Feed aggregator

Using a mix of airborne and satellite images as well as data from ground sensors, a UCLA-led research team has reconstructed how the shape and reach of the methane plumes from the 2015–16 Aliso Canyon gas blowout evolved during the 112-day disaster.

This image captured by the Copernicus Sentinel-3 mission on 26 October 2025 shows the "brightness temperature" at the top of Hurricane Melissa as it barreled through the Caribbean Sea toward Jamaica, where it is expected to make landfall.

Irrigation canal maintenance in western Nebraska is taking a giant step forward thanks to an innovative, non-invasive method by Husker geoscientist Mohamed Khalil to check canal integrity. His sophisticated time-lapse analysis pinpoints canal seepage and structural settlement far more accurately and efficiently than traditional approaches—using a technology that can have wide-ranging uses statewide for agriculture, industry and natural resources management.

This story was originally published by Knowable Magazine.

When the biggest earthquake in more than a decade rattled Russia’s remote Kamchatka Peninsula in July, seismologists around the world knew within moments. For earthquakes big or small, sensors around the globe detect the tremors and relay that information to researchers, who quickly analyze the observations and issue alerts.

Now artificial intelligence is poised to make almost everything about earthquake research much faster—and to rewrite researchers’ very understanding of how earthquakes happen.

“Machine learning opened a whole new window.”

By using a subfield of AI called machine learning, some scientists are identifying up to millions of tiny, previously unseen earthquakes in data gathered from seismically active places. These new and improved databases are helping researchers to better understand the geological faults along which quakes happen, and can help to illuminate the risks of future quakes. Some scientists are even using machine learning to improve their forecasts of how many aftershocks may rattle a location that has just experienced a large and damaging earthquake.

More broadly, researchers hope that machine learning, with its ability to crunch through huge amounts of information and learn from the patterns within, will reveal fresh insights into some of the biggest mysteries about earthquakes, including how a quake unfolds in its first devastating seconds.

“Machine learning opened a whole new window,” says Mostafa Mousavi, a seismologist at Harvard University.

Shaking Earth, Exploding Data

Earthquakes happen when geological stress builds up in the ground, such as when two plates of Earth’s crust grind alongside one another, as they do at California’s San Andreas Fault. At some point, the stress reaches a critical threshold and the fault ruptures, breaking the rock and causing seismic energy to ripple outward and shake the ground.

The San Andreas fault, seen here as a dramatic slash across the Carrizo Plain in Southern California, is an example of a geologically active area where seismologists are using AI to better understand earthquake patterns. Credit: John Wiley, Wikimedia Commons, CC BY 3.0

That energy is recorded by seismometers and other instruments around the world, which are positioned in great numbers in geologically active areas like California and Japan. The data feed into national and international systems for tracking earthquakes and alerting the world. The amount of data has exploded in recent years as seismologists find new ways to gather information on ground movements—like detecting seismic signals over fiber optic networks, or using the accelerometers built into smartphones to create a phone-based earthquake warning network.

Just a decade or two ago, much of the analysis of seismic signals was done by hand, with scientists working as quickly as possible to assess recordings coming in from their observing networks. But today, there are just too many data points. “Now the only—almost—way that you can deal with the seismic data is to go to automatic processing,” says Mousavi, who coauthored a 2023 article in the Annual Review of Earth and Planetary Sciences on machine learning in earthquake seismology.

One of the most common uses of machine learning in seismology is measuring the arrival time of seismic waves at a particular location, a process known as phase picking. Earthquakes generate two kinds of seismic waves, known as P and S waves, that affect the ground in different ways and show up as different types of squiggles on a seismogram. In the past, a seismologist would analyze data arriving from seismic sensors and hand-select what they gauged to be the start of P waves or S waves on those seismogram plots. Picking the starts of those waves accurately and precisely is important for understanding factors such as where exactly the earthquake hit. But phase picking is very time consuming.

An earthquake’s energy appears as a squiggly waveform on measurements made by seismometers. The first type of signal to arrive is a ground movement known as a P wave, followed by a type known as an S wave. Picking where the waves first arrive on a seismometer reading is an important part of understanding an earthquake’s impact—this has typically been done by human seismologists but in recent years the process has been made much quicker by incorporating machine-learning algorithms. Credit: Knowable Magazine, adapted from USGS.gov

In the past few years, seismologists have been using machine learning algorithms to pick seismic phases much faster than a human can. There are a number of automated methods that can do phase picking, but machine learning algorithms, which have been trained on huge volumes of data on past quakes, can identify a wide variety of signals from different types of tremors in a way that was not possible before. The practice is now so standard that the term “machine learning” is no longer stated in the titles of research papers, says Mousavi. “By default, everybody knows.”

AI-based phase picking is faster than phase picking by humans and at least as accurate, Mousavi says. Seismologists are now working to expand these tools to other types of seismic analysis.

Expanding Quake Catalogs

One area that has already seen big discoveries is the use of machine learning to expand earthquake catalogs—basically, lists of what earthquakes happened where in a particular region. Earthquake catalogs include all the quakes that seismologists can identify from recorded signals—but AI can find exponentially more tremors than human scientists can.

Essentially, machine learning can trawl through the data to identify small earthquakes that people don’t have the ability or time to flag. “Either you don’t see them by eye, or there’s no time to go and look at all those tiny events,” says Leila Mizrahi, a seismologist with the Swiss Seismological Service at ETH Zürich. Often, these tremors are obscured by background noise in the data.

Tiny earthquakes are important as a window into how larger earthquakes begin.

In a pioneering 2019 study in Science, researchers used an AI algorithm that matched patterns of seismic waves to identify more than 1.5 million tiny earthquakes that happened in Southern California between 2008 and 2017 but had not been spotted before. These are itty-bitty quakes that most people wouldn’t feel even if they were standing on top of them. But knowing they exist is important in helping seismologists understand patterns of behavior along a geological fault.

