A new paper Fidan et al. (2026) demonstrates that wealth and the rate of land-cover change play a key role in determining the occurrence of fatal landslides in mountain areas. These factors are statistically more significant that precipitation and topography.

A fascinating new paper (Fidan et al. 2026 – this paper is both open access and published under a Creative Commons licence – hurrah!) has just been published in the journal Science Advances that explores rates of land-cover (in the paper, the authors use the term land-use – land-cover) change as a factor in determining fatal landslides in mountains globally. I must admit to some degree of personal interest in this paper, although I am neither an autor nor a reviewer, as it brilliantly uses the dataset that Melanie Froude and I collated on global landslide fatalities (see Froude and Petley 2018). I’m delighted to see our data being used in this way (and please do contact me if you want a copy of the spreadsheet).

Fidan et al. (2026) explores a range of factors that might influence the occurrence of fatal landslides from the perspective of either increased vulnerability (poorer people may live in more vulnerable locations for example) or increased landslide likelihood (land-cover change might increase the likelihood of a landslide being triggered, for example).

The fascinating result lies in land-cover change. The authors have looked at  approximately 60 years of land-cover changes in mountainous areas across 46 countries. Unsurprisingly, there is substantial change, especially in low- and lower-middle–income countries, often involving the loss of forest (which as a first order estimation, may buffer against slope failures), although the pattern is far more complex of course. Fidan et al. (2026) find that a key metric is the rate of change of land-cover, and that this is linked to the rate of population growth (perhaps unsurprisingly). Countries with high rates of population growth also show high rates of change of land-cover.

In many ways, the most interesting figure in this study is in the Supplementary Information. It is a complex diagram, but it’s worth more detailed analysis:-

The main map (A) shows mountain areas with high rates of land-cover change (orange), high density of fatal landslides (blue) or both (black). The left hand graph (B) shows the relationship between the landslide density and the rate of change of land-cover – here, higher rates of land-cover change are associated with a higher density of fatal landslides. The right hand graph is the same data as in (B), but with each point coloured according to the income level of the country. High income countries have a lower fatal landslide density. Thus, as the authors conclude, wealth and land-cover change appear to control fatal landslide density.

There is a really surprising element to this study, which I think requires more consideration. I think I should allow the authors themselves to express this finding, from the abstract:-

“Our statistical analyses show that land-use – land-cover changes have a substantially greater influence on the density of fatal landslides and landslide fatalities than physical factors such as topography and precipitation, especially in lower-income countries.”

As landslide researchers, we almost always default to topography and precipitation as being key in landslide occurrence. There are sound reasons for doing so. But statistically, the rate of land-cover change plays a more important role in mountain areas, especially in poorer countries.

This has (or should have) major implications for the way that we consider and manage landslide risk in such areas.

References

Fidan, S. et al. 2026. Wealth and land-cover change govern landslide fatalities on world’s mountains. Science Advances 12, eaec2739. DOI: 10.1126/sciadv.aec2739.

Froude M.J. and Petley D.N. 2018. Global fatal landslide occurrence from 2004 to 2016. Natural Hazards and Earth System Science 18, 2161-2181. DOI: 10.5194/nhess-18-2161-2018.

Text © 2026. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

Months after wildfires eliminate vegetation that holds hillside sediment together, debris flows—destructive landslides that carry bulky material down once-stable slopes—can devastate infrastructure, taking out roads and buildings in their wake.

Though the U.S. Geological Survey (USGS) creates hazard predictions used to warn communities of the risk of these postfire debris flows, those predictions haven’t fully considered how recovering vegetation reduces risk over time—until now.

A new study published in Geosphere presents a new way to calculate postfire debris flow risk that takes vegetation recovery into account. The USGS will begin using the new method this wildfire season to create more accurate maps of debris flow hazard in the years after a fire.

“I’m so appreciative that the focus on how the debris flow hazard changes over time after fire is being addressed,” said Nancy Calhoun, a geologist and postwildfire debris flow program manager at the Washington Geological Survey who was not involved in the new study. Calhoun said she relies on the USGS hazard assessments for virtually everything her job requires.

