Paper No. 9
Presentation Time: 10:30 AM

A PROBABILISTIC APPROACH TO POST-WILDFIRE DEBRIS-FLOW VOLUME MODELING


DONOVAN, Ian P., Department of Geology and Geological Engineering, Colorado School of Mines, Golden, CO 80401 and SANTI, Paul, Department of Geology & Geological Engineering, Colorado School of Mines, Golden, CO 80401, idonovan@mymail.mines.edu

As populations continue to move into more mountainous terrain, a greater understanding of the processes controlling debris flows has become important for the protection of human life and property. The potential volume of an expected debris flow must be known to effectively mitigate hazards it may pose, yet an accurate estimate of this parameter has proven very difficult to obtain. Previous models to predict debris-flow volume have very broad prediction intervals. To this end, a probabilistic method for the prediction of debris-flow volumes has been developed using a database from the Western United States. The database includes 48 debris flows with over 2500 measurements of yield rates at various points along the debris flow paths. A number of geomorphological, climatic, biological, and geotechnical basin characteristics were considered, and single and multiple regression analyses were used to identify those with the strongest correlation to debris-flow yield rates. Next, cluster analysis was used to identify important groupings of similar sets of characteristics and to identify significant break points in otherwise continuous ranges of values. For each parameter, a probability density function was developed for each of the groups identified in the cluster analysis. The characteristics of the basin to be modeled can then be defined in terms of these functions and a Monte Carlo simulation can be used to combine each of these functions to create a single probability distribution curve for total debris-flow volume. This approach will be validated by testing the model on basins with debris flows of known volumes; given the range of precision of input values, the known volumes are expected to lie within two standard deviations of the mean predicted volume.