1 Introduction

According to EM-DAT (The International Disaster Database), since 1900 there have been more than 16,000 natural disasters worldwide that have resulted in 32 million deaths and nearly 5.9 trillion dollars in economic, social and environmental losses (EM-DAT 2022). The trend of increasing frequency and damage caused by natural disasters, particularly the occurrence of storms, floods, droughts and landslides, has been linked to climate change (Knutson et al. 2010; Cook et al. 2018; Froude and Petley 2018; Kimuli et al. 2021). The continuous growth of the world population, together with patterns of economic development, which have given rise to uncontrolled urbanisation, leading many people to inhabit areas highly exposed to risk (Holcombe et al. 2016).

The compilation of data about any type of natural hazard (e.g. earthquakes, volcanic eruptions, storms, landslides, floods and tsunamis), including location, date of occurrence, triggering factors, economic and/or social impact, constitutes a primary approach to assess risk (Glade et al. 2001, Devoli et al. 2007, Van Den Eeckhaut and Hervás 2012). Moreover, reliable disaster databases are crucial for assessing the geographic distribution, temporal trends, the impact of a particular natural hazard type (Glade et al. 2001), and for evaluating the effectiveness of implemented measures to mitigate risk over time.

Natural hazards databases have different characteristics depending on their purpose and can be prepared based on various criteria and spatial and temporal scales. At the international, regional and national levels, there are several examples of natural and technological disasters databases. The EM-DAT is the most widely referenced database on disasters occurred worldwide (Panwar and Sen 2020). This database is compiled from diverse sources and includes data on natural and technological disasters occurred since 1900. However, EM-DAT employs stringent criteria for an event to be considered as a disaster and to be included in the database. An event must accomplish one of the following conditions: (1) occurrence of ten or more fatalities; (2) existence of 100 or more affected people; (3) declaration of a state of emergency; and (4) request for international assistance. Since the criteria considered by EM-DAT are rather restrictive, many events that have occurred on the Azores, which caused significant economic and social losses, are not included in this database. The EM-DAT reports six natural disasters for Azores, including only one landslide event. Furthermore, there are errors in the information about natural disasters that occurred in the Azores, including inaccuracies in event dates and the number of fatalities.

Due to the geographic and geodynamic setting of the Azores archipelago, there is much geomorphologic evidence and numerous documental references that mention the occurrence of landslides over time, triggered by different factors (e.g. earthquakes, volcanic eruptions and rainfall). Landslides have been affecting São Miguel Island since its settlement in the mid-fifteenth century. Historical accounts have documented several landslide events with significant socio-economic impacts. On 22 October 1522, an earthquake with a maximum intensity X (EMS-98) (Silveira et al. 2003) triggered several landslides over the island, one of which was a debris flow of about 6.75 × 106 m3 that covered an area of 4.5 km2 (Marques et al. 2009). These landslides killed 5,000 people and destroyed the entire village of Vila Franca do Campo, the island’s capital at that time (Frutuoso 1981). Through time, smaller landslides have been responsible for many deaths and generated major socio-economic impact in São Miguel. On 31 October 1997, about 1,000 landslides were triggered by a heavy rainfall episode (Gaspar et al. 1997), causing the death of 29 people in the parish of Ribeira Quente (Povoação municipality) together with seven injuries and 69 people made homeless. An economic loss of 21.3 million euros was estimated for this event (Cunha 2003).

