the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 22 Jun 2026
The 2022 Rainfall-Triggered Landslide-Tsunami Disaster in Pilar, Abuyog, Leyte, Philippines and its implications for coastal hazard assessment
Abstract. On 12 April 2022, a rainfall-triggered landslide in Barangay Pilar, Abuyog, Leyte, Philippines generated destructive tsunamis upon reaching the coast, resulting in one of the deadliest landslide disasters in the country’s history. The event occurred during prolonged and intense rainfall associated with the interaction of a low pressure area, Severe Tropical Storm Megi (Agaton), and Typhoon Malakas (Basyang). The cascading rainfall–induced landslide–tsunami hazards caused 54 fatalities, injured 49 individuals, and left 33 persons missing and presumed dead, with tsunami inundation responsible for a substantial proportion of the losses. We reconstructed the event chronology and failure mechanisms using an integrated dataset comprising satellite imagery, unmanned aerial vehicle (UAV) surveys, field mapping, sedimentological analyses, and three-dimensional terrain modelling, eyewitness accounts, social media videos, and incident reports. Results indicate that two landslides with a total volume of around 5.5 x 105 m3 were mobilized as debris flows. The landslides were composed of volcanic materials that originated from a fault-controlled ridge south of Pilar that reached the coast of Leyte Gulf. The initial and larger debris flow generated tsunami waves with runup heights of about 12 m above sea level, devastating residential areas along the coast and overwhelming people fleeing the landslide. Analysis of landslide–tsunami coupling indicates that wave generation was governed primarily by high-momentum impact conditions, with debris flow velocities of up to 25 m/s and short source-to-shore distances limiting energy dissipation prior to coastal entry. These conditions produced strongly supercritical flow and demonstrate that relatively moderate-volume landslides can generate significant near-field tsunamis when velocity and proximity are favorable. This study is the first well-documented subaerial landslide-generated tsunami in the Philippines and provides a data-rich example of a rainfall-induced landslide–tsunami cascade in a tropical island setting. The Pilar disaster highlights the extreme hazard posed by rainfall-triggered landslides in steep coastal volcanic terrains and underscores the need to explicitly incorporate landslide–tsunami cascade scenarios into hazard assessment, early warning systems, and land-use planning in similar coastal regions.
Competing interests: Richard Ybanez is a PhD student of NHESS Executive Editor Dr. Bruce D. Malamud in the University of the Philippines National Institute of Geological Sciences. Dr. Alfredo Mahar Francisco A. Lagmay is the main supervisor of Mr. Ybañez in the same research. No other competing interest is known to the authors.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(7469 KB) - Metadata XML
-
Supplement
(34051 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-2543', David Tappin, 17 Jul 2026
-
AC1: 'Reply on RC1', Likha G. Minimo, 18 Jul 2026
We summarized our response to RC1 in this table.
RC 1 AC It’s an interesting paper with a great deal of important information, but in its present form it’s hard to see the wood for the trees.
Thank you for your comments! We’ll work on the details to make the presentation clearer. Paper strengths are the mapping of the landslide area and the images presented and the eyewitness interviews. This is a great data set and the basis of the paper. However, the interpretations are poorly shown in the key figures (e.g. 3), which are too small and with too low resolution to see the features described in the text – e.g. Lines 200-201 “…the Pilar landslide originated from a steep, fault-cut volcanic ridge 590m SSW of Barangay Pilar (Fig. 3A)”. The ridge is not identified on Fig 3A, and the resolution is too low to distinguish the features. Lines 202-203 “These features were more defined in the post-landslide DSM (Fig. 3B). Line 204 “Field mapping identified prominent landslide scarps and arcuate failure surfaces upslope of the deposit fan (Fig. 3B)”. I disagree; these features cannot be discerned on Fig 3B.