In particular, Mousavi says, tiny earthquakes are important as a window into how larger earthquakes begin. Large earthquakes may happen along a particular fault once every century or more—far too long a time period for scientists to observe in order to understand the rupture process. Tiny quakes behave much the same as big ones, but they happen much more frequently. So studying the pattern of tiny quakes in the newly expanded earthquake catalogs could help scientists better understand what gets everything going. In this way, the richer catalogs “have potential to help us to understand and to model better the seismic hazard,” Mousavi says.

Expanded earthquake catalogs can also illuminate the structure of geological faults below a region much better than before. It’s like going from a simplistic sketch of how the faults are arranged to a painting with more photorealistic details. In 2022, a team led by seismologist Yongsoo Park, then at Stanford University, used machine learning to build an expanded catalog of quakes in Oklahoma and Kansas between 2010 and 2019, many of them induced by oil and gas companies injecting wastewater into the ground. The work illuminated fault structures that weren’t visible before, allowing the scientists to map the faults more precisely and to better understand seismic risk.

This dramatic example shows the power of machine learning to enhance scientists’ knowledge of earthquakes. Top is a depiction of an earthquake swarm that occurred near Pawnee, Oklahoma, in September 2016. Each dot represents an earthquake measured by seismometers (with yellow representing quakes that occurred early in the swarm, and red representing quakes that occurred later). Bottom is the same earthquake swarm, but in this case when scientists used machine learning to pinpoint additional, smaller quakes in the observations. The enhanced earthquake catalog shows far more detail of where the quakes occurred, including illuminating the underlying geological faults. Credit: Courtesy of Yongsoo Park

Park and his colleagues showed that 80 percent of the larger earthquakes that happened could have been anticipated based on the smaller earthquakes that occurred before the big ones. “There is always a possibility that the next major earthquake can occur on a fault that is still not mapped,” says Park, who is now at Los Alamos National Laboratory in New Mexico. “Routinely capturing smaller earthquakes might be able to reveal such hidden faults before a major earthquake happens.”

Scientists are applying this approach around the globe. Researchers in Taiwan, for instance, recently used machine learning to produce a more detailed catalog of a magnitude 7.3 tremor in April 2024 that killed at least 18 people on the island and damaged hundreds of buildings. The study, reported at a seismology meeting in April 2025, found the AI-based catalog to be about five times more complete than the one produced by human analysts, and it was made within a day rather than taking months. It revealed new details on the location and orientation of geological faults—information that can help officials better prepare for how the ground might move in future quakes. Such catalogs “will become the standard in every earthquake-prone region in the future,” says team leader and seismologist Hsin-Hua Huang of Academia Sinica in Taiwan.

Forecasting is Still a Problem

So far, AI hasn’t been as successful in tackling another of seismology’s biggest challenges—forecasting the probability of future quakes.

The field of earthquake forecasting deals with general probabilities—such as the chances of a quake of magnitude X happening in region Y over time period Z. Currently, seismologists create quake forecasts using mathematical analyses of past earthquakes, such as a statistical method that relies on observations of how past earthquakes triggered subsequent quakes. This approach works well enough for specific tasks, like understanding how many aftershocks may rattle a region after a Big One. That sort of information can help people in a disaster zone know whether it’s safe to return to their houses or whether more aftershocks might be on the way, threatening to collapse more buildings.

But this kind of analysis can’t always accurately capture the real seismic risk, especially along faults that only rarely yield big quakes and thus aren’t well represented in the seismic record. Seismologists are testing AI-based algorithms for earthquake forecasting to see if they might do better, but so far, the news is tepid. In their best performances, the machine learning analyses are about as good as the standard methods of quake forecasting. “They are not outperforming the traditional ones yet,” says Mousavi, who summarized the state of the field in an August 2025 article in Physics Today.

Overall, though, seismologists see a bright future in using AI to understand earthquakes better.

In one of the more promising experiments, Mizrahi has been trying to use AI to speed up producing aftershock forecasts in the crucial minutes and hours after a large earthquake hits. She and a colleague trained a machine-learning algorithm on the older statistical method of quake forecasting, then unleashed it on its own to see how the AI would do. It did perform much faster than the older, non-AI approach, but there’s still more work to do. “We’re in the process of evaluating how happy we are with it,” says Mizrahi, who published the findings last year in Seismological Research Letters.

In the future, researchers hope to speed up these types of forecasting analyses. Other areas of seismology could eventually benefit, too. Some early research hints that machine learning could be used in earthquake early warning, for instance estimating exactly how much the ground will move in the seconds after an earthquake has started nearby. But the usefulness of this is limited to the few parts of the world that have early warning systems in place, like California and Japan.

Park also cautions about relying too much on machine learning tools. Scientists need to be careful about maintaining quality control so they can be sure they are interpreting the results of any AI analysis correctly, he says.

Overall, though, seismologists see a bright future in using AI to understand earthquakes better. “We’re on the way,” Mizrahi says.

—Alexandra Witze, Knowable Magazine

This article originally appeared in Knowable Magazine, a nonprofit publication dedicated to making scientific knowledge accessible to all. Sign up for Knowable Magazine’s newsletter. Read the original article here.

A research team from the Nanjing Institute of Geology and Paleontology of the Chinese Academy of Sciences (NIGPAS), along with collaborators, has found that long-term orbital variations occurring over million-year timescales may have served as the "pacemaker" for Earth's ancient oxygenation pulses. Their findings were recently published in Geophysical Research Letters.

In 2020, the scientists and engineers behind NASA’s InSight lander were optimistic. The mission was performing spectacularly, and it had no end in sight. Then, its power began to fade. Fine Martian dust was relentlessly piling on top of its solar panels, blocking sunlight. Mission operators had anticipated this but hoped that occasional wind gusts or passing dust devils would sweep the panels clean. Such fortuitous cleaning had prolonged the lives of earlier robotic explorers, such as the Spirit and Opportunity rovers. But for InSight, no such wind ever came, and its batteries slowly ran out of juice. InSight fell silent in December 2022.

