Landslide Fatality Occurrence: A Systematic Review of Research Published between January 2010 and March 2022

Abstract

Landslides triggered by rainfall kill people worldwide, and frequent extreme events that are expected to be an effect of climate change could exacerbate this problem. This review aims to identify recent research, highlighting both the dynamics of landslide accidents and the characteristics of victims. From SCOPUS and WOS databases, using the PRISMA (preferred reporting items for systematic reviews and meta-analysis) approach, 25 articles written in English, published in the January 2010–March 2022 period and focused on landslide fatalities, were mined. The selected articles recognized a worldwide underestimation of landslide fatalities and analyzed landslide mortality from three perspectives, indicating the importance of this topic for a multidisciplinary research community. The papers focused on (a) fatal landslides and their geographic distribution, seasonality, trends, and relationships with socioeconomic indicators; (b) landslide fatalities and their behaviors and the dynamics of accidents; and (c) clinical causes of death or injury types, aiming to improve emergency rescue procedures. The gaps that emerged include (a) the insufficient reuse of valuable fatality databases; (b) the absence of simple take-home messages for citizens, practitioners, schoolteachers, and policymakers, aiming to set educational campaigns and adaptation measures; and (c) the lack of joint research projects between researchers working on landslides and doctors treating victims to provide complete research results that would be able to actually reduce landslide mortality.

1. Introduction

Landslides are widespread geomorphological phenomena that can occur on slopes located anywhere in the world. Most landslides are natural phenomena affected by landscape evolution and governed by triggering mechanisms primarily related to either rain or earthquakes. In addition to the economic impact, landslides dramatically affect human life by killing or injuring thousands of people worldwide each year. The human toll changes according to the area, as observed in one of the world maps created by the European Commission’s Directorate-General for European Civil Protection and Humanitarian Aid Operations (2022), indicating the fatal landslides that occurred during a short recent period (1 August–31 December 2020) (Figure 1).

According to the World Health Organization (https://www.who.int/health-topics/landslides#tab=tab_1 (accessed on 17 May 2022), between 1998 and 2017, landslides affected an estimated 4.8 million people and killed more than 18,000 individuals. In permafrost areas, the rising temperatures related to climate change are expected to trigger more landslides because as permafrost melts, slopes can become unstable (WHO, 2022). Similarly, the expected increasing frequency and intensity of extreme rainfall events in conjunction with population growth, deforestation, rapid urbanization, and unplanned development in landslide-prone areas, are likely to increase landslide fatalities worldwide [1].

Considering climate change, it is crucial to improve our knowledge with respect to human–landslide interactions to understand the dynamics of landslide accidents; the clinical causes of death; the demographics and vulnerabilities of victims according to countries, regions, and ethnicity; and the trend of both the total number of fatalities and the number of fatalities per landslide to “Advance Culture of Living with Landslides”, to plagiarize the title of a well-known book [2]. Research findings can play a key role in reducing landslide impacts on human life by supplying knowledge that should be:

(a)

transferable to civil protection agencies and emergency management offices to shape effective adaptation, defensive, and operative measures;

(b)

usable to raise public awareness through customized informative campaigns and tailored guidelines for both citizens and students of primary and secondary schools.

Because of the numerous perspectives from which science analyzes landslides, in addition to the continuous developments in knowledge, it is difficult to collect all the most recent and advanced research findings into a single paper. For this reason, the aim of the present review is to identify and summarize the primary findings of recent multidisciplinary scientific literature on this important social topic.

The paper is structured as follows: Section 2 describes the materials and methods used in the review; Section 3 summarizes the primary results by comparing different perspectives on the factors identified as the main drivers of landslide mortality; Section 4 discusses the findings raised from the literature analysis, highlighting both critical points and knowledge gaps and identifying the path for future research toward a better understanding of the problems and the identification of adequate solutions; and finally, Section 5 presents the primary conclusions.

2. Materials and Methods: Criteria for the Literature Search

The present review addresses fatal landslides (FATLAN), defined as landslides triggered by rainfall and causing one or more landslide fatalities (LANFAT). The term landslide encompasses all types of mass movements, including rock falls and topples, debris flows, soil slips, earthflows, rockslides, rock avalanches, shallow- and deep-seated slides, and complex and compound slope failures. LANFAT are people killed by direct impact from landslides and landslide debris who died either immediately due to vital organ injuries or after burial due to traumatic asphyxia. The long-term effects of landslides on human health are beyond the scope of this review.

From the SCOPUS and WOS databases, articles with the following features were mined:

• written in English;
• published between 1 January 2010 and 30 March 2022;
• titles include the following key words: landslid* and fatalit*; landslid* and mortality; landslid* and deat*; fatal* and landslid*; disaster* and fatalit*; and disaster* and deat*.

A systematic search in relevant databases is often performed to obtain literature reviews on topics such as the impact on people from natural hazards [3] and nurse disaster preparedness [4], in addition to the impact on human health from floods [5] and droughts [6] and on social vulnerability to floods [7], flood risk management [8], and the occurrence of flood fatalities [9].

Keyword searching was performed on the title of the articles to exclude papers not focused on landslide fatalities and merely reporting common sentences such as “landslides kill many people” or “landslide mortality is increasing”. The paper selection was performed according to the following criteria.

• Inclusion criteria: articles addressing FATLAN and their spatial and temporal distribution; articles concerning LANFAT characteristics (i.e., age, gender, and behaviors) and/or the clinical causes of death.
• Exclusion criteria: articles focusing on landslides triggered by earthquakes or human activities; papers on long-term effects on people’s health (i.e., in large rockfalls, the health effects on rescue teams caused by severe dolomitic dust exposure [10]).