“We’re glad to have a way that we can help our partners moderate those situations where the hazard has decreased,” said Andrew Graber, a geologist at the USGS Landslide Hazards Program and lead author of the new study.

Assessing Hazard, Again

After a wildfire, the USGS creates hazard maps that incorporate information about soil type, steepness, and burn severity (how much vegetation has been lost) to show where the risk of a debris flow may be elevated.

Then, the agency distributes this guidance to the National Weather Service, which uses it to set rainfall thresholds: levels of rainfall at which a debris flow becomes likely. State, county, and city agencies use those rainfall thresholds to issue warnings or take action when rainfall is imminent, for example, by closing highways or triggering evacuations.

The methods used to create the USGS maps, however, historically relied on a snapshot of the burned area taken just after the fire, and the maps weren’t updated to reflect conditions as vegetation grew back and began holding soil in place again.

That led to situations where public safety decisions were made on the basis of outdated maps and rainfall thresholds. For example, concern over debris flows after the 2020 Grizzly Creek Fire in Colorado led to several closures of Interstate 70 in 2022, but the debris flows never happened.

“What [the original assessments] didn’t capture is how the vegetation came back,” Graber said. “That left us with some uncertainty when we started to get further away in time from the fire.”

To test an improved method for these hazard assessments, Graber and the research team incorporated satellite imagery of 12 burned areas that showed the degree of vegetation recovery right after the fire, 1 year after the fire, and 2 years after the fire. Then, they tested their new method by comparing its predictions to rainfall and debris flow data from the 12 burned areas.

The updated method better reflected what had actually happened after the fires, reducing the number of unnecessary warnings without missing real-world debris flows.

Risk Recalibration

The USGS plans to begin using their new workflow to create hazard maps for some higher-profile fires during the coming wildfire season.

That’s exciting for Calhoun. As part of her job, she’s in constant contact with emergency managers who periodically ask how worried they should be about debris flows in areas that burned years ago. “It’s a really important question: Are we still worried about this burn scar?” she said.

Right now, Calhoun has no data to point to in the years after a fire to give an updated answer to that question. Using the new method from Graber and the research team, she will.

“Because they’re using satellite [imagery] and repeatable quantitative methods to look at these burn scars over time, we’ll actually be able to say something useful and informed about vegetation recovery,” she said.

Having a deeper understanding of how debris flow risk evolves over time is especially important because debris flows themselves are becoming a greater risk to the public as a result of increasingly intense wildfires and rainstorms. In addition, more accurate assessments can reduce warning fatigue, which occurs when too many false alarms lead to people ignoring or opting out of alerts.

Graber hopes he and the USGS will continue to improve their methods for assessing debris flow hazards by collecting more debris flow data across the country and improving the underlying equation for hazard assessments so that it better reflects the unique conditions of different ecosystems in the United States. USGS researchers also published a new study in March presenting a method to generate maps of where debris flows might travel if they do occur.

“It’s a big year for USGS’s useful postfire products,” Calhoun said.

—Grace van Deelen (@gvd.bsky.social), Staff Writer

This news article is included in our ENGAGE resource for educators seeking science news for their classroom lessons. Browse all ENGAGE articles, and share with your fellow educators how you integrated the article into an activity in the comments section below.

Citation: van Deelen, G. (2026), A new approach can better predict debris flow hazards years after fires, Eos, 107, https://doi.org/10.1029/2026EO260160. Published on 19 May 2026.

Text © 2026. AGU. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

A new study (Sun et al. 2026) shows that in six earthquakes in China between 2010 and 2022, landslides and rockfalls were responsible for at least half of the total fatalities.

It is well-established that landslides are a major cause of loss of life in earthquakes in mountainous areas. The seismology maxim that “it is not earthquakes that kill people, it’s collapsing buildings” does not apply in its pure form in mountains – landslides also kill large numbers of people.

