A web-based GIS (web-GIS) database of the scientiﬁc articles on earthquake-triggered landslides

. Over the last 2 decades, the topic of earthquake-triggered landslides (EQTLs) has shown increasing relevance in the scientiﬁc community. This interest is conﬁrmed by the numerous articles published in international, peer-reviewed journals. In this work we present a database containing a selection of articles published on this topic from 1984 to 2021. The articles were selected through a systematic search on the Clarivate™ Web of Science™ Core Collection online platform and were catalogued into a web-based GIS (web-GIS), which was speciﬁcally designed to show different types of information. After a general analysis of the database, for each article the following aspects were identiﬁed: the bibliometric information (e.g. author(s), title, publication year), the relevant topic and sub-topic category (or categories), and the earthquake(s) addressed. The analysis allowed us to infer general information and statistics on EQTLs (e.g. relevant methodological approaches over time and in relation to the scale of investigation, most studied


Introduction
Earthquakes are one of the most threatening and devastating natural hazards worldwide and cause significant loss of life and damage to human structures and infrastructures.Although most of the damage is related to the partial or full collapse of buildings caused by ground shaking, in many cases other ground effects (e.g.tsunamis, liquefaction, landslides) can significantly increase the impact of the seismic event (e.g.Jibson and Harp, 2012).In particular, earthquake-triggered landslides (EQTLs, also "co-seismic landslides") are responsible for approximately 70 % of all earthquake-related fatalities not directly caused by ground shaking (Marano et al., 2010).Casualties caused by EQTLs are generally related to the collapse of buildings induced by downslope movements, which frequently tends to cover up previous structural damage caused by seismic shaking (Bird and Bommer, 2004).EQTLs often interrupt road networks and other transportation infrastructures, hampering rescue, supply, and recovery activities (Allstadt et al., 2022).EQTLs can also partly or completely block river channels, inducing downstream floods (e.g.Fan et al., 2012) and, in the longterm, causing changes to the drainage-basin characteristics (Keefer, 1999).For all these reasons, EQTLs are an important component of natural hazard assessment in seismically active areas, and in recent years the study of collateral seismic hazards has become an issue of increasing relevance (Wasowski et al., 2011), also in the framework of multi-risk assessments, accounting for potential cascading effects.
In this article we describe the results of a systematic search of scientific papers concerning EQTLs published in peerreviewed international journals in the past decades.Searching on the Clarivate™ Web of Science™ Core Collection online platform, we have collected 810 articles which were organized in a specific web-based GIS (web-GIS) database for the analysis.The identified articles deal with EQTL themes from numerous points of view, focusing on a wide range of aspects, such as spatial distribution in relation to earthquake location, triggering and propagation mechanisms, or impact on human activities.The aim of this work is to provide a tool which shows the geographical distribution of the articles published on this topic classified using different types of Published by Copernicus Publications on behalf of the European Geosciences Union.
In the database, the information defined for each article is the following: publication year (B1), author(s) (B2), article title (B3), and journal name (B4).The first article of our search was published in April 1984 (Keefer, 1984), and only 20 articles had been published on the topic until 1999 (Fig. 1).An increasing trend of the number of published articles can be noted in the following years, particularly from the late 2000s, with a peak in 2019 and 2020 (99 and 100 articles, respectively).The 810 identified papers were published in 160 different peer-reviewed journals.The Geological Society of America Bulletin has published the lowest number of articles but for the longest period, considering that the first article and last article were published in 1984 and 2020, respectively (Fig. 2).Engineering Geology, Landslides, and Geomorphology have published the largest number of articles (i.e.233, which represents 28.8 % of the total), covering a period from the mid-1990s-early 2000s until February 2021.Natural Hazards, Journal of Mountain Science, and Bulletin of the Engineering Geology and the Environment have published 105 papers (13.0 % of the total), while Natural Hazards and Earth System Sciences, Bulletin of the Seismological Society of America, Environmental Earth Sciences, Soil Dynamics and Earthquake Engineering, and Geosciences have published 118 papers (14.6 % of the total).It is worth noting that Geosciences has published the largest annual number of papers with 14 articles in a little more than 3 years (2018-February 2021).
Overall, 56.4 % of the articles in the literature database were published in the above-mentioned 11 journals, whose scopes include the analysis of natural hazards, geological and/or geotechnical engineering, and geomorphology.Among these journals, only two (Bulletin of the Seismological Society of America and Soil Dynamics and Earthquake Engineering) specifically deal with seismology and earthquake engineering.
The analysis of the database revealed that the average number of authors for each paper varied over the years.Ex-Table 1. Summary statistics of the categories and sub-categories used in the database.The column "Counts" provides the number of occurrences as given by the authors, when applicable.MCS denotes "Mercalli-Cancani-Sieberg".The earthquake catalogues are referenced in the text (Sect.Keefer (1984), the papers published before 2000 were written, on average, by two to three authors.In more recent years, the number increased to four to five authors, which suggests a need for different expertise to perform the wide range of research activities related to EQTLs.