The selection of relevant papers was performed using the hierarchical approach proposed by Moher et al. [11], which allowed the identification of 151 papers in WOS and 146 papers in SCOPUS (Figure 2). An additional 8 papers, identified using other sources, were also added, thus reaching a total of 305 records. The elimination of duplicate records allowed the identification of 200 papers to be screened (Figure 1). Then, article screening allowed us to eliminate 137 papers that did not analyze landslide mortality, even if their titles included the search terms (i.e., [12,13,14]).

In the next step, of the 63 papers assessed for eligibility, 38 were rejected because they were not relevant to the scope of this review, and 25 papers were selected and discussed (

Table 1. Articles selected, sorted by focus, and listed in chronological order of publication. #: article number in the references section; # FL: number of fatal landslides analyzed; # LF: number of landslide fatalities; and DB: database. N.A.: not available; N.R.: not reported; and (*): not open-access. ¹ Seismic and nonseismic debris flows. ² Landslides plus rock fall.

Focus on Fatal Landslides
#	Year	Title	Area	# FL	# LF	Period	Data Sources	DB Availability
[15]	2014	Debris flows and their toll on human life: a global analysis of debris-flow fatalities from 1950 to 2011	World	213 1	77,779	1950–2011	Academic publications and newspapers	N.A.
[16]	2015	Regional trends and controlling factors of fatal landslides in Latin America and the Caribbean	Latin America and Caribbean	611	11,631	2004–2013	Durham Fatal Landslide DB	N.A.
[17]	2016	Fatal landslides in Europe	27 European countries	476	1370	1995–2014	European landslide DB (ELS-DAT)	N.A.
[18]	2018	Global fatal landslide occurrence from 2004 to 2016	World	4862	55,997	2004–2016	Global Fatal Landslide DB (Durham Fatal Landslide DB)	N.A.
[19]	2018	Spatial and temporal analysis of a fatal landslide inventory in China from 1950 to 2016	China	1911	28,139	1950–2016	Fatal Landslide Event Inventory of China	https://0-doi-org.brum.beds.ac.uk/10.1007/s10346-018-1037-6 (*)
[20]	2018	Trend and spatiotemporal distribution of fatal landslides triggered by nonseismic effects in China	China	463	4718	2004–2016	Chinese Geological Environment Monitoring Institute	https://0-doi-org.brum.beds.ac.uk/10.1007/s10346-018-1007-z (*)
[1]	2019	The human cost of global warming: Deadly landslides and their triggers (1995–2014)	World	3876	163,658	1995–2014	Global Landslide Database	N.A.
[21]	2021	Spatiotemporal variations of fatal landslides in Turkey	Turkey	389	1343	1929–2019	FATALDOT (Fatal Landslide DB of Turkey)	https://0-doi-org.brum.beds.ac.uk/10.1007/s10346-020-01580-7 (*).
[22]	2022	Changes in the factors contributing to the reduction of landslide fatalities between 1945 and 2019 in Japan	Japan	N.R.	N.R.	1945–2019	Aggregation of preexisting DBs	N.A.
Focus on landslide fatalities
[23]	2011	Causes of death and demographic characteristics of victims of meteorological disasters in Korea from 1990 to 2008	South Korea	N.R.	403	1990–2008	National Emergency Management Agency	N.A.
[24]	2016	Natural hazard fatalities in Switzerland from 1946 to 2015	Switzerland	N.R.	159 2	1946–2015	DB of fatalities caused by natural hazard processes in Switzerland	N.A.
[25]	2016	Mortality patterns of hydrogeomorphologic disasters	Portugal	N.R.	237	1865–2010	DISASTER DB	riskam.ul.pt/disaster/en (accessed on 25 July 2022)
[26]	2017	Landslide societal risk in Portugal in the 1865–2015	Portugal	291	238	1865–2015	DISASTER DB	riskam.ul.pt/disaster/en (accessed on 25 July 2022)
[27]	2017	The Vulnerability of People to Damaging Hydrogeological Events in the Calabria Region (Italy)	Calabria (Italy)	N.R.	68	1980–2016	DB of people affected by Damaging Hydrogeological Events in Calabria	http://0-dx-doi-org.brum.beds.ac.uk/10.17632/99knpdb6yp.1 (accessed on 25 July 2022)
[28]	2018	Gender, age, and circumstances analysis of flood and landslide fatalities in Italy	Italy	-	1292	1965–2014	Italian DB of landslide and flood fatalities	http://osf.io/Kamgc (accessed on 25 July 2022)
[29]	2020	Analysis of landslide-induced fatalities and injuries in Bangladesh: 2000–2018	Bangladesh	204	727	2000–2018	DB of fatal landslides in Bangladesh	N.A.
[30]	2020	Human Vulnerability to Landslides	Various countries	-	-	-	DB of 334 individuals in 38 landslides	https://www.designsafe-ci.org/data/browser/public/designsafe.storage.published//PRJ-2866 (accessed on 25 July 2022)
[31]	2021	Fatalities associated with the severe weather conditions in the Czech Republic, 2000–2019	Czech Republic	N.R.	N.R.	2000–2019	DB of fatalities of severe weather in the Czech Republic	N.A.
[32]	2021	People vulnerability to landslide: risky behaviors and dangerous conditions by gender and age	Italy	283	1039	1970–2019	Italian DB of landslide fatalities	N.A.
[33]	2021	Historical landslide fatalities in British Columbia, Canada: trends and implications for risk management	British Columbia, Canada	N.R.	390	1880–2019	British Columbia Fatal Landslide DB	https://www.frontiersin.org/articles/10.3389/feart.2021.606854/full#supplementary-material (accessed on 25 July 2022)
[34]	2022	Fatal landslides in Colombia (from historical times to 2020) and their socioeconomic impacts	Colombia	2351	37,959	1912–2000	SIMMA, DESINVENTAR, UNGRD and CNMM	https://data.mendeley.com/datasets/xbrc8gvby9/1 (accessed on 25 July 2022)
Focus on clinical causes of landslide fatalities
[35]	2012	Landslide fatalities: A study of six cases	India	1	6	Case study	Unnatural Death Cases	-
[36]	2016	Injury Patterns after the landslide disaster in Oshima, Tokyo, Japan on 16 October 2013	Japan	1	49	Case study	Analysis of survivors with severe trauma in the hospital	-
[37]	2017	Medicolegal death scene investigations after natural disaster and weather-related events: a review of the literature	-	-	-	-	Literature review	-
[38]	2022	The role of PMCT for the assessment of the cause of death in natural disaster (landslide and flood): a Sicilian experience	Italy	-	-	2 cases	-	-