We can enlarge Fig.3A to maximize the width of the page (with margins) for the drone orthophoto, and provide an inset pre-landslide DEM that highlights the geomorphological features of the study area. We’ll label these features in the map. There are three debris flows, DF1a, DF1b and DF2, but there no figures showing these, so I found it hard to understand the distribution of the different debris flows – they need be shown on a map overview? DF1a and DF1b are described in the text but not distinguished on the figures – just DF1. The models in Figure 4 are much too small to distinguish between them. DF1a and DF1b were distinguished by some eyewitnesses; others described DF1 as one flow event that continued as the T1 was triggered as DF1 initially hit the water. In the field, we could not distinguish between DF1a and DF1b; these deposits are also mostly buried under DF2 (supported by Fig.7). We labelled both DF1 and DF2 in Fig.4, but we’ll add these extents in the Fig. 3 revision. There are no figures showing the results of the geological models from Leapfrog Geo (Lines 172-173); especially the changes in heights down the valley from erosion at the top to deposition at the bottom initiation, entrainment/transport and deposition areas need to be shown on a map and in figures. There is no validation presented for the modelling – Lines 177-178 – “Model outputs were cross-checked against field observations and mapped deposit extents to ensure consistency”.
Leapfrog Geo models are shown in Fig. 3B and Fig.5, we’ll add that reference to the figures in the text to better guide the readers. The elevation differences are shown in the Supplementary File 2 as mentioned in line 170-171. We’ll add the thickness measurements from the fieldwork that were used to create the 3D deposit model (currently Fig. 3B). We'll also add the photos of the outcrops from which these were measured as a Supplementary file. The different landslides are composed of different sediments DF1a and DF1b – grey and DF2 brown – why, how do these relate to their source volcanic sediments and the location of these? We clarified these in the description of the initiation zone (5.4.1, lines 269-271). Both the eastern and western side of the zone has the sequence of “light gray source rock that becomes reddish (oxidized) and weathered into brown soil towards higher elevation. Near the top of the soil mantle is a distinct boulder-rich layer upon which vegetation was anchored, as seen both on the slumping units and the scarp.”
We illustrated 3 different models that could fit the varying eyewitness descriptions and the composition of the deposits (lines 274-287). In all these models, DF1 could have more materials from the gray source rock unit (Fig.4A), while the DF2 deposit and the slumping unit could have more from the overlying sediments and the reddish soil mantle.
Regarding the three alternative models how do these relate to the fault cut volcanic ridge - - Line 201 – “…..the Pilar landslide originated from a steep, fault-cut volcanic ridge 590m SSW of Barangay Pilar (Fig. 3A)”. This is not identified on any figure (as I note above) so it is hard to understand the interpretation of the different three models, (which are difficult to discern on Figure 5). In the interpretation of the three models, A failed on the western crown and B and C on the eastern crown – how do these different models relate to the fault cut volcanic ridge? All these models have the source rock fractures near the western edge of the failed volume of DF1 or DF1a. Both DF1 and DF2 passed through the fault identified at S6 (shown in Supplemental File 3). We’ll improve the resolution of Fig. 3 to present this better. Different terms used – the landslides are complicated but the presentation in the paper is not helped by the different terms used. On Figure 3 there are DF1 and DF2, then on Figure 4 there is added the ‘entrained volume which looks like DF2 and a slumping unit? Entrained and slumping units were only indicated in Figure 5. Figures 3 and 4 show post-landslide imagery and interpretation of the deposits in the study area, while Figure 5 was based on the pre-landslide topography, delineating the possible source areas of the slope failures, according to the eyewitness accounts and field observations. There are sections on the rainfall, local geological and geomorphological context, then the landslide event chronology, then the landslide morphology – a bit back to front? Key to the interpretations is video evidence, but this is not shown. The sequence of results presentation was according to the sequence of each method applied:
• Documentary/reports review led to the rainfall data interpretation
• Remote Sensing led to the local geological and geomorphological context analysis
• Interviews and video review led to Landslide-Tsunami event chronology
• Fieldwork, laboratory analysis and 3D modelling led to Landslide and tsunami descriptionsA link to the publicly available video of the second debris flow event, as uploaded online by the owner (Bureau of Fire Protection, 2022) was provided in the reference (https://www.facebook.com/watch/?v=2119315464893710). The rest of the online copies were only posted by secondary sources (vloggers who got the videos from eyewitnesses). We opted to show the annotated frames because the videos were taken by witnesses moving away from the landslide tsunami.
