the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Jan 2025
Cities near volcanoes: Which cities are most exposed to volcanic hazards?
Abstract. Cities near volcanoes expose dense concentrations of people, buildings, and infrastructure to volcanic hazards. Identifying cities globally that are exposed to volcanic hazards helps guide local risk assessment for better land-use planning and hazard mitigation. Previous city exposure approaches have used the city centroid to represent an entire city, and to assess population exposure and proximity to volcanoes. But cities can cover large areas and populations may not be equally distributed within their bounds, meaning that a centroid may not accurately capture the true exposure. In this study, we suggest a new framework to rank global city exposure to volcanic hazards. We assessed global city exposure to volcanoes in the Global Volcanism Program database that are active in the Holocene by analysing populations located within 10, 30, and 100 km of volcanoes. These distances are commonly used in volcanic hazard exposure assessment. City margins and populations were obtained from the Global Human Settlement (GHS) Model datasets. We ranked 1,106 cities based on the number of people exposed at different distances from volcanoes, the distance of the city margin from the nearest volcano, and by the number of nearby volcanoes. Notably, 50 % of people living within 100 km of a volcano are in cities. We highlight Jakarta, Bandung, and San Salvador, as scoring highly across these rankings. Bandung, Indonesia ranks highest overall with over 8 million people exposed within 30 km of up to 12 volcanoes. South-east Asia has the highest number of exposed city populations (~162 million). Jakarta (~38 million), Tokyo (~30 million), and Manila (~24 million) having the largest number of people within 100 km. Central America has the highest proportion of its city population exposed, with Quezaltepeque and San Salvador exposed to the most volcanoes (n=23). Additionally, we ranked the 1,283 Holocene volcanoes by the city populations exposed within 10, 30, and 100 km, the number of nearby cities, and distance to nearest city. Tangkuban Parahu, San Pablo Volcanic Field, and Tampomas score highly across these rankings. Notably, Gede-Pangrango (~48 million), Languna Caldera (~8 million), and Nejapa-Miraflores (~0.8 million) volcanoes have the largest city populations within 100, 30, and 10 km, respectively. We developed a web app to visualise all the cities with over 100,000 people exposed. This study provides a global perspective on city exposure to volcanic hazards, identifying critical areas for future research and mitigation efforts.
- Preprint
(2255 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on nhess-2024-219', Amiel Nieto Torres, 06 Mar 2025
The manuscript "Cities near volcanoes: Which cities are most exposed to volcanic hazards?" By Elinor S. Meredith et al., and the associated databases constitute a very important and up-to-date contribution to the assessment of exposure to volcanic hazards at a global level. Together they form a tool that will allow other researchers to better understand how cities are exposed to volcanic activity and for decision makers to better plan cities. Overall, the manuscript is well laid out in its introduction and is built on recent relevant work. The methodology is generally clear and detailed. The results are well presented and explained. There are very few typos that were pointed out in the attached file.
Here are some suggestions that I think can help improve the manuscript:
There are some citations that are not in the reference list. It is suggested to review the citations and references in depth. The figures, tables and appendices are adequate, although I suggest carefully reviewing table 3, I made comments in the PDF file.
There are some minor comments in the discussion section, for example: The manuscript rightly mentions the limitation that the methodology only considers city polygons (≥ 1500 inhabitants/km2 and a total population size of at least 50,000) and does not consider sub-urban and rural areas. I think it is necessary to broaden the discussion a bit to the fact that the ranking of volcanoes could change significantly if this population were considered. In most volcanic disasters the victims are in rural communities, for example the 1982 eruption of the Chichon volcano in Mexico, caused the death of at least 2 thousand people, all living in small rural communities and the effects in cities were minor. This is a limitation of the methodology and should be clearly expressed.
Also, it is worth adding a few lines about the fact that the methodology does not consider the probability of eruption. The fact that in El San Salvador is found the city with the largest number of volcanoes within 100 km does not mean that it is the city where the hazard is higher. This is a caveat of the methodology that it would be good to establish.
Regarding the volcanic fields. Without a doubt, this is one of the most important challenge, for which no existing methodology has been able to address satisfactorily. Because the site of the next eruption is not known with precision, considering the center of the field or the site of the most recent eruption should be considered as a weakness of the methodology, when dealing with such extensive areas, the uncertainty about the number of people exposed to volcanic hazards is high. Also, most volcanic fields generally present basaltic-type volcanism, with a distribution of its products and volume generally quite restricted, so the relevance of considering the 100 km buffer in these areas could be discussed.
The conclusions repeat a bit of the results and the discussion, which makes them slightly redundant. Aspects of other works are also cited, which is not pertinent.
Specific comments and corrections can be found within the PDF file.
-
CC1: 'Reply on RC1', Elinor S. Meredith, 10 Apr 2025
Publisher’s note: this comment is a copy of AC1 and its content was therefore removed on 11 April 2025.
Citation: https://doi.org/10.5194/nhess-2024-219-CC1 -
AC1: 'Reply on RC1', Elinor Meredith, 11 Apr 2025
Thank you for taking the time to provide comments on the manuscript. Below we provide our response to the comments:
“There are some citations that are not in the reference list. It is suggested to review the citations and references in depth. The figures, tables and appendices are adequate, although I suggest carefully reviewing table 3, I made comments in the PDF file.”
