Discrepancy between countries most mentioned in earthquake-related articles and those most affected by major earthquakes since 1965: A bibliometric analysis

Abstract

Background:

Major earthquakes are commonly visualized using Geographic Information Systems (GIS), with bubble maps scaled according to magnitude, depth, or casualty counts. This study investigates the hypothesis that countries most frequently mentioned in earthquake-related academic articles (CMEAs) correspond to those most impacted by significant seismic events.

Methods:

Data on 27,100 major earthquakes (magnitude ≥ 5.5) from 1965 to March 2025 were obtained from the U.S. Geological Survey (USGS). Earthquake magnitudes were visualized using GIS bubble maps, and temporal trends were analyzed based on magnitude and year. In parallel, 24,974 earthquake-related articles published between 2015 and 2024 were retrieved from the Web of Science Core Collection (WoSCC). Ten key metadata elements were analyzed to identify the top 10 CMEAs. Kano diagrams were used to assess the relations between these countries, with articles mentions and publications, and those most affected by major earthquakes. Additionally, a bibliometric analysis using slope graphs was conducted to identify the most prominent article elements exhibiting upward publication trends.

Results:

Key findings include the strongest earthquakes recorded were magnitude 9.1 events in Banda Aceh, Indonesia (2004), and off the coast of Tohoku, Japan (2011); earthquake frequency peaked in the years 2007 and 2010; China contributed the highest number of articles (6137; 24.57%), while the United States had the highest h-index (99 for the U.S. vs 78 for China); the correlation between the number of publications and the countries most severely affected by historical earthquakes was 0.232 (t = 1.167, P = .255), while the correlation between article mentions and those countries was 0.169 (t = 0.664, P = .517).

Conclusion:

This study does not support the hypothesis that countries most frequently discussed in earthquake-related literature correspond to those most affected by major earthquakes. However, the integrated use of hypothesis testing, slope graphs, Kano diagrams, and bibliometric summaries offers a robust framework for future research exploring publication trends in the context of natural disasters.

Key Points.

1. Article mentions do not align with countries most affected by major seismic events (correlation = 0.169, P = .517).

2. GIS bubble maps. Kano diagrams, and bibliometric slope graphs were used to visualize 27,100 earthquakes and 24,974 related articles.

3. China led in publication volume, while the U.S. had the highest citation impact (h-index = 99).

1. Introduction

Earthquakes are the most feared among all-natural disasters because they seem to strike suddenly, without any forewarning.^[1] Such disaster events have frequently occurred and caused enormous damage,^[2] e.g., the magnitude 7.3 Landers, California, Earthquake 1992,^[3] the 2004 Banda Aceh earthquake in Indonesia (Mw = 9.1),^[4] the magnitude 8.0 Wenchuan earthquake 2008 in China,^[5] the magnitude 9.1 Tohoku earthquake off the Pacific coast of Japan in 2011,^[6] and the recent 7.7 Myanmar earthquake in 2025.^[7–10]

1.1. Prior reviews on disaster research and publication trends

Prior reviews have highlighted the global burden of earthquake-related disasters and the role of bibliometric methods in disaster science. For instance, Wang et al (2018) provided a comprehensive review of research developments following the 2008 Wenchuan earthquake, emphasizing the evolution of disaster-related studies.^[11] Similarly, Liu et al (2021) conducted a bibliometric analysis of dengue outbreaks, demonstrating the utility of network-based approaches for trend detection in public health crises.^[12] However, to our knowledge, no study has systematically explored whether countries frequently discussed in earthquake-related articles align with those most impacted by seismic events. This study addresses that gap through a bibliometric and geospatial approach.

