UNIVERSIDAD Y SOCIEDAD | Revista Científica de la Universidad de Cienfuegos | ISSN-2218-3620

Vol 18 | No.5 | septiembre-octubre | 2026 | Publicación continua | e6384

Original Article

Bryan Steven Cortez Chichande^1*

 bcortezc@uteq.edu.ec

 https://orcid.org/0000-0001-6834-4764

Carlos Luis Muñoz Mendoza¹

 carlos.munozm@uteq.edu.ec

 https://orcid.org/0009-0006-4039-5766

Randy Jimmy Rodríguez Padilla¹

 randy.rodriguez2015@uteq.edu.ec

 https://orcid.org/0009-0003-3793-5037

¹Universidad Técnica Estatal de Quevedo. Los Ríos. Ecuador.

*Corresponding author.

Abstract:

The present study analyzes the human and economic losses caused by natural and technological disasters worldwide during the period 2000–2026, using data from the EM-DAT database. A multivariate approach is adopted, utilizing the HJ-Biplot technique, which facilitates the simultaneous representation of variables and disaster types in a limited space, thereby enabling the identification of complex relationships and patterns. In addition, the K-means clustering algorithm is employed to categorise events according to their impact characteristics, thereby facilitating a more precise segmentation of diverse disaster types. The results identify three main groups: disasters with high economic impact, characterized by high total damage, reconstruction costs, and aid contributions; disasters with high human impact, associated with higher levels of mortality and people affected; and a group with relatively low impacts in both dimensions. Moreover, specific analyzes of human and economic variables reveal differences in patterns of social impact and recovery costs, which allows a more profound understanding of impact dynamics and providing useful information for risk management and decision-making.

Keywords: Natural and technological disasters, Economic and human losses, HJ-Biplot, Clustering.

Resumen:

El presente estudio analiza las pérdidas humanas y económicas causadas por desastres naturales y tecnológicos en todo el mundo durante el período 2000–2026, utilizando datos de la base de datos EM-DAT. Se adopta un enfoque multivariante, utilizando la técnica HJ-Biplot, que facilita la representación simultánea de variables y tipos de desastres en un espacio limitado, lo que permite identificar relaciones y patrones complejos. Por otro lado, se emplea el algoritmo de agrupamiento K-means para clasificar los eventos según las características de su impacto, lo que permite una segmentación más precisa de los diversos tipos de desastres. Los resultados identifican tres grupos principales: desastres con un alto impacto económico, caracterizados por un elevado daño total, costos de reconstrucción y aportaciones de ayuda; desastres con un alto impacto humano, asociados a mayores niveles de mortalidad y de personas afectadas; y un grupo con impactos relativamente bajos en ambas dimensiones. Además, los análisis específicos de las variables humanas y económicas revelan diferencias en los patrones de impacto social y costos de recuperación, lo que permite una comprensión más profunda de la dinámica del impacto y brinda información útil para la gestión de riesgos y la toma de decisiones.

Palabras clave: Desastres naturales y tecnológicos, Pérdidas económicas y humanas, HJ-Biplot, Agrupamiento.

Received: August 14, 2026

Accepted: September 15, 2026

Published: October 12, 2026

Introduction

In recent decades, there has been a global increase in the frequency and intensity of natural and human-induced disasters, the effects of which have resulted in significant human and economic losses. It is evident that a multitude of phenomena, including but not limited to earthquakes, storms, floods, wildfires and industrial accidents, have exerted a considerable influence on both socio-economic systems and ecosystems. This has had a profound impact on the quality of life of millions of people across different continents. The consequences of such situations are manifold, including loss of life, damage to infrastructure, disruption of essential services and impacts on local and global economies.

The existence of global databases, such as EM-DAT, facilitates analysis of these events from a multivariate perspective, incorporating both human and economic impact. This database has been developed for the purpose of collating structured information on disasters worldwide. The database includes variables such as the number of deaths, the affected population and economic losses. The use of these variables facilitates comparative studies between regions and time periods (Delforge et al., 2025). The scale and diversity of the recorded events highlight the need to strengthen risk planning, prevention and management processes, as well as to develop analytical approaches that enable the identification of patterns, the prioritisation of resources and a better understanding of the global dynamics of disasters.

