Dr. Denys Dutykh

In 1117 an earthquake shook northern Italy, and a tsunami up to about 50 m high swept Lake Iseo. With Muhammad Naveed Zafar, @psabatier.bsky.social and colleagues, we show in @natcomms.nature.com that human land use had overloaded the delta that failed 🧵 #Tsunami #NaturalHazards

phys.orgHow farming helped cause two medieval tsunamis in an Italian lake—and why the danger has returnedOn Jan. 3, 1117, a strong earthquake shook northern Italy. A manuscript of the 12th century, kept today in the Vatican Library, recorded what came next: "The lakes also rose so much that very few of t...

How farming helped cause two medieval tsunamis in an Italian lake, and why the danger has returned phys.org/news/2026-10... via @sciencex.bsky.social

7/ The danger has not gone: eight centuries of erosion have rebuilt the delta. A collapse today would send a wave to the northern shores within a minute, too fast for any warning. Thick layers under Lakes Garda and Como suggest the same story there.

8/ Thank you @khalifauniversity.bsky.social, @cnrs.fr and Savoie Mont Blanc University for their support, and the French research ministry for Naveed's PhD grant. Open paper and simulation data: www.nature.com/articles/s41... doi.org/10.6084/m9.f...

5/ In the model, the wave of 1117 climbs about 50 m on the shores next to the delta and floods up to 2.7 km inland; that of 1222 reaches about 13 m. Monte Isola shelters the southern shores: at Iseo the water rises by about 5.5 m.

6/ History agrees. A chronicle of the 12th century, kept in the Vatican Library, records that after the earthquake of 1117 "the lakes rose so much that very few of those living in the villages nearby escaped". In Iseo, the church of Sant'Andrea was rebuilt soon after.

3/ Earthquakes made the delta front collapse twice: 62 million m³ in 1117, and 10 million m³ in 1222, once a century of erosion had rebuilt it. Cores and seismic profiles studied at EDYTEM and @unibe.ch show both as the thickest layers in the deep basin.

4/ We simulated the slides and the waves they raised. In the model of 1117, the wave reaches the northern shores within a minute. Monte Isola, the island in the middle, splits it in two, and at Iseo, at the far end, the water starts to rise after 6 to 7 minutes.

2/ Lake Iseo lies at the foot of Val Camonica, a large valley of the southern Alps. Since Roman times, forest clearing, farming and grazing have tripled the sediment that its river, the Oglio, brings to the lake. The delta at the river mouth grew steep and overloaded.

About 11,700 years ago, as the glaciers were leaving the Alps, an underwater landslide fell into Lake Aiguebelette. Did it raise a tsunami? With Muhammad Naveed Zafar, @psabatier.bsky.social, Mathilde Banjan and Jihwan Kim, we rebuilt it in JGR Solid Earth (2024, @agu.org) 🧵 #Tsunami

7/ Why then? As the glaciers melted, the crust freed of their weight rebounded and earthquakes became more frequent; one of them most likely shook the slope loose. The lake has recorded no other event of this size in the 11,700 years since, so the tsunami risk there today is low.

8/ Thank you @khalifauniversity.bsky.social for supporting the work, and @enseignementsup-recherche.gouv.fr for Muhammad Naveed Zafar's doctoral scholarship. The paper is open access, and the simulation data are public: doi.org/10.1029/2023... doi.org/10.5281/zeno...

5/ The waves hit hardest on the steep shore facing the slide: there the water stands up to 3.7 m deep on ground that is normally dry. These numbers come from today's lake floor and shores, which have changed in 11,700 years: they reconstruct the event, they do not measure it.

6/ A lesson for modellers: the wave is about 1.5 km long, while the lake is at most 70 m deep, 22 times less. Waves that long all travel at nearly the same speed, so a more complete model, in which speed depends on wavelength, gives almost the same waves as the simpler one.

3/ We let the 767,000 m³ of mud slide in a model where it stays rigid until pushed hard enough, then flows. Its yield strength, 60 Pa, was set so that the slide ends where its deposit lies. It crosses the lake eastwards, turns south to the deep basin and stops within 12 minutes.