1.2. Using the intensity attenuation model to measure the seismic hazard

Most major earthquakes were displayed with the Geographic Information System (GIS) to map bubbles sized by individual magnitudes, depths, or death tolls/injuries.^[13–16] Besides the “Richter” magnitude scale (labeled ML),^[17] there are numerous seismic magnitude scales that are used for measuring the extent of earthquake magnitude, e.g., Richter original local scale,^[18] the Russian surface-wave MLH scale,^[19] the Japanese Meteorological Agency magnitude scale,^[20] Moment magnitude scale,^[21] and Body-wave magnitude scales.^[22] The intensity attenuation model expressed by Eq. (1)^[23] below intrigues us to compare earthquakes with visualizations using the intensity score that denote the peak ground acceleration (PGA, i.e., equal to the maximum ground acceleration that occurred during earthquake shaking at a location, earthquake ground motion is often defined in terms of PGA.

$${\displaystyle \begin{array}{l}{\displaystyle \mathrm{I}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{y}\mathrm{s}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{e}=3.32+1.44\mathrm{M}\mathrm{L}-3.34\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{R},}\end{array}}$$ (1)

where ML is local magnitude and R is hypocentral distance in kilometers.

We are thus motivated to illustrate significant earthquakes in comparison using visualizations. That is, the earthquake intensity score is determined by both magnitude and hypocentral distance (in km) to measure the seismic hazard. The first research question is which year has the highest seismic hazard based on the weighted Intensity score in Eq. (1).

The second question is which earthquake has the highest seismic hazard among the main earthquakes, including the magnitude 7.3 Landers, California, Earthquake 1992,^[3] the magnitude 9.1 Earthquake 2004 in Banda Aceh, Indonesia,^[4] the magnitude 8.0 Wenchuan earthquake 2008 in China,^[5] the magnitude 9.0 Tohoku earthquake off the Pacific coast of Japan in 2011,^[6] and the recent 7.7 Myanmar earthquake in 2025^[7–10] when taking magnitude and depth into account, with the PGA in Eq. (1).

1.3. Which country has the most earthquakes?

Which country has the most earthquakes is one of the frequently asked questions from the public. It can be defined from 4 perspectives^[24]:

1. Japan locates the most active seismic areas and has the densest seismic network in the world as numerous such records of the earthquake are in Japan.

2. Indonesia, in a very active seismic zone, actually has the most earthquakes with a larger size than Japan.

3. Tonga, Fiji, and Indonesia have the most earthquakes per unit area since they are all in extremely active seismic areas along subduction zones.

4. China, Iran, and Turkey have the most catastrophic earthquakes causing the most damage and fatalities in the long historical records.

Investigating the evolution of disaster science is worthwhile and can be used to improve the future execution of disaster risk management.^[11] The most studied earthquakes were TOHOKU2011 (1519 publications),^[25] followed by WENCHUAN2008 (1217),^[26] SUMATRA2004 (836),^[27] CHI-CHI1999 (640)^[28] according to the International Seismological Centre (ISC).^[29] The third research question is thus conceived on which countries were authored most in published articles in research on the first and corresponding authors affiliated countries and mentioned in abstracts using bibliometric analysis, compared to 21st-century earthquakes.^[30]

1.4. Study objectives

Two topics of earthquake and publication are involved in this study. Two approaches using the traditional event counts and intensity scores are used for comparing the countries struck most by earthquakes between the 2 approaches. The hypothesis of most countries mentioned in articles is consistent with countries having most struck earthquakes was made and required for verifications.

2. Methods

2.1. Data source

2.1.1. The earthquake topic

This data of 27,100 main earthquakes were downloaded from the Kaggle website^[26] and the U.S. Geological Survey (USGS),^[24] including the date, time, location, depth, magnitude, and source of every earthquake with a reported magnitude (labeled Mw) 5.5 or higher in years from 1965 to 2025, as shown in Supplemental Digital Content 1 (Supplemental Digital Content, https://links.lww.com/MD/P879) and Supplemental Digital Content 2 (Supplemental Digital Content, https://links.lww.com/MD/P880).

2.1.2. The publication topic

A total of 24,974 article abstracts was downloaded by searching the keyword of earthquakes in Title and topic in Web of Science core collections (WoSCC) between 2015 and 2024. Ten key metadata elements were analyzed to identify top 10 countries most frequently authored in earthquake-related academic articles (CMEAs).

2.1.3. Ethical approval waived

All downloaded data are from the USGS^[24] and have met the requirement for analyzing information from public websites. Ethical approval is not necessary for this study because neither human subjects nor personal data were accessed.

2.2. Data arrangements and representations

A Microsoft Excel visual basic application module was used to wrangle the data. Two topics of the earthquake and publication involving countries authored in articles were applied using 2 approaches: (i) the traditional unweighted event counts and (ii) the proposed weighted intensity (PGA) scores in Eq. (1).