). Noteworthily, the present review focuses on nonseismically triggered landslides, even if some of the papers addressed both seismically and nonseismically triggered FATLAN.

According to their focus, the selected papers were sorted into three groups (Table 1):

1. Focus on fatal landslides (36%). Articles analyzing FATLAN, their spatial and temporal distribution, identifying global hotspots, and searching for relationships with either the climate variability or socioeconomic indicators.

2. Focus on landslide fatalities (48%). Papers analyzing LANFAT, demographics, and behaviors of victims and investigating the scenery and dynamics of the fatal accidents.

3. Focus on the clinical causes of landslide fatalities (16%). Clinical research on victims both killed immediately and injured by landslides, performed in the frameworks of both legal and emergency medicine.

From a geographic perspective, articles primarily focused on European countries (40%) and secondarily on countries on the Asian continent (28%). The remaining papers focused on the entire world (16%) and countries in South and North America (12%), while 4% of the papers were review articles not linked to a specific study area. In the study period, no papers concerning landslide mortality on the African continent were found. The number of fatalities per landslide, reported by some of the papers, allowed us to understand how the severity of the problem changed according to the area (Table 1): for European countries, it was assessed as 3 [17], while for Latin America and the Caribbean, it was approximately 19 fatalities per landslide [16].

With respect to the temporal distribution, only 12% of the papers were published in the first part of the observation period between 2010 and 2015, while 88% of them appeared between 2015 and 30 March 2022 (Figure 3). The increasing scientific community’s attention to the problem is witnessed by the trend of the number of papers published per year: it seems constantly increasing, despite an interruption in 2019, likely because of the generalized decrease in research production suffered during the first year of the pandemic.

3. Review Results

The selected articles addressed the problem of landslide mortality with different approaches and aims, confirming that the problem represents a focus for multidisciplinary communities of researchers actively working on it. The current section of the review presents the primary findings of the analyzed papers, summarizing and comparing the different opinions on factors identified as the main drivers of landslide mortality and underlining differences and similarities detected according to climatic and social frameworks.

3.1. Gathering Data on Landslide Fatalities

The number of LANFAT analyzed was generally low in the papers in Group 3, which deeply investigated the clinical causes of death and injuries caused by single landslides, while papers in Groups 1 and 2 generally used large databases concerning LANFAT that occurred in a selected spatiotemporal frame.

The realization of LANFAT databases requires a meticulous data-gathering procedure, generally performed by the same researchers who authored the papers. Newspapers often represent the primary data source because they report crucial information on time, place, losses, and the context in which the disaster occurred, permitting the understanding of what occurred to LANFAT. In addition to newspapers, other documentary sources, such as news websites, gray literature, insurance data, and civil protection and fire brigade reports, were used as complementary sources to cross-check, confirm, and validate data [25,27,31].

In other cases, LANFAT databases were obtained by “customizing” accessible international multihazards databases or landslides databases, often merging more than one database, performing supplementary research to expand the study period, and conducting data cross-checking and validation. Haque et al. [1] updated the European landslides database (ELS-DAT) using the search engine Google to select deadly landslides in each of the 49 European countries. Sepúlveda and Petley [16] used the Durham Fatal Landslide Database (later renamed the Global Fatal Landslide Database), which was compiled through a combination of a daily internet search with predetermined keywords, research literature, and government and aid agency reports. Garcia-Delgado et al. [34] obtained LANFAT in Columbia by merging data derived from four existing databases: SIMMA (Sistema de Información de Movimientos en Masa), DESINVENTAR (Disaster Information Management System), UNGRD (National Disaster Risk Management Unit) and CNMM (National Catalogue of Mass Movements), containing the compilation of fatal events obtained from newspapers and reports of agencies involved in emergency response, such as the fire and rescue services and emergency rooms.

In some cases, data on LANFAT were extracted from national databases of the impact of natural disasters, as in the cases of (a) Salvati et al. [28], who extracted LANFAT from the catalog of Italian landslide and flood fatalities (1965–2014); (b) Pereira et al. [25], who used DISASTER, the database of social consequences of landslides and floods mentioned in Portuguese newspapers between 1865 and 2010; (c) Badoux et al. [24], who used the database of fatalities caused by natural hazard processes in Switzerland; and (d) Petrucci et al. [27], who collected fatalities in Calabria (Italy) caused by damaging phenomena such as floods, landslides, lightning, windstorms, hail, and storm surges.

