Feed aggregator

The 21 July 2025 rock avalanche is generating a lake that could have a volume of 86 million cubic metres at the point of overtopping. This poses a threat to at least seven downstream communities in east Taiwan.

Yesterday, I posted about the enormous 21 July 2025 rock avalanche in the Matia’an valley, in Wanrong township in eastern Taiwan. Coincidentally, etaiwan.news has posted an article about the landslide that includes some images of it, and that highlights the growing concerns about the potential hazard from the landslide dammed lake.

So, let’s start with the images. This is the headscarp area:

The headscarp area of the 21 July 2025 landslide in the Matia’an valley in Taiwan. Image by etaiwan.news.

The large source area is clear in the upper left of the image. Note the dust cloud from continued rockfall activity. The initial track of the landslide has left a complex topography that includes bare rock and some landslide material (especially on the right side of the image).

This image captures the lake that is forming:-

The barrier lake of the 21 July 2025 landslide in the Matia’an valley in Taiwan. Image by etaiwan.news.

Note the very substantial height difference between the lake and the top of the landslide deposit. Given that this valley was free draining before the landslide occurred, this must all be landslide material. As such it is erodable in the event of overtopping.

Finally, this is a view of the whole length of the landslide:-

The entire track of the 21 July 2025 landslide in the Matia’an valley in Taiwan. Image by etaiwan.news.

Again, note the height of the saddle formed from landslide material. The deposit appears on first inspection to be steep, which suggests it might be quite erodible.

The etaiwan.news article highlights work being undertaken by the Hualien Branch of the Forestry Conservation Department to understand the hazard. The statistics of the dam are concerning:

As of 7 August 2025At overtoppingLake volume23 million m386 million m3Lake length1,770 m2,900 m

The current freeboard is 79 m. Current inflow into the lake is 920,000 m3 per day, giving an overtopping date of mid-October at current rates (but see below).

A key issue is then the assets at risk downstream. This is a Google Earth image of the channel entering the Longitudinal Valley:-

Google Earth image of downstream assets from the landslide dam in the Matia’an Valley in Taiwan.

The article mentions that the risk will extend to:-

“will include the Matai’an Creek Bridge on the downstream Taiwan Line 9 [this is main highway on the eastern side of Taiwan], public and private river defence facilities and settlements on both sides of the river, and the administrative area covers Mingli Village, Dama Village, Daping Village, Dongfu Village in Wanrong Township, and Changqiaoli, Darongli and Shanxingli in Fenglin Township.”

Taiwan is well-placed to manage this hazard, but it is going to be a major issue in the coming weeks. Finally, as noted above, the overtopping date is estimated from current inflow rates. But, the next few weeks are the peak of the typhoon season, which can bring exceptional rainfall.

And, right on cue, Tropical Storm Podul has formed to the east of Taiwan, and is now moving westward. It is too early to tell whether this will bring heavy rainfall to the Matia’an valley (it is likely to pass by Taiwan on about 13 – 14 August), but if it does then this will accelerate the filling of the lake. Even if it does not bring heavy rainfall, the development of another typhoon that affects this area in the next two months would not be a surprise.

Return to The Landslide Blog homepage Text © 2023. 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.

Subduction, a crucial geological process on Earth, may have begun hundreds of millions of years earlier than traditionally believed.

Glaciers carved the deep valleys of Banff, eroded Ontario to deposit the fertile soils of the Prairies, and continue to change Earth's surface. But how fast do glaciers sculpt the landscape?

The SWOT (Surface Water and Ocean Topography) satellite captured the tsunami spawned by an 8.8 magnitude earthquake off the coast of Russia's Kamchatka Peninsula on July 30, 11:25 a.m. local time. The satellite, a joint effort between NASA and the French space agency CNES (Centre National d'Études Spatiales), recorded the tsunami about 70 minutes after the earthquake struck.

On New Year's Day 1995, a monstrous 80-foot wave in the North Sea slammed into the Draupner oil platform. The wall of water crumpled steel railings and flung heavy equipment across the deck—but its biggest impact was what it left behind: hard data. It was the first time a rogue wave had ever been measured in the open ocean.