The advent of standardized databases, such as EM-DAT (Emergency Events Database), has enabled access to structured and comparable information on catastrophic events occurring worldwide (Guerrero & Lacambra, 2020). The database, which is accessible to the public at https://public.emdat.be, incorporates variables related to human and economic losses. This enables comparative analyzes to be conducted between different types of disasters, regions and time periods. However, the nature of these data, characterized by the presence of multiple indicators, disaster categories and geographical diversity, poses challenges for its interpretation and analysis.

In this context, it is necessary to employ multivariate statistical techniques that allow the variables involved to be analyzed jointly and their interrelationships to be understood. The HJ-Biplot, proposed by Villardón (1986) as an extension of the classical Biplot method, is particularly suitable for this purpose, as it enables the simultaneous representation of individuals and variables in a low-dimensional space while optimising the quality of representation of both elements. This facilitates the identification of patterns, similarities, correlations and underlying structures in multidimensional datasets (Cascante-Yarlequé et al., 2025). The method has been applied in different analytical contexts in which the joint interpretation of complex multivariate structures is required (Diaz et al., 2024; Peralta-Gamboa et al., 2024; Saraiva & Caiado, 2025).

Furthermore, clustering is an unsupervised learning technique that enables observations to be grouped according to their similarity, such that elements within the same cluster are more similar to one another than to those belonging to other groups (Jain et al., 1999). In disaster analysis, this approach facilitates the identification of underlying structures and patterns without requiring predefined dependent variables. Different clustering algorithms, including K-means, hierarchical methods, DBSCAN and Gaussian mixture models, employ different criteria to identify groups within multidimensional datasets (Tin et al., 2024). Their application to disaster data facilitates the identification of relationships between characteristics such as disaster type, location, frequency and impact, providing a structured basis for comparative analysis and risk-management decision-making.

Previous studies have examined disaster impacts from different economic, environmental and risk-management perspectives. Brei et al. (2024) analyzed the effects of natural disasters on banking-sector stability, while Adam & Bevan (2020) examined post-disaster financing mechanisms in developing economies. Other studies have addressed flood vulnerability, participatory risk management and the relationships between specific types of disasters through regional or event-specific approaches (Diaz et al., 2024; Reynoso et al., 2024; Wang et al., 2021). In addition, methodological and vulnerability-oriented approaches have contributed to the analysis of disaster impacts from different perspectives (Gutierrez-Lopez, 2022; Jacobi et al., 2022). Although these studies provide relevant evidence, the simultaneous examination of multiple dimensions of human and economic losses across different disaster categories remains comparatively limited.

In this context, the aim of this study is to conduct a multivariate analysis of the human and economic losses caused by natural and technological disasters globally during the period 2000–2026. To this end, the HJ-Biplot technique is utilized to concurrently depict variables and observations, thereby facilitating the identification of relationships and patterns in the data. Furthermore, clustering techniques are employed to categorise events based on shared characteristics and to analyze their behavior in terms of their impact. The study's objective is twofold: firstly, to enhance comprehension of the global dynamics of disasters, and secondly, to generate information that facilitates comparative analysis between regions and supports decision-making in risk management.

Materials and methods

The objective of this study was to analyze the human and economic losses caused by natural and technological disasters worldwide during the period 2000–2026. To this end, a structured and reproducible workflow was designed. The process entailed the selection of pertinent variables, the management of missing values, and the implementation of multivariate statistical techniques. The analysis was conducted in its entirety within the R programming environment, with bespoke libraries employed for data management, transformation and analysis, in addition to the implementation of HJ-Biplot and clustering techniques.

Data collection

The EM-DAT (Emergency Events Database) dataset (Delforge et al., 2025) was utilized, as it is a compendium of standardized information on disasters worldwide. The database is maintained by the Center for Research on the Epidemiology of Disasters (CRED) at UCLouvain (Belgium), and it consolidates historical records from various official and scientific sources. For an occurrence to be included in EM-DAT, at least one of the following criteria must be met: The criteria for the classification of an incident as a major incident are as follows:

• The incident must result in 10 or more deaths.
• The incident must affect at least 100 people.
• The incident must result in a declaration of emergency or a request for international assistance.

In this study, an updated version of the database was utilized, which was downloaded in 2026. This database was employed to filter events that occurred globally between 2000 and 2026. The study considered variables pertaining to human losses (total deaths, people injured, affected and made homeless) and economic losses (reconstruction costs, insured losses, total losses and international aid).

