How Farming Triggered Two Medieval Tsunamis in an Italian Lake — And Why the Danger Is Back

Farming and land use helped precondition two medieval lake tsunamis at Lake Iseo (northern Italy), and the hazard has returned because the delta has rebuilt.

This is based on a 2026 Nature Communications paper (“Human- enhanced tsunami hazard in large lakes”) led by Muhammad Naveed Zafar, with co-authors including Denys Dutykh (who wrote the Phys.org piece summarizing the work). The team combined sediment cores, seismic profiles, historical and archaeological evidence, and numerical modeling of landslides and waves.

Lake Iseo sits at the foot of Val Camonica in the southern Alps. The Oglio River builds a large delta at its northern end. Since Roman times, forest clearing, farming, and grazing in the catchment have tripled sediment delivery to the lake relative to earlier Holocene levels.

This rapidly built up and over steepened the underwater delta front, making it unstable. A prior collapse had already occurred in the 6th century; centuries of accelerated erosion then reloaded it.

The two events

• 3 January 1117: A magnitude ~6.5 earthquake (near Verona) triggered collapse of ~62 million m³ of sediment. Simulations show waves reaching northern shores in under a minute, with run- ups of ~50 m near the delta and flooding up to ~2.7 km inland along the Oglio plain. Farther south (e.g., at Iseo town), run-up was ~5.5 m after 6- 7 minutes. Monte Isola partly sheltered southern shores. A contemporary Vatican Library manuscript notes lakes rising so high that few nearby villagers escaped.⁠
• Christmas Day 1222: A magnitude ~6.0 earthquake triggered a smaller collapse (~10 million m³) after the river had partially rebuilt the front. Modeled run- ups reached ~13 m.⁠

The thickest mass- wasting deposits in the deep basin (up to ~250 m depth) match these volumes and locations. Modeling treated the slides as viscoplastic mud and the long waves with shallow-water equations (waves traveled ~50 m/s or ~180 km/h in deep water). Results align with historical accounts, church rebuilds, archaeological damage, and the absence of deposits in sheltered southern areas.

After 1222, land- use pressure eased temporarily (e.g., during the Black Death) but intensified again from the 17th century and peaked in the industrial era. Eight centuries of continued human- driven erosion has rebuilt the delta front. Lake Iseo’s shores are now densely populated. A similar collapse today would reach northern towns (e.g., Pisogne, Lovere) in seconds to under a minute, too fast for effective warning in many places, with more time (minutes) farther south.⁠

Similar human- amplified risks exist elsewhere (e.g., evidence of medieval collapses in Lakes Garda and Como; analogous cases from Lake Geneva to Alaska and New Zealand). Global sediment delivery to lake deltas has risen sharply with modern land use, increasing the potential for future failures.⁠

In short, people did not cause the earthquakes, but centuries of farming and deforestation overloaded the delta, turning seismic shaking into catastrophic underwater landslides and tsunamis. The same process has reset the hazard.

Lake Iseo tsunami modeling methods (from Zafar et al., Nature Communications 2026) used a one- way coupled two- stage approach: first simulate the subaqueous landslide dynamics, then feed the resulting time- dependent bathymetric changes into a tsunami propagation, inundation model. Feedback from the waves onto the landslide is neglected, but hydrodynamic forces (viscous drag, added mass) are included in the landslide stage.

1. Reconstruction of pre- failure geometry and volumes

• High-resolution LiDAR topography and multibeam bathymetry (2002 surveys by Regione Lombardia and OGS, CARG program).
• Present- day bathymetry served as a morphological analogue for the pre-failure delta (assuming post- event sedimentation largely infilled the scars along a similar pathway).
• In GIS, post- failure morphology was reconstructed by removing accumulated sediment and restoring a plausible failure- scar shape guided by bathymetric trends and erosional features.
• Source volumes were calculated as the difference between the reconstructed post- failure surface and modern bathymetry, then iteratively adjusted until they matched the volumes of the mapped mass-wasting deposits (from seismic profiles + sediment cores) in the deep basin:
  • 1117 CE: $6.2\times {10}^7$6.2×107 m³
  • 1222 CE: $1.0\times {10}^7$1.0×107 m³
• Assumption of single- event failure supported by the sedimentology of the deposits (erosive sandy base → single upward-fining unit → homogeneous suspension- settling layer from seiches).