InSight’s demise illustrates a long-standing gap in Martian science: Researchers still know little about how winds move across the planet’s surface and interact with dust. To help fill this gap, a group of researchers has now reviewed decades of orbital imagery from two European Space Agency (ESA) spacecraft—Mars Express and the ExoMars Trace Gas Orbiter, operational since 2004 and 2016, respectively—looking for dust devils and using them as a proxy for surface winds.

Over the years, these orbiters have captured thousands of high-resolution images of Mars’s surface. Hidden within this vast dataset are countless sightings of dust devils, which drift with the prevailing winds. Because surface winds are otherwise impossible to measure directly from Martian orbit with the available instruments, tracking the motion of these vortices provides a rare window into their direction and velocity.

To measure these parameters, the researchers exploited a technical quirk of the spacecraft cameras, namely, the slight temporal delay between capturing different color layers of an image or between the right and left images in stereoscopic views. By tracking the dust devils’ movement between exposures, the team could track the velocity and directions of the winds carrying them. Their observations revealed some of the fastest surface wind speeds ever detected on Mars, challenging existing atmospheric models.

The Colour and Stereo Surface Imaging System (CaSSIS) on board the European Space Agency’s ExoMars Trace Gas Orbiter captured this dust devil tracking across the Martian surface on 28 February 2019. Credit: ESA/TGO/CaSSIS, CC BY-SA 3.0 IGO

“With dust devils, we now have a tool to measure wind velocities across the planet, across space and time,” said Valentin Bickel, a planetary scientist at the University of Bern and lead author of the study. “We get a measurement of wind speeds in a distributed way around the planet, not just in specific lander locations.”

AI-Assisted Research

Detecting dust devils in orbital images, however, is not easy. For instance, the Colour and Stereo Surface Imaging System (CaSSIS) camera on board the ExoMars Trace Gas Orbiter resolves the surface at about 4 meters per pixel, meaning that dust devils dozens of meters wide appear as tiny smudges. Finding all these dust devils in the images is something that “an army of people could do in a few months or years, but nobody can pay for that,” Bickel said.

To automate the search, Bickel and colleagues trained a convolutional neural network—a type of artificial intelligence (AI) commonly used in image recognition—to identify the dust devils. After training the algorithm with about 50 examples labeled by experts, they let it loose on their full dataset of 50,000 orbital images. “Its only function is to identify dust levels in images; it can’t do anything else. It’s very stupid,” Bickel said. However, it needed only a few hours to scan the entire collection.

“The velocities we measured are totally surprising; I didn’t think we would see so many fast dust devils on Mars.”

The neural network detected more than a thousand dust devils across nearly all Martian latitudes. Each detection offered a new data point on local surface winds. The analysis revealed that Martian surface winds are generally faster than current atmospheric models suggest—and occasionally stronger than any speeds directly recorded by landers or rovers equipped with weather instruments. For instance, the researchers detected wind speeds of up to 44 meters per second, which is substantially faster than the previous 32 meters per second mark recorded by the Perseverance rover. Scientists previously assumed that dust devils might not even have been able to form at such wind speeds, as they could be destroyed by currents, Bickel said.

“The velocities we measured are totally surprising; I didn’t think we would see so many fast dust devils on Mars,” Bickel said. “You always picture them as these slowly moving clouds of dust, but it turns out they’re like superfast, highway speed level objects. I think it’s just crazy.”

The second key finding is that fast winds are more widespread across the planet than previously thought. To showcase this, the researchers produced a map with the locations of all 1,039 dust devils detected, including the direction of motion for 373 of them, confirming that dust devils are found all over Mars, even atop the tallest volcanoes. However, dust devils tend to cluster in specific regions, for instance, in Amazonis Planitia (visible at upper left in the map), a vast area known to be covered by an extensive, fine layer of dust and sand.

Researchers created a map showing 1,039 dust devils that occurred on the Martian surface, as seen in 20 years’ worth of images from European Mars orbiters. Credit: ExoMars TGO data: ESA/TGO/CaSSIS; Mars Express data: ESA/DLR/FU Berlin; Background: NASA Viking color mosaic, CC BY-SA 3.0 IGO

“Of course,” Bickel noted, “we have a bias because we need dust devils to see [the winds], so if there’s no dust at ground level, we don’t see the wind.”

The team also observed a clear seasonal pattern: Dust devils and strong winds appear more frequently during each hemisphere’s spring and summer, typically happening around midday, when surface heating is more intense. The researchers published their findings in Science Advances.

Blowing in the Dusty Wind

Deciphering how Martian winds work is key to understanding how dust shapes the planet’s weather and climate. Wind is the main force that lifts and transports the Red Planet’s abundant dust, which in turn regulates how the Martian atmosphere absorbs and radiates heat.

Understanding dust transport is thus critical for future exploration, both robotic and human. A global map of wind patterns might have helped InSight’s engineers choose a landing site less prone to rapid dust accumulation. On a larger scale, planet-encircling dust storms that erupt every decade or so—sometimes blocking sunlight for months—remain a serious hazard for exploration.

“Wind is one of the holy grails for human exploration and to understand the Martian climate.”

“Wind is one of the holy grails for human exploration and to understand the Martian climate,” said Germán Martínez, a researcher at the Center for Astrobiology in Madrid, Spain, who wasn’t involved with the new study. “Surface winds are very important, for Martian climatology, but especially at this time for human exploration safety, and we know very little.” In that sense, Martínez said, this research is important because it provides a map of surface wind speeds and directions that we didn’t have before, even if it’s a bit coarse.

Bickel agreed that more data, and more tools in orbit, will improve understanding of the Martian wind system. In the meantime, he hopes the new map will be used to validate and improve climate and wind models of Mars.

—Javier Barbuzano (@javibar.bsky.social), Science Writer

Citation: Barbuzano, J. (2025), Martian dust devils reveal dynamic surface winds, Eos, 106, https://doi.org/10.1029/2025EO250404. Published on 28 October 2025. 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.

A study led by researchers from the University of Sydney and the University of Adelaide has revealed how the breakup of an ancient supercontinent 1.5 billion years ago transformed Earth's surface environments, paving the way for the emergence of complex life.