4/ The moving mass lifts the water by about 3 m within seconds. In half a minute the leading wave, some 1.5 m high, reaches the opposite shore. Within two minutes the waves cover most of the lake, and within three, all of it. The animation runs nine times faster than real time.

2/ The lake keeps a record. From the air, a pale shallow shelf rims its western shore, except where the slope gave way. In the deep basin, seismic profiles show the debris, 0.6 to 1.7 m thick, and a core taken by @edytem.bsky.social holds a layer 115 cm thick, dated to 11,700 years ago.

How does a flow on a rotating planet turn from regular to chaotic? With Francesco Carbone of Italy's National Research Council, in PLOS ONE (@plosone.org, 2024), we kept only five interacting waves and found two different roads to chaos. 🧵 #FluidDynamics #ChaosTheory

7/ The phases tell the same story. In the first chaos each wave stands still: its crests keep their place and at times swap with its troughs, so its phase is locked and jumps by half a turn while its amplitude is chaotic. In the second chaos the locking breaks: the waves drift.

8/ A model of five waves shows how large-scale motion loses stability, not fully developed turbulence. Thank you, Francesco, and @khalifauniversity.bsky.social for its Almesbar supercomputer. Paper and data: doi.org/10.1371/jour... zenodo.org/records/1136...

5/ The second road is different. With a force some 24 times stronger, just above C = 4.8, the motion goes from almost periodic to chaotic at once, with no cascade, and the energy of the surface now rises in irregular bursts.

6/ What differs is the part played by the surface. Under the weak force it holds about 0.01% of the energy, and the flow's own inertia drives the chaos. Under the strong force it holds about 1%: waves raised by the force push back on the flow.

3/ One number, C, sets the strength of the force. At C = 0.19 the surface breathes regularly, about once per rotation time, and the motion retraces a single loop. At C = 0.23 the breathing never repeats itself: the motion is chaotic.

4/ The first road is the period-doubling cascade, which Feigenbaum showed to be universal. The steady flow starts to oscillate at C ≈ 0.16; near C = 0.2 its loop splits in two, then four, eight, sixteen, faster and faster, until chaos.

2/ The rotating shallow-water equations describe a thin layer of fluid on a spinning planet, a classic model of oceans and atmosphere. We keep five waves in two triads, trios of waves whose wave vectors close a triangle and which trade energy. A steady force pushes one wave.

Dead satellites and rocket stages fall to Earth uncontrolled. When exactly? In Acta Astronautica, M. Khawariz Andaristiyan, Affiani Machmudah, Ridanto Eko Poetro, Wahyudi Hasbi (BRIN, ITB) and I answer with deep learning and orbital physics. 🧵 #SpaceDebris

7/ These are hindcasts: like the reconstructions published after a fall, they use tracking data up to shortly before it. Three objects are few; the real test is to predict objects still in orbit, including those tracked for less than 11 years.

8/ The equations are in the picture. Thank you, @khalifauniversity.bsky.social, which funded the work, and BRIN's Research Center for Satellite Technology for its facilities. The paper, in Acta Astronautica (@elsevierconnect.bsky.social): doi.org/10.1016/j.ac...

5/ We tested it on three objects that have fallen: a Soviet lander built for Venus in 1972, NASA's NEOWISE telescope and a Chinese rocket stage. In all three cases, the time reported or observed lies inside our window of uncertainty, ±7.8 to ±13.2 minutes.

6/ What if the coefficient is frozen at its last value? For the compact lander, the predicted fall moves by only 5 minutes. For the rocket stage, prone to tumbling, and NEOWISE, with its solar panels, it comes 36 and 31 minutes later, further from the reported times.

3/ Drag in the thin upper air brings objects down. It grows when the active Sun swells that air, and with the ballistic coefficient, drag coefficient × area ÷ mass, which shifts as an object tumbles. The three falls we studied came in the Sun's active years, 2023 to 2025.

