLN, data from the optical transient detector (OTD) located in the OrbView-1/MicroLab satellite (Christian et al. 2003) launched in April 1995, and data from the Lightning Imaging Sensor (LIS) located in the Tropical Rainfall Measuring Mission (TRMM) satellite (Kummerow and Barnes 1998). From the comparison of these datasets from OTD/LIS and WWLLN, Kucienska et al. (2010) concluded that the same annual variability was observed over Mexico and the four adjacent oceanic regions considered in the study. While the detection efficiency of WWLLN is still very low, the dataset is considered valid, since it displays the same monthly variability as the 1995–2005 period of the OTD/LIS dataset.
3. Results of the analysis of the fatalities database: 1979–2011
a. Temporal evolution
A total of 7362 fatalities were reported in Mexico as “death by lightning stroke” during the period analyzed. Because of the changes introduced in reporting mentioned in section 2, we divided the statistics into two periods, yielding 5458 deaths between 1979 and 1997 (19 years) and 1904 deaths between 1998 and 2011 (14 years). These figures result in an average yearly death rate of 287 (with a standard deviation of 43) for the period 1979–97, and a decrease to only 127 (with a standard deviation of 54) in the more recent period. It is difficult to explain the large decrease observed in the more recent years, since there were no measures introduced by the government in 1998 to try to diminish the large death toll observed between 1979 and 1997. The more relevant information to discuss is the density of deaths per million inhabitants as shown in Fig. 1. The population in Mexico has increased from 64 million in 1979 to 114 million in 2011 (a linear increase was estimated from the information reported in the official national census carried out in 1970, 1980, 1990, 2000, and 2010). The first period indicates an average of 3.75 deaths per million inhabitants (with a standard deviation of 0.89), a value considerably lower than the value of 6 deaths per million estimated by Holle (2008) as representative of the density of fatalities in the less developed countries. The average density of fatalities in the second period decreases to 1.32 deaths per million (with a standard deviation of 0.50), but it is still over 4 times the value estimated by Holle (2008) for the developed countries (0.3 deaths per million people). However, the last two years in the database (2010 and 2011) report densities of 0.59 and 0.67 per million people, only about twice the value estimated for developed countries. Note that the density in both subperiods shows a clear interannual variability, which could be the result of a number of factors, including variability in the lightning strikes due to changes in atmospheric conditions. Unfortunately, we are unable to compare the deaths by lightning with the physical phenomenon for the whole period 1979–2011, since Mexico does not have a long record of measurements of cloud-to-ground lightning strokes. The interannual variability of the deaths by lightning time series of deaths appears moderately correlated to the La Niña phase (correlation coefficient = 0.26).
The annual distribution of fatalities is presented in Fig. 2, indicating that the maximum in fatalities is observed during July and August, with more than 1600 casualties per month reported for the whole period 1979–2011, in which a total of 7362 fatalities were reported. June is third in the ranking of fatalities (1211), indicating that the majority of the deaths occur during the first half of the rainy season in Mexico (from June to October). Note that we have kept the separation between the two segments because of the different reporting, and while the total numbers have decreased in the more recent period, they both exhibit similar annual variability.
The study by Kucienska et al. (2010) constitutes the first published report of lightning activity over Mexico with 5 years of data from the WWLLN database (2006–09), presenting a preliminary climatology of lightning over Mexico and adjacent oceans. The results from that study have been updated to produce the figures presented in this study. The monthly variability of lightning over the whole country indicates a maximum during August, followed by September and July (Fig. 3). Very low activity is observed from January to March and in November and December, months that correspond to the dry season in Mexico. While April and May show lightning density comparable to October, little precipitation is observed during those months that corresponds to the transition between the dry and wet seasons, and the high lightning-to-precipitation ratio is perhaps linked to high atmospheric aerosol concentrations from the increased agricultural practices (slash and burn) before the rainy season and also industrial emissions (Kucienska et al. 2012).
