Keywords

figure a

Introduction

It has been difficult to estimate the real impact of landslides on society and the environment. Owing to its nature, a complete documentation of the impact of landslide disasters is difficult to attain as the occurrence of other hazards directly linked to their triggering mechanisms, such as precipitation, earthquakes and volcanic activity, mask landslide consequences.

As already noted in earlier studies (Guha-Sapir and Below 2002; Marulanda et al. 2010; Kron et al. 2012), existing world-wide disaster databases have been developed to meet specific requirements and objectives, therefore, their use and applicability carry both, advantages and limitations. This is of particular relevance for research on landslide disasters.

Previous studies have reported a series of discrepancies among available datasets (Petley 2012; Kirschbaum et al. 2015; Froude and Petley 2018), and although several assays have recognised the general impact of high magnitude-low frequency landslide disasters, research has yet to systematically investigate the causality and effect of high frequency-low magnitude disasters associated with landsliding.

Notwithstanding the major drawbacks of landslide disaster databases, few studies have attempted to compare quantifications of the impact of landslide disasters. As reported by Froude and Petley (2018) underestimation of number of landslide disaster events and fatalities derived from EM-DAT database has been recognised. To this regard, Petley’s comparative study (Petley 2012) found for the period 2004–2010 an underestimation of the number of fatal landslide events by ~2000% and of the death toll by 430%, whereas Kirschbaum and collaborators detected that between 2007 and 2013, underestimations accounted for ~1400% of fatal landslide events, and 331% of life losses (Kirschbaum et al. 2015).

In an attempt to further contribute to the analysis of the estimations of landslide disaster impact, records from EM-DATFootnote 1 (CRED: EM-DAT 2019) repository were considered for the different regions of the world, and data for some countries included in both EM-DAT and DesInventarFootnote 2 (DesInventar 2019) databases were also used to identify additional disparities.

Methodology

Data for this paper were produced by CRED EM-DAT and DesInventar databases. For the former, an initial period between 1900 and 2019, was considered, whereas for the latter it varied from country to country.

A map of the spatial distribution of high frequency-low magnitude landslide disaster events, including information regarding number of landslide disaster events, associated human losses and people affected, for the different regions of the world, between 1900 and 2019, was created by using the information available in the EM-DAT database (Fig. 1).

Fig. 1
figure 1

Data source EM-DAT database

Spatial distribution of high frequency-low magnitude landslide disasters in terms of occurrence, associated human losses and people affected, between 1900 and 2019.

Full size image

Considering that this paper aimed at documenting some uncertainties associated with the information regarding the impact of landslide disasters, it was beyond its scope to undertake an assessment of landslide disasters at global scale, but rather to compare the results obtained from the two sources for the selected countries in terms of number of landslide disasters, human losses and total people affected.

Selection of countries for a further comparative analysis was carried out according to data availability for each of them in both data repositories, as follows: Argentina, Bolivia, Chile, Colombia, Costa Rica, Ecuador, Guatemala, Mexico, Nepal, Nicaragua, Perú, Sri Lanka, and Venezuela.

Results

EM-DAT: High Magnitude-Low Frequency Disaster Events

Notwithstanding the limited evidence of a consistent documentation of the impact of landslide disasters around the globe, results derived from EM-DAT database indicated that between the period 1900 and 2019, a total of 788 high magnitude-low frequency landslide disasters involved 70,884 human losses worldwide and affected a population of circa 14.5 million (Fig. 1).

Regarding the regional distribution of landslide disaster occurrence for the same period, Asia accounted for more than half of the number of all landslide reported disasters (56%). The Americas had a 25% share of the number of global landslide disasters, Europe and Africa both took less than a tenth of landslide disasters occurrence (9 and 7% correspondingly) and Oceania 3% (Fig. 1).

The number of human losses in Europe, Africa and Oceania showed values of 6%, 5% and 1%, respectively. Additionally, as much as 55% of global landslide disaster casualties were from Asia, compared to 33% from the Americas. On the other hand, more than a third of the total number of people reported affected was concentrated in Asia (59%), while in the Americas was as high as 39%. Africa, Europe and Oceania accounted for figures of 1.46, 0.26 and 0.15% accordingly (Fig. 1).