Landslide databases are fundamental tools used to assess landslide susceptibility, hazard and risk. They are important for supporting investigations regarding the spatial and temporal distribution of landslides and for predicting their occurrence in future (Kirschbaum et al. 2015). Historical landslide databases have been developed for various countries and regions, mostly in Europe: such as Spain (Cuesta et al. 1999), Italy (Guzzetti et al. 2001; Niculiță et al. 2017; Piacentini et al. 2018), Ireland (Creighton 2006), the UK (Foster et al. 2008), Norway (Jaedicke et al. 2009), Switzerland (Hilker et al. 2009), Greece (Kalantzi et al. 2010), Portugal (Zêzere et al. 2014; Pereira et al. 2014), Germany (Damm and Klose 2015), Albania (Jaupaj et al. 2017) and Czech Republic (Bíl et al. 2021). In other countries, such as Nicaragua (Devoli 2005), the USA (Elliott and Kirschbaum 2007), Canada (Septer 2007; Blais–Stevens et al. 2015), Nepal (Petley et al. 2007) and China (Li et al. 2016), there are also some studies on historical landslide databases. On a global scale, some authors have focused on compiling and comparing databases, analysing the data, and engaging in discussions regarding the rules for database construction and harmonisation (e.g. Van Den Eeckhaut et al. 2013; Herrera et al. 2018; Gómez et al. 2023).

This paper aims to characterise the occurrence of landslides on São Miguel Island in the period 1900–2020 based on NATHA database (Natural Hazards in Azores; Marques 2013). The specific objectives of this work are (1) to analyse the temporal trends of landslide events; (2) to analyse the geographic distribution of landslide events; (3) to analyse the triggering factors of landslide events; and (4) to analyse the impact and the spatiotemporal distribution of fatalities caused by landslide events.

2 Study area

The Azores archipelago is located in the North Atlantic Ocean and is composed of nine volcanic islands (Fig. 1). From a tectonic perspective, the archipelago is located at the triple junction between the Eurasian, African and North American Plates. As a result of this context, seismic and volcanic events are common. However, landslides are the most prevalent type of geologic hazard, primarily due to the volcanic composition and geomorphology of the islands (Marques 2013; Silva et al. 2018).

Fig. 1
figure 1

a Geographic location and b municipalities (black lines) and morphostructural units of São Miguel Island: 1–Sete Cidades Volcano; 2–Picos Fissural Volcanic System; 3–Fogo Volcano; 4–Congro Fissural Volcanic System; 5–Furnas Volcano; 6–Povoacão Volcano and 7–Nordeste Volcanic Complex

Full size image

São Miguel Island has an area of 744.6 km2, with a maximum length of 64 km and width ranging between 8 and 15 km. The elevation ranges from 0 to 1143 m a.s.l. The island has a total of 133,390 inhabitants (INE 2021), divided into six municipalities: Lagoa, Nordeste, Ponta Delgada, Povoação, Ribeira Grande and Vila Franca do Campo (Fig. 1b). There are three active central volcanoes on the island, all with summit calderas: Furnas Volcano (Guest et al. 1999, 2015), Fogo Volcano (Wallenstein 1999; Wallenstein et al. 2015) and Sete Cidades Volcano (Queiroz 1997; Queiroz et al. 2015). Landslides commonly occur in distinct morphologic structures characterised by steep slopes, such as the inner walls of calderas volcanoes, fault scarps, sea cliffs and stream valleys, mostly composed of non-cohesive materials (pumice deposits) (Fig. 2a).

Fig. 2
figure 2

Examples of damaging landslide occurrences in a Furnas Volcano and b Povoação Volcano

Full size image

The central volcanoes are connected by the two fissural volcanic systems: Picos (Ferreira 2000; Ferreira et al. 2015) and Congro (Booth et al. 1978; Guest et al. 1999). The presence of extensive outcrops of basaltic lava flows and scoria strongly influences the morphology of the Picos Fissural Volcanic System. This region is characterised by gentle slopes and a poorly developed drainage network. Landslides in this area are mostly concentrated along the sea cliffs. Congro Fissural Volcanic System is dominated by scoria cones and associated lava flows. Additionally, there are some pumice cones, domes and maars. The entire area is covered by thick pumice fallout deposits from explosive eruptions of Fogo and Furnas volcanoes. This area corresponds to an elevated region in the axial zone with elevations exceeding 500 m a.s.l. and gentle slopes that rise as it approaches the north and south coastlines. Landslides are common in this region, particularly along coastal cliffs and in deep narrow valleys.