Since the Industrial Revolution, scientists estimate that the ocean has become around 30% more acidic from the uptake of additional anthropogenic carbon dioxide. Ocean acidification has widespread effects, including loss of coral reefs and a decline in shellfish. Current methods for measuring acidification in the ocean are point-based and labor-intensive, making large-scale, long-term monitoring challenging.

In May 2023, Italy's Emilia-Romagna region experienced devastating, if not unprecedented, floods that caused widespread damage to infrastructure, homes, businesses, and farmland. Seventeen people lost their lives, and the disaster caused an estimated €8.5 billion in damages. The persistent rainfall and resulting landslides and flooding displaced tens of thousands of residents, leaving a deep mark on the region's economy and communities.

Sinking land is quietly destabilizing urban infrastructure in India's largest cities, putting thousands of buildings and millions of people at risk, according to Virginia Tech scientists.

Various online datasets will allow a detailed understanding of Hurricane Melissa as it impacts Jamaica and then Cuba

Hurricane Melissa is now making headlines around the world in anticipation of its landfall today. As always with tropical cyclones, the picture is evolving continuously as the storm evolves. Their behaviour is highly complex.

I thought I’d highlight some useful tools for monitoring the evolution and behaviour of Hurricane Melissa. First, of course, NOAA CPHC provides a range of graphics, some of which are adaptable. This includes the forecast track of the centre of the storm, the forecast earliest arrival time of the centre of the hurricane and (most usefully in the context of landslides), the rainfall potential:

Precipitation potential for Hurricane Melissa. Graphic from NOAA as at 07:18 UTC on 28 October 2025.

Note that this is three day potential rainfall (the graphic that I posted yesterday was for four days). Jamaica is going to start to feel the full brunt of the storm today (Tuesday 28 October), and it will then move on to eastern Cuba. The latest forecast suggests that the most serious rainfall will occur in the central part of Jamaica, but that there will also be very significant rainfall in the west of the island. The change appears to be the result of a slightly later than forecast turn to the north.

The NASA Global Precipitation Measurement site provides near real time data – the best tool available for understanding the rainfall that the storm is delivering. This is the latest image showing 24 hour precipitation totals:-

24 hour precipitation accumulation for Hurricane Melissa. Graphic from NASA GPM as at 07:34 UTC on 28 October 2025.

Note that this site also provides a global landslide nowcast, but sadly the site indicates that this is not functioning. I am unsure as to why – maybe this is the effect of the government shutdown.

In terms of the landslides themselves, this map of Jamaica and Cuba provides landslide susceptibility – yet again, this is work from NASA:-

Landslide susceptibility for Jamaica and Cuba. Data from NASA.

Overlaying this with the forecast precipitation is fascinating – the east of Jamaica has the highest landslide susceptibility, but is now forecast to receive less rainfall. Central Jamaica has lower average susceptibility, but may receive more rainfall. But also remember that landslides in storms like this are often driven mostly by rainfall intensity, which is hard to forecast and very variable. There’s also a nice BGS report on landslide hazard for a catchment in Central Jamaica, which gives an idea of the scale of the issues.

In terms of news within Jamaica itself, the Jamaica Observer and the Jamaica Star will be providing local coverage.

Finally, in such situations there is a tendency in the international media to adopt a slightly condescending tone to reporting of such events in countries with lower levels of per capita GDP. Both Jamaica and Cuba have advanced disaster management systems – they are far from helpless victims. Indeed, Cuba has a remarkably successful record of managing disasters and Jamaica fared remarkably well during Hurricane Beryl last year due to its preparedness. But tropical cyclones are complex, and the impact of a Category 5 event is very much greater than that of a Category 4 storm. Even the best prepared nation struggles to cope with such a storm.

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.

When a slab slides beneath an overriding plate in a subduction zone, the slab takes on a property called anisotropy, meaning its strength is not the same in all directions. Anisotropy is what causes a wooden board to break more easily along the grain than in other directions. In rock, the alignment of minerals such as clay, serpentine, and olivine can lead to anisotropy. Pockets of water in rock can also cause and enhance anisotropy, as repeated dehydration and rehydration commonly occur at depth in a subducting slab.

Two earthquakes were recorded within hours of each other near Marion, North Carolina, and witnesses report they felt shaking miles away, according to the U.S. Geological Survey.

A research team led by Academician Jin Zhijun from the Institute of Energy, Peking University, has revealed how interactions between Earth's tectonic activity and astronomical cycles jointly shaped the planet's climate and carbon cycle during the Late Paleozoic Era (360–250 million years ago, or 360–250 Ma). The findings are published in Nature Communications, titled "Tectonic-astronomical interactions in shaping Late Paleozoic climate and organic carbon burial," offering new insights into the deep-time climate system.

The towering peaks of the Southern Andes are not just shaping the skyline of South America—they are also quietly influencing Earth's atmosphere.

Earthquakes in the American state of Utah, the Soultz-sous-Forêts region of France or in the Dutch province of Groningen should not be able to occur even if the subsurface has been exploited for decades. This is because the shallow subsurface behaves in such a way that faults there become stronger as soon as they start moving. At least that is what geology textbooks teach us. And so, in theory, it should not be possible for earthquakes to occur. So why do they still occur in such nominally stable subsurfaces?

Lake Victoria, Africa’s largest lake, supports a quarter of East Africa’s population with fish, fresh water, and critical transportation routes. It’s also one of the deadliest bodies of water in the world.

Storms, high winds, and waves frequently capsize boats, causing thousands of deaths each year.

Despite the hazard, Lake Victoria has historically lacked enough weather observation stations to provide a clear picture of weather patterns in the region. Overly general and often inaccurate forecasts have meant that those heading out on the water had little idea what weather they’d face.

In 2017, the World Meteorological Organization (WMO), the weather agency of the United Nations, began a multiyear effort to improve weather information for the lake and establish early-warning systems for life-threatening storms. Now, much of the lakeside population uses the program’s tailored forecasts, leading to an estimated 30% reduction in weather-related deaths on the lake.