The Colorado River is in trouble: Not as much water flows into the river as people are entitled to take out of it. A new idea might change that, but complicated political and practical negotiations stand in the way.

Natural weathering processes are removing CO2 from the air in a wide range of environments across continents and oceans. Until recently, these "CO2 vacuum cleaners" were often studied separately, without properly examining their complex interactions.

The Perito Moreno Glacier in Argentina—often described as one of the most stable glaciers in Patagonia—is retreating far more rapidly than previously thought, according to a paper in Communications Earth & Environment. The results show that over the last few years, the glacier has retreated by as much as 800 meters in some areas, and that it may collapse and retreat by several kilometers in the near future.

This summer, much of North China has endured widespread temperatures above 35°C. Even typically cooler, high-latitude summer retreats like Harbin in Northeast China—usually a refuge from the heat—saw temperatures soar past 35°C in late June and July. As climate change accelerates, extreme heat events will become increasingly frequent.

Source: Journal of Geophysical Research: Biogeosciences

Earth’s atmosphere transports tiny forms of cellular life, such as fungal spores, pollen, bacteria, and viruses. On their journeys, these microorganisms encounter challenging conditions such as cold temperatures, UV radiation, and a lack of nutrient availability. Previous research showed that certain microorganisms can withstand these harsh conditions and potentially reside in dormancy until being deposited in a more favorable environment. But could the atmosphere itself also be the site of an active microbial system, harboring growing, adapted, and resident microorganisms?

The study of these floating life-forms is called aerobiology, but progress in the field is difficult to make: No standardized method exists for sampling the aeromicrobiome, it’s common for microbe samples to become contaminated, and it’s challenging to replicate atmospheric conditions in a laboratory setting.

Martinez-Rabert et al. suggest that computer modeling and theoretical approaches could help to improve understanding of the aeromicrobiome. Using known information about the metabolism and bioenergetics of microbial life—especially in harsh environments—as well as the chemistry and physics of the atmosphere, specialized modeling frameworks may be able to provide insight into the aeromicrobiome.

That bottom-up modeling approach, the researchers propose, could allow them to test how changing individual elements of Earth’s atmosphere would affect the proliferation of the microbial life it contains. For instance, are microbes better suited to a “free-living” lifestyle in atmospheric gases, inside droplets, or attached to solid particles? What energy sources are available to these microorganisms? How does the acidity of atmospheric aerosols affect the ability of atmospheric microorganisms to thrive?

The group suggests that combined with data generated through sampling measurements, experiments, and observations, theoretical modeling could help researchers to assess our atmosphere’s capacity to sustain a microbial biosphere and even to learn more about how microorganisms influence the atmosphere’s chemical makeup. This work could also someday be useful for modeling how life may exist in other planetary atmospheres, the researchers say. (Journal of Geophysical Research: Biogeosciences, https://doi.org/10.1029/2025JG009071, 2025)

—Rebecca Owen (@beccapox.bsky.social), Science Writer

Citation: Owen, R. (2025), Can microorganisms thrive in Earth’s atmosphere, or do they simply survive there?, Eos, 106, https://doi.org/10.1029/2025EO250293. Published on 7 August 2025. Text © 2025. 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.

In Malaysian Borneo, demand for timber, land, and palm oil has caused profound levels of deforestation. But the lack of historic data on the region’s rainforests has left scientists without much understanding of these forests’ baseline—what did they look like before industrial logging began?

Researchers have now found records hidden in an unexpected place. A new study, published in Scientific Reports, used corals from three reefs off the coast of Malaysian Borneo to build a timeline of deforestation and erosion on the island. The timeline provides valuable information for researchers and policymakers about how coral reefs are affected by deforestation as well as what Malaysian forests looked like in Earth’s past.

The study “allows us to have an idea of what the system looked like when it was relatively undisturbed,” said Walid Naciri, a geologist and lead author of the new study; he completed the research during his doctoral studies at the University of Leicester in the United Kingdom.

Coral Clues

Scientists can track deforestation trends with satellite imagery, but no such records exist before the onset of the satellite era around 1973.

