The Garrotxa region is located in the northeastern part of Catalonia, within the province of Girona, Spain (Fig. 1A). It is characterized by a mountainous landscape and significant geological diversity, with the most distinctive feature being the Garrotxa Volcanic Zone Natural Park. This protected area represents the most well-preserved volcanic landscape in the Iberian Peninsula, comprising over fifty volcanic cones and extensive lava flows (Martí et al., 2016). Two different relief units can be distinguished in the Garrotxa region (Fig. 1A, B) and the limit between the two sectors is marked by the Vallfogona thrust that runs from west to east (Martínez-Rius et al., 1990) (Fig. 1).

The Alta Garrotxa is part of the pre-Pyrenees mountain range. It covers the northern part of Garrotxa and extends beyond it. It is formed by cliffs and karstic relief originating from the east–west Alpine antiforms with Eocene limestones and Garumnian lutites. In the core of the folds, granites and schists outcrop the Paleozoic. In this area, the valleys are very closed and steep, with canyons produced by karstic erosion and mainly dominated by tectonic structures in the form of folds and thrusts in an east–west direction produced by the Alpine orogeny (65 to 35 Ma).

The center and the south of the Garrotxa region are part of the Transversal mountain range of Catalonia, here composed of Eocene marls and sandstones. They are faulted by the Neogene (20 Ma to present) tectonic structures that form part of the European Rift. This area shows a flatter bottom and less abrupt reliefs as a result of the Quaternary volcanism and sedimentation (0.3 Ma to early Holocene) (Martí et al., 2025). The different lava flows and sediments accumulated by the past volcanic eruptions have formed the valleys of Bas, Olot, Santa Pau and Brugent. The Fluvià River traverses the region from west to east, playing a crucial role in shaping the local environment and hydrological dynamics.

The climate conditions of the Garrotxa region are classified as Mediterranean with montane influences, characterized by cold winters and mild summers. Precipitation levels are relatively high, which, together with the fertile volcanic soils, fosters the development of extensive forests, such as the beech forest of La Fageda d'en Jordà, one of the most ecologically significant ecosystems in the region (Garrotxa Tourism, 2025).

Beyond its geological and geographical characteristics, the Garrotxa region also possesses distinct socio-economic features that make it particularly suitable for a study like this. First of all, Garrotxa is the only mountainous administrative region of Catalonia with more than 20 000 inhabitants (GarrotxaDigital, 2022). Additionally, with a total area of 734.5 km², the region is highly accessible, and numerous scientific work has been conducted there, providing valuable knowledge of the zone (e.g., Bartolini et al., 2015; Varga et al., 2018; Revelles et al., 2023). 91 % of the residents of the Garrotxa region live in urban settlements (SIGMA, 2023). Olot, the regional capital, is home to 38.836 people (Idescat, 2025a), accounting for more than half of the total population of Garrotxa. The second most populated municipality, La Vall d'en Bas, has only 3221 inhabitants (Idescat, 2025b). Therefore, although Olot is a small to mid-sized city, it functions as a metropolitan hub for the region. The remaining 9 % of the population resides in dispersed settlements (SIGMA, 2023). This population distribution reflects a broader trend in Catalonia, where only the province of Barcelona accounts for almost 73 % of the total population (Gencat, 2023).

Thus, Garrotxa serves as a valuable study area, representing a larger demographic pattern. Additionally, Garrotxa has a strong economic activity. Despite being a mountainous area, 42 % of the workforce is employed in industry, making it the second-largest economic sector of the region (SIGMA, 2023). As a result, the unemployment rate remains below 10 %, slightly lower than the Catalonian average (SIGMA, 2023). Industrial growth has driven the development of major road infrastructure in the region, creating strong connections to key destinations such as Barcelona, the French border, the high-speed train station in Figueres-Vilafant, and Girona-Costa Brava Airport. Both industrial and transport infrastructure in Garrotxa can be treated as vulnerable elements, which have to be included in risk assessment and management studies.

Moreover, tourism is another important income for Garrotxa, which is attracted by its great natural beauty. In 2023, Garrotxa welcomed 192 109 tourists (overnight stays), while 559 515 people visited the Garrotxa Volcanic Zone Natural Park (Turisme Garrotxa, 2023). However, the growing number of visitors has raised concerns about overcrowding in natural areas of Garrotxa, especially in water ponds during summer and in La Fageda d'En Jordà during autumn.

