Improved Prediction of Debris-Flow Deposit Delivery in Headwaters of the OCR

Abstract:

Debris flows in montane watersheds impact fish habit, roadways and property, and remain a concern for managers and policy-makers. This research draws upon decades of research into debris-flows in the PNW, combining multiple datasets, to optimize a geomorphic landscape model using field observations of valley floor deposits.

The complexity of landscape models trying to predict debris-flows have grown increasingly Rube-Goldbergian over time, with one of the more exhaustive models developed by Miller & Burnett in 2008. Representing the stream network as a series of linked nodes, the model predicts the likelihood that a point on the stream network receives a debris-flow deposit from any upstream source. This research uses a record of 370 radiocarbon dated valley deposit locations and transit times, surveyed by Lancaster from 2006-2009, to test whether debris-flow delivery probabilities from the model are predictive of deposit locations in the field.

Analogous to the method Miller & Burnett used to calibrate their topological predictor of landslide initiation to a record of landslides from field surveys, by comparing the change in density of landslide initiation points to the change in density of total pixel area as their topological predictor changes, we calibrate the delivery probability index to a record of debris-flow deposits by comparing the change in density of debris-flow deposits to the change in density of total deposits as delivery probability changes, in order to produce a weighting function for delivery probabilities optimized using field observations.

The optimized weighting function increases monotonically with delivery probability, indicating that delivery probabilities are predictive of debris-flow deposit locations from the Lancaster surveys (see Figure A). When the delivery index is below 0.01, the relative density of debris-flow deposits on the valley floor is less than the mean. Above index values of 0.022, debris-flow deposits occur at more than three times the mean rate. Our continuing research uses the Lancaster record of deposit volumes and transit times to estimate flux at the level of the Miller & Burnett stream node, weighted by optimized delivery probability.