Data selection and pre-processing

A total of eight quantitative variables associated with human and economic losses were selected for the purpose of conducting a joint analysis of the impacts of disasters at a global level. The pre-processing was conducted within the R environment utilising a reproducible workflow, encompassing data cleansing, the selection of pertinent variables, and the preparation of the matrix for multivariate analysis.

Missing values in the numerical variables were replaced with zero, given that these variables represent counts or monetary amounts. This means that the absence of a record can be interpreted as the absence of losses. Subsequently, the data were aggregated by disaster type, thereby providing a consolidated representation of the impact of each category. A standardisation process was finally applied using scaling in order to ensure the comparability of the variables.

Table 1 presents a comprehensive list of the variables employed in the analysis. In addition to their full names, the table provides concise descriptions and abbreviated names for each variable.

Table 1: Variables used in the analysis.

Multivariate analysis and clustering

In order to conduct a comprehensive analysis of the human and economic losses associated with different types of disasters, a combined approach involving multivariate analysis and clustering techniques was applied. This methodological approach facilitated two distinct yet interconnected objectives. Firstly, it enabled the exploration of the relationships between variables and the various types of disasters. Secondly, it facilitated the identification of clustering structures based on similar behaviors. The implementation of both techniques was conducted within the R programming environment, thereby ensuring the reproducibility of the analysis.

Multivariate analysis using HJ-Biplot

The multivariate analysis was conducted utilising the HJ-Biplot technique, which was implemented via the MultBiplotR package. For this purpose, a previously standardized data matrix was utilized, in which each row represents a specific type of disaster and each column corresponds to a variable relating to human or economic losses. Standardisation was employed to eliminate scale effects between variables, thereby ensuring that all variables contributed equally to the analysis.

The HJ-Biplot facilitated the concurrent depiction of disaster categories (rows) and variables (columns) within a two-dimensional space delineated by the initial two principal dimensions, derived from singular value decomposition. This representation facilitates the joint interpretation of the data, where the proximity between points indicates similarity between types of disasters, the length of the vectors reflects the variability of the variables, and the angles between them allow correlation relationships to be inferred.

Furthermore, independent analyzes were conducted on the human and economic impact variables to ascertain whether the patterns identified in the overall analysis were consistent when each dimension was analyzed separately. This procedure enabled a more detailed interpretation of the behavior of the different types of disasters based on the nature of their impacts.

Clustering analysis

A clustering analysis was performed with the aim of identifying groups of disaster types that exhibit similar patterns in terms of human and economic losses. To this end, the K-means algorithm was applied to the coordinates of the disaster types in the first two dimensions obtained from the HJ-Biplot. This approach facilitates the reduction of dimensionality while preserving the fundamental structure of the data.

The optimal number of clusters was determined using the Elbow Method, whereby the variation in the intra-cluster sum of squares for different values of k was evaluated. Based on this criterion, a number of clusters was selected that balances the internal compactness of the groups and the simplicity of the model. In the subsequent phase of the research, each type of disaster was assigned to a specific cluster. This process enabled the identification of groups with similar characteristics.

In addition, statistical summaries were calculated for each cluster based on the standardized variables, thus enabling each group to be characterized according to its loss profile. This analysis facilitates the interpretation of the results by highlighting distinct patterns between types of disasters and complementing the information obtained via the HJ-Biplot.

Results and discussion

Experiment A - Global Analysis of Disasters

• Elbow Method for Cluster Selection

The Elbow Method was employed to determine the optimal number of clusters, with the intra-cluster sum of squares (WSS) evaluated for different values of k. As demonstrated in Figure. 1, a precipitous decline in WSS is evident between k = 1 and k = 3, subsequently transitioning into a more gradual decrease for higher values of k.

This behavior indicates the presence of an inflection point at k = 3, suggesting that this number of clusters offers an appropriate balance between the compactness of the groups and the complexity of the model. Beyond this value, the improvement in the reduction of internal variability is negligible.

Consequently, k = 3 was selected as the optimal number of clusters for the clustering analysis using the K-means algorithm.

Fig. 1: Elbow method used to determine the optimal number of clusters based on the within-cluster sum of squares (WSS) for the global disaster dataset.