2.2.1. Earthquake topic

2.2.1.1. Two approaches using unweighted magnitude and weighted PGA scores

2.2.1.1.1. A traditional approach using the unweighted counts

We visualized earthquake events on a world map with bubbles sized by the magnitude across magnitudes over the years.

2.2.1.1.2. A proposed approach using the weighted PGA scores

The intensity attenuation model^[19] was applied to calculate the weighted intensity score (denoted by PGA) for each year and earthquake of magnitude within a given year using Eq. (1).

2.2.2. Publication topic

One summary report was created for top 10 elements in each article entity.^[31] Author collaborations by country were displayed using social network analysis (SNA).^[32,33]

The SNA is the process of investigating social structures through the use of networks and graph theory. It characterizes networked structures in terms of nodes (individuals, groups of people, or objects within the network) and the ties, edges, or links (relationships or interactions) that connect them.^[33] The use of SNA was thus expected to identify the burst countries mentioned in articles based on FLCA algorithm.^[34]

The purpose of this section is to verify whether countries having the most earthquakes^[20] are consistent with countries frequently authored and mentioned in articles.

2.3. Statistical tools and data analysis

Descriptive statistical methods were used to summarize earthquake characteristics (e.g., magnitude, PGA, depth) and publication metadata (e.g., article counts, countries, h-index) in Supplemental Digital Content 1 (Supplemental Digital Content, https://links.lww.com/MD/P879) and Supplemental Digital Content 2 (Supplemental Digital Content, https://links.lww.com/MD/P880). Visual tools, including slope graphs^[31] were applied to explore trends (based on the method of gene expression analysis^[35]) and Kano diagrams^[36,37] were used to examine the correlation between article mentions and earthquake occurrence by country Given the observational and bibliometric nature of the study, inferential univariate analysis (e.g., t-tests) was conducted, as specific group comparisons or hypothesis tests between variables were required in both axes.

The pages of Hypertext Markup Language (HTML) used for Google Maps were created. All relevant information was linked to the dashboards laid on Google Maps. The instruction for conducting this study is deposited in Supplemental Digital Content 3 (Supplemental Digital Content, https://links.lww.com/MD/P881).

3. Results

3.1. Topic 1 of earthquake using visuals and statistics

A total of 27,100 main earthquakes with magnitudes ≥ 5.5 are displayed on the world map in Figure 1, using bubbles sized by magnitude. The most significant events include the 2004 Indonesia earthquake, the 2011 Tohoku earthquake in Japan, the 2010 Chile earthquake, and the 1965 Rat Islands earthquake in Alaska (US). Earthquakes are color-coded into 5 magnitude tiers: dark green (≥9.0), yellow (≥8.0), red (≥7.0), light green (≥6.0), and brown (≥5.6).

Figure 1.

Geographical earthquakes from 1965 to 2025 using world maps.

Summary statistics for these earthquakes are presented in Table 1, categorized into 10 strata based on PGA scores.

Table 1.

Summary statistics for peak ground acceleration (PGA) among 27,100 earthquakes (magnitude ≥ 5.5, 1965–2025).

Strata	Variable	PGA
	Sample size	27,100
	Mean magnitude	5.88
	Mean depth	70.61
	Year range	1965–2025
	Lowest value	−10.65
	Highest value	14.17
	Arithmetic mean	0.03004
	95% CI for the Arithmetic mean	−0.01448 to 0.07457
	Median	0.14
	95% CI for the median	0.07000 to 0.1800
	Variance	13.985
	Standard deviation	3.7397
	Relative standard deviation	124.4761 (12,447.61%)
	Standard error of the mean	0.02272
	Coefficient of Skewness	−0.6037 (P < .0001)
	Coefficient of Kurtosis	0.6597 (P < .0001)
	D’Agostino–Pearson test	Reject normality (P < .0001)
	for Normal distribution

	Percentiles		95% confidence interval
1	<2.5	<−9.49
2	2.5	−9.42	−9.4900 to −9.3400
3	5	−7.97	−8.1400 to −7.7576
4	10	−5.15	−5.2800 to −5.0333
5	25	−1.67	−1.7597 to −1.6000
6	75	3.23	3.2300 to 3.2300
7	90	3.87	3.8100 to 3.9500
8	95	4.48	4.3976 to 4.5300
9	97.5	5.17	5.1000 to 5.2600
10	>97.5	>5.26

Dataset: 27,100 main earthquakes with magnitude ≥ 5.5 (from 1965 to 2025); PGA (peak ground acceleration) reflects the maximum ground motion measured during an earthquake and is used to represent seismic hazard intensity.