Article topic
The articles collected in the database were classified into four categories considering the following main topics: regional landslide analysis -articles focusing on the analysis of EQTLs over large areas; single-landslide analysis -articles focusing on the analysis of single mass movements; other -articles presenting various types of research activities related to EQTLs, such as data presentation (e.g.Fortunato et al., 2012;Tanyaş et al., 2017;Villani et al., 2018;Rodríguez-Peces et al., 2020) and synthesis of historical information and review articles (e.g.Keefer, 1984Keefer, , 2002;;Bird and Bommer, 2004;Wasowski et al., 2011;Fan et al., 2019); not applicable -articles on topics only partially related to EQTLs (see Sect. 3.1.3).
The articles included in the "regional landslide analysis" and "single-landslide analysis" categories were further subdivided in sub-topics that characterize different aspects of the main research activity described in the paper (such as mapping, characterization, modelling).In many cases more than one sub-topic has been attributed to a single article.For the regional landslide analysis the sub-topics can be summarized as follows: 1. regional mapping -recognition of the spatial distribution of EQTLs; 2. regional landslide descriptive statistics -statistical analysis of the main physical and geometrical features of EQTLs (e.g.number, extension, volume, run-out, type of movement, involved material); 3. regional susceptibility/hazard assessment -modelling and zonation of susceptibility/hazard scenarios for EQTLs; 4. regional risk modelling -assessment of the effects on human activities of EQTLs occurring over large areas; 5. regional landslide comparison -analysis focused on the comparison between different landslide inventory maps, with the aim of evaluating general rules about earthquake-induced landslide occurrence.
For the single-landslide analysis, the sub-topics are as follows: 1. single-failure mapping -reconstruction of the surficial geometry; 2. single-failure geotechnical characterization -definition and description of the geotechnical parameters; 3. single-failure geophysical characterization -definition and description of specific parameters, which can be assessed using geophysical investigations (e.g.stiffness, shear modulus, moisture conditions); 4. single-failure modelling -reconstruction of singlelandslide events triggered by earthquakes; 5. single-failure impact/risk modelling -analysis focused on the assessment of the effects induced by a single EQTL on human activities (e.g.damaging of buildings); 6. single-landslide comparison -analysis focused on the comparison between single landslides triggered by earthquakes.