• K-means Clustering Results

The K-means algorithm was implemented with k set at 3, with the objective of identifying clusters of disasters based on a combination of human and economic losses. As demonstrated in Table 2, Cluster 1, consisting predominantly of flood and storm, demonstrates elevated values for economic variables such as total damage (TDG = 3.26), insured damage (ID = 2.89) and aid contributions (AC = 2.21), in addition to elevated values for affected population and homelessness (NA = 2.65, NH = 2.89). This indicates a multifaceted impact, albeit one that is predominantly characterized by economic losses. Cluster 2, as indicated by earthquake, demonstrates the highest values for the number of deaths (TD = 4.69), people injured (NI = 1.87) and reconstruction costs (RC = 5.36), indicating disasters with severe human impact and high recovery costs. Conversely, Cluster 3 encompasses the preponderance of disaster categories, exhibiting predominantly negative or near-zero values across all variables. This suggests a minimal impact, both in terms of human well-being and economic consequences. This segmentation facilitates the establishment of a clear distinction between disasters that are characterized by a high economic impact, a high human impact, and a low overall impact.

Table 2: Cluster summary based on standardized variables.

• HJ-Biplot with Clustering Interpretation

The HJ-Biplot, when employed in conjunction with clustering analysis, facilitates the integration of visual representations of the relationships between loss variables and the various categories of disasters, in addition to the previously identified clustering structure. As demonstrated in Figure. 2, the initial two dimensions capture the majority of the variability present within the dataset, thereby ensuring an adequate representation in the two-dimensional plot.

The graph demonstrates that economic variables, including total damage (TDG), reconstruction costs (RC) and aid contributions (AC), exhibit vectors with a comparable orientation, thereby signifying a substantial correlation between them. In a similar manner, variables associated with human impact, including the number of deaths (TD) and the number of injuries (NI), demonstrate a close relationship, as evidenced by the angular proximity of their respective vectors.

With respect to the distribution of the clusters, Cluster 1 (represented by Flood and Storm) is located in the region associated with high values of economic variables, particularly TDG and AC, thus confirming its characterization as disasters with a high economic impact. Conversely, Cluster 2, as illustrated by earthquake, is positioned in the direction of variables associated with human losses, particularly TD and NI, thereby demonstrating its significant impact on the population.

Fig. 2: HJ-Biplot with K-means clustering showing the relationship between disaster types and human and economic loss variables.

It is evident that Cluster 3 includes a variety of disaster types, which are predominantly concentrated around the core of the plot. This observation signifies that the majority of the variables in question are likely to exhibit low or moderate values. This pattern suggests that the impact of these events is more limited in both human and economic terms. The integration of the HJ-Biplot with clustering facilitates the clear identification of three distinct patterns: disasters dominated by economic losses, disasters with a high human impact, and disasters with a lower overall impact. This representation facilitates the interpretation of complex relationships between variables and types of disasters, providing an overview of their behavior.

Experiment B - Human Impact Analysis

• Elbow Method for Cluster Selection

In order to ascertain the optimal number of clusters in the analysis of human impact variables, the elbow method was applied by evaluating the within-cluster sum of squares (WSS) for different values of k. As demonstrated in Figure. 3, there is a pronounced decrease in the WSS between k = 1 and k = 3, followed by a more gradual reduction for higher values. This behavior indicates an inflection point at k = 3, suggesting that this number of clusters provides an appropriate balance between group compactness and model complexity; it was therefore selected as the optimal value for the clustering analysis.

Fig. 3: Elbow method for determining the optimal number of clusters based on the within-cluster sum of squares (WSS) for human impact variables

• K-means Clustering Results

The K-means algorithm was implemented with k set at 3, with the objective of identifying clusters of disasters based on their human impact. The results of the study indicate the presence of three distinct clusters, each exhibiting unique characteristics (Table 3). Cluster 1 demonstrates elevated values for the number of deaths (TDE = 2.32) and people injured (NI = 2.84), indicating a significant impact in terms of mortality and injuries, although with a comparatively lower impact on the displaced population. Cluster 2 is distinguished by negative values across all variables, indicative of a comparatively low human impact in comparison to the other groups. Conversely, Cluster 3 is distinguished by elevated values in the number of individuals affected (NA = 2.88) and those rendered homeless (NH = 1.82), signifying disasters with a substantial social impact, particularly with regard to displacement and its repercussions on the population. This segmentation facilitates the differentiation between disasters with high mortality, low-impact disasters, and disasters with a high social impact.

Table 3: Cluster summary based on standardized human impact variables.