Citation: https://doi.org/10.5194/egusphere-2026-2543-AC1
-
AC1: 'Reply on RC1', Likha G. Minimo, 18 Jul 2026
-
RC2: 'Comment on egusphere-2026-2543', Sebastian Watt, 15 Sep 2026
This paper presents an important and comprehensive description and analysis of a landslide-tsunami disaster in the Philippines, the first documented event of its type in the region, and one of very few well described such events globally. The dataset provides valuable insights into an important and little studied hazard type that presents particular preparedness and management challenges. The conclusions highlight the potential risk associated with cascading landslide-tsunami hazards, and provide lessons that can inform consideration of these risks in the Philippines and globally, representing an important case study given that there are very few well documented landslide-tsunami disasters. The paper draws on an interesting array of data, including qualitative and quantitative information, combining interviews, eyewitness accounts, online sources and field observations, and demonstrating the value of drawing on a diverse dataset to interpret such events.
My comments are all relatively minor, but highlight several points in the text, presentation and provision of data that could be improved or clarified, including a few areas where further information would enhance the interpretation and provide context useful for future studies in comparable environments. These are ordered by section below.
Abstract: It would be helpful to provide specific information in the abstract on distances and timescales (rather than just describing these as short pathways) to highlight the challenges associated with events of this type (e.g. potential for preparedness and warning).
Introductory to methods sections:
Line 70: The description of previous potential landslide-generated tsunamis is unclear in this sentence. It would be helpful to expand on this slightly to clearly explain the previous potential landslide generated events in the Philippines, their location and scale, and notes on the source (particularly if this is uncertain). Perhaps this would be best presented as a short table, to place the event described here into a national context.
Did the interviews follow a specific set of questions? If there was a formal set of questions, this would be useful to share as a supplementary document, as it could act as a useful guide for similar studies of this type and also help interpret the way in which information was obtained.
Line 128: “distinguishing” rather than “establishing”?
Line 130: It’s unclear how h was derived to establish run-up heights of the debris flow (the implication is that these are debris flow heights, not tsunami run-up, unless I’ve misunderstood). I understand that this marks the maximum inundation, but how was the baseline established, and what was the reference point? (from the channel, or from the deposit surface?).
L141: I can understand that total destruction could be a marker of tsunami damage, but such damage patterns could also be produced by debris. Presumably, though, the overall spatial pattern of damage also guided the interpretation (i.e. spatial relationship to the debris flow margins, or the landward damage pattern observed in the area impacted by the tsunami)?
Geological context of the landslides: At several places, the volcanic geology is alluded to as significant, but it is hard to find a clear, comprehensive description of the source area. The detail is a bit disparate, mentioned at several places, but not providing a clear picture of this important element of the event.
At Line 75, the reference to “volcanic terrain” implies that the volcanic nature of the setting is the important element, but this statement is not really explored. It would be informative to pick up this point in the discussion to consider if volcanic slopes, or more specifically, particular volcanic or volcaniclastic rock types, are particularly prone to mass wasting of the Pilar-event type, and what the exposure of such settings is in coastal environments of the Philippines.
The geology is described again around Line 90, but the structural elements are not properly explained (are the rocks in the landslide source area steeply dipping, faulted etc.?). Faults are referred to in the area, but is there any information on the specific nature of the source region in terms of fracture density or deformation?