We will make the edit suggestions provided in the PDF and we will go through to check the reference list. For Table 3 we will edit the ranking as suggested. There is a comment in the results section about what total exposed city population means and why it is within 100 km if we had stated in the introduction how hazard deposits may extend beyond this. In the methodology we had justified why we still used the 100 km buffers. Lines 242-244: as these are commonly used to assess volcanic exposure based on typical maximum distances of primary volcanic hazards (Ewert, 2007; Brown et al., 2017; Biass et al., 2024). To clarify, we will add to lines 291-292: Here, we define an exposed city population as the population of cities located within 100 km of a volcano.
“There are some minor comments in the discussion section, for example: The manuscript rightly mentions the limitation that the methodology only considers city polygons (≥ 1500 inhabitants/km2 and a total population size of at least 50,000) and does not consider sub-urban and rural areas. I think it is necessary to broaden the discussion a bit to the fact that the ranking of volcanoes could change significantly if this population were considered. In most volcanic disasters the victims are in rural communities, for example the 1982 eruption of the Chichon volcano in Mexico, caused the death of at least 2 thousand people, all living in small rural communities and the effects in cities were minor. This is a limitation of the methodology and should be clearly expressed.”
We constrained the methodology to the densest part, as impact here will likely have the greatest effect on the city. As you state, the 1982 eruption of El Chichón had minor effects on the city itself but great effects on the suburbs. We think that future research could include or individually assess suburban populations and that more study is needed on how interconnected cities and suburbs are in terms of impacts. We do agree that including the suburban areas of the city will increase exposure, so we will include the following sentences to lines 579-583: If suburban and rural populations were also considered, the ranking of volcanoes would likely change, as these areas may have higher exposure to volcanic hazards. Past volcanic eruptions, such as the 1982 El Chichón eruption in Mexico, have primarily affected rural communities rather than city centres. Further research is needed to understand how the interconnectedness of cities and suburbs, influences impact.
“Also, it is worth adding a few lines about the fact that the methodology does not consider the probability of eruption. The fact that in El San Salvador is found the city with the largest number of volcanoes within 100 km does not mean that it is the city where the hazard is higher. This is a caveat of the methodology that it would be good to establish.”
Yes, we chose not to include the probability of eruption and instead look at the number of volcanoes as an indication of the hazard. As global-level hazard assessment has its own limitations and uncertainties, we state in the text that localised hazard assessment should be conducted once the key exposed cities are identified. We will add this sentence to Lines 603-606: For example, the presence of multiple volcanoes within 100 km, as in San Salvador, does not directly equate to higher hazard, as eruption probabilities vary. While our approach identifies cities with high volcanic exposure, localised hazard assessments are essential for a more precise evaluation of the threat. We will add this phrase in bold here to Line 593: Future work could classify volcanoes by probability of eruption, last eruption, VEI range, or tectonic setting to better understand the specific types of potential hazard they pose to nearby cities.
“Regarding the volcanic fields. Without a doubt, this is one of the most important challenge, for which no existing methodology has been able to address satisfactorily. Because the site of the next eruption is not known with precision, considering the center of the field or the site of the most recent eruption should be considered as a weakness of the methodology, when dealing with such extensive areas, the uncertainty about the number of people exposed to volcanic hazards is high. Also, most volcanic fields generally present basaltic-type volcanism, with a distribution of its products and volume generally quite restricted, so the relevance of considering the 100 km buffer in these areas could be discussed.”
The reviewer picks up the important limitation of using point sources for volcanic fields, as eruptive products may extend beyond this distance. As explored in our paper, as well as in Biass et al. (2024), in general, using concentric radii is a conservative approach. The reviewer suggests changing volcanic fields on line 207 to distributed volcanism however, this sentence is directly referring to the term used in the GVP. However, we will change volcanic fields to distributed volcanism on line 600 and also add this sentence: Lines 208-210: Using a single point to represent distributed volcanism introduces uncertainty, as the precise location of future eruptions is unknown. We will also add the phrase in bold here into Line 601: Additionally, research could explore alternative methods to the traditional volcano buffers, considering approaches that account for the spatial variability of distributed volcanism, such as those used in Nieto-Torres et al. (2021), and of shield volcanoes.
“The conclusions repeat a bit of the results and the discussion, which makes them slightly redundant. Aspects of other works are also cited, which is not pertinent.”
We will make the conclusion more concise and focus on the contributions of this study and their implications instead of methods.
Citation: https://doi.org/10.5194/nhess-2024-219-AC1
-
CC1: 'Reply on RC1', Elinor S. Meredith, 10 Apr 2025
-
RC2: 'Comment on nhess-2024-219', Luis E. Lara, 05 May 2025
The manuscript ‘Cities near volcanoes: Which cities are most exposed to volcanic hazards? by Meredith et al. is a nice piece of work that moves the population exposure assessment to volcano hazards to a next level. In fact, the real spatial distribution of population in urban areas seems to be especially relevant when large cities (even megacities) are threatened by natural hazards, the latter causing an impact that depends on the distance to the source.