2. Landslide modeling (BingClaw)

• Two- layer, depth- averaged (2HD) viscoplastic debris- flow model BingClaw.
• Rheology: Herschel- Bulkley (suitable for cohesive subaqueous landslides).
• Layers:
  • Upper plug layer (moves en bloc, no internal shear).
  • Lower shear layer (deforms once shear stress exceeds yield strength).
• Mass balance solved over the full depth; momentum equations solved separately per layer with a Godunov- type finite- volume scheme.
• Key rheological parameters chosen to best match observed runout distance, deposit thickness, and spatial distribution (runout is most sensitive to yield strength):
  • 1117 CE: yield strength 25 Pa, bulk density 1600 kg m⁻³
  • 1222 CE: yield strength 7 Pa, bulk density 1650 kg m⁻³ (lower yield strength reflects higher water content and coarser sediment fraction)
  • Other fixed parameters (from literature and validation studies): fluidity index $n=0.5$n=0.5, added mass 0.1, frictional drag 0.01, pressure force 1.0, reference strain rate 100 s⁻¹.
• Output: time- evolving landslide thickness and runout until final deposition in the deep basin (validated against seismic profiles and cores). The 1222 CE slide traveled farther and faster due to lower strength and more accommodation space.

3. Tsunami modeling (GeoClaw)

• Time-dependent bathymetric changes from BingClaw drive wave generation.
• Model: GeoClaw: 2- D depth -averaged nonlinear shallow- water equations (NSWE).
• Features: shock- capturing Godunov- type finite- volume scheme, adaptive mesh refinement, effective wet or dry interface handling for inundation.
• Grid resolution: ~10 m rectangular cells.
• Manning roughness coefficient: 0.03.
• Lake level: modern reference (185.44 m a.s.l.); no significant level change inferred over the last ~2 kyr.
• Waves are long (~5.5 km wavelength for 1117 CE, ~22× maximum lake depth of 250 m), so NSWE is appropriate; water moves nearly as a coherent sheet.
• Long- wave speed $\approx \sqrt{gh}\approx 50$≈gh​≈50 m s⁻¹ (~180 km h⁻¹) in deep water.

4. Additional analyses and sensitivity

• Froude number ($Fr=U\mathrm{/}\sqrt{gh}$​, where U is landslide velocity magnitude): calculated to assess generation efficiency (most efficient near $Fr\approx 1$Fr≈1). Tracked temporally for the 1117 CE event.
• Dispersion sensitivity: comparison of nondispersive NSWE vs. a more complete (Boussinesq- type) model; results were nearly identical at the same times, supporting the NSWE approximation for these long waves.
• Initiation- depth sensitivity: alternative scenarios with the 1117 CE failure starting deeper (45 m or 75 m below surface) produced lower maximum run-ups (~22 m or ~11 m instead of >50 m).
• Validation: modeled deposit geometries match observations; tsunami results consistent with historical accounts (Vatican manuscript), archaeological damage (e.g., church of Sant’Andrea), and absence of deposits in sheltered southern areas.

Simulations used the Clawpack framework (BingClaw and GeoClaw).

Simulation data (outputs, Makefiles, plotting scripts for both events, including optional Boussinesq runs for 1117 CE) are available on Figshare. The approach has been validated on other events (e.g., Storegga, Alpine lake slides).

These methods combine geological constraints with physics- based numerical modeling to reconstruct the two medieval events and assess modern hazard.

Human-enhanced tsunami hazard in large lakes

Human-induced erosion increases sediment delivery to lake deltas, raising their risk of collapse. At Lake Iseo (northern Italy), roughly two millennia of intensified land use (deforestation, agriculture, pastoralism) tripled sediment flux to the lake. This over steepened the Oglio River delta and preconditioned it for failure.