L217: The statement of the source geology here notes that the area is “underlain” by volcaniclastics. Please clarify if this was the rock that actually failed, or if this is the unit underneath the rock that failed (the wording is ambiguous). Did the failure occur on a boundary between two lithologies, or within this volcaniclastic lithology? From the description, it sounds like the failed material was a very poorly sorted volcaniclastic unit (clasts up to boulder sized). Can the authors provide any further information on the nature of this unit and its extent, which may help understand why it was prone to failure? A few further details are provided at L270 (and at L260, where this is described as a weathered volcano). Further information is given again later on, where the coarse clastic unit appears more confined, so the overall description of the source lithology becomes quite confusing to the reader. Because the properties of DF1 and DF2 are notably different (Fig 8 and as described in the text), this seems particularly important to understand what failed initially, the nature of those materials, and how they may have been affected by the intense rainfall events. It would be better to consolidate all these descriptions of the source geology into one place, to provide a detailed description of the failure regions’s stratigraphy and properties.
Other potential triggering factors: Two intriguing elements of the failure emerge in the discussion (knowledge about tension cracks along the strike of the failure plane; and the occurrence of an earthquake the previous day), but are not developed at all. Furthermore, the timing of the event is surprising (perhaps?) in being around two days after the peak rainfall intensity. Alongside an evaluation of the properties of the initial failure, it would be informative if the authors could develop this slightly in the discussion.
The statement at L221 about previous tension cracks along the line of the failure is very interesting but only very briefly mentioned. How were these cracks noticed in 2021? Were these reported by local residents or linked to concerns around ground deformation or seismicity, or did they just happen to be observed during a routing survey? Did the survey cover the area that was ultimately the source of the landslide?
L457: The occurrence of a shallow earthquake is briefly mentioned, a day before. This would coincide with the peak of the rainfall and seems a potentially important additional triggering factor. Can details be provided of the magnitude and location of the earthquake, and likely shaking intensity in the source region?
Eyewitness observations: Table 1 and L232: The first line in the table and the statement described at L232 is surprising, in implying that small-scale feature (cracks in the slope and gradual movement) could be observed from offshore. Does this imply that the deformation was therefore large scale and visible from some distance (presumably from around 1 km away from the source, if I’m reading the scale in fig 3 correctly), or does it raise questions about the nature of what was observed, or is there perhaps some uncertainty about timing, and could this observation be coincident with the start of the landslide?
Field observations:
L261: This sentence is a bit unclear in terms of the geological relationships in the source area (“landslide units” – are these the fresh deposits?).
The distribution of tsunami run-up information in Fig. 9 is very interesting. How specifically were the heights in Malaguicay and Bahay obtained. Is this from eyewitness data, or field measurements? If it’s the latter, how was this evident so long after the event? Describing these details would be helpful for future post-event surveys. Additionally, were the eyewitness height observations in Pilar consistent with field measurements, or was one or other systematically higher (again, useful information to help interpret different sources of obtaining height information).
For the distal sites, was there any evidence of a deposit left by the tsunami? There are very few such deposits described in the literature, so understanding if any indicators were preserved, and what these were like, would be valuable to future studies and assessments.
Discussion: The discussion provides an insightful wider context and comments on many elements of the event that have wider implications for this hazard type and the risk from relatively small-volume coastal landslides. A few comments:
L460: Very interesting wider context, with 11 landslides reaching the coast as part of this event. Can any evaluation be made of the pathway distances of these other landslides and their total volume, to evaluate why the Pilar event provided an effective tsunami source whereas the other events did not? Is the source-rock type of these other events known? And was the Pilar event larger, or simply nearer the coastline? Is there any information about the timing of these other events, and did they also lag the peak rainfall intensity?