• HJ-Biplot with Clustering Interpretation

The HJ-Biplot, when employed in conjunction with clustering analysis, facilitates the visualisation of the relationships between human impact variables and disaster types. As demonstrated in Figure. 4, the variables 'number of deaths' (TDE) and 'number of injuries' (NI) demonstrate a comparable trend, thereby indicating a high correlation between the two. In a similar vein, the variables 'people affected' (NA) and 'people made homeless' (NH) demonstrate a close relationship, thus revealing a pattern associated with the social impact of disasters.

Fig. 4: HJ-Biplot with K-means clustering showing the relationship between disaster types and human impact variables.

With regard to the distribution of the clusters, Cluster 1 is located in the direction of the TDE and NI variables, thus confirming its characterisation as disasters with a high impact on mortality and injuries. Cluster 3 is located in the region associated with NA and NH, indicating events with greater social impact and population displacement. Conversely, Cluster 2 is concentrated near the origin, reflecting low values across all variables and, consequently, a reduced human impact. This distribution facilitates the clear differentiation of disaster types according to the nature of their impact on the population.

Experiment C - Economic Impact Analysis

• Elbow Method for Cluster Selection

In the determination of the optimal number of clusters in the analysis of economic variables, the elbow method was implemented through the assessment of the within-cluster sum of squares (WSS) for varying values of k. As illustrated in Figure. 5, a pronounced decrease in the WSS is observed between k = 1 and k = 3, subsequently followed by a more gradual reduction for higher values. This behavior indicates an inflection point at k = 3, suggesting that this value provides an appropriate balance between cluster compactness and model complexity. Consequently, k = 3 was selected as the optimal number of clusters for the clustering analysis.

Fig. 5: Elbow method for determining the optimal number of clusters based on the within-cluster sum of squares (WSS) for economic variables.

• K-means Clustering Results

The K-means algorithm was implemented with k set at 3, thereby facilitating the identification of clusters of disasters according to their economic impact. As demonstrated in Table 4, the findings indicate the presence of three distinctly separate clusters. Cluster 1, on the other hand, encompasses the majority of disaster types and is distinguished by negative values across all variables, signifying a minimal economic impact. Cluster 2, predominantly characterized by Storm, demonstrates remarkably elevated values for insured damage (ID = 5.24) and total damage (TDM = 4.69), signifying a substantial economic impact associated with this category of event. Cluster 3, comprising Earthquake and Flood, shows the highest values for reconstruction costs (RC = 2.81) and aid contributions (AC = 3.50), reflecting disasters with high recovery costs and significant international intervention. This segmentation facilitates the differentiation of disasters according to their economic impact, with one group showing low economic impact, another characterized by high insured and total damage, and a third by a high reconstruction costs and aid contributions.

Table 4: Cluster summary based on standardized economic variables.

• HJ-Biplot with Clustering Interpretation

The HJ-Biplot, when employed in conjunction with clustering analysis, facilitates the integration of visual representations of the relationships between economic variables and the various categories of disasters. As demonstrated in Figure. 6, the variables reconstruction costs (RC) and aid contributions (AC) demonstrate a comparable orientation, thereby indicating a positive relationship between the two. Conversely, total damage (TDM) and insured damage (ID) exhibit a divergent orientation, thus reflecting an alternative pattern of economic impact. With regard to the distribution of the clusters, Cluster 3, comprising Earthquake and Flood, is located in the direction of RC and AC, indicating high reconstruction costs and financial support. It is evident that Cluster 2, as depicted by Storm, is situated in the direction of ID and TDM. This suggests a predominance of direct and insured economic damage. Conversely, Cluster 1, which exhibits a low economic impact relative to the other groups, is concentrated near the origin.

Fig. 6: HJ-Biplot with K-means clustering showing the relationship between disaster types and economic impact variables.

Discussion

The results show that the impact of disasters does not follow a uniform pattern when human and economic losses are analyzed together. The overall analysis made it possible to distinguish three profiles: one group associated mainly with economic losses, another characterized by a greater human impact, and a third with comparatively lower figures in both dimensions. This differentiation provides a complementary perspective to studies that have examined the effects of disasters from specific economic dimensions. Brei et al. (2024), for example, analyzed the consequences of natural disasters on the stability of the banking sector in the Caribbean, whereas (Adam & Bevan, 2020) studied post-disaster financing mechanisms in low-and middle-income countries. Compared with these approaches, the findings of this study consider both human and economic indicators together and enable a distinction to be made between impact profiles across different categories of disasters.