b. Spatial distribution
Figure 4 presents the spatial distribution of fatalities for Mexico, at the state level. As already noted, the total death toll in the period 1979–2011 is 7362, and only 7 of the 32 states account for 60% of all fatalities. The states with the largest number of fatalities are Estado de México (1777), Michoacán (721), Oaxaca (515), and Guanajuato (406). The state of Estado de México is located in the central plateau and Michoacán borders with it to the west (both shown in brown in Fig. 4). The average population density for Mexico as a country is only 57 inhabitants per square kilometer, but the different states vary widely in density. Estado de Mexico has the second largest population density at 659 km−2, because it surrounds the federal capital (Mexico City, which has a density of 5920 km−2). However, Estado de Mexico also includes many other less populated counties located away from the capital. As we will see later, the largest number of victims is not observed in the most populous counties that surround the capital, but rather in the rural areas in the western region of the state. The three other states with the largest numbers of fatalities in decreasing order have the following population densities: Michoacán, 74 km−2; Oaxaca, 41 km−2; and Guanajuato, 179 km−2 (see Fig. 4 to locate these states).
The spatial distribution of lightning for Mexico (Fig. 5) identifies the location of the areas with the highest flash lightning density, showing a localized maximum in the southern part of the state of Veracruz, close to the southern region of the Gulf of Mexico. Two other regions of relative maxima are observed on the Pacific coast and in the southwest corner of the state of Estado de México (on the border of Michoacán and Guerrero). This latter region is where a recent radar study has shown that the largest storms (in terms of areal and volume extent and echo-top heights derived from the radar reflectivity database) develop in the region of the central Mexican plateau and surrounding mountains (Novo and Raga 2013). Michoacán has the second largest death toll by lightning with a statewide population density of 74 km−2. We will see that the majority of deaths in Estado de Mexico and Michoacán occur in this region of deep convection, even though the counties are not very densely populated.
The state of Oaxaca (shown in red in Fig. 4, population density = 41 km–2) is located on the southern Pacific coast and encompasses the Isthmus of Tehuantepec. Large convective storms and systems are frequent in this region during the rainy season, because of the northward movement of the intertropical convergence zone and tropical disturbances (e.g., easterly waves) that travel through the region. Even though convection and precipitation are frequent in the region, the distribution of lightning (see Fig. 5) indicates only a very small region of the state bordering Veracruz with high lightning activity. So, the natural threat is not very large and the population density is lower than the national average, but this state, one of the poorest in Mexico, has the third highest death toll by lightning.
In fourth place in terms of number of fatal victims is the state of Guanajuato (shown in orange in Fig. 4, population density = 179 km−2), which borders Michoacán to the south and similarly to the case of Oaxaca (population density = 41 km−2); there is no particular indication that the threat of lightning would be large in the region (see Fig. 5), suggesting that the large death toll may be related not to the actual threat but to other factors, and we postulate here that the factor that leads to the large fatality density is the vulnerability of the population. However, note that the total number of deaths by lightning in Oaxaca and Guanajuato is only between one-third and one-fourth of the casualties reported in the state of Estado de México, which is the state with the second largest population density after Mexico City (with a very low fatality density).
c. Social patterns: Age and gender
Figure 6 shows the number of fatalities as a function of gender, keeping separate the two reporting periods mentioned in section 2. The number of female fatalities is small compared to the number of male fatalities, amounting to about only 21% of the total deaths for the whole period analyzed. This smaller fraction of female victims may be related to the type of outdoor activities carried out by the different genders, as we will explore and argue in section 4.