The Nordeste Volcanic System and the Povoação Volcano are the oldest geologic formations of the island (Duncan et al. 2015). The Nordeste Volcanic System corresponds to a shield volcano composed mainly of basaltic lavas, which was subjected to extensive erosion that occurred both during and after the volcano activity. Povoação Volcano has an early lava shield followed by explosive trachytic activity and caldera collapse. Landslides affecting the Nordeste Volcanic System occur in areas where weathered basaltic lava flows (s.l.) are combined with steep slopes and very deep narrow valleys where intense fluvial erosion occurs. In Povoação Volcano, landslides are associated with the non-cohesive materials, predominantly pumice, which underwent significant erosion both during and after its volcanic activity (Fig. 2b).

According to the Köppen–Geiger climate classification (Köppen 1936), the Azores has a temperate climate (C). Due to the spatial distribution of the islands, the climate can be classified, from east to west, as a transition between the Cs and Cf subtypes. São Miguel Island exhibits a climate with distinctly Atlantic characteristics, where two well–defined seasons can be identified based on the precipitation patterns: a wet season from October to March and a dryer season from April to September. The mean annual precipitation on São Miguel Island ranges from 1,200 to 2,000 mm near the coast and reaches up to 4,800 mm in the higher inland areas of the island (Fernandes 2004).

3 Design of the NATHA database

3.1 Data collection

The NATHA database (Marques 2013) is a repository of documents relating to natural disaster events that have occurred on the Azores. The database contains information regarding earthquakes, volcanic eruptions, degassing phenomena, landslides, storms, floods and tsunamis that have occurred since the settlement of the Azores in the mid-fifteenth century. The data collection process was based on historical accounts, scientific articles, theses, technical and scientific reports and newspapers.

In this study, the data were obtained exclusively from newspapers. Like in similar studies using newspapers as the primary source of data (e.g. Zêzere et al. 2014), the registered landslide events caused impact, such as fatalities, people made homeless, damage to buildings or road blockages. More than 55,500 newspaper specimens were consulted. Two newspapers were selected as the primary press archive: Açoriano Oriental and Correio dos Açores (Fig. 3). Both newspapers have a regional coverage and are published on São Miguel Island. Açoriano Oriental is the oldest newspaper in the Azores and Portugal and is still published today. It has been published weekly since 18 April 1835 and daily since 1 January 1979. Correio dos Açores has been published daily since 1 May 1920. In addition, six local weekly newspapers were surveyed to complement or validate certain events (Table 1).

Fig. 3
figure 3

Examples of newspaper specimens reporting landslide disaster events: a article from Açoriano Oriental describing a landslide that killed seven people in 1942; b newspaper cover of Açoriano Oriental describing a landslide that caused 29 fatalities in 1997 and c newspaper cover of Correio dos Açores describing a landslide that caused three fatalities in 2013

Full size image
Table 1 Newspaper consulted to build the NATHA database
Full size table

3.2 Database structure

The bibliographic data in the NATHA database are organised according to the type of consulted documents (Fig. 4). Whenever possible, extracts from the catalogued documents were included as images, obtained through photography or by scanning the original documents.

Fig. 4
figure 4

NATHA database structure

Full size image

Each natural event included in the NATHA database consists of three groups of data: (1) information, (2) characterisation and (3) impact. The information field contains the date, time and the type of event (earthquake (EA), volcanic eruption (ER), degassing phenomena (DG), storm (ST), landslide (LD), flash flood (FF) and tsunami (TS)). In the case of landslides, an event refers to a specific date that was characterised by slope instability occurrence, which may involve multiple landslides. The characterisation field contains specific details about each natural event, such as the type of landslide and the triggering factor. The impact field includes spatial information about the event, including the island, municipality, parish and locality. It also includes data on fatalities, injuries, destroyed buildings, people made homeless and costs associated with the event, when available.