Corals can be used as a proxy: They build themselves in alternating dark and light bands that, like trees, correspond to seasonal changes. As they grow, they absorb elements from seawater.

Corals’ ratio of barium, an element mostly found on land, to calcium, a common element in the ocean, can indicate the sediment content in river discharge that has reached a reef. Increased river discharge is an indicator of excess erosion, a consequence of deforestation. By analyzing the ratio of barium to calcium in a given band, scientists can determine the seawater composition in a given year.

“It’s a very well known impact of deforestation that you have less soil stability and more soil erosion ultimately arriving into the coastal ecosystem.”

Previous work by Naciri indicated that the barium-calcium ratio in corals matched river discharge data from the Baram River in Malaysian Borneo from 1985 to 2015. “We thought, okay, this is a pretty good analogue…so let’s have a look at the entirety of the record,” Naciri said.

The research team wanted to see whether they could use that ratio to construct a record of deforestation before 1985. They selected three separate reefs in the Miri-Sibuti Coral Reef National Park off the coast of Borneo. The three reefs, named Eve’s Garden, Anemone Garden, and Siwa, were located at different distances from the main sources of sediment: the Baram and Miri Rivers.

The team took cores of each reef and analyzed the ratios of barium to calcium. The resulting records were almost exactly what Naciri expected—cores from each of the three reefs showed a relatively flat trend until barium spiked beginning in the mid-20th century. The spike showed up later in the reefs farther from the island.

“We were pretty sure this was due to deforestation, because it’s a very well known impact of deforestation that you have less soil stability and more soil erosion ultimately arriving into the coastal ecosystem,” Naciri said. “But we needed a bit more proof.”

The team found archival forestry records that matched the mid-20th-century spike, indicating that the onset of industrial deforestation occurred around 1955. A previous analysis of land use across all of Southeast Asia from 1700 to 1990 showed a similar trend.

Naciri and his colleagues also ruled out other potential causes of the barium spike. They analyzed seawater at various distances from the Baram River to show that the barium was coming from the river rather than from leaching groundwater, for example.

The study’s authors did careful analytical work to rule out other interpretations of the data, said Dominik Fleitmann, a paleoclimatologist at the University of Basel in Switzerland who was not involved in the new study. Having three sites at varying distances from the island “adds confidence to the general reconstruction,” he said. 

The Far-Reaching Effects of Deforestation

The findings highlight how the impacts of on-land ecosystem degradation trickle down to coral ecosystems. Coral reefs rely on photosynthesis. Any influx of sediment into a reef clouds the water and harms coral’s ability to grow and even affects its ability to fight off diseases. 

“This is one more impact of deforestation that is really not talked about.”

“This is one more impact of deforestation that is really not talked about,” Naciri said. 

He urged other scientists studying deforestation to think outside of their usual study spaces and consider how marine ecosystems might be affected. “It’s an even worse problem than we think it is,” he said. 

Fleitmann said the results clearly show that “we are well above the natural variability in terms of soil erosion increases” and that the data could be useful to show policymakers the impacts of deforestation. 

The study could also be the beginning of a larger network of coral cores that could be used to build a more systematic reconstruction of deforestation records across larger areas of interest, such as the Australian or East African coast, he said. “What is the natural baseline, and how far away are we?”

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

Citation: van Deelen, G. (2025), Coral cores pinpoint onset of industrial deforestation, Eos, 106, https://doi.org/10.1029/2025EO250289. Published on 7 August 2025. Text © 2025. 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.

Critical mineral lithium—the lightest of all metals—had long eluded geologists by slipping through the cracks of traditional analysis.

Editors’ Highlights are summaries of recent papers by AGU’s journal editors. Source: Journal of Geophysical Research: Solid Earth

Previous studies of the microearthquakes (MEQs) produced from Enhanced Geothermal Systems (EGSs) have focused on the initial phase of high‑pressure “stimulation.” Chamarczuk et al. [2025] track what happens during normal operation, the phase in which plants will spend most of their lives.