Figure 7 shows the number of fatalities as a function of gender and age, again keeping separate the two periods mentioned in section 2. There are a couple of noteworthy points that emerge from these graphs. First, the distributions are remarkably similar for both periods, but note that the vertical axes have very different scales, so that in the more recent period (1998–2011) the maximum value is only about one-fourth of the older period (1979–97). Second, the peaks of the distributions correspond to boys between 15 and 19 years old and girls between 10 and 14 years old for the period 1979–97. While there is a slight increase in deaths of boys between 10 and 14 in the more recent period (1998–2011), there is a marked shift toward older female fatalities, with a secondary peak observed for 20–24-year-olds. These results indicate that a large fraction of fatal victims in Mexico are boys between the ages of 10 and 19, with about 45.8% of all male victims younger than 25 years old. Deaths in that age segment constitute 34.5% of all fatalities (male and female).
A natural phenomenon, like lightning, becomes a very serious risk when there is a vulnerable population (Wilches-Chaux 1993), which can then lead to a large death toll. It is necessary to explore the vulnerability of the society and to determine how it may become a main factor in leading to casualties. To evaluate the different aspects that may define the social vulnerability, we will concentrate on the state of Estado de México, which accounts for the largest number of fatal victims (1777 out of 7362). We will discuss physical geography, demographics, education level, and local economy to assess social vulnerability.
The state of Estado de México (refer to Fig. 4), centrally located in the country and surrounding the federal capital (Mexico City), covers an area of 22 357 km2 and as mentioned before has the second largest population density in the nation (659 km−2). Its topography consists of a high plateau (on average more than 1.5 km above sea level) and partially includes two mountainous regions: the Sierra Madre del Sur (oriented roughly parallel to the Pacific coast) and the transvolcanic axis (roughly aligned east–west, where the highest volcanoes in Mexico are located). These mountainous regions are important in triggering deep convective storms, and during the rainy season there is a relative maximum of lightning density located in the southwest corner of the state (see Fig. 5).
The state of Estado de México is politically divided into 125 counties (INEGI 2006), with a total population in 2005 of 14 007 495, divided into 6 832 822 males and 7 174 673 females (INEGI 2011b). About 76% of the population is located in large urban areas, with 27% of its population located in just 3 of the 125 counties, those that surround Mexico City.
In terms of the education level of the population, 6% has no formal education at all and 10% has not finished elementary school. Only 27% of the population has completed middle school and only 15% has graduated from high school. These figures indicate the low levels of formal education of the population of this state. Moreover, there is also a component of native population in the state, mainly from four ethnic groups: Mazahua, Otomí, Nahuatl, and Mixteca (about 250 000 in total). While this number is small given the total population of the state, it is a fact that the native population has more difficulty accessing and completing formal education and that they mostly inhabit rural areas.
Since the state surrounds Mexico City, most of the economic activities that contribute to the gross national product (GNP) are related to manufacturing (28%) commerce (22%), and services (21%) (INEGI 2006). While the statewide statistics only indicate that agriculture contributes 1.3% to the GNP, it should be pointed out that agricultural activities for self- or local consumption are very widespread in rural areas and would not appear as a quantifiable figure in the GNP.
We now concentrate on the counties where the largest fatality density is observed, those that are located in the western part of the state. Table 1 presents a list with the 11 counties (out of 125 counties in the state) that are located from south to north along the western boundary of the state. It is interesting to note that while agriculture is only a marginal activity statewide (only 1.3% of the GNP), in these 11 counties, agriculture is the most important economic activity (INEGI 2006). The main crops are maize, beans, chili, tomatoes, and squash, all of which are staple foods of rural Mexico and clearly for local or self-consumption, given the small areas sown (Table 1). These crops require mainly unskilled manual labor (perhaps even involving children in the fields) and depend on natural rains (not irrigated). In particular, note that the county San Felipe del Progreso has 70% of its surface devoted to agriculture.