The Garrotxa region, characterized by its volcanic geomorphology, mountainous terrain, and Mediterranean climate, is exposed to multiple natural hazards. Although the volcanic activity in the region remains dormant (Martí et al., 2025), other geological and hydrometeorological threats, including seismicity, landslides, flooding, and wildfires, present ongoing risks to both the environment and local communities (Vilaplana et al., 2008; Ajuntament d'Olot, 2022).

Seismic activity, while generally moderate, is influenced by the proximity to the Pyrenean tectonic boundary and the presence of Neogene-Quaternary active normal faults (ICGC, 2006; Bolós et al., 2015). Despite large-magnitude earthquakes in the region are relatively infrequent, moderate and low seismicity located along the main Neogene faults that cross the study area are common (Secanell et al., 2004; Ajuntament d'Olot, 2006; Jiménez and García, 2008; Departament d'Interior i Seguretat Pública, 2021). This seismicity, while rarely destructive, highlight the underlying geodynamic instability of the region (IGN, 2017; ICGC, 2020).

The steep topography, combined with intense seasonal precipitation and soil saturation, contributes significantly to landslide susceptibility (Palau et al., 2023). The many slopes, composed of weathered volcanic materials and unconsolidated sediments, are particularly prone to mass-wasting events following prolonged rainfall or seismic activity (ICGC, 2026; Copons, 2008a). Landslides can disrupt transportation infrastructure, alter river courses, and threaten residential areas in both the highlands and foothill zones. Additionally, rockfall hazard is well-established at Castellfollit de la Roca. This municipality is characterized by a prominent, columnar-jointed basalt cliff standing above less cohesive alluvial and pyroclastic layers, whose differential erosion and fractures contribute to frequent rockfall hazards (Abellán et al., 2011; Bassols and Calm, 2018; Janeras et al., 2023).

Hydrometeorological hazards, particularly flooding, represent another major concern. The Fluvià River and its tributaries frequently experience episodes of rapid water level rise, particularly during extreme precipitation events (Escuer, 2008; Departament d'Interior i Seguretat Pública, 2015; SIGMA, 2017). The combination of steep catchments and variable rainfall patterns promotes flash flooding, which can impact urban areas such as Olot and pose significant threats to agricultural productivity (ACA, 2024). The increasing incidence of extreme weather events, attributed to ongoing climatic shifts, has further exacerbated flood risks in recent decades (Barriendos et al., 2019; Llasat, 2021; Sharma et al., 2021).

Soil erosion, closely linked to hydrological instability, is particularly pronounced in the volcanic soils of Garrotxa, which are highly susceptible to weathering and displacement (Palou et al., 2010; Planagumà-Guàrdia et al., 2022). Factors such as deforestation, changes in land-use practices, and overgrazing have intensified soil degradation, especially on inclined terrain, leading to loss of arable land and increased sedimentation in river systems.

Wildfires remain a persistent risk, particularly in densely forested areas dominated by Mediterranean oak and beech woodlands (Consell Comarcal de la Garrotxa, 2021; Departament d'Interior i Seguretat Pública, 2024). The dry summer season, coupled with increased temperatures and periodic droughts, creates favorable conditions for fire ignition and rapid spread (Pausas and Fernández-Muñoz, 2012; Turco et al., 2019; Duane et al., 2021). The expansion of forested areas due to rural depopulation has further increased fuel availability, heightening wildfire risks (Vila-Subirós et al., 2014).

Ground subsidences are concentrated in the eastern part of the study area, particularly in the municipality of Besalú. This trend has been documented by previous investigations on the evaporite karst system developed within the geological complex of the Fluvià Valley (Copons, 2008b; Gutiérrez et al., 2016).

Regarding volcanic hazard in the Garrotxa volcanic field, the area is considered active, with the most recent period of volcanic activity covering the last 250 ka (Martí et al., 2025). The most recent eruption in this zone has traditionally been dated to the early Holocene (11–13 ka), although recent studies suggest it may have occurred as recently as ∼8 ka (Revelles et al., 2023; Martí et al., 2025). Volcanic activity in Garrotxa has ranged from effusive Hawaiian-style eruptions to more explosive Strombolian episodes, with numerous phases of phreatomagmatic activity (Martí et al., 2011). This volcanism is structurally controlled by the main regional normal faults formed during the Neogene extensional phase that affected the area (Bolós et al., 2015). Bartolini et al. (2015) identified five distinct hazard levels in the Garrotxa volcanic field, based on simulations of lava flows, pyroclastic density currents (PDCs), and fallout scenarios.