Earthquake information
The 810 identified papers describe ground effects related to 136 earthquakes.For each earthquake we collected a set of data using the ANSS Comprehensive Earthquake Catalog (ComCat), implemented by USGS (http://earthquake.usgs.gov/earthquakes/search/, last access: 1 October 2022) as the main source of information.In the catalogue, the available information depends on the date of occurrence.For earthquakes that occurred before 1900 (18 events), it was possible to attribute only a broad localization of the epicentre, an estimation of the date of occurrence, and the macroseismic intensity, which allows us to quantify the shaking level from observations of the effects on buildings, environment, and people (Masi et al., 2020).For earthquakes that occurred after 1900, it was instead possible to obtain more detailed information and it was possible to associate with each seismic event the data listed in Table 1: earthquake ID (D1), the name (D2), the time of occurrence (D3) in the format yy-mm-dd-hh:mm-ss UTC (coordinated universal time), and the country of occurrence (D4).A distinction was made between single and multiple events (D5): an earthquake was considered multiple when the seismicity near the main shock was characterized by at least 10 events with M w (moment magnitude) ≥ 5.In the database we have also indicated the focal mechanism (D6).Specifically, the main deformation style of the earthquake-generating fault was defined using the Kaverina-type double-couple (DC) classification diagram (Kaverina et al., 1996), which considers seven different fault mechanisms: normal, normal-strike-slip, reverse, reverse-strike-slip, strike-slip, strike-slip-normal, and strikeslip-reverse.From the Global Centroid Moment Tensor Catalogue (GCMTC) (https://www.globalcmt.org/,last access: 1 October 2022), we retrieved the moment tensor solution (D7) for 97 out of 136 investigated earthquakes.We used the moment tensor solution as input for the FMC software (Álvarez-Gómez, 2019), which allows us to automatically generate a Kaverina-type DC classification diagram.In the cases where the moment tensor solution was missing, we defined the focal mechanism from the specific literature if available (D8).The .Percentage of articles with respect to the main topics (a) and sub-topics for the regional landslide analysis (b), singlelandslide analysis (c), and not applicable (d) categories.The percentage is calculated with respect to the total number of articles included in the considered category (in brackets).RSHA: regional susceptibility/hazard assessment; RM: regional mapping; RLDS: regional landslide descriptive statistics; RLC: regional landslide comparison; RRM: regional risk modelling; SFMD: single-failure modelling; SFGTC: single-failure geotechnical characterization; SFMP: single-failure mapping; SFGPC: single-failure geophysical characterization; SLC: single-landslide comparison; SFIRM: single-failure impact/risk modelling; HHL: historical/Holocene landslides; WEL: water environment landslides; GDFL: ground deformation, failure, and liquefaction processes; NA: not applicable; VAE: vegetation and EQTLs; RIL: rainfall-induced landslides; LE: landscape evolution.
different types of magnitude, i.e.M L -Richter local magnitude (D9), M s -Richter surface-wave magnitude (D10), M w -moment magnitude (D11), and the Mercalli-Cancani-Sieberg (MCS) intensity scale (D12), were also added into the database, together with the epicentre coordinates (latitude -D13 and longitude -D14) and the hypocentre depth (D15).For each record, we specified the source of information (D16).As mentioned above, most of the information is derived from the ANSS catalogue, but for several historical earthquakes, we used the National Centers for Environmental Information World Data System (NCEI WDS) Global significant Earthquake Database (https://www.ngdc.noaa.gov/hazel/view/hazards/earthquake/search, last access: 1 October 2022), which includes over 5700 earthquakes from 2150 BCE to the present.

Structure and features of the web-GIS database
A dedicated web-GIS (http://194.119.218.119/it/map/a-web-gis-database-of-the-scientific-articles-on-e/qdjango/ 54/, last access: 1 January 2023) was implemented to manage and consult the articles where geographical data (i.e.country and/or earthquake information) are available.The web-GIS was implemented using G3W-SUITE (https://g3wsuite.it/en/g3w-suite-publish-qgis-projects/, last access: 1 January 2023), a modular client-server application that fully integrates the QGIS Python application programming interface (API) and allows us to publish and manage QGIS cartographic projects.The web-GIS contains two geographic layers and two tables.The first geographical layer refers to the earthquakes, which are located according to their epicentre (latitude -D13 and longitude -D14); the second refers to the countries based on the "world_boundaries" shapefile (https://public.opendatasoft.com/explore/dataset/world-administrative-boundaries/export, last access: 1 January 2023).The two tables ("articles_earthquake" and "articles_country") are associated with the corresponding geographic layers through a one-to-many relationship.
The web-GIS has a user-friendly interface consisting of different panels.In the left panel it is possible to explore and use the following information and tools: In the right panel there are commands for zooming in and out, measuring distances and areas, and uploading new layers, as well as a snapshot tool.In this panel, the user can also raise queries (selecting the "i" icon) on specific earthquakes or countries and visualize the main associated information.
3 Preliminary analysis of the literature database In the following sub-sections, we present and discuss a preliminary analysis on the main topics and sub-topics (Sect.3.1) and on the earthquakes addressed and described in the articles (Sect.3.2).In addition, in Sect.3.3 we have evaluated the information on the topics addressed and the listed earthquakes for inferring general and specific aspects of the EQTL scientific literature.

Analysis of article topic
The preliminary analysis of the articles included in the database allowed us to recognize different themes and commonalities and to group them according to main topics and sub-topics (see Sect. 2.1.2).In many cases, we assigned more than one main topic and/or sub-topic to a single article in relation to the variety of the analyses and activities performed.
As regards the main topics, Fig. 3a reveals that the most represented category is regional landslide analysis, with more than half (53.8 %) of the articles, followed by singlelandslide analysis (27.2 %), not applicable (19.5 %), and other (3.4 %).