4 Results

4.1 Temporal trends of landslide events

The annual distribution of landslide events that caused impact on São Miguel Island during the time period 1900–2020 is presented in Fig. 5. A total of 236 landslide events occurred on São Miguel Island were catalogued. When examining the temporal trends of these events, it is evident that there is no consistent increment and/or pattern in their distribution over time. However, two time periods can be distinguished: 1900–1995 and 1996–2020.

Fig. 5
figure 5

a Temporal and b cumulative distribution of landslide events occurred on São Miguel Island in the period 1900–2020

Full size image

In the first period (1900–1995), the landslide events were less frequent than in the second period (1996–2020). The first time period represents 79% of the entire time series but includes only 42% of the registered events. The linear regression for the period 1900–1995 indicates an annual increase rate of 1.07 in landslide events (Fig. 5b). However, it is noteworthy that the periods of 1930–1949 and 1980–1989 exhibit a higher concentration of events.

The second time period covers 21% of the total tome series and contains 58% of the total landslide events. The rate of landslide events shows a significant increase of 5.74 events per year, which is five times higher than in the period 1900–1995 (Fig. 5b). During the period 1996–2020, landslide events were recorded every year. Moreover, more than ten events were identified in the years 1996, 2007, 2010 and 2013, which were very wet years.

The cumulative distribution of landslide events that occurred in each municipality of São Miguel Island highlights some differences between municipalities (Fig. 6). During the period 1900–1995, Povoação registered the highest number of landslide events. Povoação and Ribeira Grande exhibit a similar increase in the number of events per year (0.35 and 0.32, respectively). In comparison, Ponta Delgada, Vila Franca do Campo, Lagoa and Nordeste registered a lower number of events, with increase of 0.27, 0.17, 0.07 and 0.09 events per year, respectively.

Fig. 6
figure 6

Cumulative distribution of landslide events that occurred in each municipality of São Miguel Island in the period 1900–2020

Full size image

The 1996–2020 period was characterised by a significant increase in the number of events registered in all municipalities of the island. Ponta Delgada exhibits the highest increase (1.81 events per year), which is the approximately seven times higher than the first period. Similarly, Ribeira Grande and Povoação show substantial increase (1.76 and 1.59 events per year, respectively), which are about five times higher than the 1900–1995 period. Vila Franca do Campo, Lagoa and Nordeste also experienced an increase in the number of events during the 1996–2020 period, although lower when compared to the other municipalities.

4.2 Spatial distribution of landslide events

All municipalities of São Miguel Island have been affected by landslide events during the period 1900–2020. Povoação is the municipality with the highest number of landslide events, accounting 76 events (25% of the total). It is followed by Ponta Delgada (71 events, 23%), Ribeira Grande (69 events, 22%) and Vila Franca do Campo (47 events, 15%). In comparison, the municipalities of Nordeste and Lagoa registered a lower number of landslide events (26 and 21, respectively), corresponding to 8% and 7% of total landslide events, respectively.

The spatial distribution of these events across the municipalities normalised by 10 km2 is shown in Fig. 7. Povoacão is the municipality where most landslide events occurred (more than 6 per 10 km2). Ponta Delgada, Ribeira Grande Vila Franca do Campo and Lagoa registered from 3 to 6 events per 10 km2 and Nordeste is the municipality with less landslide events (less than 3 per 10 km2). It is worth noting that the distribution of landslide events summarised in Fig. 7 represents neither the susceptibility nor the density of landslides of each municipality, but rather the recurrence of events that had some kind of impact and, as such, were referred to in the newspapers.