In recent years, climate change has played a significant role in amplifying the frequency and intensity of extreme weather phenomena in the region, including heatwaves, droughts, strong wind events, and intense rainfall episodes while at the same time contributing to a marked decline in snowfall events (OPCC and CTP, 2018; Cortès et al., 2019; METEOCAT, 2024). These changes have compounded existing hazards, necessitating a more comprehensive approach to risk assessment and mitigation strategies in the Garrotxa region.

The dataset spans 1 January 1900–31 December 2023, chosen to ensure a century-scale temporal window suitable for trend and recurrence analysis while maintaining reliable documentation. Six hazard types were included: earthquakes, landslides, rockfalls, floods, wildfires, and ground subsidences. Selection was guided by (1) their risk relevance in Garrotxa, (2) data availability within the study period, and (3) their potential interactions in a multi-hazard framework. In reference to this last criterion, relevant scientific literature was reviewed (e.g., Gill and Malamud, 2014; Zscheischler et al., 2020; López-Saavedra and Martí, 2023; Lee et al., 2024; UNDRR and ISC, 2025) to identify cascading or triggering relationships that could plausibly occur in a region with the characteristics of Garrotxa, such as earthquake-induced landslides, rainfall-triggered landslides, or landslide-induced flooding resulting from the formation of natural dams and subsequent overbank flow. Volcanic activity, though geologically important, was excluded because its most recent eruption (∼8 ka BP; Revelles et al., 2023) falls outside the historical timeframe.

The dataset was intentionally structured using a single-hazard approach, in which each hazard occurrence is recorded individually rather than being classified a priori as part of a multi-hazard event. This design choice minimizes subjective interpretation during data compilation and preserves event-level information, allowing potential hazard interrelationships to be explored subsequently through data-driven analyses of temporal and spatial co-occurrence.

To guide the data collection strategy, we first identified the main variables commonly used to characterize the selected natural hazards. This step aimed to ensure that the dataset would be compatible with existing hazard modelling approaches and could support future analyses.

The identification of relevant variables was based on three complementary sources: (1) a review of existing hazard models to identify commonly used input parameters (e.g., IBER for floods (Bladé et al., 2014), RockGIS for rockfalls (Matas et al., 2017), Cell2Fire for wildfires (Pais et al., 2019), FSLAM for landslides (Medina et al., 2021; Guo et al., 2022), and the Peak Ground Acceleration QGIS plugin (Núñez-Murillo, 2017)); (2) a review of scientific literature and additional expert consultation; and (3) the previous experience of the authors in natural hazard research.

Once the relevant variables had been identified, they were conceptually classified into four analytical categories (Table 1): (1) internal or constitutive variables, representing intrinsic physical or geological properties influencing hazard occurrence; (2) external or environmental variables, describing external conditions that may trigger or modulate events; (3) process-development variables, capturing the temporal or dynamic evolution of events; and (4) non-influential or meta variables, corresponding to descriptive or administrative attributes related to data documentation.

This classification was used as a conceptual framework to guide the compilation of the historical dataset. However, during data collection it became evident that many theoretically relevant variables were not consistently reported in historical sources such as press archives or oral testimonies. As a result, the final dataset includes a combination of theoretically informed variables and surrogate variables (e.g., “affected_municipality”, “np_ini_date”, “np_fin_date”) adapted to the level of detail available in the available sources.

Data were collected from three complementary source categories (Tables 2 to 4): (1) inventoried datasets from official institutions (e.g., ICGC seismic catalogues, ACA flood inventories, and the wildfire database of the Catalan Government), obtained via formal data-sharing requests; (2) written sources, including historical newspapers and scientific literature, retrieved through digital archives (e.g., Arxiu Municipal de Girona, ReGira) using bilingual keyword searches (Catalan/Spanish); and (3) oral and citizen-science sources, gathered through two workshops (Montagut and La Vall d'en Bas) involving several participants with local and technical knowledge. Participants annotated printed maps and provided recollections, particularly regarding major floods and slope instabilities.