Using a distributed acoustic sensing (DAS) cable in a monitoring well and on‑site processing, the authors built a two‑month MEQ catalog through stimulation, crossflow testing, and five load‑following cycles. During those cycles, seismicity rose and fell with subsurface fluid pressure, then settled toward an equilibrium between injections. Event locations formed a cloud whose growth matched a simple diffusion model, which points to pressure migration as the main earthquake triggering mechanism.

These observations suggest that operators have the ability to control seismicity through careful management of injection rates and fluid pressure. They also demonstrate that affordable, real‑time monitoring is feasible for future commercial projects.

Citation: Chamarczuk, M., Ajo-Franklin, J., Nayak, A., Norbeck, J., Latimer, T., Titov, A., & Dadi, S. (2025). Insights into seismicity associated with flexibly operating enhanced geothermal system from real-time distributed acoustic sensing. Journal of Geophysical Research: Solid Earth, 130, e2025JB031634. https://doi.org/10.1029/2025JB031634

—David Dempsey, Associate Editor, JGR: Solid Earth

Text © 2025. 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.

Author(s): Shabbir A. Khan and Atsushi Fukuyama

Linear absorption of electromagnetic wave injected in a hot plasma is usually associated with non-normal incidence; here, it is shown that absorption can take place at normal incidence as well. By developing a kinetic model based on integral form of dielectric tensor in the presence of static electr…

[Phys. Rev. E 112, L023202] Published Thu Aug 07, 2025

Changes in water levels beneath Earth's surface caused by glacier snowmelt and rainfall could be responsible for triggering small but frequent earthquakes in New Zealand's central Southern Alps, according to new research led by The Australian National University (ANU), published in the journal Geochemistry, Geophysics, Geosystems.

On 21 July 2025, a very large rock avalanche occurred in the mountains of Hualien County, Taiwan. Initial measurements suggest that this ran out over about 6 km.

On 21 July 2025, an extremely large rock avalanche occurred in the administrative area of Wanrong Township in Hualien Count in Taiwan. This event was detected on seismic data and it has been described on Facebook by Chen-Yu Chen. In the days before the landslide, southern Taiwan had been affected by heavy rainfall associated with the passage of Tropical Storm Wipha.

The crown of the landslide is at [23.72645, 121.29021]. A rough measurement suggests that it is in the order of 6 km long and 2 km wide. The location is steep and rugged – this is a Google Earth image of the site of the landslide:-

Google Earth image of the site of the 21 July 2025 landslide in Wanrong township, Taiwan.

As the image above shows, the area affected by the rock avalanche is exceptionally steep (even by Taiwan standards) and deeply dissected, suggesting regular landslide activity. I will return to this theme in a future post.

This is a Planet Labs image of the site, draped onto the Google Earth DEM, captured on 25 July 2025. So far, this is the only image of the site that I have been able to access – this part of Taiwan is exceptionally cloudy at this time of the year. Whilst some of the landslide is covered in cloud, most is visible.

Planet Labs image, draped onto the Google Earth DEM, showing the site of the 21 July 2025 landslide in Wanrong township, Taiwan. Satellite image copyright Planet Labs, used with permission. Image dated 25 July 2025.

Of particular note is the large-scale of the event, the long runout and the large amount of dust on the adjacent slopes. Note also the lake that has started to develop – it is reported on Facebook that the hazard associated with this is being managed.

The crown of the landslide is at about 2,450 metres and the toe is at roughly 700 metres, so this has a vertical extent of about 1,750 metres.

Here is an initial slider of the before and after images of the landslide:-

This is probably the largest landslide in Taiwan by volume since the Tsaoling rock avalanche and the Chiufengershan rock avalanche, both triggered by the Ch-Chi earthquake in 1999. However, the runout of the Wanrong landslide is, I think, larger than both of these landslides. I do not have a volume estimate at this point.

In the autumn, it is likely that clear imagery will become available of this exceptional landslide. However, Taiwan is likely to be affected by further heavy rainfall in the coming weeks, so the landslide might evolve further.

Reference and acknowledgement

Many thanks to Brian Yanites of Indiana University Bloomington for highlighting this event, and for his work on the landslide.