Regional landslide analysis category
The two most numerous sub-categories in the regional landslide analysis category (Fig. 3b) include articles which discuss "regional susceptibility/hazard assessment" (RSHA) analyses and "regional mapping" (RM) activities (47.0 % and 43.8 %, respectively).The third sub-category ("regional landslide descriptive statistics" -RLDS) presents statistical descriptions of EQTLs data and contains about a quarter of the articles (25.2 %).Two further sub-categories discuss "regional landslide comparison" (RLC) and "regional risk modelling" (RRM) and contain only few articles (11.9 % and 5.7 %, respectively).
In the sub-category RSHA (205 articles), many of the articles use and describe modified versions of the Newmark (1965) model (e.g.Jin et al., 2019) or statistical methods such as logistic regression (e.g.Polykretis et al., 2019), artificial neural networks (e.g.Tian et al., 2019), or fuzzy logic (e.g.Razifard et al., 2019).In the set of articles that introduce modified versions of the Newmark model, the EQTL scenario related to a specific seismic event usually implies the exceedance of pre-established co-seismic displacement thresholds (Romeo, 2000), while in the group of statistical models, the outcome generally consists in susceptibility maps resulting from the weighting of different environmental factors (e.g.slope angle, geology, vegetation cover, ground shaking intensity).In both cases the analysis of the EQTL phenomenon concerns almost exclusively the triggering stage, while propagation and deposition processes are seldom examined (e.g.Guo et al., 2014).In addition, no specific methodological variations are presented with respect to the investigated landslide type (e.g.rockfall, earthflows).
The high number of papers included in the RM subcategory (191 articles) highlights the importance of landslide inventories for analyses over large areas.In this respect, it is important to point out the presence of different online catalogues which report information on EQTLs for recent and historical earthquakes.At a global scale, it is worth mentioning the Open Repository of Earthquake-Triggered Ground-Failure Inventories (Tanyas et al., 2017), which reports EQTLs and liquefaction effect data for 363 earthquakes, or the Earthquake Environmental Effects Catalogue (Guerrieri et al., 2015), which collects information about environmental effects triggered by specific seismic events, whose intensity is expressed according to the ESI (Environmental Seismic Intensity) scale (Michetti et al., 2007).
Of the 191 RM articles, 81 can also be classified within the third sub-category RLDS, since they include not only a description of one or more EQTL inventories but also a statistical analysis of the features of the mapped landslides (e.g.Roback et al., 2018).Conversely, 66 other RM articles are exclusively focused on mapping procedures, which are performed with different methods and techniques such as field investigation (e.g.Martino et al., 2017), analysis of aerial imagery (e.g.Saito et al., 2018), and satellite data (e.g.Hu et al., 2019).
As regards the RLC and RRM sub-categories, the relatively low number of articles (52 and 25, respectively) suggests the high specificity of the themes.In the first case, the comparison between landslide inventories can be performed for evaluating potential differences among earthquakes occurring in the same area (e.g.Jibson et al., 2020) or for in-  ferring general rules (Tanyaş and Lombardo, 2020) and/or correlations, e.g.magnitude vs. affected area (e.g.Marc et al., 2017) or ground motion vs. landslide size (e.g.Jibson and Tanyaş, 2020).In the case of RRM articles, the analyses are focused on specific anthropic elements such as buildings, roads, and railways (e.g.Vega and Hidalgo, 2016) or can be framed within a wider risk assessment (e.g.Martino et al., 2020).