In summary, regardless of the strategy used, data gathering is a fundamental step in this type of research, and regardless of the approach used, the valuable databases obtained are the result of long processes demanding meticulous work.

3.2. Landslides Are Dangerous to Human Life

What are the characteristics that make a landslide dangerous to human life? As can be argued, speed is one of the basic factors that determines landslide impacts on people, as rapid landslides may surprise people, giving them no time to decide what to do. From the perspective of emergency management, the rapidity of a landslide may restrict the anticipation time of an effective response, and result in less time for both warning and emergency procedure activation (i.e., road closures and evacuations). This was confirmed in 27 European countries, where between 1995 and 2014, the largest amount of human losses was caused by rapid phenomena such as rock falls and debris flows, in addition to landslides for which the typology was not identified [17]. Similarly, in Portugal (1865–2015), falls and flows were responsible for the majority of LANFAT [26]. Limiting the focus on debris flows, from 1950 to 2011, on a global scale, Dowling and Santi [15] highlighted that debris flows triggered by rainfall were the most frequent and less dangerous for human life (median fatality rate of 9 people per event) than less frequent nonrainfall-induced debris flows (median fatality count > 500 people).

3.3. The Geographic and Seasonal Distribution of Fatal Landslides

The geographic distribution of FATLAN depends on the interactions between (a) the geomorphologic and hydrologic factors predisposing landslide occurrence and (b) human exposure to landslide risk, as landslides occurring in unpopulated areas do not have effects on human life. The identification of FATLAN hot spots is affected by the possible underestimation of LANFAT, which depends on (a) failed inclusion of FATLAN causing a few fatalities, (b) absence of data on LANFAT occurring in remote areas and/or in remote periods, and (c) uncertainty of total LANFAT in multihazard events (i.e., landslides and floods occurring simultaneously). Considering these limitations, for the 2004–2016 period, Froude and Petley [18] identified Asia as the dominant geographic area of nonseismic FATLAN at the global scale. At the scale of the European continent, in the 1995–2014 period, the most seriously affected country was Turkey (335 LANFAT), followed by Italy (283), Russia (169), and Portugal (91) [17]. Nevertheless, several factors must be considered to assess the meaning of these figures: for example, by generally dividing the number of LANFAT for a “static” parameter as the surface of the country, it seems evident that LANFAT divided by the country area and multiplied by 10,000 supplies higher values in small countries, such as Portugal (9.86) and Italy (9.36), with respect to larger countries, such as Turkey (4.27) and Russia (0.09).

The seasonal distribution of rainfall-induced FATLAN depends on the interplay between (a) geological–geomorphological stability factors, (b) climatic characteristics of the area, and (c) urban settlement configurations. On a global scale, FATLAN seem to occur most frequently between June and December in the Northern Hemisphere and between December and February in the Southern Hemisphere [1]. At the European scale, LANFAT are reported throughout the year: during summer and fall, they occur mostly in central and northern Europe, while in southern Europe, they mainly occur in the winter and spring periods. The seasonal pattern is likely dominated by snow melt (February) and rainfall (April to July), when in July, there is the highest peak of fatal landslides [17]. In China, between 2000 and 2016, FATLAN primarily occurred between April and September (82.15%), which is consistent with the monthly precipitation regimen [19]. In Colombia, a strong correlation between the climate variability phenomenon known as the El Niño Southern Oscillation (ENSO) and FATLAN is observed, particularly during those years when strong La Niña (cold phase of ENSO) events occur [34].

3.4. The Trend of Landslide Fatalities

Some of the papers focused on the trend of LANFAT throughout different study periods and attempted to explain the meaning of its oscillations. In China, high-impact FATLAN, killing more than 30 people, was on the rise during 1950–1999 and declined from 2000 to 2016, likely because of an increase in landslide mitigation investments. In contrast, low-impact FATLAN causing fewer than 10 fatalities demonstrated a significant increasing trend, particularly between 2000 and 2016, related to (a) the improvement of the availability of landslide data online, (b) an increase in extreme precipitation events, and (c) the effects of land urbanization [19].

According to Shinohara and Kume [22], the factors contributing to the decrease in LANFAT changed with time because the defensive measures evolved according to the degree of maturity of the nation. Japan, for example, reduced the number of LANFAT between 1945 and 2019 due to changes in building structure, increases in the number of people evacuated, and increases in the area of mature forests.

In Portugal, there is an absence of any exponential growth in time of both landslide cases and landslide mortality, and the temporal trend of the landslide mortality rate did not indicate a decrease. Most fatalities occurred inside buildings, despite the increasing quality of the buildings and the adoption of construction techniques and codes, likely because these improvements are not enough to resist very rapid landslides such as falls and flows, particularly when buildings are located on (or very close to) landslide-prone slopes [26].

In British Columbia (Canada), LANFAT are rare, and despite a fivefold increase in the population, they decreased from five fatalities per year in 1960 to one fatality per year in 2019. This may be due to a change in settlement patterns, transitioning from a greater prevalence of rural communities in mountainous areas to a densification of urban areas away from landslide hazards [33].

3.5. Demographic and Socioeconomic Vulnerability to Landslides

According to the vulnerability perspective, the impact of natural disasters is not completely “natural”; rather, it is determined by people’s unequal exposure to risks, which depends on different socioeconomic systems, and then the differences in mortality are explained by biological, demographic, social, and cultural vulnerabilities (gender, age, caste, and race), economic vulnerability (class) and physical vulnerability (housing structures) [39].