All sources were georeferenced, standardized, and entered into a unified relational structure. Each record includes metadata on provenance and reliability.

Raw data were processed in two stages: (1) Standardization and merging: Temporal data reformatted to DD-MM-YYYY, and spatial coordinates converted to UTM Zone 31 N (EPSG:25831); (2) Duplicate detection and cross-validation: Automated string and coordinate matching in Python identified potential duplicates (e.g., two reports of the same flood). All candidates were manually checked against source texts. Discrepancies between institutions (e.g., AEMET vs METEOCAT rainfall units) were resolved through conversion scripts. A more detailed description of the data processing procedures is provided in the Supplement.

As for data cross-validation and quality assurance, three methods were applied: (1) Spatial verification: events plotted in QGIS and visually checked against topographic basemaps; (2) Expert review: geologists from SIGMA and the Natural Risks Assessment and Management Service (NRAMS) verified 100 randomly selected entries for plausibility; and (3) Uncertainty documentation: the fields “np_ini_date” (non-precise initial date) and “np_fin_date” (non-precise final date) are used to record uncertainty arising from the absence of exact dates in the written and oral sources consulted. This multi-layer validation ensures that even heterogeneous, century-spanning information attains consistent standards for subsequent analysis.

The finalized Garrotxa natural hazard event dataset contains 1049 records covering six hazard types: Wildfires (264), Earthquakes (246), Floods (242), Ground subsidences (137), Landslides (114), and Rockfalls (46) (Fig. 4). The final table has a total of 45 categorized variables spanning location, timing, physical descriptors, triggering factors, and metadata fields for quality control. About 73 % of events derive from inventoried sources and 27 % from non-inventoried written or oral materials (Fig. 5). Temporal coverage is uneven, with a clear increase in reporting after 1980, reflecting improved monitoring and media documentation. This bias does not necessarily indicate higher hazard frequency but improved detectability and institutional recording capacity.

The dataset is publicly available at https://digital.csic.es/handle/10261/394924 (Lagresa et al., 2025). Table S1 in the Supplement provides information on the coordinate reference system used for geolocating the different natural hazard events, as well as the codes, descriptions, units (where applicable) or examples for each variable collected across the various natural hazards.

Although the primary purpose of the dataset is descriptive, a simple exploratory analysis was performed to illustrate its potential for multi-hazard investigation. We examined temporal and spatial co-occurrence among hazard events using a simple yet realistic procedure, intended solely for demonstration purposes.

First, heuristic assumptions regarding potential temporal and spatial relationships between triggering hazards and subsequently triggered hazards were defined. These assumptions were informed by a preliminary inspection of the dataset, prior geological knowledge of the study area, and by existing literature describing interactions between natural hazards (e.g., Gill and Malamud, 2014; UNDRR and ISC, 2025).

Based on these considerations, plausible temporal and spatial windows were defined to explore potential trigger–response relationships within the dataset. These thresholds do not represent definitive physical limits but were selected as heuristic values to illustrate how the dataset can be used to explore possible multi-hazard interactions. The assumed trigger–response relationships and the corresponding temporal and spatial windows are summarized in Table 5. Wildfire and ground subsidences were not considered as triggering hazards for other events in this exploratory analysis. Ground subsidences were only evaluated as potentially triggered hazards. Wildfires were excluded from the interrelationship analysis due to their predominantly anthropogenic origin in La Garrotxa, which limits the interpretability of temporal and spatial associations within the adopted analytical framework. Therefore, a total of 912 events out of the total of 1049 were finally considered for this analysis.

Out of the 912 events considered, 79 (approximately 8.4 %) participated in at least one potential interrelated occurrence. Table 6 presents a sample of the observed relative frequencies of these interrelations. Percentages are calculated with respect to the total number of events of each potentially triggered hazard type and indicate the proportion of those events that were temporally and spatially associated with a given trigger hazard, according to the adopted criteria. For example, 28.6 % of landslide events were temporally and spatially associated with flood occurrences within a 7 d window and a 2 km distance threshold.

In addition to hazard–hazard interrelations, a complementary exploratory analysis was conducted to examine associations between selected natural hazards and meteorological conditions. For this purpose, a set of simple meteorological thresholds was defined, and hazard occurrences were evaluated with respect to whether they coincided with these conditions within the relevant temporal context.