Still, a dearth of weather data persists across the continent. Because of ongoing economic depression, conflict, and disruptive weather patterns, Africa has gone decades without observational networks that meet international standards.

Today, the continent has the least dense weather observation network in the world. The average density of stations meeting WMO standards is 8 times lower than the WMO recommended level; more reporting surface stations exist in Germany than across all of Africa, according to the WMO and the World Bank.

The lack of observations often leaves communities without early warnings of natural disasters, cripples forecasts that farmers rely on, and leaves scientists who are studying global climate and weather with a major data gap.

In 2019, the need for improved weather networks around the world was recognized at the Eighteenth World Meteorological Congress in Geneva, Switzerland. There, the WMO’s 193 member states and territories established the Global Basic Observing Network (GBON), an international agreement that specifies requirements for weather data collection, including which parameters to measure, when and at what resolution to measure them, and how to share the data between countries.

With encouragement from the WMO and smaller organizations, national meteorological agencies in Africa are also recognizing the need for enhanced weather and climate services and planning for them, said Zablon Shilenje, the technical coordinator of services for the WMO’s Regional Office for Africa.

That recognition, combined with increased economic investment, has slowly led to weather stations being added to networks throughout Africa.

“The situation now has improved a lot,” said Frank Annor, a water resources expert at TU Delft and Kwame Nkrumah University of Science and Technology in Ghana. But the continent is still far from meeting GBON reporting standards. And in the face of an ever more variable climate, additional investments and improved ways of working together are needed.

“There is a huge gap, and we need to work on it,” Shilenje said.

Scarce Stations

Climate models used by scientists and forecasts created by meteorologists fundamentally rely on current and past weather data, particularly temperature and precipitation measurements, to extrapolate patterns into the future. “Any time the historical information isn’t perfect, that’s going to cause potential issues,” especially for estimating the impacts of climate change, said Pascal Polonik, a climate scientist at Stanford University.

Forecast accuracy declines as the number of observations drop. That’s particularly problematic when an entire region or large swaths of it have little to no observational data—as is the case in many parts of Africa.

“We lack the ground data. That data is not being ingested into models, so then when you do predictions, your predictions are less accurate.”

“We lack the ground data. That data is not being ingested into models, so then when you do predictions, your predictions are less accurate,” Annor said.

There wasn’t always such a lack of station density. In the first half of the 20th century, African countries’ networks were growing on par with those in other parts of the world, though they never reached the same densities as in places like North America. But now, Africa has less than one third of the weather stations that it once had.

Social and political conflict is one reason for the decline. One 2019 analysis of temperature records available in sub-Saharan Africa found that a country’s civil conflict risk was negatively correlated with both the number and density of weather stations contributing to its temperature record.

Some conflicts or social upheavals have had an outsized effect on monitoring networks. During and after the 1994 genocide in Rwanda, for instance, the average number of actively reporting weather stations in the country dropped from more than 50 to less than 10. Nearly 15 years passed before station coverage returned to preconflict levels. In another instance, station density in Uganda declined from a peak of about 500 stations following independence in the 1960s to less than 100 by 2001. A civil war in Ethiopia beginning in 2018 resulted in a sharp decline in reporting weather stations in the northwest part of the country, where much of the fighting took place.

“You can see from one year to another how unrest can affect station density,” said Tufa Dinku, a senior research scientist at the Columbia Climate School in New York.

The ongoing conflict in the northwestern part of Ethiopia may have led to a decrease in reported weather data from stations in the same area. Credit: Tola et al., 2025, https://doi.org/10.3389/fclim.2025.1551188, CC BY 4.0

Beyond conflict and the challenges of establishing a stable national government, a lack of economic resources has also contributed to the drop in weather station density. In the late 1980s and 1990s, Africa entered an economic depression that made it difficult for states to update their weather observational systems with technology on par with that used by countries in Europe and North America.

“African countries were not able to recover their meteorological [networks],” said David Mburu, who worked for more than 30 years as a meteorologist at the Kenya Meteorological Department.

Weather itself is partly to blame for slow economic development: Climate variability has caused frequent droughts, floods, heat waves, and land degradation, Dinku wrote in a 2006 article exploring the challenges of managing climate change on the continent.

The places in the world that are most affected by climate change tend to overlap with places in the world that are economically poor and, often, also data poor, Polonik said.

Shilenje said climate change only adds to the challenge of maintaining networks, affecting the durability of the instruments and equipment used to make observations. “There is a strong correlation between climate change and the ability to maintain a stable observing network on the continent,” he said.

A Dearth of Data

The reporting weather stations that do exist are often located along roads or concentrated in urban areas, meaning they’re not dispersed well enough to give an accurate reflection of weather across a whole country or region, Dinku said. Weather station coverage tends to be worse in rural areas, where better weather and forecasting information is most needed.

The dearth of observational stations has far-reaching consequences for Africans and those doing business there. Farming, for example, makes up about 15% of the continent’s economy. Without accurate forecasts, farmers are left without the information they need to make decisions about how to keep their livelihoods afloat.

“If you don’t know when it’s going to rain or how much to expect, then how do you go about your agriculture activities? That is the problem,” Annor said.

Inaccurate accounting of rainfall also means some farmers struggle to get their insurance claims paid, he said. “People then don’t want to invest in insurance again,” Annor said. “What that means is that people take calculated risk: They minimize the amount of food they can grow so they can minimize their risk.”

“The data from Ethiopia is not just for Ethiopia. The more observations you have in Africa, the better forecast we’ll have anywhere else in the world.”

The observations lost over the past few decades didn’t just limit forecasts then: The holes in the data will exist forever, always needing to be filled with reanalysis data—climate modeling that interpolates data on the basis of available observations—whenever a meteorological service or scientists want to analyze trends in a country’s weather and climate.

Lost observations also mean policymakers have no long-term data to use to plan adaptation strategies. “The resilience of people in the communities is reduced, and people become very vulnerable,” Annor said.

It’s not just local residents who suffer the consequences of a lack of data, either. Weather patterns in Africa play a role in the genesis of Atlantic hurricanes and spark Saharan dust storms, which can travel thousands of kilometers and affect global atmospheric processes.

