Settling, Swirling, Sticking: Clustering-Driven Interactions In Volcanic Particle Flows

Particle-laden volcanic flows are hazardous across a wide range of settings, from dispersing ash clouds to pyroclastic density currents (PDCs). Their impacts depend not only on bulk mass loading and particle size, but also on how particles self-organise in space. Yet, many studies and hazard models are built on bulk- or layer-averaged properties, so concentration inhomogeneities within the flow are poorly constrained. A key missing piece is clustering (preferential concentration): particles concentrate into dense regions separated by voids, creating sharp local contrasts that can alter settling, generate short-lived sedimentation pulses, and enhance particle–particle interactions even when mean concentrations are low. We investigate these processes using controlled laboratory experiments that isolate clustering and its effects in sustained, free-falling columns of volcanic ash. We vary particle size distributions and mass release rates to span particle volume fractions ≈10^-5–10^-2, encompassing conditions relevant to dispersing clouds and ash-laden regions within PDCs. High-speed laser imaging and particle tracking resolve instantaneous particle positions and velocities. We quantify clustering with Voronoi tessellation, measure settling velocity variability, and estimate a collision-rate proxy from local particle statistics to link spatial organisation to encounter likelihood. Results suggest that clustering can create strong local concentration contrasts, whose intensity can enhance particle–particle interactions and increase the potential for collisions, aggregation, and turbulence-modulated settling. Importantly, peaks in the collision-rate proxy are not explained by velocity variability alone, indicating that spatial organisation shortens effective interaction length scales and increases encounter frequency. These findings link dilute turbulent suspensions to enhanced fallout and collision/aggregation potential, and they highlight the need for hazard models to capture local concentration contrasts, not just bulk-mean concentrations.