However, the actual number of people killed by landslides in earthquakes is poorly understood. This is largely due to the challenges of collecting reliable information in the aftermath of a major earthquake, when the focus is on rescue and recovery rather than data collection. For this reason, many studies of landslide fatalities do not include seismically-triggered events. This is true of my own work.

However, a study has just been published (Sun et al. 2026) in the journal Natural Hazards Review that starts to address this issue. The paper nominally examines fatalities from all causes from earthquakes in China from 2001 to 2022. However, the authors note that the data has low reliability until 2010, so I’ll focus on the period from 2010 to 2022. I also note that the authors use the term “geological hazards“, which is a little broader than landslides. I should note that the paper isa broad look at fatalities from earthquakes – there is a much richer range of analyses than I will cover here.

In the period from 2010 to 2022, Sun et al. (2026) identified 14 earthquakes in which geological hazards caused loss of life. In some cases, the impacts were substantial. Thus, the M=6.5 3 August 2014 earthquake at Ludian in Yunnan led to 134 fatalities and 40 people missing from geological hazards from a total of 728 fatalities (c.24 % of the total), whilst the 5 September 2022 M=6.8 earthquake at Luding in Sichuan led to 76 geological hazard fatalities and 25 missing from a total of 118 fatalities (c.86% of the total). In six of the 14 examples, geological hazards caused at least 50% of the fatalities.

Sun et al. (2026) highlight that “fatalities from geological hazards concentrate in geologically complex, mountainous provinces, i.e., Sichuan, Yunnan, Gansu, Guangxi, and Guizhou”. They note that even small events can trigger fatal landslides – for example, six people were killed in a rockfall triggered by a M=4.3 earthquake in Guizhou in 2010, whilst a M=2.8 aftershock from the Yanjin earthquake in 2006 triggered a rockfall that killed a person.

This is an incredibly useful study. It starts to shed light on the impact of landslides in large earthquakes. It is not the definitive study, and questions remain – not least, the pattern of landslide losses in very large earthquakes, like the 2010 Wenchuan event, in which landslides were ferocious. But it forms the basis for such investigations, starting to fill a major gaps in our understanding.

Reference

Sun, B. et al. 2026. Causes Analysis of Earthquake-Related Deaths in Mainland China 2001–2022. Natural Hazards Review, 27 [2]. https://doi-org.ntu.idm.oclc.org/10.1061/NHREFO.NHENG-2458

Text © 2026. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

Critically endangered Tapanuli orangutan population falls after heavy rain and landslides, fuelled by climate crisis, in North Sumatra

Extreme rainfall and landslides fuelled by the climate crisis killed 7% of the remaining population of the world’s rarest great ape, a study has found, prompting fears for the species’ survival.

The research suggests 58 out of the remaining 800 critically endangered Tapanuli orangutans (Pongo tapanuliensis) were killed after more than 1,000mm (39in) of rain fell over four days in Indonesia’s North Sumatra province in November 2025. This equates to 11% of the local population and 7% of the entire species.

In April 2026 I recorded 36 fatal landslides causing 90 fatalities, the lowest monthly total for 2026 to date.

This is my regular update for the number of fatal global landslides, focusing on March 2026. As usual, this data has been collected in line with the methodology described in Froude and Petley (2018) and in Petley (2012). References are listed below – please cite these articles if you use this analysis. Data presented in these updates should be treated as being provisional at this stage as I will reanalyse them prior to formal publication, and other events will emerge.

The headline figures are as follows:

March 2026: 36 fatal landslides causing 90 fatalities;

This is an interesting result, unusually showing that fatal landslides in April were substantially lower than for any of the preceding months in 2026. This is the updated annual chart by month:-

Loyal readers will know that I like to present the running total using pentads (five day blocks). This is the cumulative total pentad graph to the end of Pentad 24 (which captures all of the events to the end of April):-

Thus, whilst April 2026 was unexceptional compared with the previous months of this year, the number of fatal landslides was still above the long term mean. Overall, 2026 continues to run extremely hot, exceeding even the record-breaking year of 2024.