The approach used in the manuscript is clear and easy to reproduce, with meaningful results in terms of a ranking of exposed cities. However, further discussion could be included regarding:
- The significance of the city polygons when urban areas are inherently dynamic, and when transition from urban to rural zones is essentially diffuse (e.g., Chen et al., 2018; Sorensen and Danielle, 2020).
- The reason for using buffers at 10, 30 and 100 km from active volcanoes, commonly used in population exposure analysis but perhaps not thus useful to capture the decreasing density of population from the centroid, especially between 30 and 100 km, where probably most the large cities have their bounds. Why not using a continuous scale with intervals at 10 km?
No further comments in the pdf
-
AC2: 'Reply on RC2', Elinor Meredith, 09 May 2025
Thank you for taking the time to provide comments on the manuscript. Below we provide our response to the comments:
“The approach used in the manuscript is clear and easy to reproduce, with meaningful results in terms of a ranking of exposed cities. However, further discussion could be included regarding: 1. The significance of the city polygons when urban areas are inherently dynamic, and when transition from urban to rural zones is essentially diffuse (e.g., Chen et al., 2018; Sorensen and Danielle, 2020).”
We agree with the reviewer that the boundaries between urban and rural are diffuse and can change through time. These important considerations are being addressed as part of our separate ongoing research. For the scope of the current study, we use these city polygons to represent areas of highest population density compared to surrounding areas, with the assumption that these are the areas where impacts will be likely most severe and concentrated. We are also consistent with the GHS-UCDB dataset, which also refers to these areas as “urban centres”. To add a bit more discussion on this topic, we added the following sentences to the future work part of lines 573-575: "The delineation between urban, peri-urban, and rural areas is not always clear (Chen et al., 2018; Sorensen and Labbé, 2020). Urban centres can sprawl into surrounding peri-urban and rural regions, and through time, urban centres may merge together and densify.” To capture more about the dynamic population, we also added this phrase: "Our approach could be expanded or used to explore more than just cities, by assessing changing land-use patterns, transient populations using different population datasets (e.g., LandScan ambient population dataset: Lebakula et al., 2024)”
"1. “The reason for using buffers at 10, 30 and 100 km from active volcanoes, commonly used in population exposure analysis but perhaps not thus useful to capture the decreasing density of population from the centroid, especially between 30 and 100 km, where probably most the large cities have their bounds. Why not using a continuous scale with intervals at 10 km?”
Thanks for this valuable observation. The choice of the 10, 30, and 100 km buffer distances was based on the methods of prior volcanic hazard and exposure analyses (e.g., Aspinall et al., 2011; Brown et al., 2015), so it can be used comparatively with these studies. These thresholds broadly correspond to the average of maximum extents of life-threatening volcanic hazards at different eruption magnitudes. For example, VEI ≤3 eruptions typically affect areas within ~10 km, VEI ≤4 eruptions up to ~30 km, and VEI ≤5 eruptions up to ~100 km (Biass et al., 2023; Ewert, 2007). These are based on past observations of hazard extents (e.g., Newhall and Hobblitt, 2002). Therefore, we feel this provides a meaningful framework for assessing potential impact zones, as they can be related to eruption sizes. Although we acknowledge that this approach does not fully capture the spatial gradient in population density, particularly between 30 and 100 km where large urban areas are located. We are exploring in our follow up study using 10 km interval buffers to assess more about the population density distribution, although this falls out the scope of this study. This sentence is added to the future work, Lines 579-581: “Our approach could be expanded or used to explore more than just cities, by assessing changing land-use patterns, transient populations using different population datasets (e.g., LandScan ambient population dataset: Lebakula et al., 2024), or gradations of population density around volcanoes using different buffer distances.”References used in this reply:
Aspinall WP, Auker MR, Hincks TK, Mahony SH, Pooley J, Nadim F, Syre E, Sparks RSJ, Bank TW (2011) Volcano hazard and exposure in track II countries and risk mitigation measures – GFDRR volcano risk study. The World Bank, p 309
Biass S, Jenkins SF, Hayes JL, Williams GT, Meredith ES, Tennant E, Yang Q, Lerner GA, Burgos V, Syarifuddin M, Verolino A (2023). How well do concentric radii approximate population exposure to volcanic hazards?. Bulletin of volcanology, 86(1), 3.
Brown SK, Auker MR, Sparks RSJ (2015) Populations around Holocene volcanoes and development of a Population Exposure Index. In: Loughlin S, Sparks S, Brown S et al (eds) Global Volcanic Hazards and Risk. Cambridge University Press, pp 173–222
Ewert J (2007) System for ranking relative threats of U.S. volcanoes. Nat Hazards Rev 8:112–124
Newhall C, Hoblitt R (2002) Constructing event trees for volcanic crises. Bull Volcanol 64:3–20Citation: https://doi.org/10.5194/nhess-2024-219-AC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 341 | 123 | 14 | 478 | 15 | 14 |
- HTML: 341
- PDF: 123
- XML: 14
- Total: 478
- BibTeX: 15
- EndNote: 14
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1