Year: 2026
Type: article
Abstract: Debris-flow inventories usually record where and when events occur but often compress source initiation, transformation and runout into one whole-event label. This is problematic in high mountains, where rainfall, melt, ground-ice thaw, rock–ice failure, lake release and channel entrainment can occupy different causal roles within a single cascade. We propose an evidence-graded six-stage attribution architecture that records conditioning context, dynamic preconditioning, proximate initiator, release mechanism, transformation and routing/runout separately. Water source and external forcing are treated as attributes rather than stages, and each stage retains its evidence basis, corroboration, plausible alternatives and unresolved status. The framework is designed to reduce prediction-time data leakage, align monitoring signals with the decisions they can support, and distinguish background climate influence from event-scale initiation. We outline benchmark inventories and prospective tests against binary baselines to determine when stage-resolved attribution improves interpretability, transferability, calibration or forecast updating.
Source: Perspectives in Social Sciences
DOI: 10.63468/pss.205
Language: English
Subfield: Management, Monitoring, Policy and Law
Field: Environmental Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7236231730