Fig. 7
figure 7

Spatial distribution of landslide events (number per 10 km2) in each municipality of São Miguel Islandin the period 1900–2020

Full size image

4.3 Triggering factors of landslide events

The landslide events were classified based on their triggering factors, which include rainfall, earthquake, sea erosion and anthropic. Those cases where it was not possible to discriminate the triggering factor were classified as unknown. Rainfall was the triggering factor responsible for 70% of the registered landslide events on São Miguel Island. Sea erosion was the trigger of 8% of the events, followed by anthropic factors (4%) and earthquakes (2%). The relative frequency of these triggering factors in each municipality of São Miguel Island is depicted in Fig. 8, with rainfall unsurprisingly dominating as the primary factor across all municipalities.

Fig. 8
figure 8

Relative frequency of the triggering factors in each municipality of São Miguel Island

Full size image

4.4 Analysis of the rainfall effect on landslide events

To gain better understanding of its impact on the occurrence of rainfall triggered landslide events, a detailed analysis of rainfall was made. For each municipality of the island, a specific meteorological station (MS) was selected for the rainfall analysis: Sete Cidades MS (Ponta Delgada), Chã da Macela MS (Lagoa), Fogo III MS (Ribeira Grande), Praia MS (Vila Franca do Campo), Monte Simplício MS (Povoacão) and Algarvia MS (Nordeste). Figure 9 illustrates the monthly average rainfall distribution registered in these meteorological stations for the period 2010–2020. The monthly rainfall distribution reveals a seasonal pattern, in which the 'rainiest season' lasts from October to March, while the summer period is the 'driest season', with minimal rainfall levels occurring in July. This depiction of the monthly rainfall distribution provides valuable insights into the temporal variations in precipitation on the island, which are vital for comprehending the underlying conditions contributing to landslide events.

Fig. 9
figure 9

Monthly rainfall average for different meteorological stations in São Miguel Island

Full size image

Figure 10 shows the monthly distribution of landslide events triggered by rainfall on São Miguel Island in the period 1900–2020. These events primarily take place during the rainiest months of the year, particularly between November and March, constituting 78% of the recorded events. The highest number of events occurred in December, followed by January and February with roughly the same number. The number of events registered from May to August is much lower, as this is the less rainy period along the year.

Fig. 10
figure 10

Monthly distribution of landslide events on São Miguel Island in the period 1900–2020

Full size image

Figure 11 illustrates the monthly distribution of landslide events triggered by rainfall and their associated fatalities on São Miguel Island from 1900 to 2020. The data reveals that the maximum number of fatalities occurred in October and March. This is primarily attributed to two extreme events that took place in 1942 and 1997. Notably, the number of fatalities is not directly proportional to the number of landslide events. While December, January, February and March account for approximately 60% of the total events, they only represent around 30% of the fatalities. This indicates that the severity of the events during these months, in terms of loss of life, may be relatively lower compared to other periods, but additional data are needed to confirm this hypothesis.

Fig. 11
figure 11

Monthly distribution of landslide events triggering by rainfall and respective fatalities occurred on São Miguel Island in the period 1900–2020

Full size image

While there is a relationship between the monthly rainfall average and the occurrence of landslide events triggered by rainfall, the same does not hold for the number of fatalities resulting from landslide activity. In fact, fatal events are typically the consequence of landslides triggered by extreme rainfall episodes characterised by high intensity and short duration, which can occur during months with low monthly rainfall average. Therefore, it is crucial to consider not only the total rainfall amount but also the intensity and duration of rainfall events when assessing the potential impact of landslides on fatalities.

To gain better understanding of the influence of rainfall on the temporal trend of the number of landslide events, an analysis of the daily precipitation for the period 1900–2020 was conducted using data from the Ponta Delgada Meteorological Station (the only centenary station on São Miguel Island). Figure 12 shows the cumulative annual precipitation and the number of days per year with daily rainfall exceeding 65 mm. This threshold value is based on the rainfall intensity-duration threshold equation for landslide triggering calculated for the Povoação municipality (Zêzere et al. 2015), normalised by the mean annual precipitation of Ponta Delgada Meteorological Station. The daily rainfall threshold of 65 mm was exceeded 29 times during the period 1900–1995 (0.3 per year in average). In the most recent period (1996–2020) the same threshold was exceeded 17 times (0.7 times per year in average). Notably, the year 1997 was an exceptional one, with the daily rainfall threshold exceeded four times. The accumulated annual precipitation does not show a clear trend. It is possible to identify an increase in the late 1930s that remains constant over the years, with drier and wetter years. The values of cumulative annual precipitation show a decrease over the years 2013 to 2020.