The link between total population per county and deaths is explored through the data in Table 2, for the largest 35 counties in the state listed in decreasing order of population. Note that the counties with the larger populations are not necessarily those that report the most fatalities. The entries in bold correspond to all the counties listed in Table 1, where the percentage of area dedicated to agriculture in each of them is shown. Note that most of those counties (highlighted in bold) have a fatality density of more than 30 per million inhabitants per year, which is extremely large, more than an order of magnitude of the average value for the whole country (2.72). In particular, the county San Felipe del Progreso is highlighted in bold and italics because even though it has about one-seventh of the population of the largest county (Toluca, where the state capital is located), it has more than twice the number of fatalities in the period 1979–2011; moreover, it reports the largest death rate per million inhabitants per square kilometer (71) and on average, seven deaths per year in the period. This county has 70% of its surface devoted to agriculture (Table 1) and is located in the west, but its location does not coincide exactly with the relative maximum in lightning density (located farther south). It appears that fatalities are not strictly related to the magnitude of the threat but are also a result of the main activities (e.g., agriculture) of the local population. Its population in 2005 was a little over 100 000 but only 178 physicians in 35 medical centers appear in the census for the county, which is indicative of low access to public health facilities for the rural population. Figure 8 shows that there is no correlation between the natural threat (in terms of lightning flashes) and the fatalities in the 35 counties listed in Table 2.
One last aspect to be considered at the county level is the one that links the low level of education of the population with its increased vulnerability. Rural areas have lower population but the education level is also lower, and it is there where the larger fraction of native population lives as well. The lower levels of education are very likely related to the lack of knowledge about the physical phenomenon, its dangers, and the simple actions needed to be taken in cases of lightning threat. The population is also more susceptible to misinformation about the phenomenon and, because of more traditional and religious beliefs, to relating it to supernatural aspects. Newspaper articles that report deaths of youngsters and adolescents recurrently mention the phrase “acts of God” and that “He gives life and He takes it.” Families that are affected by fatalities claim that “nobody should be blamed” for those deaths. But the newspapers often also report that the children took refuge under a large tree when the storm started and that they were either playing soccer or working in the fields, actions most likely related to their deaths. It is ignorance of the basic protection measures that is to blame for certainly a large number of the fatalities (El Universal 2001).
Clearly, the loss of life associated with lightning is perceived by society at large and by the government as something that is beyond their power to prevent. However, in other countries the actions of paramedics and first responders to a person struck by lightning are crucial to the victim’s survival.
Other countries also carry out assessments of the damage to infrastructure and property as a result of lightning strikes. In Canada, for example, while having a very low fatality density, the government estimates losses of about CAD $0.6 billion–$1 billion (Mills et al. 2010). Unfortunately, there are no equivalent data of economic losses gathered in Mexico for comparison. However, since the largest fatality densities are found in rural areas with low population density, if the data on economic losses were available, then it would likely be small, given that the large fatality rates are observed in rural counties, where subsistence agriculture is the main activity.
Finally, it is clear that the loss of these young people over the years is not seen as a disaster to society. The lack of attention paid by the government to small and moderate disasters as opposed to a huge disaster (e.g., earthquake), may also play a role in the large number of fatalities, particularly in rural areas. Small or moderate disasters have a slow impact and occur at the local/county level. These deaths are not perceived as disasters, but they can accumulate over time and become significant. Some of the statistics of deaths presented for the counties in the southwestern portion of the state of Estado de Mexico indicate that there are indeed extremely large fatality densities per million inhabitants and that they are related to the vulnerability of the population. However, it seems that because the population considers the deaths to be related to supernatural forces, it does not demand that the government take action to prevent them. The majority of the population in Mexico lives in major urban centers, and hence, there has been a decrease in the individual risk exposure. However, we have shown that the large death toll in Mexico is related to agricultural activities and large fatality densities are seen in rural counties, with more than half (and up to 70%) of their area dedicated to agriculture, implying that the population working in the fields would be very exposed to the natural threat. Moreover, the largest fatality densities per million inhabitants occur in mostly rural regions with low education levels and low access to medical facilities and mostly affect children and adolescents.
Cloud-to-ground lightning constitutes a real natural threat to populations everywhere and in particular in Mexico during the wet season (June–October). This study presents the first evidence of the relationships between the spatiotemporal distribution of the phenomenon and fatalities due to lightning strikes. While there are regions where the natural lightning threat is clearly larger than in other regions, the fatalities are not necessarily related to the threat but also appear to be the result of a combination of factors. In fact, the largest lightning density is observed in the southeastern portion of the country and there are very few fatalities reported there.