Single-landslide analysis category
The largest number of articles (153 out of 220) within the single-landslide analysis category (Fig. 3c) is included in the "single-failure modelling" (SFMD) group.These articles thoroughly investigate the triggering (e.g.Zhang et al., 2018) and/or the propagation process (e.g.  by the authors, e.g.laboratory experimental testing (e.g.Pu et al., 2020), numerical modelling still represents the most widely used technique.Over the years, numerous types of increasingly sophisticated models have been developed (Jibson, 2011), which have allowed the simulation of complex physical processes related to EQTLs, for example progressive slope failure induced by strain-softening behaviour (e.g.Islam et al., 2019), dynamic fragmentation (e.g.Zhao and Crosta, 2018), and pore-water pressure variation on the sliding surface (Huang et al., 2019).
The SFMD sub-category also comprises 39 out of 68 articles included in the second sub-category ("single-failure geotechnical characterization" -SFGTC).In fact, numerical modelling often requires a detailed definition of the geotechnical parameters of the landslide body, which can be measured directly in the field (e.g.Gratchev and Towhata, 2010) or estimated through laboratory tests and parametric analyses (e.g.Dang et al., 2016).It is important to specify that, in many cases, the authors defined the input parameters of the numerical model simply based on literature values referring to the same or a similar landslide event (e.g.Nian et al., 2020).
Similarly, as described for regional scale, "single-failure mapping" (SFMP) sub-category includes all those articles dealing with mapping of a single-landslide body (e.g.Bozzano et al., 2008).Within this sub-category it is also possible to find almost all the articles belonging to the single-failure geophysical characterization (SFGPC).This point suggests how geophysical investigations are often associated with the detailed reconstruction of the geometry of the landslide (e.g.Havenith et al., 2002).
As regards "single-landslide comparison" (SLC) and "single-failure impact/risk modelling" (SFIRM), we identified a relatively low number of articles (12 and 11, respectively), as in the case of regional-scale analyses.In the first instance, the comparison concerns landslides that occurred during the same earthquake (e.g.Nakamura et al., 2014) or in response to other events that occurred in different regions (e.g.Aydan, 2016).In SFIRM articles, the effects of an EQTL on human activities can be estimated after a specific event (ex post assessment, e.g.Cui et al., 2012) or a priori through the reconstruction of specific single-landslide risk scenarios (e.g.Mousavi et al., 2011).

Not applicable category
As mentioned above, the category "not applicable" contains articles which address and discuss themes only partially related to EQTLs.Considering the not negligible number of articles (158), we performed their analysis and evaluation.On the contrary, we decided to disregard the "other" category, since the low number of articles ( 28) did not allow us to infer general aspects and significant commonalities.
The examination of the articles classified as not applicable led us to identify different sub-categories (Fig. 3d).The most numerous ("historical/Holocene landslides" -HHL) contains 53 articles (33.5 %) dealing with landslides that occurred in historical times (i.e.before 1900) and, more generally, during the Holocene.In the case of historical failures, only generic information about the triggering earthquake is available, while in the second one (landslides that occurred during the Holocene), it is often difficult to recognize with certainty the earthquake as a triggering factor.In these articles, the authors often use terms that refer to the possible nature of the landslide trigger, such as "was probably triggered" (Pérez-López et al., 2019) or "most likely triggered" (Lv et al., 2014).In this group, several authors also describe dating techniques, such as dendrochronological (e.g.Struble et al., 2020), lichenometric (e.g.Pérez-López et al., 2019), and isotopic analyses (e.g.Kojima et al., 2014).For landslide events that occurred in historical times, archival records often represent the main source of information (Koukouvelas et al., 2020).Back-analysis of such events was also used to assess the intensity and the location of historical earthquakes as described, for instance, by Rodríguez-Peces et al. (2011) for the 1755 Lisbon (Portugal) and 1884 Arenas del Rey (Spain) earthquakes.
Beyond the historical landslides, we also decided to include in the not applicable category all those articles dealing with landslides related to the geological hazard chain potentially resulting after an earthquake (Fan et al., 2019).In this respect, we identified three sub-categories: 1. water environment landslides (WEL) -this subcategory comprises 47 articles (29.7 %) specifically focused on the secondary effects of EQTLs in water environments, such as landslide-induced tsunamis (e.g.Takagi et al., 2019) and floods related to river landslide dams (e.g.Fan et al., 2012).In the case of historical earthquakes, EQTLs in water environments are generally used as evidence for confirming past tsunamis (e.g.Kitamura et al., 2020) and river damming events (e.g.Ehteshami-Moinabadi and Nasiri, 2019) or, more generally, the occurrence of one or multiple earthquakes in a specific area (e.g.Goto et al., 2010).2. rainfall-induced landslides (RIL) -the nine articles (5.7 %) included in this sub-category were classified as not applicable, since an earthquake represents a predisposing factor for subsequent rainfall-induced failures, i.e. post-seismic landslides (Tanyas et al., 2021).In general, these papers deal with comprehensive landslide hazard assessments (e.g.Hong and Adler, 2007) or cascading effects' evaluations (e.g.Tunas et al., 2020).
Another sub-category of articles classified as not applicable includes works mainly focused on other earthquake-induced ground effects which can occur concurrently with EQTLs.In particular, 22 articles (13.9 % of the total) describe and discuss EQTLs together with co-seismic "ground deformation, and failure, and liquefaction processes" (GDFL) (e.g.He et al., 2020).However, it is important to stress that liquefaction https://doi.org/10.5194/nhess-23-1789-2023Nat.Hazards Earth Syst.Sci., 23, 1789-1804, 2023 is a relevant and important earthquake-induced ground effect which frequently occurs in flat areas and thus independently of EQTLs.Many scientific articles are specifically focused on this topic, but they are not listed in our database, since the word "liquefaction" is not one of the search keywords.The analysis of "vegetation and EQTLs" (VAE) is addressed by 15 articles (9.5 %) and represents another subtopic.These articles are focused on the stabilizing effect of vegetation recovery at landslide sites after earthquake events (Yang et al., 2018) or, conversely, on vegetation change/alteration induced by EQTLs (e.g.Cheng et al., 2012).
Finally, 21 articles were classified as fully not applicable (NA), since the topic addressed is not related to EQTLs, such as the dynamic response of slopes based on in situ monitoring systems (Moore et al., 2011;Lenti et al., 2015) or the evaluation of the hydrological response of landslide bodies under seismic loading (Beyabanaki et al., 2016;O'Brien et al., 2016).

