Temporal dynamics and predictability of sediment plumes from a deep-sea mining vehicle under different source configurations

Plumes generated by deep-sea mining vehicles are multiphase flows in which hydrodynamic adjustment, entrainment, transport, settling, and particle state evolution act over competing time scales. How source configuration reorganizes these dynamics and their predictability remains unclear. We analyzed 22 stationary collection (SC), mobile collection (MC), and transit (TR) events using temporal analysis, ExtraTrees models, and recursive simulation. The modes represented a fixed collection source, collection combined with motion, and motion without active collection. Approximate velocity to concentration lags were 110 s for SC and MC and 170 s for TR, with broad uncertainty intervals. Particle irregularity decreased in five of six responsive MC events and increased in 10 of 11 TR events in the mapped records. Across withheld events, velocity carried appreciable predictive importance for concentration increments in the moving modes. ExtraTrees reduced velocity root mean square error (RMSE) by 11.1%–17.6% at 30–120 s, while derived flux RMSE decreased by 1.4%–17.5% at 10–120 s. Concentration and particle properties retained persistence as the final predictor, with particle scores influenced by sampling resolution. Recursive validation and initial state tests constrained the 95 scenarios to hypothetical comparisons. Together, the results show that the tested source configurations differ in measured temporal responses and fitted predictive dependence, while recursive scenario trends remain conditional on model validation.