“The data from Ethiopia is not just for Ethiopia,” Dinku said. “The more observations you have in Africa, the better forecast we’ll have anywhere else in the world.”

A lack of observational stations leaves scientists without sufficient data to answer research questions, too. For instance, sparse rainfall observations limited a full assessment of whether and how climate change influenced heavy rainfall and flooding around Lake Kivu that killed at least 595 people in Rwanda and the Democratic Republic of Congo in 2023.

“The scarcity and inaccessibility of meteorological data…meant we couldn’t confidently evaluate the role of climate change,” scientists from World Weather Attribution wrote in a summary of their attempt to analyze the event.

Low-Cost Stations as a Solution

In 2006, Oregon State University hydrologist John Selker ran into a similar data problem. He was working in Ghana, attempting to measure how much rainfall trees intercept. He and his collaborators found themselves stymied by a lack of rainfall measurements that kept them from completing the analysis they had planned.

“It was really shocking,” Selker said, adding that the only rainfall data they seemed to be able to find were sparse datasets that they had to apply for access to.

Selker and his colleague at the Delft University of Technology, Nick van de Giesen, brainstormed a solution: a low-cost weather station that could transmit meteorological and hydrological data over cell networks. They called their new project the Trans-African Hydro-Meteorological Observatory, or TAHMO.

With TAHMO, “the question was, What can we do now to improve on the density of stations to ensure that we can have reliable data from Africa that can both help feed the global models and [create] local models that are as accurate and useful as the ones that we have in the U.S. and EU countries?” said Annor, TAHMO’s CEO.

To date, TAHMO has worked with national liaison agencies (most frequently, national meteorological agencies) to install more than 750 stations in 23 countries and has collected more than 7 billion total observations. The stations, owned and installed by TAHMO, measure standard weather parameters such as precipitation, wind speed, wind direction, relative humidity, temperature, and solar radiation. Often, TAHMO approaches national meteorological agencies with a proposal to install stations, though some countries have asked TAHMO for assistance, too.

The data from TAHMO stations are shared directly with each country’s liaison agency. Each agreement between TAHMO and a country allows TAHMO to make these data available for any researchers interested in using them in peer-reviewed studies. It also gives a country the right to halt data collection, if it chooses.

“Policymakers are supposed to be guided by climate scientists, and the climate scientists can only authoritatively talk about that if they have quality data.”

Mburu, the longtime Kenyan meteorologist, became one of TAHMO’s main contacts in that country, helping to establish a relationship between the organization and the Kenya Meteorological Department. Now semiretired, he is a consultant for TAHMO at the organization’s headquarters, located in Nairobi. In Kenya, he said, TAHMO stations have been the most reliable forecasting system over the past decade.

Data from TAHMO stations have given Kenya’s Meteorological Department significant insight into what causes flooding, especially in Nairobi County, said Paul Murage, a climate scientist at the department who also trains other meteorologists at the WMO regional training center in Nairobi. Flash flooding has become a significant issue in the city; Murage recounted a day in March 2024 when the Nairobi Expressway, a major roadway, was impassable during heavy rains.

Murage said having rainfall data from TAHMO stations empowers his agency to persuade policymakers that better, climate-proofed infrastructure is needed. “Policymakers are supposed to be guided by climate scientists, and the climate scientists can only authoritatively talk about that if they have quality data,” he said.

TAHMO stations were included in the High Impact Weather Lake System Project (HIGHWAY), the WMO project to improve early-warning systems across Lake Victoria.

Another U.S.-based project, called 3D-PAWS (for 3D-Printed Automatic Weather Station), works with national meteorological agencies in developing countries to establish 3D printing fabrication facilities, install 3D printed, low-cost observational stations, and train local staff to maintain their own stations long term.

The group has worked with six African countries—Kenya, Malawi, Senegal, Uganda, Zambia, and Zimbabwe—and has deployed more than 250 stations. Prior to being dissolved in July 2025, the U.S. Agency for International Development (USAID) was a major 3D-PAWS funder, connecting the organization with countries via requests from national meteorological agencies in Africa.

Staff from the Zimbabwe Meteorological Services Department install a 3D-PAWS (Printed Automatic Weather Station) tipping bucket rain gauge at the Kutsaga Research Station in 2024. Credit: Paul Kucera

The goal is for each country to eventually run its network completely on its own, said Paul Kucera, an atmospheric scientist at the University Corporation for Atmospheric Research (UCAR) who codeveloped the 3D-PAWS program with his colleague Martin Steinson. They designed the original 3D printed stations themselves and incorporated feedback from international partners in newer iterations of the design.

Each partner country owns the stations once they’re installed. Though the initial training and installations are supported by grants to 3D-PAWS, the expectation is that each country’s own staff will incorporate the costs of operation and maintenance into its annual budgets after a few years.

Barbados, one of the countries outside Africa that 3D-PAWS works with, now has a self-sufficient team independent of the 3D-PAWS group that provides feedback to 3D-PAWS on how to improve their operations. Kucera hopes Kenya will be the first African country to achieve the same level of independence.

The 3D-PAWS data are typically open to the public via a free database system. Near-real-time data from 3D-PAWS stations in Kenya and Zimbabwe are also sent to the Famine Early Warning Systems Network (FEWS NET), a program established by USAID to provide early-warning systems for famine. Kucera’s own research group aims to use the data to develop other tools, such as automatic early alerts for weather events.

Many national governments in Africa are investing in climate services, too, Shilenje said. He’s seen an increase in the number of countries adopting national frameworks for climate services and putting plans in motion to improve weather networks.

As one example, the Tanzania Meteorological Authority installed five new weather radars in 2024, bringing the countrywide total to seven and giving Tanzania the greatest number of radars in East and Central Africa. It’s “quite a significant investment” for a single African country to make, Shilenje said.

