The Ground Moves in Chains. Hazard Models Need to Follow the Chain.

Water at basin scale

In a preprint on groundwater modelling, the G-ADOPT solver enabled seasonal, basin-scale, fully three-dimensional solutions of the full Richards equation on unstructured, terrain-following meshes. Vertically lumped coarse-space preconditioners reduced wall-clock time by a factor of two to four compared with single-level methods in near-saturated profiles. The solver was applied to a 30,000 square kilometre model of the Lower Murrumbidgee floodplain. The advance is computational, but its purpose is practical: water managers need models that can run at basin scale and follow terrain, rather than models that are accurate only on small, simplified domains. Speed matters here for a specific reason. A model that takes too long to run cannot be used for seasonal planning or for repeated scenarios, so a two- to four-fold speedup in near-saturated profiles is not a technical footnote. It is what makes basin-scale simulation usable as a planning tool rather than a research demonstration. m:…3118

A landslide preprint shows why the physics inside such models cannot be simplified away. Erosion-induced momentum production and system acceleration were essential for accurate landslide mobility modelling, erosion rate alone was insufficient to determine mobility, and neglecting erosion-induced reduced inertia led to underestimation of landslide mobility. A model that treats a landslide as a block sliding downhill, without accounting for how erosion itself adds momentum and changes inertia, will systematically understate how far and how fast material travels. That understatement is exactly what puts roads, settlements and infrastructure in the wrong place on a hazard map. The model therefore reframes the question for hazard assessors: not how fast material erodes, but how erosion changes momentum and inertia as the slide moves, because those are the quantities that determine where it stops. m:…6538

A cascade, not a single event

The 26 August 2026 Bhote Koshi event in Nepal shows what happens when hazard screening assumes a single trigger. A multi-source quantitative reconstruction found the event was a complex rock-ice avalanche-debris-flow-flood cascade rather than a simple glacial lake outburst flood. The source-to-border segment of the mass flow reached approximately 188 kilometres per hour. A public warning was issued 38 minutes after the source seismic signal, leaving only five minutes of lead time before station failure at the Betrawati site. A warning system built around one lake, one trigger and one timeline could not keep up with a chain that changed form as it moved. m:…5411

The costs in the reconstruction make the modelling gap concrete. Total disaster effects were USD 2.695 billion, with recovery needs of USD 4.775 billion, of which 54% was attributed to the energy sector. Energy infrastructure, which often sits in valleys and along rivers precisely where cascades travel, bore more than half the recovery burden. This suggests that hazard assessment for energy assets should be organised around process chains, with groundwater, slope and flood processes modelled together, rather than around isolated hazards assessed one at a time. m:…5411 m:…3118 m:…6538

What still needs testing

These are preprints and a reconstruction, not validated operational systems. The groundwater solver was demonstrated on one large floodplain, the landslide model is analytical, and the Bhote Koshi reconstruction explains one cascade after the fact. The test that matters is prospective: whether chain-aware models, run before an event, give earlier and more accurate warnings and better asset placement than single-hazard models in the same terrain. m:…3118 m:…6538 m:…5411