Discrepancies Between Databases on Selected Countries

Larger differences than those previously reported have been obtained when comparing landslide disaster data for thirteen countries on national basis between EM-DAT database and DesInventar. Although EM-DAT offers figures from 1900 to 2019, periods of time of analysed records presented here varied according to time frames of national disaster data bases comprised within DesInventar. Considering that contrasting the two databases by using percent points was skyrocketing, comparisons were made in orders of magnitude.

Costa Rica was the country with highest disparities regarding number of disasters, as in EM-DAT only one event was registered, whilst in DesInventar 3902 disasters were included between 1968 and 2019. Similar results were found for Colombia, Bolivia, Ecuador and Venezuela where records from EM-DAT showed 45, 8, 14 and 3 disasters respectively, while numbers in DesInventar were three order of magnitude higher with 53,605, 5498, 7899 and 1509 disasters, correspondingly. Nicaragua was the country with the lowest disparities; one disaster was included in EM-DAT and 47 in DesInventar. Annual average of landslide disasters based on the information provided by DesInventar was highest in Colombia with 515 events, and the lowest in Nicaragua, with 2 disaster events. When information for all countries was summed up, discrepancies between data represented three orders of magnitude. Total number of disasters estimated by EM-DAT was of 147, whereas in DesInventar it was as high as 82,944 (Table 1).

Table 1 Comparison between the number of landslide disasters registered in EM-DAT and DesInventar databases for selected countries
Full size table

The contrast concerning human losses was highest in Nicaragua, and the lowest in Guatemala, where 29 and 716 disasters were included in EM-DAT correspondingly, whereas 2044 and 1210 were registered in DesInventar, respectively (Table 2).

Table 2 Comparison between human losses associated with landslide disasters registered in EM-DAT and DesInventar databases for selected countries
Full size table

Total human losses for all selected countries differed in one order of magnitude. EM-DAT records included a death toll of 11,304, while in DesInventar was as high as 84,901 (Table 2).

In terms of the number of people affected by landslide disasters, the highest annual average of 289,316 was registered in Colombia. Guatemala and Mexico followed with 128,419 and 71,944 persons respectively. Results of the analysis yielded the highest difference of four orders of magnitude between DesInventar and EM-DAT for Mexico. Data collected for Sri Lanka, Costa Rica and Colombia involved three orders of magnitude, whereas in the case of Guatemala was of two orders of magnitude. The variation for Venezuela, Chile, Bolivia, Nicaragua and Argentina came about an order of magnitude, whilst in the case of Ecuador, Nepal and Perú was the same order of magnitude. Combined data for all countries produced a discrepancy of an order of magnitude between both databases (Table 3).

Table 3 Comparison between total people affected associated with landslide disasters registered in EM-DAT and DesInventar databases for selected countries
Full size table

Discrepancies in Latin-American Countries Between Databases (1970–2013)

Owing to availability of data, the period 1970–2013 was considered for further analyse nine Latin-American countries: Argentina, Bolivia, Colombia, Costa Rica, Chile, Ecuador, Mexico, Peru and Venezuela. According to DesInventar, highest number of disaster landslides occurred in Colombia (8484), Costa Rica (3407), and Peru (3382). Yet the differences of data between DesInventar and EM-DAT databases was of four, and three orders of magnitude for Costa Rica, and Venezuela, respectively, whereas for all other countries was two. Whereas in EM-DAT, a total of 108 disasters were recorder for all the countries, 22,373 disasters were included in DesInventar. This means a difference of two orders of magnitude when data for all countries was added on each database (Fig. 2).

Fig. 2
figure 2

Discrepancies regarding the number of landslide disasters for selected Latin-American countries between databases

Full size image

Annual average of number of disasters was highest in Colombia (N = 193), followed by Costa Rica (N = 77) and Peru (N = 77), and Ecuador (N = 61). Lowest values were estimated for Argentina and Chile with annual averages of six and nine disasters, respectively (Fig. 2).

Data concerning fatal victims showed a disparity of an order of magnitude between both databases for Peru, Mexico, Venezuela, Argentina and Costa Rica, whereas records included for Colombia, Ecuador, Bolivia and Chile were in the same order of magnitude. Joint records for all countries showed a difference of one order of magnitude. According to DesInventar, highest figures of human losses occurred in Peru, Colombia and Mexico, amounting 21,799, 5784 and 2701, with yearly averages of 495, 131 and 61, respectively.