Global Support

Larger, international efforts also provide assistance. In 2021, the WMO launched the Systematic Observations Financing Facility (SOFF), a partnership between the WMO, United Nations Environment Programme, United Nations Development Programme, and United Nations Multi-Partner Trust Fund meant to finance advancements in weather observational networks in developing countries through grants. SOFF is also part of the U.N. Early Warnings for All initiative, a project aiming to ensure that everyone on Earth is protected from natural disasters by early-warning systems by 2027.

SOFF, now 3 years old, has partnered with 24 African countries to support station building, radiosonde launches, and continued maintenance of these networks. SOFF, like TAHMO and 3D-PAWS, emphasizes continued support after the initial installation of stations. SOFF does this via a peer-to-peer network of national meteorological agencies. Twenty-eight agencies worldwide have expressed interest in acting as peer mentors, many to African agencies, said Mario Peiró Espí from the partnership office at SOFF.

The concept has seen successes. Peiró Espí recounted a recent interaction with a staff member at the Austrian meteorological agency: “He said the guys in South Sudan write to him all the time to check on questions that before, they didn’t have anyone to check in with. They didn’t have any number to call and say, ‘Hey, we are facing this challenge, we don’t know how to solve it, can you help us?’”

Nine of SOFF’s African partner countries have entered the organization’s investment phase, during which national meteorological agency staff install stations and launch radiosondes with SOFF support. Mozambique, a coastal nation that frequently faces destructive floods and tropical cyclones, is one of those countries.

During a flooding event in 2000, Mozambique lost the majority of its weather stations. A $7.8 million grant from SOFF is helping the country’s national meteorological agency to recover the network by establishing 6 new land-based weather stations, upgrading 15 existing stations, and launching 4 airborne stations.

Farther north, Chad faces a dire lack of weather data, too—as of October 2024, Chad was reporting just 3% of the surface weather observations required by the GBON agreement. SOFF is working with the country toward the goal of installing or upgrading at least 34 weather stations.

Markus Repnik, SOFF’s director, feels strongly that the world should think of improvements in Africa’s observational networks not just as assistance to Africa but as a global public good. The world is dependent on African meteorological agencies for accurate forecasts everywhere, he said. It’s as much as 20 times more valuable to install a single station in a data-poor area of Africa than to add one to a European network, he said.

While 3D-PAWS, TAHMO, and SOFF focus on station building and radiosonde launching, other groups lend additional support. Beyond investments in SOFF, the WMO is spending roughly $56 million in Africa on climate service projects such as those used to support improved food security and health outcomes.

Dinku’s research group has an additional solution: the so-called ENACTS (Enhancing National Climate Services) approach.

Via ENACTS, Dinku and his colleagues work with countries’ national meteorological and hydrological services to create more comprehensive and usable precipitation and temperature datasets. For rainfall, they combine a country’s meteorological station data with satellite rainfall estimates to improve the coverage of the dataset. For temperature, they use reanalysis to fill in missing past data points.

ENACTS prioritizes satellite and modeling work over installing new stations because new stations can’t provide the decades of past data needed to provide reliable forecasts or understand climate trends now. ENACTS has been implemented in 15 African countries at the national level.

An Upward, Uncertain Trend

Thanks to these efforts and others, the number and density of reporting weather stations in Africa continue to tick slowly upward. Philanthropic donors are beginning to understand the importance of a robust, global weather observation system, and the need for improvements has gotten recent exposure on the world stage. But there’s still a long way to go before observational networks on the continent reach the density of those in Europe or North America.

One ever-present barrier is funding. Many African countries still lack the financial resources to improve their meteorological services, according to a recent WMO report. Even SOFF, which has been able to mobilize more than $100 million in grants in the 4 years it’s existed, faces a “very challenging fundraising environment,” Peiró Espí said. SOFF needs an additional $200 million by the end of 2026 to meet the needs of the countries it’s working with.

Facing such fundraising challenges, SOFF plans to announce a new funding mechanism, the SOFF Impact Bond, at COP30 (the 30th Conference of the Parties) in Belém, Brazil. The bond will “make resources available upfront…while allowing donors to spread contributions over a longer period,” according to SOFF.

A changing political landscape in the United States could pose obstacles, too. This summer, the Trump administration officially shut down USAID and said many of its programs would be absorbed by the U.S. State Department. Kucera said 3D-PAWS is still waiting to hear whether the changes will affect its fiscal year 2026 funding, but the group is being “cautiously optimistic” and working to diversify its funding sources.

Fragmentation of efforts also slows progress. Africa is full of fragmented investments, Repnik said. “In each and every country, you have a hodgepodge of investments.”

“The future benefits [of more investment] will be immense.”

This hodgepodge leads to scenarios like the one that Dinku witnessed in Kenya, where the national meteorological agency receives data from a handful of different types of weather stations provided by various nongovernmental organizations and intergovernmental organizations, all with different data reporting systems. Shilenje’s seen this too: “You have different companies providing different sets of equipment,” he said. “It may be working very well, but compatibility is a challenge.”

To help with that issue, Dinku and a colleague created a data tool that allows users to access, process, perform quality control on, and visualize data from all their different weather station systems.

The WMO is working to solve the fragmentation issue as well, including efforts to improve national meteorological agencies’ digital capacity, a software tool to homogenize data, and the African Partner Coordination Mechanism, a platform by which nongovernmental organizations, intergovernmental organizations, and companies can exchange plans and objectives to ensure that everybody is working toward the same goal.

Still, collaboration and coordination are an uphill battle, Dinku said. “The last two or three decades, we have been talking about coordination. But everybody talks about coordination, and then they go about doing their own thing.”

Climate change only adds urgency to the efforts. As the climate warms, it will become even more variable. Risky decisions that farmers make about when, where, and what to plant will come with higher consequences than they once did. Mitigating the effects of climate change “will not happen without proper climate services,” Mburu said, adding that a robust observational network is critical to drastically reduce the impacts of climate change in Africa.

“The future benefits [of more investment] will be immense,” he said.

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

Citation: van Deelen, G. (2025), Building better weather networks, Eos, 106, https://doi.org/10.1029/2025EO250386. Published on 27 October 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.