We now start to enter the crucial period of much higher global fatal landslide occurrence. Whilst in the long term dataset this acceleration typically occurs in June (or even July), in recent years it has happened in May, as the 2024 line shows. I will watch with great interest to see what happens this month.

As I always stress, the occurrence of fatal landslides prior to the South and East Asia rainy seasons is not a predictor of what will happen in that period. Interestingly, the WMO is forecasting a below average summer monsoon in South Asia.

References

Froude, M. and Petley, D.N. 2018.  Global fatal landslide occurrence from 2004 to 2016.  Natural Hazards and Earth System Sciences 18, 2161-2181.

Petley, D.N. 2012. Global patterns of loss of life from landslides. Geology 40 (10), 927-930.

Text © 2026. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

New evidence from the Natural Hazards Commission – Toka Tū Ake (NHC) shows that landslides are now New Zealand’s most costly natural hazard.

New Zealand is a country that is prone to a range of natural hazards. Located on a series of major fault systems, earthquakes cause high levels of loss. The country is also volcanically active, with occasional tragedies. Heavy rainfall brings floods.

To share the cost of these perils, following the 1942 Wairarapa earthquakes, the New Zealand government established the Earthquake Commission (EQC) in 1945, initially focusing on earthquakes and war damage, but subsquently expanded to cover other natural hazards.

In the subsequent years, the EQC has evolved into the Natural Hazards Commission – Toka Tū Ake (NHC), with a purpose “to reduce the impact of natural hazards on people, property, and the community”. Essentially it operates as a financial pool, with home owners paying a levy on top of their insurance to generate the fund. In the event of a loss, the fund pays for the rebuild costs up to a cap (currently NZ$300,000); the remainder is then covered by the property’s insurance. Claims are funded directly from the pool, with reinsurance cover and ultimately a government guarantee in place to ensure that there are sufficient funds.

In reality, NHC does much more than this, acting to manage and settle claims, and to understand the range of hazards to which New Zealand is prone.

In the last few days, a range of media outlets in New Zealand have been reporting new data from NHC about losses from natural hazards in New Zealand. This is the headline from 1News:

“Landslides are New Zealand’s most expensive natural hazard – and new data reveals a sharp rise in damage claims and growing risks to homes, infrastructure and communities.”

In total, since 2021 NHC has received 13,000 landslide claims and has paid out NZ$322 million (US$191 million). New Zealand is seeing an abrupt increase in landslide losses, driven primarily by increasingly frequent high magnitude rainfall events. NHC is urging property owners to undertake preventative maintenance and to be aware of the limitations of EQC cover.

In common with many other places, these landslide hazards represent a major challenge to New Zealand. The landscape has many dormant landslides that are being reactivated by these increased rainfall events, and many new failures are also occurring. But, generating reliable risk maps for landslides remains a major challenge. This needs to be a major research focus in the coming years. It will require better understanding of triggering events (rainfall and earthquakes primarily); of the initiation processes within the slope; of runout / debris mobility; and of vulnerability and consequent losses. It is probably true to say that in all of these areas, landslide research lags behind that of earthquakes and floods, primarily because of a lack of long term investment.

In many countries, landslides are not an insured risk for this reason. On its own, this will be a major challenge that must be addressed. For those countries in which landslides are insured, we need quickly to get up to speed.

Text © 2026. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

To date news reports suggest two fatal landslides with a combined toll of 17 people.

There are various news reports trickling in about the landslides triggered by the 8 June 2026 M=7.8 earthquake offshore Mindanao in the Philippines. As usual, the remote locations of many of the landslides means that the information is a bit hit and miss at this point.

To date, the most serious event appears to have occurred at a community called New Aklan, located in Glan, Sarangani. It appears that New Aklan is at: [5.7705 N, 125.3356]. News reports indicate that 13 people were killed, although there are also indications of additional fatalities in this area.