As during floods [9], even in the case of landslides, males are more vulnerable than females. Several papers reporting data on gender confirmed the predominance of male fatalities: in Italy [28], as in Columbia [34] and Korea [23], to quote some cases. Unfortunately, in developing countries, there is a lack of data on the sex of LANFAT due to difficulty in collecting data in high-impact multiple or sequential landslides [29]. The predominance of males can be related to a stronger exposure for males at risk due to work duties, particularly to a higher proportion of males working outdoors in emergency management and first aid services [28]. In Portugal, in the 1865–2015 period, male victims were more than double the female victims, possibly due to cultural reasons related to the social role of the family provider that exposes men to more hazardous occupations than women [26]. In Italy, male landslide fatalities occur frequently outdoors along roads, primarily involving drivers, indicating a specific dangerous death contingence preferentially occurring in daylight; conversely, female landslide fatalities occur more frequently indoors [32].

The socioeconomic indicators confirm that populations are disproportionately affected: FATLAN largely affect developing countries, characterized by significant poverty levels, weaker health care systems, and more corrupt governments. This can be clearly perceived, for example, by comparing the median number of fatalities caused by debris flows in developing (23 fatalities) versus developed countries (6 fatalities) [15]. On the Asian continent, Froude and Petley [18] highlighted that landslides become fatal most frequently in countries with lower gross national income. In Columbia, departments with low income and high levels of corruption and inequality are typically more affected by FATLAN [34].

Santi et al. [40] defined debris flows as “disasters of social vulnerability”, causing more fatalities in populations with economic, space, and influence/power restrictions. Economic and space restrictions force these populations to live in debris flow-prone areas, where the topography may confine the population in times of disaster, while political power limitations prevent these populations from receiving the resources to address debris flow risk. Socioeconomic indicator analyses most frequently compare landslide fatalities with the following parameters:

• GDP per capita;
• number of hospital beds per capita;
• government corruption index;
• maternal mortality rate;
• life expectancy at birth;
• number of general technical journal articles per capita [15].

According to Pollock and Wartman [30], in economically developing nations, individuals are up to twice as vulnerable to landslides as those in developed nations, and the rate of landslide victim rescues (15%) in developing nations is lower than that in developed nations (20%), perhaps because of fewer resources and personnel for emergency response.

3.6. Places and Circumstances of Deaths

With respect to the places of fatal accidents, papers analyzed the dualism of urban versus rural areas. In rural areas, landslide risk for human life can be assumed to be higher because (a) rural/remote areas are characterized by the lack of fast-responding units for rescues and emergency management (to issue alerts such as building evacuation or road closures); (b) the low population density diminishes the chances of receiving help first from laypeople; and (c) the distance from emergency management centers extends the times for first aid arrivals to help injured/buried people. In developed countries, characterized by a higher level of disaster preparedness, larger use of warning systems, and more advanced rescue procedures, the number of fatalities per landslide event is tendentially lower. The first minutes after a landslide are critical. Among persons rescued and included in the database of people affected by rapid landslides in buildings compiled by Pollock and Wartman [30], 77% were first located by neighbors and 11% by emergency response personnel. Due to their proximity, neighbors located and began to rescue victims more rapidly than emergency personnel, who often had to travel miles to reach the landslide site.

In contrast, highly urbanized areas are themselves a risk, due to the greater human exposure characterizing them. This is particularly true in settlements with low living standards, typical of most developing countries; here, low standards in terms of mitigation, control, and human settlement characteristics lead to numerous fatalities for each landslide. Sepúlveda and Petley [16] confirmed that in Latin America and the Caribbean, the presence of informal settlements in urban areas increases the number of LANFAT due to the effects of poverty and marginalization. Similarly, in Asia, Froude and Petley [18] highlighted that FATLAN generally cluster around cities. In contrast, in the European landscape (1995–2014), FATLAN are widely distributed throughout the countries, even if the concentration is higher in mountainous areas [17].

With respect to the circumstances of death, the most frequent element analyzed was the location where the victim was killed, often analyzed in conjunction with other elements such as gender and age of the fatalities. In the Italian catalog of LANFAT, fatal accidents occur more frequently indoors than outdoors, killing people mainly at home or in buildings (partially or totally collapsed) or dragged by earth, debris, or mud [28]. In Japan, LANFAT occur mostly inside buildings because the ratio of indoor to outdoor space is higher than that in many other countries [22]. Female fatalities occur more often indoors, while male fatalities occur outdoors along roads (Figure 4), particularly caused by debris flows and mudflows trapping people in vehicles or striking pedestrians [23,32].

3.7. Hazardous and Protective Behaviors

In contrast to floods, for which inappropriate behaviors have largely been reported by the literature [9], only a few papers have reported such behaviors in the case of landslides, such as Pereira et al. [26]: “men often assume risky behaviors outdoors and act with a false sense of security when they are driving vehicles”. In the Italian catalog with respect to LANFAT, hazardous behaviors were not reported: people most frequently were caught by surprise by the landslide and were apparently unaware of the imminent risk, as confirmed by the limited number of victims who attempted to escape [28]. A specific study classified damage to people in Calabria (Italy) (1980–2016) into three severity levels: fatalities (people who were killed), injured (people who suffered physical harm), and involved (people who were present at the landslide but survived without injuries). They did not identify hazardous behaviors either in fatalities or in people who were injured or involved in landslides. Moreover, involved people were younger than both injured people and fatalities, suggesting that younger people demonstrated greater promptness in reacting to dangerous situations [27].

Pollock and Wartman [30] identified two hazardous key actions that placed individuals at greater risk:

(a)

opening a door out of curiosity, that is, a human response to move toward unknown or unfamiliar phenomena, whether to identify a potential threat or out of curiosity; in these cases, people can be swept away by a surge of debris;

(b)

sheltering behind or beside large furniture: deaths caused by blunt force trauma, even among indoor victims, suggest that unsecured furniture is a significant contributor to landslide mortality.