Source: Geophysical Research Letters

When a slab slides beneath an overriding plate in a subduction zone, the slab takes on a property called anisotropy, meaning its strength is not the same in all directions. Anisotropy is what causes a wooden board to break more easily along the grain than in other directions. In rock, the alignment of minerals such as clay, serpentine, and olivine can lead to anisotropy. Pockets of water in rock can also cause and enhance anisotropy, as repeated dehydration and rehydration commonly occur at depth in a subducting slab.

It is well known that an earthquake generates both a compressional wave and a shear wave. If the shear wave passes through anisotropic rock, it can split into a faster shear wave and a slower one with different polarizations.

Although seismologists routinely measure the shear wave splitting in subduction zones by analyzing recorded seismic waveform data, it is challenging to pinpoint where splitting occurs along the wave propagation path.

In the past, researchers have investigated the circulation of Earth’s interior for answers, in particular in the mantle wedge region above and below the slab. However, Appini et al. suggest a different explanation: that, contrary to popular wisdom, it is the downgoing slab that causes most of the shear wave splitting.

The researchers tested their theory using recordings of 2,567 shear waves from the Alaska-Aleutian subduction zone. They found that the way the waves split as they propagate through the slab varied by earthquake location and that these variations were consistent with the anisotropy observed in the dipping slab. They also used a forward model to predict that the splitting pattern will differ depending on the direction the shear wave comes from, which was verified by data observation. Previously, scientists thought the variation in splitting patterns was due to complex mantle flows.

Furthermore, a dipping anisotropic slab also explains why deep earthquakes within a slab have unusual seismic wave radiation patterns. Other recent findings also hint that the composition of subducting plates causes anisotropy, the authors write.

If the slab holds most of the anisotropy, instead of the mantle wedge or subslab region, this finding has far-reaching consequences that could fundamentally change established ideas on how mantle dynamics work and how rock deforms, the authors suggest.

These results drive home the plausibility that slab anisotropy is an understudied component of seismology and geodynamics, the authors say. (Geophysical Research Letters, https://doi.org/10.1029/2025GL116411, 2025)

—Saima May Sidik (@saimamay.bsky.social), Science Writer

Citation: Sidik, S. M. (2025), New earthquake model goes against the grain, Eos, 106, https://doi.org/10.1029/2025EO250403. Published on 27 October 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.

Hurricane Melissa is bearing down on Jamaica, with many areas likely to see over 500 mm of rainfall. The impacts could be extremely significant.

Hurricane Melissa has strengthened substantially over the weekend, and is now on course to track across Jamaica in the next couple of days. Various media agencies have identified the threats that this storm poses to a country with high vulnerability. As always, NOAA has excellent tracking charts for this storm.

The current forecast track will take the storm directly across Jamaica:-

The forecast track of Hurricane Melissa. Graphic from NOAA as at 07:52 UTC on 27 October 2025.

NOAA also provides data on forecast precipitation (rainfall):-

Precipitation potential for Hurricane Melissa. Graphic from NOAA as at 07:52 UTC on 27 October 2025.

There is a great deal of uncertainty in this type of forecast – the final totals will depend upon the track, the rate at which the storm moves, the intensity of the storm (and how that changes as a result of the contact with the land mass) and orographic effects. But much of Jamaica is forecast to receive over 500 mm of rainfall, and some parts may receive more than 750 mm.

Now, the average annual rainfall in Jamaica is 2,100 mm for the island as a whole, and much more in some places, so this must be seen in context. However, as I have noted often before, in most cases the dominant medium through which tropical cyclone losses occur in water (even though windspeed often grabs the headlines). As the Google Earth image below shows, the island is characterised by steep slopes – this is a recipe for channelised debris flows:-

Google Earth image showing the landscape of eastern Jamaica.

There is active preparation underway in Jamaica, including evacuations, and in Hurricane Beryl last year this was a success. However, we know that many people choose not to move, and this storm is on a different scale.

In the immediate aftermath, the initial focus will inevitably be on the capital, Kingston, as this is where the reporters are likely to be located. Watch out for news from the east of the island though, especially on the coast and on the southern and eastern sides of the mountains. In severe storms, communications are often lost, so in this case no news may well be probably bad news.

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.

SummaryThe complex three-dimensional (3D) geometry of active faults plays a crucial role in controlling earthquake location, extent, and rupture behavior, making the accurate representation of fault models essential. Fault structures are typically interpreted manually from relocated hypocenters and interpolated to generate 3D fault surfaces. However, this process is often non-unique and uncertain due to the uneven spatial distribution of earthquake hypocenters, the subjectivity of manual interpretation, and the complexity of non-planar faults. To address these challenges, we developed a method that combines adaptive threshold hierarchical clustering with quantitative evaluation to automatically and effectively construct 3D models of seismogenic faults. This method utilizes the nearest neighbor index (NNI) to determine whether seismic activity exhibits clustering characteristics indicative of fault structures. Adaptive threshold hierarchical clustering is subsequently applied to identify small earthquake clusters associated with each fault. High-density 3D automatic slicing ensures robust fitting of fault lines, and in combination with surface rupture data, discrete smooth interpolation (DSI) is used to construct a 3D fault model. For each fault, we calculate distances from small earthquake clusters to the 3D fault structure and analyze their spatial distribution using kernel density estimation (KDE) to optimize the model for a near-symmetric distribution of small earthquake clusters on both sides of the fault. We applied this method to the 2013 Ms 7.0 and 2022 Ms 6.1 earthquakes in southern Longmenshan, Sichuan, China, refining the 3D seismogenic fault models for both events. Additionally, we constructed a 3D fault model for the 2019 Ridgecrest Mw 7.1 earthquake sequence using the same approach. The results indicate that this method is applicable to both individual faults and multiple intersecting fault systems. Compared to traditional manual modeling approaches, our method significantly enhances the identification of small earthquake clusters, reduces reliance on manual interpretation, increases modeling efficiency, and minimizes errors. This innovative modeling technique advances the 3D geometric construction of complex active faults and is adaptable to a wide range of seismic research applications.