A further four people are missing under a landslide at Sitio Buhangin, Barangay Patuco, Sarangani. Patuco is in the area of [5.4770, 125.4859]. This appears to have been a failure on a coastal cliff:-

Over the next few days, satellite imagery should become available that will help identify the landslide impacts, but in the meantime Dan Shugar has identified some (using Planet imagery, I’d imagine):-

Text © 2026. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

In the early morning of 10 August 2025, a mountainside collapsed into the waters of Tracy Arm Fjord in southeastern Alaska.

Nobody was harmed by the rockslide or tsunami, but cruise ships were scheduled to visit the fjord later that morning. If the collapse had happened just a few hours later, it could have been disastrous.

“While the [South Sawyer] Glacier is in the fjord, it’s supporting those valley walls, like the buttresses on a cathedral,” said Daniel Shugar, a geomorphologist at the University of Calgary who led the study. “As that glacier retreated over the last few decades, it retreated just past the spot that did fail. It’s like if you have a kid and they said they cleaned their room but really all they did was throw everything in the closet. As soon as you open that door, everything falls out.”

In other words, the glacier that carved the fjord in the first place was also holding its slopes in place, and the ice’s retreat under warming temperatures exposed rock that became vulnerable to crumbling. The proximate cause of the landslide might have been something else—as Shugar noted, rainfall is plentiful in that part of Alaska, which could have weakened the fjord’s walls further—but it might also have been a combination of small, individually insignificant factors. In any case, the removal of that glacial “closet door” was what made the collapse and tsunami possible.

The researchers shared their findings at a press briefing on Wednesday at the European Geosciences Union 2026 General Assembly.

Debuttressing and Slope Instability

Unlike Lituya Bay, which resulted from an earthquake, Tracy Arm provided very little seismic warning before the slope collapsed, requiring forensic work to determine what caused it.

Shugar noted that South Sawyer Glacier had retreated by roughly 500 meters in the spring of 2025 alone, on top of the general trend of shrinking and thinning over the decades. And it’s not alone: Interferometric synthetic aperture radar (InSAR) images taken by satellites indicate that many slopes in Alaska and beyond are in motion, pointing to potential future danger.

“Not every single one, but it seems like a huge majority of [shifting slopes] are above the lower parts of thinning glaciers,” Shugar said. He described this phenomenon as “debuttressing,” as in losing the glacial buttress holding a slope up. He added, “I think in the next 5 years or so, we’ll probably have a much better understanding of just how and how quickly slopes respond to that debuttressing.”

Threats, Hazards, and Climate Change

Most tsunamis are set in motion by earthquakes and travel across the open ocean, wreaking their destruction when they reach shallower water near coasts; the word “tsunami” means “harbor wave” in Japanese. The Tracy Arm tsunami joined the ranks of other landslide-driven tsunamis, like the ones in Taan Fiord (Alaska) and Dixon Fjord (Greenland), in being linked to human-driven climate change. Beyond the immediate impact of the waves, this category of hazard requires rethinking potential risks from abrupt catastrophes like debuttressing as well as slower effects such as sea level rise.

“The risk to any particular cruise ship [from a tsunami] on any particular day is very low,” Shugar said. “We were unbelievably lucky that the [tsunami] occurred with the timing that it did, and not 5 hours later. The risk certainly still could be increasing as we build new settlements, new mining camps, or new oil and gas infrastructure.”

Both Shugar and Stearns highlighted the importance of learning lessons from Tracy Arm and related events.

—Matthew R. Francis (@BowlerHatScience.org), Science Writer

Citation: Francis, M. R. (2026), The forensics of a skyscraper-sized tsunami, Eos, 107, https://doi.org/10.1029/2026EO260140. Published on 6 May 2026.

Text © 2026. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

Source: Water Resources Research

Wildfires can increase flooding risks in and downstream of burned areas by removing vegetation and disturbing hydrologic processes. As the climate changes, the severity of both wildfires and heavy rainfall events is increasing, meaning flooding is likely to become more severe in the near future. Better understanding how, and by how much, wildfires change flood risk is important for disaster and infrastructure planning for communities around the country.