L534: The maximum run-up for the Anak Krakatau 2018 tsunami, evident from trim lines on both Rakata and Sertung, nearby islands, was 80-90 m (see, for example, Grilli et al., 2021, Fig. 11: Modeling of the Dec. 22nd 2018 Anak Krakatau volcano lateral collapse and tsunami based on recent field surveys: Comparison with observed tsunami impact - Marine Geology). The point is still reasonable, that relatively small landslides can still cause highly damaging tsunamis, but the Anak Krakatau failure was several orders of magnitude larger and, despite the open water setting and greater distances to the nearby impacted shores, the run-up was substantially higher. This should also be corrected in Table 3.
A further volcanic landslide tsunami in a lake context that could be included for comparison, also likely triggered by prolonged intense rainfall, is that at Yate volcano in 1965 (with an approximately 60 m run-up), see: Watt et al., 2009, Landslide and tsunami hazard at Yate volcano, Chile as an example of edifice destruction on strike-slip fault zones | Bulletin of Volcanology | Springer Nature Link
Figures and supplementary data:
Figure 1: The location map could be clearer. The left panel is very dark, and the detail in the other two panels is difficult to see, especially the text, labels and particularly the legends.
Figure 3: This would benefit from being larger, particularly key details around the source area and tsunami impact zone. It’s hard to see features at the current size.
A pathway profile would also be very helpful, because there isn’t a clear elevation map of the source area provided (the only mention in the text is of a source elevation of 200 m, but to understand the terrain, a profile along the channel is really needed).
The supplementary files provide a useful further set of figures, mainly interpreted images and photographs. These would likely be better compiled into a single document, as they all represent additional interpretation of field observations.
For long-term recording of observational data, the field sites and details of run-up measurements (for both the debris flow and tsunami) should be provided as a supplement to the paper (with coordinates, types of measurement, height and inundation etc.). There are very few datasets of this type, especially for landslide tsunami, and such data would enhance the value of this for future studies (these data should be associated with the paper to ensure accessibility, even if the broader datasets (drone footage etc.) are in a separate repository).
Sebastian Watt (Sept 2026)
Citation: https://doi.org/10.5194/egusphere-2026-2543-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 180 | 109 | 13 | 302 | 33 | 10 | 6 |
- HTML: 180
- PDF: 109
- XML: 13
- Total: 302
- Supplement: 33
- BibTeX: 10
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Overview
The paper addresses the cause of the April 12th, 2022, rainfall-triggered landslide and tsunami disaster in Pilar, Leyte, Philippines, addressing the event chronology, scientific insights, and recommendations for hazard assessment and disaster risk reduction. It is based on detailed fieldwork, eye-witness interviews and video footage, data analysis, and comparison with global events.
The landslide was generated by prolonged, intense rainfall (from a low-pressure area, Severe Tropical Storm Megi, and Typhoon Malakas) and there were 54 fatalities, 49 injuries, and 33 missing persons presumed dead, most from the tsunamis. The event was reconstructed from satellite imagery, UAV surveys, field mapping, sedimentological analyses, 3D terrain modelling, eyewitness accounts, social media videos, and incident reports. The fieldwork included interviews with survivors and responders.
There were two major landslides (total volume ~550,000 m³) composed of volcanic materials originating from a fault-controlled ridge and reaching the Leyte Gulf coast as debris flows. The first and larger debris flow generated tsunami waves with runup heights up to 12 meters above sea level, devastating coastal areas and overwhelming evacuees.
The tsunamis caused great destruction, with houses and infrastructure swept away, and the coastal landscape transformed with the shoreline advancing by up to 95 meters. The majority of Barangay Pilar was destroyed, and the community was permanently relocated. The loss of life was due to short warning times and high landslide flow velocities. The event is the first well-documented subaerial landslide-generated tsunami in the Philippines providing important information on the tsunami hazard in the country.
The disaster exposes the limitations of treating landslides and tsunamis as independent hazards; cascading and compound hazards can amplify disaster impacts. It demonstrates how rainfall-driven landslides in steep, volcanic coastal terrains can rapidly cascade into high-impact tsunamis, with devastating consequences for communities. The event provides lessons for improving hazard assessment, early warning, and disaster risk reduction in similar environments.