In the specific analysis of human variables, differences were observed between events associated with higher levels of mortality and injuries and those primarily linked to the number of people affected and made homeless. This distinction indicates that the magnitude of the human impact cannot be interpreted solely on the basis of mortality, as certain disasters can have significant effects on the population even when the number of fatalities is comparatively lower. This interpretation is consistent with studies that have addressed vulnerability and the effects of disasters from territorial and social perspectives. Diaz et al. (2024) analyzed vulnerability to flooding using geostatistical models, whilst Reynoso et al. (2024) proposed participatory approaches to risk management in local contexts. Although these studies employ different scales and methodologies, they agree on the need to consider various dimensions when characterising the effects of disasters on the population.

Furthermore, the economic analysis made it possible to distinguish between events primarily associated with total damage and insured losses, and those linked to higher reconstruction costs and aid contributions. In particular, storms were more strongly associated with total and insured damage, whilst earthquakes and floods were linked to higher reconstruction costs and aid contributions. These results show that economic consequences can take different forms depending on the type of disaster and the indicator considered. In this regard, the comprehensive analysis by Wang et al. (2021) on the relationship between cyclones and floods also highlights the importance of studying phenomena that may exhibit related behavior in conjunction. However, the present analysis broadens this perspective by incorporating different categories of disasters and economic variables within a single multivariate framework.

From an analytical perspective, the combination of the HJ-Biplot and K-means clustering made it possible to link the multivariate structure of the data to the identification of groups of disasters with similar characteristics. The results obtained from the three analyzes show that the joint representation of variables and categories facilitates the identification of relationships that may be less obvious when each indicator is examined independently. This approach is related to the work of Gutierrez-Lopez (2022), which focuses on the technical analysis of disasters, and that of Jacobi et al. (2022), which centers on the study of vulnerabilities from a qualitative perspective. In contrast to these approaches, the present study employs a multivariate quantitative perspective that integrates human and economic losses and allows for the simultaneous comparison of different categories of disasters.

Overall, the results indicate that the characterisation of disasters depends on the dimension of impact under consideration. The patterns identified in the overall analysis are complemented by the differences observed when human and economic variables are studied separately. In this way, the analysis is not limited to establishing which categories incur the greatest losses, but also enables the predominant nature of those losses to be distinguished. This differentiation may contribute to a more structured interpretation of the information available in EM-DAT and serve as a reference for subsequent analyzes aimed at risk assessment and management.

Conclusions

The multivariate analysis of EM-DAT data for the period 2000–2026 enabled the identification of differentiated patterns in the human and economic losses associated with natural and technological disasters worldwide. The combined application of HJ-Biplot and K-means clustering provided a joint representation of the variables analyzed and facilitated the classification of disaster types according to their impact profiles.

The global analysis identified three differentiated groups characterized by predominant economic impact, high human impact, and comparatively lower overall impact. Seismic events were associated with high levels of mortality and reconstruction costs, whereas storms and floods showed stronger relationships with economic damage and aid contributions. The remaining disaster types generally presented comparatively lower values across the dimensions analyzed.

The separate analyzes of human and economic variables provided a more detailed characterisation of these patterns. Human-impact variables distinguished disasters associated with high mortality and injury levels from those characterized by larger numbers of affected and homeless people. Economic variables, in turn, differentiated disasters associated with insured losses from those characterized by high reconstruction costs and aid contributions. These findings demonstrate the usefulness of combining multivariate representation and clustering to characterize the multidimensional impact of disasters.

Future research could incorporate a temporal perspective to assess changes in these patterns across different periods and extend the analysis through geographical, climatic and socio-economic variables. The application of alternative clustering and dimensionality-reduction techniques could also allow the stability of the identified structures to be evaluated. These extensions could contribute to the development of analytical tools for disaster-risk assessment, planning and decision-making.