Bold value indicates P < .05.

The 2 years in 2007 and 2010 had the high frequencies. The annual distribution of earthquakes (Mw ≥ 5.5) from 1965 to 2025 is shown in Figure 2. The highest average PGA values (1.4) were recorded in 2013 and 2014.

Figure 2.

Distribution of earthquakes (Mw ≥ 5.5) from 1965 to 2025.

Significant individual events include:

• The 1992 magnitude 7.3 Landers earthquake in California^[3] with a PGA of 14.17,

• The 2004 magnitude 9.1 Banda Aceh earthquake in Indonesia,^[4] PGA 4.95,

• The 2008 magnitude 8.0 Wenchuan earthquake in China,^[5] PGA 4.69,

• The 2011 magnitude 9.1 Tohoku earthquake off Japan’s Pacific coast,^[6] PGA 5.06,

• The recent 2025 magnitude 7.7 earthquake in Myanmar,^[7–10] PGA 6.4.

3.2. Publication topic

3.2.1. Publications across countries, journals and over years

China contributed the highest number of articles (6137; 24.57%), while the United States had the highest h-index (99 vs 78 for China), as shown in Figure 3.

Figure 3.

Summary reports of 24,974 earthquake-related articles in recent 10 yr.

The journals with the most number of publications on the topic of the earthquake were Journal of Geophysical Research: Solid Earth, Seismological Research Letters, and Bulletin of the Seismological Society of America in Figure 3.

The highest number of articles were published in years 2014, 2011, and 2013 (Fig. 3).

Thirty-one significant article elements showing increasing and decreasing trends are in Figure 4. Slope graphs for 31 significant article elements showing increasing and decreasing trends are in Figure 5, such as China, India, Russia, and Iran, with increasing publication counts in past 10 years.

Figure 4.

Thirty-one significant article elements showing increasing and decreasing trends.

Figure 5.

Slope graphs for 31 significant article elements showing increasing and decreasing trends.

Author collaborations led by China, the U.S. and the U.K. are shown in Figure 6A, compared to countries (led by Japan and Turkey) most frequently mentioned in abstracts, as shown in Figure 6B.

Figure 6.

Author collaborations led by China, the U.S. and the U.K.

3.3. Kano diagrams used to present correlations between article mentions and earthquake occurrences by country

The correlation between the number of publications and the countries most severely affected by historical earthquakes was 0.232 (t = 1.167, P = .255), while the correlation between article mentions and those countries was 0.169 (t = 0.664, P = .517), as shown in Figure 7. Nonetheless, the correlation between the number of publications and article mentions was significant with R = 0.642 (t = 8.295, P < .0001).

Figure 7.

Applying Kano diagrams to reveal the correlation between values in x and y axes.

4. Discussion

We observed that the strongest earthquakes recorded were both magnitude 9.1 events: one in Banda Aceh, Indonesia (2004), and the other off the coast of Tohoku, Japan (2011); the highest frequency of earthquakes occurred in the years 2007 and 2010; China produced the most earthquake-related publications (6137 articles; 24.57%), while the United States achieved the highest h-index (99 for the U.S. vs 78 for China), indicating greater citation impact; correlations between affected countries and publication counts and article mentions were 0.232 (t = 1.167, P = .255) and 0.169 (t = 0.664, P = .517), respectively.

The hypothesis that countries most frequently mentioned in articles align with those most affected by earthquakes was not supported in this study.

4.1. What this knowledge adds to what we already knew

A 7.7 magnitude earthquake struck Myanmar on 28 March, killing over 2700. Aid is restricted by the ruling junta, while damaged health facilities face severe shortages of medical supplies.^[8] The PGA, classified in the tenth stratum of Table 1, reached 6.4.