Table 7 summarizes the most characteristic hazard–meteorological associations identified. Percentages represent the proportion of events of each hazard type that occurred under the specified meteorological conditions. These associations are intended to illustrate potential links between hazards and meteorological drivers, rather than to establish causal relationships.

Patterns observed in Tables 6 and 7 are broadly consistent with commonly reported multi-hazard sequences and hazard-meteorology in Mediterranean mountain regions (e.g., Guzzetti et al., 2007; Turco et al., 2018; Rodríguez-Peces et al., 2020). In particular, rainfall-driven processes appear to dominate, with intense precipitation episodes frequently associated with flood occurrences and slope instabilities (Table 7). Seismic activity is occasionally followed by localized rockfalls or minor ground subsidences, although associated magnitudes are generally low (ML <4) (Table 6). In addition, wildfire occurrences tend to coincide with periods characterized by limited precipitation over several consecutive days and elevated temperatures, as reflected in the adopted meteorological thresholds (Table 7). These associations are consistent with the regional climatic setting but should be interpreted cautiously, as they do not imply direct causality.

The temporal distribution of identified interrelations reflects a recency bias, with approximately 66 % of associations occurring after 2011. This pattern likely results from both an increased reporting of hazard events and a substantial improvement in detection and documentation capabilities in recent decades.

A key limitation of the present analysis remains the uneven quantity of available data across events. Interrelationships are more readily identified for well-documented episodes, such as major storms (e.g., the Gloria event in January 2020), where several hazards were widely reported and therefore recorded as separate entries in the dataset. As a result, potential interactions become easier to detect in such cases, highlighting the importance of data completeness for robust multi-hazard analysis.

Although preliminary and based on synthetic aggregation, these findings demonstrate the capacity of the dataset to support exploratory multi-hazard association and causal-chain analyses once more complete and probabilistic information becomes available.

Visualization in QGIS (Fig. 6) reveals several areas where different hazards spatially overlap, including: (1) the Fluvià River valley, where flood and landslide are both present; (2) the Bas and Santa Pau valleys, combining volcanic substrates and steep slopes conducive to slope instability; and (3) the Alta Garrotxa range, showing numerous wildfire and rockfall events. Such clustering supports the identification of priority areas for integrated risk mitigation rather than single-hazard interventions.

Concerning temporal distribution, Fig. 7 shows an apparent rise in hazard occurrence after 2000. While partly due to improved reporting, this trend coincides with observed regional increases in extreme rainfall frequency and summer heatwave duration, aligning with findings from the Catalan Meteorological Service (Meteocat, 2024). The dataset therefore also offers an empirical basis for studying climate-hazard linkages at the regional scale.

The Garrotxa natural hazard event dataset fills a unique spatial and thematic gap between global/national repositories (e.g., EM-DAT, SHELDUS, Risk Data Hub) and local administrative records. Compared with EM-DAT, which lists only disasters causing major economic losses or fatalities, our dataset captures all scales of events, including minor but frequent slope movements and small wildfires that shape local vulnerability. In contrast to SHELDUS (county-level U.S. data) or the Copernicus Emergency Management Service (European-scale), the Garrotxa dataset reaches municipality-level geospatial resolution (<250 m). Moreover, it integrates non-institutional information such as historical newspapers and citizen-science memories, rarely present in official inventories.

From a scientific perspective, these differences translate into three clear contributions: (1) Enhanced resolution and completeness, supporting validation of susceptibility models that require fine-scale inputs; (2) Inclusion of cultural memory, linking physical events with community perception, crucial for risk-communication studies; and (3) Interoperable structure, as variables follow common vocabularies (e.g., INSPIRE, UNDRR Hazard Glossary), facilitating cross-dataset integration at European scale. Thus, the dataset moves beyond a purely archival purpose and becomes an enabling infrastructure for multi-hazard data-driven research.