Based on stories of survivors of 38 rapid landslides, these authors [30] identified and classified six “positively deviant” actions that can be crucial factors for survival (

Table 2. “Positively deviant” actions to increase possibility of surviving a landslide, both before and during a landslide (from: [30], modified).

BEFORE	BE INFORMED ABOUT POTENTIAL HAZARDS	Prior experience represents greater preparation, perception, and ability to cope with it. Talk to people who have experienced hazards.
	MOVE BEDROOMS UPSTAIRS, OR TO THE DOWNHILL SIDE OF THE HOME	Avoid placing bedrooms on the uphill side of the home, closest to potential hazards. During the night, occupants are unaware of an imminent threat. If moving bedrooms is unfeasible, then move beds away from exterior walls.
DUING	ESCAPE VERTICALLY	Mortality rate decreases for those above the ground floor of a landslide-impacted structure, even when the entire home is destroyed.
	IDENTIFY UNFURNISHED AREAS	Closets, bathrooms, and interior hallways can offer protection: these small spaces are unlikely to collapse due to the density of structural elements and are generally free from unsecured furniture, which could pin or crush a person.
	OPEN DOWNHILL DOORS AND WINDOWS	Fluid landslide debris can bury and suffocate occupants or develop enough pressure to rip apart the structure. Opening downhill doors or kicking out windows allows debris to flow through the home. However, individuals who did so risked being swept out of their homes.

).

3.8. The Clinical Causes of Death

Despite the increasing social emphasis on disaster preparation and response, there seems to be little increase in expert knowledge about how fatal accidents actually develop [35]. The number of disaster-related deaths recorded by vital statistics departments often differs from that reported by different agencies, such as the National Oceanic and Atmospheric Administration–National Weather Service Storm Database and the American Red Cross [37]. Generally, in the chaotic aftermath of disastrous landslides, data on people are ephemeral (people relocated, dispersed, or dead in hospitals), very challenging to collect, and tendentially characterized with low accuracy.

From the perspective of emergency rescue and care, the earliest possible rescue and prompt treatment are vital for survivors, and the recognition of injury patterns in each disaster setting is crucial for effective preparedness and management of the emergency phase [36].

From the perspective of forensic medicine, it is crucial to collect data at the death scene because these are the building blocks for identifying the cause and manner of death for different purposes, i.e., for insurance. Because disaster-related death scenes can be chaotic, it is crucial to provide death scene investigators with guidance about a consistent approach for collecting and reporting specific data [37]. Generally, the exact cause of death is an information available only in research carried out by forensic surgeons, which is performed in cases of FATLAN with a limited number of victims. Therefore, in landslides with many victims, medical investigations such as postmortem computed tomography might be an alternative to an autopsy and would allow us to assess the cause of death, providing fast information on the type and distribution of lesions [38].

Homma et al. [36], investigating the effects of the 2013 Oshima landslide (Japan), highlighted that timeliness in the rescue and treatment of injuries are crucial. The landslide affected 49 people: 34 triaged as black (deceased/expectant: injuries incompatible with life or without spontaneous respiration) and 15 triaged as both red (immediate: severe injuries but high potential for survival with treatment) and yellow (delayed: serious injuries but not immediately life-threatening). These last 15 people were quickly transported to the medical facility, with severe chest or pelvic injury and traumatic asphyxia in 62.5% of patients, and they survived, while a delay in starting initial treatment might have led to their death. Among persons triaged as black, there were no survivors as early as 1 day after the incident, while in disasters such as earthquakes, survivors can still be rescued several days after the event.

A landslide on the National Highway-53 (India) hit a minibus with six passengers: chest injury, head injury with skull fracture, and laceration of the brain associated with intracranial hemorrhages were the causes of death on the scene. In only one case, external injuries were not significant, and mechanical asphyxia due to clogging of the respiratory tract with soil particles was the cause of death [35]. In this case, information on the type of landslide could explain the differences between the five fatalities: four victims appear to have been hit by rock fall (able to cause multiple fractures) and one seems to have been a victim of mudflow (asphyxia). In a database focusing on rapid landslides (velocity > 5 m/s), falling within the categories of debris flow, flow slide, rock/debris avalanche, or debris slide, with the majority being highly fluidized, the 77 deaths for which the primary cause of death was known, traumatic injury was the immediate cause of death in 86% of cases, while 10% of individuals died by mechanical asphyxiation, and 4% died by drowning [30].

4. Discussion

As typically occurs in reviews, this review is also affected by some gaps that essentially depend on the criteria for article selection. First, this review did not analyze materials published outside of the WOS and SCOPUS platforms; this could cause gaps, but conversely, this ensured that only scientific articles validated through the peer review process were selected. Second, the selection of the highlights from each paper was, to some extent, “biased” by the author’s personal assessment of the relative importance of the factors defining the landslide mortality framework.

In addition to the described limits, the present review has two advantageous qualities:

Novelty. This review presents the most recent scientific research on people–landslide interactions and supplies an updated framework of factors currently investigated by different scientific communities.

Perspective. This review underlines the knowledge gaps that should be filled to reach a deeper understanding of people–landslide interactions, which is useful to increasing awareness in citizens, practitioners, and policymakers, in addition to improving community resilience and reducing future landslide mortality.

4.1. Critical Points

This section summarizes the critical points that emerged from the literature review and possible ways to bypass them to support advancements in knowledge on people–landslide interactions and the development of operative indications for decreasing landslide mortality.