Analysis of earthquake information
As mentioned in Sect.2.1.3,the 810 articles collected in the database discuss and present ground effects of 136 earthquakes, mainly located in Italy, Japan, the USA, and China (Fig. 4).The earthquakes analysed in the articles agree with the typical distribution of the main seismicity at the global scale (Fig. 5), with most of the events located along the boundaries of the tectonic plates.About 80 % of the world's seismicity is situated along the circum-Pacific margin, corresponding to subduction zones, where earthquakes are caused by thrust-type and transcurrent-type faults (Toriumi, 2021).The focal mechanisms of the 136 earthquakes are predominantly associated with reverse and reverse-strike-slip faults (52 events, which correspond to 38.2 % of the total) and, secondly, to faults with a prevalent strike-slip component (45 events, which correspond to 33.1 % of the total) (Table 1 and Fig. 6).Moreover, the reverse-faulting mechanism characterizes not only most of the identified events but also the strongest (M w ≥ 8) ones, in agreement with the general observation that the largest earthquakes mainly occur in subduction zones (Funiciello et al., 2020).
As regards the intensity, 46 earthquakes (33.8 % of the total) have M w ≥ 7 (Table 2).Although it may seem like a large number, the data collected between 2000 and 2021 (https://www.usgs.gov/programs/earthquake-hazards/lists-maps-and-statistics, last access: 1 October 2022) indicate that the average yearly number of M w ≥ 7 earthquakes in the world is 15.This implies that strong earthquakes occurring worldwide are more numerous than those effectively analysed by the scientific community, at least in the framework of EQTLs.This can be explained considering that, in principle, the investigated earthquakes are those more impactful for human activities.In the specific case of EQTLs, scientific interest generally focuses on earthquakes occurring in populated areas or nearby.In this respect, most of the ar- ticles included in the database deal with earthquakes that occurred in the Asian inland and surrounding areas (Fig. 5).The devastating 2008 Wenchuan earthquake (M w 7.9) is certainly the most studied event (180 articles), while the 1999 Chi-Chi earthquake (M w 7.7) is the second-ranked one, although with a much lower number of articles (59) (Table 3).The most studied earthquake which did not occur on the Asian continent is the 1994 Northridge earthquake (M w 6.7) with 13 articles.In general, earthquakes that occurred in China are the most studied events in the framework of EQTLs (i.e.256 articles distributed among 13 earthquakes), followed by those located in Japan (i.e.78 articles for 16 earthquakes) (Fig. 7a-b).It is interesting to note that the earthquakes that occurred in Italy, albeit more numerous (i.e.19 events), are not as studied, since only 40 articles address these events, with a maximum value of 10 articles for the 2016 Amatrice earthquake (M w 6.2) (Fig. 7c).This point can be explained considering that earthquakes tend to arouse scientific interest mainly at national level.In this sense, only 2 out of 40 articles do not include Italian authors.As regards the 16 earthquakes that occurred in the United States (Fig. 7d), only the 1994 Northridge earthquake has been examined in more than 5 articles (Table 3).Thus, if we consider that only 30 articles of the database focus on earthquakes in the USA, we can assert that these events, despite their high magnitudes (average M w 7.1), have been analysed to a lesser extent, at least in relation to EQTLs.
Finally, it is interesting to highlight that 18 out of 136 earthquakes have been labelled as historical (i.e.before 1900).Only 20 articles focus on these events, with a maximum value of 4 papers for the 1786 Moxi earthquake (Ta- ble 4).Based on what is explained in Sect.3.1.3,these articles have been classified as not applicable, since they concern historical landslides.Specifically, 6 articles can also be included in the sub-category "water environment landslides", while only 1 article, which describes the effects of the 1811-1812 New Madrid earthquakes (Tuttle and Barstow, 1996), can be linked to the ground deformation, failure, and liquefaction processes sub-category.