This study was divided into 3 topics. In the first topic with unweighted magnitudes, we see most major earthquakes were displayed with GIS to map bubbles sized by the magnitude^[9,33–35] as we did in Figure 1, similar to the result in the website.^[38]

The intensity attenuation model^[17] was applied to the first topic with weighted intensity (PGA) scores which define what ground motion should be expected at Location A due to an earthquake of known magnitude at Location B (or epicenter). High seismicity areas such as California (in the US, with PGA = 14.17 in 1992) and Japan (with PGA = 5.06, 2011) have a tendency sufficiently high to pose a potential hazard to sensitive installations such as dams and chemical plants.^[17] This research has made it possible to calculate the level of hazard for main earthquakes (i.e., a large earthquake around magnitude 5.5 or more.) since 1965. The recent 2025 magnitude 7.7 earthquake in Myanmar,^[7–10] which recorded a PGA of 6.4, attracted significant international attention.

Prior to the development and deployment of seismographs (starting around 1900), magnitudes can only be estimated based on historical reports of the extent and severity of damage.^[39] Earthquakes that occurred in remote areas prior to the advent of modern instrumentation in the early to mid-1900s were not well-reported, and the locations and magnitudes of such events are often unknown. Therefore, the apparent increase in large earthquake frequency over the last few centuries is unlikely to be accurate.^[8] The time frame of main earthquakes since 1965 is reasonable and downloaded from the Kaggle website.^[22]

In the second topic, the scientific visualization of bibliometric analysis has been achieved using many kinds of tools used to investigate the evolution of scientific issues,^[11,40–42] including CiteSpace,^[40] VoSviewer,^[43] Pajek,^[44] and CitNetExplorer.^[45] The geospatial networks of citations could be adequately depicted by Google Maps as we did in this study with Figure 2.

In the third topic, the hypothesis that countries contributing most frequently to earthquake-related articles are also those most affected by earthquakes was not supported (R = 0.169, and R = 0.23, P > .05, for article mentions and publications, respectively). This finding, which reveals a lack of alignment between research activity and seismic exposure, is rarely highlighted in traditional bibliometric analyses, where such hypotheses are seldom subjected to empirical validation.

As previous studies did not provide instructions for extracting countries from earthquake occurrences and article mentions,^[46,47] 2 novel R scripts^[48,49] provided in Supplemental Digital Content 1 (Supplemental Digital Content, https://links.lww.com/MD/P879) were applied in this study. A total of 17 and 27 countries – excluded from article mentions and author-affiliated countries, respectively – did not align with the consistency of earthquake occurrences shown in Figure 7. Additionally, the R script for generating the Kano diagram^[50] was used to visualize the classification results.

Several research questions were proposed in the Introduction section. The use of hypothesis-driven validation is recommended for future bibliometric studies to enhance readers’ understanding of research trajectories and thematic development.

4.2. What this study contributes to current knowledge

4.2.1. Integration of seismic intensity and GIS visualization

This study advances earthquake hazard assessment by combining unweighted magnitudes with weighted intensity scores (PGA) derived from an attenuation model, enabling a more precise comparison of seismic hazards across time and geography using GIS-based visualizations.

4.2.2. Link between earthquake frequency and research productivity

By cross-analyzing earthquake events and scientific publication data through the 2 novel R scripts,^[48,49] the study reveals a weak correlation between countries most affected by major earthquakes and those contributing most to earthquake-related research – a novel validation rarely addressed in traditional bibliometric studies.

4.2.3. innovative use of geospatial bibliometric mapping

The study introduces an interactive Google Maps-based approach to visualize bibliometric networks, author collaborations, and seismic distributions, enhancing the interpretability and accessibility of global research trends in earthquake science.

4.2.4. Kano diagrams reveal misalignment between seismic impact and research focus

The Kano diagrams (Fig. 7) visually illustrate the discrepancy between the countries most frequently mentioned in earthquake-related literature and those most affected by major seismic events. The asymmetric distribution of data points along the diagonal axes indicates a lack of alignment between research activity (Y-axis) and seismic exposure (X-axis). Notably, only a few countries occupy the upper-right quadrant (high research attention and high impact), while many cluster near the origin, suggesting limited representation of highly affected regions in scholarly output. These findings reinforce the central conclusion of this study: that research attention does not proportionally reflect actual earthquake impact across countries.