Planet Team 2025. Planet Application Program Interface: In Space for Life on Earth. San Francisco, CA. https://www.planet.com/

Return to The Landslide Blog homepage Text © 2023. 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.

SummaryWe used the multiple-frequency seismic tomography method to image the upper mantle beneath Northeast Brazil by processing P-wave broadband seismograms in six frequency passbands. The data comprised 87 896 relative traveltime residuals for P and PKIKP phases, simultaneously inverted to obtain 3D models of P-velocity anomalies. We conducted resolution tests using checkerboard patterns with horizontal dimensions of 312×312 km and 390×390 km. For the 312×312 km structures, we observed good horizontal recovery beneath the areas of the Borborema province and northern São Francisco craton at the depths of 136 and 226 km, with the areas of the São Luís craton and Parnaíba basin also presenting recovery, albeit with reduced amplitudes. For the 390×390 km structures we observed good horizontal and amplitude recovery throughout the entire study area. Our model was unable to recover the sharp vertical transitions between the anomalies. We present the preferred model for Parnaíba basin and Borborema province. The model shows a fragmented basement for the Parnaíba basin, with two strong high-velocity anomalies consistent with the Parnaíba and Granja blocks, and another slight high-velocity anomaly north of the basin, consistent with the São Luís craton. To the west of the basin, the Parnaíba block appears separate from another high-velocity anomaly associated with the Amazonian craton. A strong high-velocity anomaly south of the Borborema province is interpreted as the northern portion of the São Francisco craton. The São Francisco craton anomaly presents strong high-velocity anomalies, interpreted as a thickening of certain portions of the craton, separated by a weaker positive anomaly, interpreted as the Paramirim Aulacogen. The Borborema province is characterized by a low-velocity anomaly. The central and northeastern portions of this anomaly presented even lower velocities, which was interpreted as lateral flow in the asthenosphere, originating from the passage of a plume to the north of the province. A low-velocity anomaly located west of the Borborema province strikes roughly NE-SW and separates the São Francisco craton from the Parnaíba block and the Amazonian craton. It is interpreted as the Transbrasiliano Lineament. To test the capability of our data to resolve the limits of large-scale structures, we created four synthetic models simulating the presence of different cratonic nuclei. The models show good horizontal recovery, with the fourth model, based on our findings, presenting the best correlation between the real and recovered models. Seismicity in the study region is mainly correlated to low-velocity anomalies.

SummaryThe reliable classification of seismic events is crucial to precise seismic cataloging and robust hazard evaluation. Recent advances in deep learning have achieved great success in seismic event identification, leveraging their exceptional ability to automatically extract and recognize features. However, existing deep learning approaches to seismic classification rely exclusively on deterministic models, which cannot quantify epistemic uncertainty, preventing the estimation of prediction confidence that is critical for reliability evaluation. To address this issue, in this paper, we develop two uncertainty-aware deep learning models for earthquake (EQ) vs. explosion (EP) classification using the DiTing 2.0 artificial intelligence training dataset: a Bayesian convolutional neural network (BCNN) and a dropout-based CNN (DropCNN). We also implement a conventional deterministic CNN as a baseline model for comparative analysis. The experimental results demonstrate that both the BCNN and DropCNN can achieve a classification accuracy comparable to the conventional CNN, while providing additional uncertainty metrics for estimation confidence of prediction. Crucially, their uncertainty scores increase markedly in terms of encountering misclassifications or out-of-distribution samples compared to correct classifications, enabling automatic rejection of unreliable predictions based on the uncertainty threshold setting, triggering human verification or alternative discrimination methods. We then apply the trained models to analyze suspicious explosion events in the DiTing 2.0 dataset. The BCNN and DropCNN results exhibits strong agreement, consistently identifying 79 EP and two EQ events and flagging the remaining samples as uncertain classifications needing further verification. Our findings demonstrate that deep learning methods incorporating uncertainty estimation not only maintain a high accuracy in seismic event discrimination but also provide uncertainty estimation. This capability significantly enhances the model’s reliability and decision-making value in practical applications.