The first point to reflect on is the choice between the use of publicly available multihazard databases or the realization of a specific database of landslide fatalities at the scale of the study area.

International multihazard databases are ready to use, but (a) they include fatalities caused by more than one type of disaster (i.e., landslides and floods are sometimes grouped together), (b) they report only cases where the number of fatalities exceeds a prefixed threshold (e.g., more than 10 fatalities in EM-DAT, https://www.emdat.be/, accessed on 10 May 2022), and (c) they do not report information on fatalities except for their number. In contrast, building a landslide fatality database (either collecting data or merging existing datasets) is a longer process that may discourage researchers who opt for ready-to-use multihazard databases.

The choice between the two kinds of databases must be performed consciously to understand the significance of the results, while also considering that the number of landslide fatalities is frequently quoted in the introduction of landslide studies as a proxy of landslide dangerousness. Therefore, it is important to be careful in reporting numbers that could be widely cited in the literature. Some researchers, when acting as reviewers, have the questionable belief that international multihazard databases (such as EM-DAT) are, in every case, the best data source because their robustness is ensured by international bodies funding their maintenance. Nevertheless, as is true when focusing on disasters, this is no longer confirmed when the focus is on individuals killed by disasters, which are not the focus of disaster databases. Some branches of the scientific community should therefore demonstrate a broader open-mindedness on the use of specific LANFAT databases because they can demonstrate aspects that otherwise would remain undisclosed.

Parameters that can suffer bias due to database incompleteness are both the location of FATLAN hot spots and the long-term trend of LANFAT.

Due to the current impossibility of obtaining a worldwide database without the underestimation of LANFAT, as affirmed by most reviewed papers, the location of hot spots for fatal landslides can be considered both partial and ephemeral. In some sectors of the globe, the absence of investigations and data can bias the results because the absence of data on landslide fatalities does not indicate an absence of fatal landslides. Moreover, hot spot location is a portrait of the studied period; if its validity can be considered affordable for the investigated interval, it should be continuously updated to reflect the current situation. To solve these problems, global data collection should be based on an official international network of data collectors working at the country scale, using local affordable data sources written in the language of people performing the research, and covering the entire country during the entire study period. Most importantly, no thresholds in the number of fatalities should be set to avoid bias in the final result due to the absence of FATLAN killing a number of people under the threshold. Therefore, this can bias the total count of LANFAT, particularly in developed countries, characterized by very frequent cases of landslides causing from one to a few fatalities.

Similarly, the trend of landslide fatalities is a datum frequently quoted in the literature that must be assessed with utmost prudence for the following reasons: (a) data gathering does not ensure a constant quality and quantity of data throughout long periods [17], (b) data reliability rapidly declines moving from recent to older landslides, (c) trend of LANFAT varies both in space and in time because of the combined effects of climate change and socioeconomic dynamics of the area, and (d) trends obtained using specific LANFAT inventories can be positive because they included landslides that had several victims who were not counted in studies based on global databases (i.e., EM-DAT discarded cases with fewer than 10 fatalities).

These problems can be partially mitigated by assessing the trend in intervals of 10 or 20 years, characterized by similar data availability. For periods before the massive advent of the internet, for example, data sources are significantly scarce and report only the most destructive events, neglecting cases with low-medium impacts in terms of human life (i.e., a few LANFAT). Then, if compared to those periods, the abundance of recent data could be mistaken for an exceptional increase in LANFAT, while it is merely an exceptional increase in data availability. Moreover, working on shorter periods has the advantage of detecting the onset of effects caused by either changes in people’s habits and behaviors or lifestyle evolution, as described with respect to hazardous and protective behaviors in Section 4.2.

The database sharing is strictly related to the previous points and, in a certain sense, represents its solution. Detailed LANFAT databases, extending to large temporal windows at either regional or national scales, represent the building block for promising affordable research results. Among the 25 selected articles, excluding 4 papers based on case studies, the database was freely available for only 8 papers (32%), while for 3 papers (12%), the database did not have open access, and for 10 papers (40%), the database was not publicly available. This represents a factor that slows down the possible advancement of the research because each research group must start from scratch to build its own database, while in contrast, database reusability should be ensured to offer the basis for new data validation and analyses or even to allow the use of data from different scientific communities with different disciplinary viewpoints.

A further perfectible theme is the lack of interdisciplinary research: this review demonstrated that people–landslide interactions are the focus of both researchers working on landslide mechanisms and doctors focusing on people’s health, but not one of the papers presented a collaboration between these two communities, although they both aim, to a certain extent, to landslide mortality reduction. Experts on landslide mechanisms could proficiently collaborate with specialists in the medical sector to investigate the impact of different types of landslides, attempting to identify similarities and differences to be included in civil protection recommendations. Similarly, some research of the medical sector could be placed into a clearer framework by providing the correct classification of the type/types of phenomena that caused fatalities. Therefore, to analyze fatalities caused by phenomena having borderline characteristics, i.e., between landslides and floods, despite being useful in terms of emergency management, this cannot provide clues for emergency preparedness, because landslides and floods are very different from the geomorphological, civil protection, and risk management perspectives and affect different planimetric sectors of the territory. Moreover, characterizing landslide type, velocity, geometry, spatial evolution, and the state of its debris (i.e., if it is solid or fluidified by high-water contents) can help to understand what can be expected in terms of injuries and, accordingly, improve emergency management protocols.