Lowest number of fatal victims occurred in Costa Rica (N = 207), Argentina (N = 266) and Bolivia (N = 409). Total human losses estimated by EM-DAT for all countries were 8006, whereas in DesInventar were as high as 33,924 (Fig. 3).

Fig. 3
figure 3

Discrepancies regarding the number of human losses associated with landslide disasters for selected Latin-American countries between databases

Full size image

Resulting estimates in terms of affected people between DesInventar and EM-DAT indicated four, three and two orders of magnitudes for Mexico, Costa Rica and Colombia, correspondingly. Likewise, in the case of Chile, Venezuela, and Argentina a difference of an order of magnitude was identified, whilst, the difference found for Peru and Ecuador did not reach one order of magnitude. In contrast to all above mentioned data, records for Bolivia showed a difference of one order of magnitude between EM-DAT and DesInventar given the fact that records included in the former were of 170,653, whereas in the later corresponded to 30,237. The main difference of such data was the inclusion in EM-DAT of 165,000 people affected on February 1st, 1994 by landslides that were not accounted for in DesInventar. A difference of an order of magnitude between DesInventar and EM-DAT databases was estimated for overall data for all countries analysed. On the other hand, based on DesInventar, highest records of affected persons were recognised for Mexico, Colombia and Chile with totals of 3,165,546, 2,656,600 and 651,521, besides annual averages of 71,944, 60,377 and 14,807, correspondingly (Fig. 4).

Fig. 4
figure 4

Discrepancies regarding the number of total people affected associated with landslide disasters for selected Latin-American countries between databases

Full size image

Discussion and Concluding Remarks

The research data in this work was drawn from two main sources, which provide information regarding both, high magnitude-low frequency, and high frequency-low magnitude events. It has often proved difficult to separate out the landslide disaster impact from those caused by its triggering mechanisms. Due to information availability constraints, this paper cannot provide a thorough world-wide comparison of landslide disaster databases. Owing to disparities and lack of systematic information on the impact of landslide disasters at global, national, subnational and local scales in-depth landslide disaster risk assessment is impeded. It can be argued, however that engineers and Earth scientists have been keen to understand the dynamics of landsliding, through instrumentation, monitoring and mapping, as well as contributed to a great extent to the establishment of Landslide Early Warning Systems. Nonetheless, specific landslide disaster risk drivers have not been identified and hence, appropriately addressed. Current research and strategies regarding landslide disaster risk reduction and management reflect such weaknesses. Consequently, at a more everyday level, it would be important to systematically document, understand, and analyse the complex dynamics of the increasing vulnerability conditions of communities exposed to landsliding, particularly, but not exclusively, those strongly intertwined to the occurrence of high frequency-low magnitude events (Fig. 5).

Fig. 5
figure 5

Increasing vulnerability and exposure to landsliding, Teziutlán, Puebla, México

Full size image

A considerable amount of care needs to be taken to accurately document landslide risk and landslide disaster occurrence, so that a comprehensive understanding of the social construction of landslide disaster risk by all relevant stakeholders can be encouraged (Blaikie et al. 1994; Oliver-Smith et al. 2016, 2017). This would allow to recognise and deal with dynamic pressures or drivers of landslide disaster risk, in addition to enhancing the availability of consistent landslide hazard, vulnerability and exposure data, to fill the gaps of landslide knowledge (Alcántara-Ayala et al. 2017) that can be ‘useful, usable and used’ (Boaz and Hayden 2002).

At a general level, the analysis presented here raises a series of questions in need of detailed exploration from an integrated disaster risk research perspective (IRDR 2013) in regard not only of the temporal and spatial distribution of landslide disasters, but concerning vulnerability and exposure to landslides, as the main ingredients for the creation of landslide disaster risk. This type of research could provide useful insights into the ways in which landslide disaster risk research can be used for policy making and practice to promote disaster risk reduction and management (Alcántara-Ayala et al. 2015; Cutter et al. 2015).

Finally, when properly documented, and owing to the mounting vulnerability of communities exposed to landslides due to lack of adequate territorial management, from which main landslide disaster risk drivers derive from, coupled with climate change, sooner than later the impact of landslide disasters worldwide could exceed that of the disasters triggered by floods.