SummaryThe central portion of the 2019 ± 2 Ma Vredefort (South Africa) impact structure comprises a 40-50 km diameter central uplift of Archean basement rocks surrounded by a 15-20 km wide collar of late Archaean to early Proterozoic Witwatersrand Supergroup sedimentary and volcanic rocks. The collar is characterized by a ring of strongly negative (←5 500 nT) aeromagnetic anomalies surrounding much of the structure where the strata dip steeply to overturned. To better understand the origin of this magnetic feature, we undertook a ground survey along 20 transects (340 km) in the Vredefort structure using a three-axis fluxgate magnetometer mounted on a mountain bicycle. Upward continuation of our profiles to 150 m matches the aeromagnetic data in shape and amplitude. From the bicycle measurements, we pinpointed the rocks responsible for the extremely negative anomalies. Field observations and microfabric analyses of the rocks from six outcrops substantiated that the magnetic signal correlates with 10-100 m thick metamorphosed banded iron formations (BIFs) at the base of the supergroup as the main producer of the anomalies. Paleomagnetic samples collected from the rocks at the surface that produce the most intense anomalies (up to -22 000 nT) have extremely high natural remanent magnetization intensities (up to > 1000 A·m−1) likely arising from lightning strikes. Stepwise demagnetization and rock magnetic experiments establish a new protocol to distinguish samples that escaped remagnetization from lightning and possess the established 2.02 Ga paleodirection at Vredefort. From a suite of thermoremanent magnetization (TRM) experiments, the best estimate for the paleofield intensity at the time of impact was 52 μT, corresponding to an average remanence of 32.5 A·m−1. The results of the TRM experiments together with the paleodirection enabled us to successfully model the prominent negative anomalies in the metasediments only when accounting for the post-impact orientation of the BIFs. We interpret the strongly negative magnetic anomalies in the collar region as being formed directly after crater exhumation and uplift of the rocks. This interpretation implies that Bushveld-related metamorphism at 2.06 Ga created the up to mm-sized magnetite and garnet crystals in the BIFs, which resided at temperatures higher than the Curie temperature of magnetite (580°C) until the impact rapidly brought the BIFs close to the surface, where magnetite cooled to acquire a thermal remanence in the 2.02 Ga field.

SummaryMagnetic fabric analysis of dikes is a powerful technique when assessing magma transfer processes. This study presents an integrated analysis combining magnetic susceptibility and anisotropy of magnetic susceptibility (AMS), magnetic mineralogy, geochemistry, and new ⁴⁰Ar/³⁹Ar dating of dikes intruding formations ranging from the Lower Cretaceous to the Miocene on the island of Maio, in the Cabo Verde archipelago. We show that the dikes, dated at ≈ 9.2 Ma, intruding the younger Miocene Casas Velhas formation, display a Ti-rich titanomagnetite composition, higher whole-rock TiO2 content and very high magnetic anisotropy. They are clearly distinguished from the dikes, ranging in age from ≈ 9.3 to 11.3 Ma, intruding older formations, which show a predominantly Ti-poor titanomagnetite composition with multiple magnetic phases, lower whole-rock TiO2 concentration, higher range of magnetic susceptibilities and very low anisotropy. Magnetic fabric is predominantly normal with no significant imbrication relative to the dike margins. Numerical analysis of fabric shows a dominant coaxiality between the magnetic lineation and the preferred orientation of opaques and phenocrystals suggesting that magnetic lineation is, therefore, the proxy of the magmatic flow axis orientation. Based on the orientation of the magnetic fabric, we infer that magmatic flow within the studied dikes is predominantly vertical. The differences observed between the younger dikes and all other dikes may be related to magma sourced from distinct magma chambers. One, probably shallow, underneath the Casas Velhas fm in the southwest of the island, which would explain the very high values of magnetic anisotropy and the inferred vertical flow, and another located in a central position in the island, responsible for the dikes intruding the older formations. The location of such magma reservoirs and the dikes ages suggest a hypothetical migration with age of the magmatic sources that fed the dikes from the central part of the island to the southwest region. The magnetic and mineralogical heterogeneities of the dikes intruding older Lower Cretaceous formations may also be a result of a wider age range of the intrusions.