Combined analysis of article topics and earthquake information
Based on the above information, we have performed combined analyses to evaluate the following: (i) the main topics and sub-topics addressed over time and (ii) the type of scientific evaluation performed for specific earthquake events.Figure 8 reveals that from the second half of the 2000s, the regional landslide analysis articles have always been more numerous than single-landslide analysis ones, with an almost constant proportion during the entire period.This confirms the increasing use, in the last few decades, of remote sensing data, especially satellite imageries, for the generation of EQTL inventories (Fan et al., 2019).This point, alongside the development of the calculation skills and performance of advanced computers, greatly enhanced and expedited the analysis of terrain data, especially over large areas.
As regards the analysis of specific earthquakes events, we selected the first five earthquakes listed in Table 3 (i.e. 1999Chi-Chi, 2004Niigata-Chūetsu, 2008Wenchuan, 2013Lushan, and 2015 Gorkha), for which more than 20 articles were published.The first three occurred more than 10 years ago and can be considered suitable for evaluating potential topic trends over the years.
In terms of the main topics, in the years immediately after the Chi-Chi earthquake, only single-landslide analysis articles were published (Fig. 9).On the contrary, in the first years after the Niigata-Chūetsu and Wenchuan earthquakes, https://doi.org/10.5194/nhess-23-1789-2023 Nat. Hazards Earth Syst.Sci., 23, 1789-1804, 2023 regional landslide analysis studies were also performed.For the Wenchuan earthquake, the number of regional landslide analysis articles is even greater than that of single-landslide ones until 2014, i.e. 6 years after the event.The increasing trend of regional landslide analysis articles is further confirmed by the 2013 Lushan and 2015 Gorkha events.This result suggests that, in recent years, researchers have performed regional analyses as a priority with respect to slopescale studies.Such a preference may be put in relation to the increasing availability of remote sensing data and highperformance computational tools, which enhance the execution of regional-scale analyses in a short time.On the contrary, the collection of data generally employed for slopescale analyses can be extremely time-consuming, thus affecting the timing of publication.Within regional landslide articles we also noticed that, in general, "mapping" (RM) is the prevailing activity in the first period after the event, while "modelling" (RSHA) tends to increase over time.RM articles can also still be published several years after the event: these works can include updates of existing inventories (e.g.Chen et al., 2020) or ex novo mapping for in-depth analyses of landslide activity (e.g.Liu et al., 2020).
As regards single-landslide analysis articles, modelling (SFMD) is the prevailing activity and shows an increasing trend over the years (Fig. 9).Alongside modelling, geotechnical characterization (SFGTC), and mapping (SFMP) are generally carried out, especially in the first years after the Chi-Chi and Niigata-Chūetsu earthquakes, while other types of sub-topics are substantially secondary.In the case of Wenchuan, different SFMP articles have also been published in recent years (e.g.Cui et al., 2017).In this respect, although more than 60 000 landslides were triggered during the event (Gorum et al., 2011), only a few failures have been investigated individually.This leaves room for the verification of the real triggering and conditioning mechanisms, which as a matter of fact rely only on a few studies.Among these, the Daguangbao landslide is the most studied (e.g.Luo et al., 2020), probably due to its great size (7.5 × 10 8 m 3 ).As regards the Lushan and Gorkha events, very few articles have been published in the framework of the single-landslide analysis topic; thus, it was not possible to observe any specific trend.