Geological records and numerical simulations show that earthquakes in 1117 CE (Mw 6.5) and 1222 CE (Mw 6.0) triggered major subaqueous delta collapses.

These generated tsunamis with maximum run- ups of approximately 50 m and 13 m, respectively. Historical accounts (Vatican Library manuscript) and archaeological evidence support the reconstructions.

Since the last collapse, eight centuries of continued human- driven erosion has rebuilt the delta. Shoreline populations are now highly vulnerable to future earthquake- triggered failures.

A global lake synthesis indicates that human- driven erosion has reached its highest levels since the Industrial Era, progressively overloading delta’s worldwide and potentially increasing the likelihood of similar events.

Main Results from Lake Iseo

• Sediment loading: Since Roman times, human activities tripled sediment flux relative to earlier Holocene background levels. The delta front became steep and overloaded. A prior collapse occurred in the 6th century; further accumulation reloaded it.
• Collapse volumes:
  • 1117 CE: ~62 × 10⁶ m³ (thickest deposit ~7 m in the deep basin).
  • 1222 CE: ~10 × 10⁶ m³ (deposit ~2 m thick), after rapid rebuilding in the intervening century.
• Tsunami modeling:
  • 1117 CE: Waves >50 m near the northern shores; reached shores within ~1 minute; inundation up to ~2.7 km inland along the Oglio plain. At the town of Iseo (~7 min later), ~5.5 m run- up. Monte Isola partially sheltered southern areas.
  • 1222 CE: Peak waves ~13 m; inundation up to ~1 km inland.
• Validation used sediment cores, seismic profiles, viscoplastic landslide modeling, and nonlinear hydrodynamic (shallow- water) tsunami simulations. Results match deposit geometries and historical and archaeological evidence (e.g., damage and rebuilding of the church of Sant’Andrea in Iseo; minimal impact and no deposits in the sheltered southern Torbiere del Sebino area).

The work demonstrates a persistent, human- amplified mechanism for generating large lacustrine tsunamis. Similar overloaded deltas and thick mass- wasting deposits exist in other lakes (e.g., Garda, Como).

Modern high sediment delivery rates mean comparable failures may become more likely under seismic loading. Hazard assessments for densely populated lake shores in tectonically active regions should incorporate long- term anthropogenic sediment dynamics, not only seismic sources.⁠

The paper integrates geophysics, sedimentology, history, archaeology, and numerical modeling (landslide rheology and tsunami propagation and inundation). Supplementary materials include cores, seismic profiles, and additional figures.

Corresponding author: Muhammad Naveed Zafar.

Journal: Nature Communications, volume 17, Article number: 10531 (2026)

DOI: 10.1038/s41467-026-78272-8

Published: 15 October 2026

Authors: Muhammad Naveed Zafar,
Pierre Sabatier,
Denys Dutykh,
William Rapuc,
Raphaël Paris,
Andrea Piccin,
Pascal Vuillemin &
Flavio S. Anselmetti

Abstract

Worldwide, human-induced erosion enhances sediment delivery to lake deltas, increasing their susceptibility to collapse. In Lake Iseo, northern Italy, two millennia of intensified land use tripled sediment flux, oversteepening the delta and preconditioning it for failure. Here we show from geological records and numerical simulations that earthquakes in 1117 CE (Mw 6.5) and 1222 CE (Mw 6.0) triggered major subaqueous delta collapses, generating massive tsunamis with maximum run-ups of ~50 m and ~13 m, respectively. Historical accounts from the Vatican Library and archeological evidence support these reconstructions. Since the last collapse, eight centuries of human-induced erosion have rebuilt the delta, leaving shoreline populations highly vulnerable to future earthquake-triggered failures. A global lake synthesis showed that human-driven erosion has reached its highest levels since the Industrial Era, progressively overloading deltas worldwide; thus, similar catastrophic failures may become more likely in the future. These findings reveal a persistent, human-amplified mechanism of lacustrine tsunami generation and intensity with major implications for hazard assessment.