The dataset forms the first step in a chain leading to a comprehensive multi-hazard risk analysis. Figure 8 (described below) outlines the sequential steps and data requirements in this process. Step 1, Hazard Event Dataset (Observed Data), uses an empirical record of past events, as detailed in this study. Step 2, Susceptibility Mapping, identifies the relative likelihood (low to high classes) of a hazard occurring at a specific location, based on intrinsic and environmental factors such as slope, lithology, and rainfall. Step 3, Hazard Mapping, integrates susceptibility data with the probability of an event occurring within a given time frame, such as flood inundation probability for 50- or 100-year return periods. Step 4, Exposure Assessment, overlays spatial data to identify exposed elements like people, buildings, infrastructure, and ecosystems. Step 5, Vulnerability Analysis, quantifies potential damage or loss for each exposed element. Step 6, Multi-Hazard Integration and Risk Mapping, combines hazard layers using weighting or network models to represent cascading or compound effects. Finally, Step 7, Decision Support and Policy Feedback, produces outputs that inform risk management plans, emergency preparedness, and climate adaptation strategies.

A susceptibility map expresses the relative predisposition of a location to event initiation, while a hazard map quantifies the probability of a given event magnitude occurring within a specified period. Not all hazard types have both products routinely: landslides, for example, often have susceptibility maps but rarely fully probabilistic hazard maps.

For the Regional County Council and SIGMA, the dataset provides the first integrated evidence base of recorded natural-hazard events in Garrotxa. Previously, disparate archives hindered coordinated action. Now, cross-hazard visualization reveals which municipalities face overlapping threats. For example, the overlap of flood and landslide records along the Fluvià River underscores the need for joint hydraulic–geotechnical planning, while patterns linking wildfire records to periods of limited precipitation and elevated temperatures inform vegetation-management priorities.

To maximize long-term utility, the dataset should be synchronized with existing real-time monitoring networks (meteorological and seismic) so that new events feed automatically into the dataset. Developing an open-data API would further enable public transparency and early-warning integration.

Beyond Garrotxa, this initiative demonstrates a replicable governance model: a partnership between research institutions and local administrations that aligns with the National Plan for Adaptation to Climate Change (2021–2030) in Spain.

One of the main challenges encountered during dataset compilation was the scarcity of pre-inventoried data from official records, particularly for geological hazards such as landslides, rockfalls, and ground subsidences. Consequently, alternative sources, including historical newspapers and input from local experts, had to be consulted. Earthquakes were the only hazard type for which all relevant information could be retrieved from existing inventories, due to the presence of an established seismic monitoring network in the region.

Gathering information beyond official inventories proved time-consuming, but enabled the identification of additional events and improved data completeness. The integration of data from different institutional sources also required methodological adjustments, as similar datasets were often stored using different formats that had to be standardized before inclusion in the dataset.

Another challenge arose from the varying nature of the consulted sources. While inventoried datasets generally provided structured and precise parameters, written or oral sources often contained incomplete or subjective information, requiring additional verification through aerial imagery or expert consultation.

Human influence also had to be considered for certain hazards, including wildfires, floods, rockfalls, and landslides. This highlights the importance of accounting for human–environment interactions in hazard datasets. Finally, the largely manual nature of the data collection process underscores the need for future automation, for example through tools capable of extracting structured information from digitized archives or other semi-structured sources.

In terms of scalability, we believe that the approach presented here could be adapted to regions with characteristics similar to Garrotxa, as well as to other geographical contexts. Because hazards are recorded individually, the dataset structure can be readily extended to additional hazard types or adapted to different regional settings. However, the transferability of this approach may be constrained by data availability and by the heterogeneity of existing data sources and formats.

The scope of this paper is primarily limited to the presentation of the natural hazard dataset, which represents a foundational step toward broader multi-hazard risk assessment. In addition, a simple exploratory analysis is provided to illustrate how the dataset may be applied in multi-hazard investigations; however, this example is intended for demonstration purposes only and does not constitute a comprehensive framework for analyzing hazard interrelationships.

The dataset currently includes events related to six types of natural hazards. Although these hazards were selected based on their relevance to the study area and the availability of reliable information, the exclusion of other hazards may limit the scope of future multi-hazard analyses.

The time span covered by the dataset (1900–2023) also introduces limitations. Some significant events, or even entire hazard types with longer recurrence intervals, such as volcanic activity, may not be represented. Consequently, the available record may not fully capture long-term hazard dynamics, particularly for rare but high-impact events.

Finally, no formal quantification of uncertainty has yet been conducted to assess the quality of the resulting dataset, particularly with respect to reliability, validity, and provenance (Koedel et al., 2022).