Comments
It’s an interesting paper with a great deal of important information, but in its present form it’s hard to see the wood for the trees.
Paper strengths are the mapping of the landslide area and the images presented and the eyewitness interviews. This is a great data set and the basis of the paper. However, the interpretations are poorly shown in the key figures (e.g. 3), which are too small and with too low resolution to see the features described in the text – e.g. Lines 200-201 “…the Pilar landslide originated from a steep, fault-cut volcanic ridge 590m SSW of Barangay Pilar (Fig. 3A)”. The ridge is not identified on Fig 3A, and the resolution is too low to distinguish the features. Lines 202-203 “These features were more defined in the post-landslide DSM (Fig. 3B). Line 204 “Field mapping identified prominent landslide scarps and arcuate failure surfaces upslope of the deposit fan (Fig. 3B)”. I disagree; these features cannot be discerned on Fig 3B.
There are three debris flows, DF1a, DF1b and DF2, but there no figures showing these, so I found it hard to understand the distribution of the different debris flows – they need be shown on a map overview? DF1a and DF1b are described in the text but not distinguished on the figures – just DF1. The models in Figure 4 are much too small to distinguish between them.
There are no figures showing the results of the geological models from Leapfrog Geo (Lines 172-173); especially the changes in heights down the valley from erosion at the top to deposition at the bottom initiation, entrainment/transport and deposition areas need to be shown on a map and in figures. There is no validation presented for the modelling – Lines 177-178 – “Model outputs were cross-checked against field observations and mapped deposit extents to ensure consistency”.
The different landslides are composed of different sediments DF1a and DF1b – grey and DF2 brown – why, how do these relate to their source volcanic sediments and the location of these?
Regarding the three alternative models how do these relate to the fault cut volcanic ridge - - Line 201 – “…..the Pilar landslide originated from a steep, fault-cut volcanic ridge 590m SSW of Barangay Pilar (Fig. 3A)”. This is not identified on any figure (as I note above) so it is hard to understand the interpretation of the different three models, (which are difficult to discern on Figure 5). In the interpretation of the three models, A failed on the western crown and B and C on the eastern crown – how do these different models relate to the fault cut volcanic ridge?
Different terms used – the landslides are complicated but the presentation in the paper is not helped by the different terms used. On Figure 3 there are DF1 and DF2, then on Figure 4 there is added the ‘entrained volume which looks like DF2 and a slumping unit?
There are sections on the rainfall, local geological and geomorphological context, then the landslide event chronology, then the landslide morphology – a bit back to front? Key to the interpretations is video evidence, but this is not shown.
Conclusions, Recommendations
Great data, great figures, but presentation is poor and results presented confusing. Three landslides identified but no figures showing their location and relationships. There seems to be a lot of video evidence, that would be good to see. Three alternative models are presented, but do not seem to relate to the location of their origin on the fault-cut volcanic ridge? No figures to support the relationship(s) between the modelling and the field observations and mapped deposit extents. The order of presentation - chronology before landslide description is back to front. Geographical relationships between the landslides not shown, so impossible to understand. Can the different sedimentary composition of the landslides, DF1a and DF1b – grey and DF2 brown – be related to different volcanic source sediments and the location of these?
Fundamental to the paper is a description and interpretation of the landslides, and this is missing. More figures on the landslides, their distribution and thickness, and their relationships are required and a fuller presentation of the results of geological models together with justification of these results from the landslide field mapping. The overall results from the research need improved explanation and justification. Figure 3 is critical in introducing the context of the interpretations but the images here are poorly annotated and too low resolution to identify their description in the text – so higher resolution images and descriptions required here. Is it possible to include some of the video evidence?
Overall, this promises to be a significant paper, but major revision is required to answer my comments and recommendations before publication.