References

Adam, C., & Bevan, D. (2020). Tropical cyclones and post-disaster reconstruction of public infrastructure in developing countries. Economic Modelling, 93, 82–99. https://doi.org/10.1016/J.ECONMOD.2020.07.003

Brei, M., Mohan, P., Perez Barahona, A., & Strobl, E. (2024). Transmission of natural disasters to the banking sector: Evidence from thirty years of tropical storms in the Caribbean. Journal of International Money and Finance, 141, 103008. https://doi.org/10.1016/J.JIMONFIN.2023.103008

Cascante-Yarlequé, R., Galindo-Villardón, P., Guevara-Viejó, F., Vicente-Villardón, J. L., & Vicente-Galindo, P. (2025). HJ-BIPLOT: A Theoretical and Empirical Systematic Review of Its 38 Years of History, Using Text Mining and LLMs. Mathematics, 13(12), 1913. https://doi.org/10.3390/MATH13121913

Delforge, D., Wathelet, V., Below, R., Sofia, C. L., Tonnelier, M., van Loenhout, J. A. F., & Speybroeck, N. (2025). EM-DAT: the Emergency Events Database. International Journal of Disaster Risk Reduction, 124, 105509. https://doi.org/10.1016/J.IJDRR.2025.105509

Diaz, N. D., Lee, Y., Kothuis, B. L. M., Pagán-Trinidad, I., Jonkman, S. N., & Brody, S. D. (2024). Mapping the flood vulnerability of residential structures: Cases from The Netherlands, Puerto Rico, and the United States. Geosciences, 14(4), 109.https://doi.org/10.3390/GEOSCIENCES14040109

Guerrero Compeán, R., & Lacambra Ayuso, S. (2020). Disasters and Loss of Life: New Evidence on the Effect of Disaster Risk Management Governance in Latin America and the Caribbean. https://doi.org/10.18235/0002781

Gutierrez-Lopez, A. (2022). Methodological Guide to Forensic Hydrology. Water (Switzerland), 14(23), 3863. https://doi.org/10.3390/w14233863

Jacobi, P. R., de Arruda Filho, M. T., & de Pierro, B. (2022). Ambiente e Sociedade em Tempos de Emergência Climática: Do Resgate Histórico ao Momento Atual. Fronteira: Journal of Social, Technological and Environmental Science, 11(3), 35–46. https://doi.org/10.21664/2238-8869.2022V11I3.P35-46

Jain, A. K., Murty, M. N., & Flynn, P. J. (1999). Data clustering: A review. ACM Computing Surveys, 31(3), 264–323. https://doi.org/10.1145/331499.331504

Peralta-Gamboa, D. A., Espinel-Obregoso, F. P., Hernández-Dominguez, C. S., Cervantes-Díaz, D. J., & Espinoza-Solis, E. J. (2024). HJ-Biplot for the study of unemployment and violence: A socio-economic and criminological analysis in Ecuador. Journal of Infrastructure, Policy and Development, 8(15), 7990.

Reynoso Vanderhorst, H. D., Pathirage, C., & Proverbs, D. (2024). Navigating Flood Resilience: Challenges, Solutions, and Lessons Learnt from the Dominican Republic. Water (Switzerland), 16(3), 382. https://doi.org/10.3390/W16030382

Saraiva, C., & Caiado, J. (2025). Global development patterns: A clustering analysis of economic, social and environmental indicators. Sustainable Futures, 10, 100907. https://doi.org/10.1016/J.SFTR.2025.100907

Tin, T. T., Hao, Y. J., Yeou, Y. C., Mooi, L. S., Yew, G. T., Olugbade, T. O., & Aitizaz, A. (2024). Natural Disaster Clustering Using K-Means, DBSCAN, SOM, GMM, and Mean Shift: An Analysis of Fema Disaster Statistics. International Journal of Advanced Computer Science and Applications, 15(9), 667–681. https://doi.org/10.14569/IJACSA.2024.0150968

Villardón, M. P. G. (1986). Una alternativa de representación simultánea: HJ-Biplot. https://api.semanticscholar.org/CorpusID:191105893

Wang, L., Gu, X., & Beck, H. E. (2021). Cyclones and Global Floods from an Observation-Simulation Evaluation: Contributions and Long-Term Changes. Water (Switzerland), 13(21), 2965. https://doi.org/10.3390/W13212965

Conflicts of Interest:

The authors declare that they have no financial, academic, or personal conflicts of interest that may have influenced the conduct or results of this research.

Author Contributions

Bryan Steven Cortez Chichande: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing.

Carlos Luis Muñoz Mendoza: Conceptualization, Formal analysis, Writing – review & editing.

Randy Jimmy Rodríguez Padilla: Conceptualization, Formal analysis, Writing – review & editing.

University and Society publishes its articles under a Creative Commons license https://creativecommons.org/licenses/by-nc-nd/4.0/deed.es