Fig. 12
figure 12

Number of days per year with daily rainfall exceeding 65 mm (blue bars) and the cumulative annual precipitation (orange curve)

Full size image

4.5 Impact of landslide events

The landslide events catalogued in the NATHA database were responsible for 82 fatalities, 41 injuries, 305 people made homeless and 66 buildings partially or totally destroyed. The temporal distribution of fatalities caused by landslide events is shown in Fig. 13a. In average, 0.7 fatalities were recorded per year, and the ratio between deaths and total events (mortality index) is 0.35. The obtained ratio is strongly influenced by the outlier event of 31 October 1997. When this outlier event is excluded from the analysis, the mortality index drops to 0.22. The events with the highest number of fatalities occurred on 14 October 1942 and 31 October 1997 and were both triggered by very intense rainfall. The first event occurred in the parish of Furnas (Povoação municipality) and led to seven fatalities. The second event took place in the parish of Ribeira Quente (Povoação municipality) and led to 29 fatalities.

Fig. 13
figure 13

a Temporal and b normalised cumulative distribution of fatalities caused by landslide events occurred on São Miguel Island

Full size image

Figure 13b shows the increasing number of fatalities per year on São Miguel Island, normalised by 10,000 inhabitants based on the number of residents per decade. Linear regression functions were computed for two periods: 1926–1995 and 1996–2020. The period 1900–1925 (no fatalities) and the events of 14 October 1942 and 31 October 1997 (outliers) were excluded from the analysis. The annual mortality increased by 0.03 fatalities/year/10,000 inhabitants during the period 1926–1995. In the most recent period (1996–2020), the annual mortality is twice as high increased by 0.06 fatalities/year/10,000 inhabitants.

The distribution of impact varies among the municipalities of São Miguel Island despite the relative similarity in the number of landslide events in Povoação, Ribeira Grande and Ponta Delgada (Fig. 14). Povoação stands out with a high number of events that resulted in fatalities and significant damage. Specifically, Povoação recorded 48 fatalities, accounting for approximately 59% of the total fatalities documented in the database. Ponta Delgada stands out for the number of buildings destroyed and the number of people made homeless as a result of landslides. The other municipalities on the island registered a more limited impact.

Fig. 14
figure 14

Number of fatalities, injuries, buildings destroyed and people made homeless caused by landslide events in each municipality of São Miguel Island

Full size image

The type of impact caused by landslide events, as described by newspaper reports, is important to understand the variation of the number of events through time. Figure 15 shows the temporal distribution of landslide events according to their impact. For the period 1900–1995, the database contains 59 landslide events that causing road blockages (0.6 per year on average) and 40 events causing fatalities, injuries, buildings destroyed and people made homeless (0.4 per year on average). For the most recent period (1996–2020), the database contains 109 landslide events causing road blockages (4.4 per year on average) and 28 events causing fatalities, injuries, buildings destroyed and people made homeless (1.1 per year on average).

Fig. 15
figure 15

Temporal distribution of landslide events according to their impact

Full size image

The F–N plot in Fig. 16 illustrates the relationship between the frequency of landslide events and the corresponding number of fatalities on São Miguel Island. This plot serves as a tool for comparing the effects of natural disasters with established or acceptable criteria for societal risk assessment, as explained by (Guzzetti 2000). Based on the results and applying the most used risk acceptable criteria (Fell et al. 2005), it can be concluded that the societal risk associated with landslide events on São Miguel Island is considered unacceptable.