More than 7300 people died by lightning stroke in the period 1979–2011 in the country, resulting in an average yearly death rate of 287 ± 43 for the period 1979–97, and a decrease to only 127 ± 54 in the more recent period (1998–2011). Considering the population growth in the country, the first period results in an average of 3.75 deaths per million inhabitants (with a standard deviation of 0.89), a value considerably lower than the value of 6 deaths per million estimated by Holle (2008) as representative of the density of fatalities in the less developed countries. The average density of fatalities in the second period decreases to 1.32 deaths per million (with a standard deviation of 0.50), but it is still over 4 times the value estimated by Holle (2008) for the developed countries (0.3 deaths per million people). The last 2 years in the database (2010 and 2011) report densities of 0.59 and 0.67 per million people, only about twice the value estimated for developed countries.
A total of 60% of the fatal victims occur in only 7 out of the 32 states in Mexico, with the largest fraction occurring in the state of Estado de México (24%). Within this state, the fatal victims are primarily found in certain counties (not the ones with the largest populations densities, which are located surrounding Mexico City), mainly located in the western and southwestern portions of the state, where the dominant agricultural activities may be exposing the population to the natural threat. Moreover, those counties have less than 15% of the state population, have more native population with a low formal education level, and have a fatality density over an order of magnitude higher than the national average.
The results have indicated a clear bias in the fatal victims toward boys and young males (under the age of 25), with more than 45% of the total deaths in that segment of the population. While female deaths constitute a small fraction of the total number, the under-25 age segment also has the largest number of fatal victims.
Socioeconomic activities at the county level in the counties that report the largest numbers of fatal victims suggest that the vulnerability of the population stems from low educational levels and from performing agricultural activities outdoors, which increase their exposure to the phenomenon (mainly during the first half of the rainy season, where more activities in the growing fields are required). Young males usually help in the agricultural activities in the open fields and would be exposed to the threat of lightning. We could argue that the older population, having had more experience, may have learned to protect themselves while working outdoors. The large difference between young male and young female deaths may also be related to the practice of sports (predominantly soccer), carried out by young males rather than females.
The lower levels of formal education in the rural areas very likely translate to a lack of knowledge about the physical phenomenon, its dangers, and the simple actions needed to be implemented in cases of lightning threat. The population is also more susceptible to misinformation about the phenomenon and, because of more traditional and religious beliefs, to relating it to supernatural aspects and to not demanding action from the authorities.
We argue that the government has not recognized that these fatalities could be preventable, because it constitutes a small disaster (indeed, not even considered a small disaster in the political discourse), one characterized by slow impact that builds over time and happens mostly in rural counties, where the population density is low. It would be easy for the government to implement strategies to diminish the fatal aspect of the threat, such as information on the phenomenon and concrete actions on how to protect oneself, as is done in the case of earthquakes. Even though the lightning bolt is accompanied by loud thunder, Mexican society and its government have yet to recognize this silent killer that seems invisible to them.
The authors are grateful to two of the anonymous reviewers for their comments and suggestions, which helped improved the manuscript. The authors also wish to thank the World Wide Lightning Location Network (http://wwlln.net), a collaboration among over 40 universities and institutions, for providing the lightning location data used in this paper. Finally, they gratefully acknowledge Dr. R. Holzworth and his team at the University of Washington for careful data monitoring and constant improvement of the retrieval algorithm. This study was partially funded by Projects SEP-Conacyt 62071 and 154729 and Semarnat-Conacyt 23499.
A comment/reply has been published regarding this article and can be found at http://journals.ametsoc.org/doi/abs/10.1175/WCAS-D-14-00046.1 and http://journals.ametsoc.org/doi/abs/10.1175/WCAS-D-15-0006.1