4.3. Strengths of this study

We have not yet found an article to verify countries struck most by earthquakes consistent with those authored frequently in articles. There are 3 strengths in this study:

• The study confirms a lack of correlation between earthquake-prone countries and those most frequently mentioned in academic literature using Kano diagrams.

• GIS visualizations and bibliometric tools were integrated to explore global earthquake trends from 1965 to 2025.

• China led in publication volume, while the U.S. had the highest h-index.

4.4. The most-cited articles

The top-ranked article was titled “Convolutional neural network for earthquake detection and location” authored by Perol et al, in 2018.^[46] It was cited 604 times. This study introduces ConvNetQuake, a scalable AI-based convolutional neural network for detecting and locating earthquakes from single waveforms. Applied to Oklahoma’s induced seismicity, it identified over 17 times more events than official catalogs and operates significantly faster than traditional methods, advancing seismic hazard assessment capabilities.

4.5. Limitations and suggestions

Despite the findings shown above, several potential limitations require further research efforts in the future. First, the sample of this study only comprised articles in WoSCC. It should not be generalized to other databases such as the Scientific Citation Index (Thomson Reuters, New York) and Scopus (Elsevier, Amsterdam, The Netherlands). As such, the most-cited articles and the countries authored in articles are barely determined by the publications indexed in WoS.

Second, there might be some biases when extracting countries mentioned in articles though multi-countries mentioned have been considered and based on Keywords with earthquake in Title and topic. Because only WoS provides Keyword plus involving earthquake for following analyses, other databases used for the relevant studies should be cautious or search for other text-mining techniques to extract earthquake-related articles.

Third, we recommend using SNA to partition clusters. SNA is not limited to a specific software used in this study because many such kinds of software are in use in academics.^[40,43–45] The style of the visual representations might be somewhat different, but the principle and algorithm of partitions for cluster analysis are similar.^[34,51]

Fourth, although our suggestions are limited to main earthquakes within years between 1965 and 2025, other time spans or topics can be applied in the future, such as a hurricane, flu-like outbreaks, HIV/AIDS, flood, Typhon, and avian influenza, all of which can be explored using the technique of keyword-based country extraction, such as the 3 R scripts^[48–50] provided in this study. The correlation between the number of publications and article mentions was significant (R = 0.642, t = 8.295, P < .0001), suggesting that the relationship between these 2 variables warrants further investigation in future bibliometric studies focused on article characteristics.

Fifth, the formula in Eq. (1) is not unique. Many researchers developed various formulas to compute the intensity scores,^[47] which are worth comparing with each other in the future.

Sixth, Turkiye and Turkey were both included in the top 10 countries shown in Figure 1, despite being identical, which may lead to duplicated representation.

Seventh, this study adopts a bibliometric observational design rather than an umbrella review type, as it aims to quantitatively analyze metadata from original articles – rather than synthesize findings from existing systematic reviews or meta-analyses. As such, it does not assess the quality of evidence or outcomes reported in individual studies, which may limit its applicability for drawing clinical or policy recommendations.

Eighth, this study did not include univariate or multivariate inferential analyses, as the dataset comprised aggregated bibliometric and geospatial metadata without group-based experimental design. The focus was on trend visualization and descriptive characterization, which may limit inferential conclusions. Future studies using individual-level or structured outcome data could enable statistical hypothesis testing for deeper analysis.

Finally, numerous scientometrics were used for evaluating the authors’ individual research achievements. We merely analyzed countries involved in this study, such as the first and corresponding authors affiliated to countries. The kind of studies related to earthquakes can be further done using bibliometric analysis in the future.

5. Conclusions

This study demonstrates that integrating seismic intensity modeling with bibliometric and geospatial analyses provides a comprehensive view of earthquake hazards and research trends. By validating the mismatch between seismic activity and scientific output, and introducing interactive visual tools, the study offers a novel framework for enhancing global earthquake risk understanding and preparedness.

Acknowledgments

We thank Coding Service for the English language review of this manuscript.

Author contributions

Conceptualization: Wan-Ching Kao.

Data curation: Julie Chi Chow.

Formal analysis: Willy Chou.

Investigation: Lifan Chen.