Final remarks and conclusions
In this work, we present a comprehensive database of the main scientific articles published in the last 4 decades on the EQTL theme.After the identification of 810 papers, the database was compiled in a GIS environment, specifying, for each article, different types of information.Great effort was dedicated to the identification of the earthquakes addressed and to the articles' grouping with respect to the main topics and sub-topics relevant to EQTLs.In this sense, we highlighted that more than 50 % of the articles focus on re-gional analyses which, in turn, are equally distributed among modelling and mapping sub-topics.On the contrary, singlelandslide analysis works, which represent just over a quarter of the total number of articles, deal more with modellingtype activities and, secondarily, with geotechnical characterization of the landslide body.As regards the earthquakes addressed, although a great part of the 136 identified events occurred in four different countries (Italy, Japan, the United States, and China), earthquakes which took place in China are by far the most studied events (256 articles), although 180 out of these 256 articles analyse the 2008 Wenchuan earthquake.Finally, by analysing the topics addressed over time for the five main earthquakes (1999Chi-Chi, 2004Niigata-Chūetsu, 2008Wenchuan, 2013Lushan, and 2015 Gorkha), we pointed out an increasing trend of regional landslide analysis articles in the last few years, alongside a growing use of modelling approaches for both regional-landslide-scale and single-landslide-scale analyses.However, it is important to stress that the considerations presented in this work just represent general inferences resulting from the analysis of the collected articles.More specific observations would require an in-depth, critical analysis of the articles, which is beyond the scope of the present study.In fact, this work can be considered a starting point for further analysis and investigations and is proposed to experts and the general reader interested in a broader view of EQTLs.
As regards future activities, it is important to note that the database can be updated very promptly.In this respect, authors invite all readers to report not only newly published articles on EQTL themes but also articles that have been published in the past and were not included in the database due to shortcomings in the use of keywords during the literature search on the Web of Science database.Future versions of the database can also be improved through integration with other sources of information, such as the ISC Event Bibliography.In conclusion, the web-GIS database presented here can represent a powerful tool for performing cross-correlated literature searches focused on specific main topics and sub-topics in relation to given earthquake events and/or study areas.

Figure 1 .
Figure 1.Analysis of the literature database listing 810 articles in the 37-year period from April 1984 to February 2021.The source of the article search was the Web of Science™ (formerly a Thomson Reuters™ product, now part of Clarivate™).The graph shows the number of articles per year (vertical blue bars, left y axis) and their cumulated number (solid blue line, right y axis).

Figure 2 .
Figure 2. Ranking of top journals in terms of number of articles published on EQTLs, according to the literature database.The number of published articles is reported in brackets next to the journal name.The colour of horizontal bars indicates the number of articles in six classes.

Figure 3
Figure3.Percentage of articles with respect to the main topics (a) and sub-topics for the regional landslide analysis (b), singlelandslide analysis (c), and not applicable (d) categories.The percentage is calculated with respect to the total number of articles included in the considered category (in brackets).RSHA: regional susceptibility/hazard assessment; RM: regional mapping; RLDS: regional landslide descriptive statistics; RLC: regional landslide comparison; RRM: regional risk modelling; SFMD: single-failure modelling; SFGTC: single-failure geotechnical characterization; SFMP: single-failure mapping; SFGPC: single-failure geophysical characterization; SLC: single-landslide comparison; SFIRM: single-failure impact/risk modelling; HHL: historical/Holocene landslides; WEL: water environment landslides; GDFL: ground deformation, failure, and liquefaction processes; NA: not applicable; VAE: vegetation and EQTLs; RIL: rainfall-induced landslides; LE: landscape evolution.

Figure 4 .
Figure 4. Number of earthquakes listed in the database for each country.The magnitude of each event is reported in the coloured bar.For each country, the lines with the black dot show the first and the last earthquake.

Figure 5 .
Figure 5. Spatial distribution of the 136 earthquakes identified in the literature search.Colour indicates the magnitude, while the size of the circle is proportional to the number of articles dealing with each seismic event.Countries are classified based on the number of articles collected in the database.

Figure 6 .
Figure 6.Definition of the focal mechanism of 97 out of 136 investigated earthquakes according to the Kaverina-type DC classification diagram.Several earthquakes are not plotted due to the lack of information concerning P , T , and B centroid moment tensor axes.

Figure 7 .
Figure 7. Spatial distribution of the earthquakes listed in the database for four countries, i.e.China (a), Japan (b), Italy (c), and the USA (d).Colour indicates the magnitude, while the size of the circle is proportional to the number of articles dealing with the earthquake.Countries are classified based on the number of articles collected in the database.

Figure 8 .
Figure 8.The graph shows the temporal distribution of the main topics addressed by the authors in the 37-year period from April 1984 to February 2021.

Table 2 .
Number of earthquakes identified for different intervals of moment magnitude (M w ).

Table 3 .
List of the main characteristics of the earthquakes analysed in more than five articles.

Table 4 .
List of historical earthquakes collected in the database.