Fig. 16
figure 16

Frequency vs consequences (F–N plot) for landslide events that caused fatalities on São Miguel Island

Full size image

5 Discussion

The NATHA database was established in 2013 and serves as a repository for documental information regarding various natural hazards that have occurred in Azores. The data used in this work were collected by consulting approximately 55,500 regional and local newspapers. The temporal coverage of these newspapers is not uniform, and the period from 1900 to 2020 is not consistently represented in terms of the number of available newspapers. The Açoriano Oriental newspaper provides coverage for the entire studied period, while the Correio dos Açores newspaper has records starting from 1920 onwards. The early years of the time series are supported by the existence of six local newspapers, which contribute to the comprehensiveness of the data during that period. However, the data collection based on press archives have some limitations (Ibsen and Brunsden 1996; Cuesta et al. 1999; Guzzetti and Tonelli 2004): (i) many archives only record landslides with higher magnitude or that caused significant impact; (ii) the spatial coverage of the events is sometimes compromised, because the information sources do not describe the complete affected locations; (iii) it is not always possible to determine the triggering factor of the landslides, account for the costs or obtain the number of fatalities, injuries, homeless and destroyed infrastructures; (iv) changes in the quality and quantity of the news published by newspapers may have occurred along time, some related to the dictatorial regime of the country during the period 1933–1974; and (v) the journalist's experience and perception, as well as semantics in their descriptions introduce additional uncertainties.

A total of 236 landslide events were documented on the São Miguel Island during the period from 1900 to 2020, resulting in an average 1.9 events per year. The temporal distribution of events allows for the identification of two distinct periods: 1900–1995 and 1996–2020. The first period recorded a relatively low number of events (99), averaging 1.0 event per year. In contrast, the second, shorter in time, had a higher number of events (137), averaging 5.5 events per year. The results obtained are consistent with what has been observed in global landslide databases, which indicate an increase in the number of landslides in recent decades (Gómez et al. 2023).

The increasing number of landslide events seems to be related to a change in the precipitation regime. The analysis of daily rainfall showed that since 1996 the number of days exceeding the daily rainfall threshold of 65 mm was more than double in comparison to the initial period (0.3 per year in 1900–1995; 0.7 per year in 1996–2020). Given that rainfall is the primary trigger for landslides on São Miguel Island, this alteration in the rainfall pattern can, at least in part, account for the rising number of landslide events catalogued in the database. It is worth noting that the annual cumulative precipitation did not increase in the period 1996–2020, and therefore the increase in the number of events is due to the occurrence of extreme precipitation episodes, characterised by high intensity and short duration. This observation aligns with climate change scenarios that anticipate a heightened likelihood of extreme precipitation events (IPCC 2021). For instance, Ratnayake and Herath (2005) show that the increasing trend in rainfall intensity increases the number of landslides in Sri Lanka, and Pereira et al. (2014) show that the temporal trend of landslide activity in the north of Portugal follows the rainfall regime. Furthermore, an analysis of the impact of the landslide events reported in newspapers revealed that the rising trend in the number of events during the most recent period (1996–2020) can be attributed in part to landslide events primarily resulting in road blockages. Although landslides events with fatalities, injuries, buildings destroyed and people made homeless also increased (0.4 events per year in 1900–1995; 1.1 events per year in 1996–2020), the landslide events that caused road blockages increased 7 times and correspond to 80% of catalogued events since 1996, while in the previous period they corresponded to 60% of all events. The increasing availability and dissemination of information (e.g. coming from releases from the Regional Civil Protection Service, which was created in 1980) partially justify the increasing the number of landslide events that were reported by newspapers during the most recent period.