Canham and Lane used streamflow data from the U.S. Geological Survey’s National Water Information System and precipitation data from the NOAA Analysis of Record for Calibration product to identify storms and quantify their effects across seven burned watersheds in the western United States.

To make the most of the limited data on flooding in the years following wildfires, the researchers created a paired-storms framework: They identified postfire peak flows (PFPFs), defined as the five highest peak flows within 3 years of a wildfire across seven watersheds. Then, for each precipitation event causing a PFPF, they looked for storms with similar characteristics (or paired storms) that occurred before the wildfire. Storm characteristics used for pairing included the season in which the storm occurred, recent precipitation, and precipitation depth, duration, and peak intensity.

The researchers found significantly elevated peak flows after wildfires in many cases, underlining the risks to communities following wildfires and validating their approach for use elsewhere.

Altogether, the authors found 26 PFPF events, including 20 with paired storms occurring before wildfires. For 75% of the postfire storms, their peak flows were 2 or more times greater than prefire peak flows. PFPFs were most likely to happen in the first year after a wildfire and typically occurred following storms that were centered upstream of the watershed centroid, were uniform in shape, and fully covered the watershed and burned area, the authors reported. They also found some evidence that the first storm in the year immediately following a fire has a higher-than-expected chance of producing a PFPF.

Future work could look more deeply at the characteristics of storms occurring over burned areas, such as storm direction and watershed recovery, and could apply the automated methods to more burned watersheds and storm events to enhance the robustness of the work, the authors say. (Water Resources Research, https://doi.org/10.1029/2025WR040693, 2026)

—Nathaniel Scharping (@nathanielscharp), Science Writer

Citation: Scharping, N. (2026), How wildfires worsen flood risk, Eos, 107, https://doi.org/10.1029/2026EO260133. Published on 30 April 2026.

Text © 2026. AGU. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.

On 19 July 2025, intense, long duration rainfall triggered over 550 landslides in Sancheong, South Korea, killing at least 10 people.

On 19 July 2025, extremely heavy rainfall triggered multiple landslides in Sancheong, South Korea. This event has been described by a new paper (Nguyen et al. 2026) just published in the journal Landslides. The paper is behind a paywall, but this link should give you access at the time of writing.

The core of the affected area is at [35.4333, 127.9111] (as usual, Landslides provides the location in degrees minutes and seconds when digital degrees is so much more useful – a pet frustration of mine!). This is a Planet Labs image of a part of the area, captured before the event. The marker is at the coordinate noted above:-

And this is the same area after 19 July 2025:-

And here is a slider to allow a comparison:-

Nguyen et al. (2026) have mapped 568 individual landslides triggered by this rainfall event, triggered by rainfall in the range of 498 – 619 mm over a c. 55 hour period. These landslides killed at least 10 people and caused damage to homes and infrastructure. It is estimated that the restoration costs are in the order of US$800 million.

In common with many other events of this type, the landslides are mainly shallow, translational failures in soil or regolith on steeper slopes. As I have frequently noted, such terrain is very susceptible to unusually intense rainfall events, which often trigger a cluster of landslides in close proximity. These often merge to form channelised debris flows. Nguyen et al. (2026) note however that their modelling indicates that it was a combination of the intensity of the rainfall and its duration that led to these failures.

As rainfall intensities increase due to climate change, we are seeing increasing numbers of these landslide clusters. I greatly welcome studies such as Nguyen et al. (2026) , which allow us to build understanding in each case.

Reference and acknowledgement

Nguyen, H.H.D., Song, C.H. & Kim, Y.T. 2026. Physically based data-driven analysis for large-scale investigation of the July 2025 rainfall-induced landslide in Sancheong, South Korea. Landslides. https://doi.org/10.1007/s10346-026-02778-x

Planet Team 20246. Planet Application Program Interface: In Space for Life on Earth. San Francisco, CA. https://www.planet.com/

Text © 2026. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.