Finally, from a general perspective, if the present paper seems not particularly innovative, it must be considered that it is a review, and merely reflects what the literature has produced on this important social topic in recent years. Perhaps the reflection that should be made is on the measure of the level of interest in the last 10 years by the landslide scientific community with respect to the impacts on human life. However, in several articles (such as those eliminated in the preliminary phases of this review), the researchers quote “landslide fatalities” even in paper titles, to obtain the attention of the reader, even if the paper focused on all but fatalities.

Therefore, one of the aims of this review is to offer an updated panoramic view of what has been done on this social topic, both in terms of data gathering and phenomenon understanding, which can be appropriately quoted with respect to landslide–people interactions, which could be more appropriate than employing trivial sentences such as “landslides kill many people”, as frequently can be seen in the literature, which seems inappropriate in scientific papers.

4.2. Knowledge Gaps

This section presents the main gaps highlighted in the analyzed literature and further examines the clues borrowed from studies on flood mortality.

One of the most important points is the absence of modeling of the dynamics of fatal accidents, in contrast to recent research on flood fatalities, for which the most frequent dynamics have been schematized [42,43]. Similarly, autoprotective strategies, which can supply practical and feasible suggestions for citizens, were reported in only one of the articles [30].

Concerning victim behaviors, it can be assumed that individuals who faced an unexpected landslide did not demonstrate hazardous behaviors and instead adopted self-protective actions, such as escaping very quickly to reduce the potential landslide impact. Nevertheless, victim behaviors were investigated by only a few papers and were based on limited cases; objectively, without witnesses, victim behavior is difficult to determine. Considering that research on this aspect could reveal important elements, efforts could focus on either injured survivors or persons who were on the scene of the landslide but managed to be untouched [27]. Particularly in old data sources, this information is a generic and concise mention of survivors; in contrast, for recent landslides, searches online can be more fruitful, thanks to the present desire of sharing personal experiences that are promptly and extensively funneled in social media. Victim behaviors can also be mined and recorded from scientific papers focusing on the perception of natural hazards and realized by means of interviews with people who experienced landslides.

Victim behaviors can change through time due to progress in technology and lifestyle. For example, in the last 20 years, mobile phones have allowed some victims to call for help and save their own lives during natural disasters or to warn other people of the onset of a landslide (protective behavior). Nevertheless, they have also permitted the creation of movies of dangerous challenges (hazardous behavior) posted on social media to stimulate a spirit of emulation in people watching them, as sometimes occurs during floods.

Thus, the absence of data on hazardous behaviors does not mean that these behaviors have not caused someone’s death. It is also possible that some hazardous behaviors are not reported because they did not cause fatalities, even if they could have done so. As an example of the latter case, I personally witnessed an episode during a scientific meeting in Cassis (France). During a break, two of the researchers attending the meeting went to visit a local beautiful rocky promontory surmounted by a castle, reaching it by walking on a road that was closed by police with evident road signs. During their walk, a small rockfall occurred, and they had to run back to avoid being injured by debris. When I asked why they went on that road despite it being closed, they answered: “there were also other people doing the same…”.

To confirm a specific behavior as effectively protective and propose it as a self-defensive measure to apply in the case of landslides, it is necessary to identify a conspicuous number of cases in which that behavior truly allowed victims to survive.

With respect to the age and gender of fatalities, it is unclear if there are specific connotations; more information about these aspects could correctly drive customized educational campaigns to reach different segments of the population.

Among the elements that could be investigated is the difference, if any, between residents and tourists, supposing that people living in a certain location have a deeper knowledge of their environment compared to persons who are not residents. Then, focused data gathering could be performed to determine whether there is an increasing risk among tourists because of their unconsciousness of dangerous places in the territory, particularly in panoramic locations such as cliffs, which are often characterized by fragile equilibrium conditions.

From a geographic perspective, the robustness of data for developing countries is not guaranteed, and almost all the papers maintained that in those countries, the number of fatalities is largely underestimated due to several problems in data gathering, ensuring inclusion of only larger disasters and discarding landslides causing only a few fatalities.

Finally, a point not analyzed that should be investigated by future research is the relationship between the different types of landslides and the types of injuries that they can cause to supply useful indications for prevention and emergency management.

5. Concluding Remarks

It is impossible to protect people from landslides by extensively realizing structural works on each slope, because landslides are part of the natural landscape evolution and are countless and nearly ubiquitous. New landslides can occur almost anywhere—even on slopes that were previously stable—due to rainfall disrupting the slope stability conditions. It can be reasonably hypothesized that the extremes related to climate change can increase landslide activity, thus threatening larger parts of the world population.

The social relevance of the problem and its possible worsening due to climate change require increasing knowledge of people–landslide interactions to serve in planning successful strategies for landslide mortality reduction and improving people’s landslide risk awareness and resilience.

Throughout 10 years of scientific literature, this review highlights the factors leading to landslide mortality observed from different perspectives, examining (a) the geographic distribution of fatal landslides and their seasonality, trend and relationships with socioeconomic indicators; (b) landslide fatalities, their demographic characteristics, their behaviors, and places and dynamics of fatal accidents; and (c) clinical causes of fatalities and types of injuries affecting survivors.

The primary research gaps concern (a) the absence/scarcity of data on landslide mortality in developing countries, causing bias in the identification of hot spots of fatal landslides and temporal trends of fatalities at the global scale; (b) the gap between scientific work and the application of research findings to improve lives, causing an absence/scarcity of indications about proactive policies integrable in civil protection procedures and self-protective actions to reduce landslide mortality; (c) the absence of joint research involving both specialists working on landslides and researchers working on the clinical effects of landslides on people; and (d) the lack/scarcity of public databases on landslide fatalities allowing data reuse to promote advancement of the research and modeling of adverse impacts of climate change on landslide mortality.