The spatial distribution of the landslide events showed differences between municipalities of São Miguel Island. Povoação, Ponta Delgada and Ribeira Grande municipalities have similar number of events in absolute terms, indicating a relatively higher occurrence in those areas. On the other hand, Vila Franca do Campo, Nordeste and Lagoa municipalities have a smaller number of events, indicating a lower frequency in those locations. The spatial distribution of these events across the municipalities normalised per unit area (10 km2), further enhances the incidence of landslide events in Povoação. The high number of landslide events in Povoação is due to the geomorphology and geology, characterised by steep slopes, such as the inner walls of calderas volcanoes, fault scarps, sea cliffs and stream valleys, mostly composed of non-cohesive materials (pumice deposits). Notably, this municipality has 53.6% of its area classified within the high and very high classes of landslide susceptibility (Marques 2013).

The rainfall was the dominant triggering factor along the island comprising over 70% of the events, similarly to what has been observed in numerous landslide national and international databases (e.g. Hadmoko et al. 2017; Lin and Wang 2018; Gómez et al. 2023). Landslide events induced by earthquakes are concentrated in the municipalities of Povoação, Ribeira Grande and Vila Franca do Campo, where the most seismicity on the São Miguel Island have been recorded (Silva et al. 2020). The low percentage of landslide events triggered by earthquakes is justified by the low frequency of earthquakes with magnitude and hypocentre location capable to trigger such events (Marques et al. 2006). Sea erosion has been responsible for several events in the municipalities of Povoação, Ponta Delgada, Ribeira Grande and Vila Franca do Campo. Landslide events prompted by storms with rough sea and strong waves have been caused damage to buildings and roads located near the sea cliffs. Landslide events triggered by anthropic actions have been occurred almost all municipalities and have been associated with activities such as mining and construction works, which have resulted in fatalities and/or injuries.

The impact caused by landslides totals 82 fatalities, 41 injuries, 305 people made homeless and 66 buildings partially or totally destroyed on São Miguel Island. Among the municipalities, Povoação stands out with a high number of events that resulted in fatalities and significant damage. Povoação recorded 48 fatalities, accounting for approximately 59% of the total fatalities documented in the database. This highlights the severity of the impact caused by landslide events in Povoação municipality compared to other areas on the island. The research conducted by Marques (2013) concludes that 22.6% of the buildings, 44.0% of the main roads and 26.7% of the secondary roads in Povoação are located in areas of high and very high landslide susceptibility, which highlights the existing high exposure in this municipality that contributes to consequences of landslide events.

6 Conclusion

The NATHA (Natural Hazards in Azores) database is a repository of documents reporting natural events occurring in the Azores. This work aimed to analyse landslide events that occurred from 1900 to 2020, catalogued based on local and regional newspapers. The temporal distribution of the landslide events shows a higher concentration of events after 1996, which is explained by the shift in the precipitation regime that was characterised by the increasing number of days with more intense precipitation values in the most recent period. For the first time, this work demonstrates the influence of climate change on landslide occurrences in the Azores. The climate change scenarios presented in this study provide evidence of an expected increase in the probability of extreme rainfall events.

Irrespective of the impact caused, the increase in the number of events catalogued in the database in the most recent period (1996–2020) is mostly due to events that caused road blockages which can be explained by the increasing availability and dissemination of information, which increases the number of newspaper reports about this type of events. The spatial distribution of events reveals that the Povoação municipality is the most affected by landslide events. This is primarily attributed to its geology and geomorphology, which increase the likelihood of such events compared to other municipalities. The impact of landslide events varies significantly among the municipalities of São Miguel Island. Notably, the Povoação municipality stands out by the high impact of landslide events, which is explained by the exposure of its building, infrastructure and road network.

The landslide events predominantly occur during the rainiest months of the year, typically between November and March, with rainfall being the primary triggering factor on São Miguel Island. Cumulative monthly precipitation is not directly related to the occurrence of fatal events. Instead, these events are associated with the occurrence of extreme precipitation episodes, characterised by high intensity and short duration. The purpose of natural hazards databases is to prevent, reduce and mitigate the risks associated with disasters. Stakeholders in civil protection and spatial planning should consider these databases to identify emergency management hotspots and select safe locations for future territorial expansion.