4/ From 11 years of tracking data, one full solar cycle, we read the history of an object's ballistic coefficient. A neural network learns to forecast it from the object's past and from space weather, and every 3 hours feeds it to a physical orbit calculation, down to 80 km.

2/ At the end of 2025, low orbit held 25,060 catalogued objects, nearly twice as many as a decade earlier (ESA). As it falls, an object still flies at almost 8 km per second: a minute of error in the time of its fall shifts where it comes down by about 460 km along its path.

If a wormhole existed, it would ring when struck, as a black hole does after a merger. In The European Physical Journal C (@springer.springernature.com, 2025), Davide Batic and I computed how a massive phantom wormhole rings. 🧵 #Wormholes #GeneralRelativity

7/ No horizon hides the throat: part of every wave comes out on the other side. Such wormholes are hypothetical: they need negative energy, and earlier studies found that a radial disturbance makes them collapse into a black hole or inflate. Our question is one of principle.

8/ The equations we solved are in the picture. Open access thanks to SCOAP3. Thank you, Davide, and @khalifauniversity.bsky.social, which supported the work. doi.org/10.1140/epjc... github.com/dutykh/Ellis...

5/ For a throat of given size, the main gravitational tone drops as the mass grows: from 1.7377 − 0.3051i with no mass to 0.3927 − 0.0954i when the Schwarzschild radius is 1.6 times the throat parameter. The pitch is 4.4 times lower, and the tone fades 3.2 times more slowly.

6/ The same ringing in time. A pulse of gravitational waves falls into a wormhole whose Schwarzschild radius equals the throat parameter. Part of it crosses to the other side, and an observer on our side records a ringing that fades at the rate of the tone 0.5705 − 0.1299i.

3/ A wave that reaches the tunnel meets a barrier near the throat. It lingers there and leaks out as a ringing: a few tones, each with a pitch and a rate of fading. The more massive the wormhole for its throat, the lower and wider this barrier.

4/ We computed the tones with a spectral method. The whole tunnel, both sides and the throat, becomes one interval, on which the wave is a sum of 400 Chebyshev polynomials, handled with 400-digit arithmetic. Unlike earlier studies, nothing is cut in the middle and matched.

2/ The wormhole of Ellis and Bronnikov is a tunnel between two regions of space, held open by a phantom field, a hypothetical matter of negative energy. Give it a mass and it turns lopsided: our side pulls toward the throat, the other side pushes away, and clocks there run slow.

Where could the next tsunami on the Makran coast of Iran and Pakistan come from? With Amin Rashidi (University of Tehran), Mohammad Mokhtari, Mehdi Masoodi, Parvaneh Faridi and Sara Kiani, we reviewed the research (Journal of Seismology and Earthquake Engineering, 2024). 🧵 #Tsunami

7/ Sediment is where my own work enters. In 2010, Frédéric Dias and I found that a sediment layer barely changes the final uplift of the seabed, but can enlarge it during the rupture, up to 1.7 times in idealised tests. How the Makran's thick sediment behaves is still open.

8/ Open questions remain: are the plates locked in western Makran, storing strain? And the seabed near the coast is mapped too coarsely for flooding studies. Thank you @khalifauniversity.bsky.social for supporting the work. Free to read: doi.org/10.48303/jse...

5/ The great earthquake on the plate boundary is not the only source. Our review gathers the others: splay faults that branch up from it, normal faults that reach down to it, landslides on steep undersea slopes, and possibly storms, which can drive a meteotsunami.

6/ Landslides appear to be the second most important source. In 100 simulated landslide tsunamis off western Makran, Chabahar in Iran and Muscat in Oman stood out as the most exposed. No meteotsunami is on record in the Gulf of Oman, but cyclones reach it, like Gonu in 2007.

3/ Tsunamis are on record in 326 BC, 1008, 1524, 1897, 1945 and 2013, not all from great earthquakes. The 1945 waves, from a magnitude 8.1 earthquake and an undersea landslide, killed several hundred people. The 2013 waves followed an inland earthquake, likely via a landslide.