1. 2012 Southwest Wildfire Hydrology and Hazards Workshop Evan Friedman and Dr. Paul Santi Colorado School of Mines 3 April 2012

2. Contents  Background  Approach  Hazard Assessment  Rainfall and Hydrologic Response  Model Validation  Sensitivity Analysis  Conclusions

3. Background  The Medano Fire burned 6000 acres at Great Sand Dunes NP during the summer of 2010  Debris-flow hazard assessment for park resource managers  Regression models to predict probability and volume of post-fire debris flows (Cannon et al., 2009)  Debris-flow monitoring used to validate models, and establish rainfall thresholds and relative timing

4. Background

5. Approach  Perform initial hazard assessment using design storms  Monitoring of rainfall and hydrologic response (rain gauges, pressure transducers, deposit surveys)  Validate predictive model results using actual storms

6. Hazard Assessment  Model parameters: Burn severity Rainfall conditions Topographic characteristics Soil properties (STATSGO US Soils Database - Soil Survey Staff, 2011) USDA Forest Service, 2010

7. Hazard Assessment  Debris-flow probability prediction: Three logistic regression models: A, B, and C (Cannon et al., 2009) 2-year and 10-year, 1-hour design storms Models did not agree on magnitude, and thus were averaged Probability rankings (low=1, medium=2, and high=3)

8. Hazard Assessment  Debris-flow volume prediction: One regression model (Cannon et al., 2009) 2-year and 10-year, 1-hour design storms Volume rankings (low=1, medium=2, and high=3)

9. Hazard Assessment  Hazard Ranking: Sum of probability and volume rankings Hazard rankings (2-3=low, 4=medium, 5-6=high)

10. Rainfall and Hydrologic Response

12.  Intensity-duration threshold: I = 12.0D -0.5 Cannon et al., 2008

13. Rainfall and Hydrologic Response  Debris-flow and flood timing:

14. Model Validation Probability Models Recorded average storm intensity values for debris flow storms

15. Model Validation Volume Model Recorded total rainfall amounts for debris flow storms  Model predictions one order of magnitude higher than measured  Previous validations: similar models over-predict volumes for relatively small basins

16. Sensitivity Analysis

17. Conclusions  Peak rainfall intensities for short periods within storms better predict debris-flow occurrence than average storm intensity in this setting  Models A and C were successful at predicting high probability of debris flows in this setting  The volume model predicts volumes within one order of magnitude higher than measured for relatively small basins in this setting  Percentage of basin area burned at moderate to high severity is the most significant variable for debris-flow probability in the western US  Probability models are sensitive to soil property variables, thus representative values from the range of STATSGO data

18. Probability Models Probability = e x /(1+e x ) Model A: x = -0.7 + 0.03a - 1.6b + 0.06c + 0.2d - 0.4e + 0.07f a = % basin area w/ slope >30% b = Ruggedness (change in basin elev./sq. root of basin area) c = % basin area burned at moderate and high severity d = Clay content (%) e = Liquid limit (%) f = Avg. storm intensity (mm/hr.) Model B: x = -7.6 - 1.1a + 0.06b + 0.09c - 1.4d + 0.06e a = Ruggedness (change in basin elev./sq. root of basin area) b = % basin area burned at moderate and high severity c = Clay content (%) d = Organic matter (%) e = Avg. storm intensity (mm/hr.) Model C: x = 4.8 + 0.05a + 0.2b - 0.4c - 1.5d + 0.07e a = % basin area burned at moderate and high severity b = Clay content (%) c = Liquid limit (%) d = Hydrologic group (based on soil infiltration rate and depth to confining layer) e = Average storm intensity (mm/hr.)

19. Volume Model Ln V = 7.2 + 0.6(Ln A) + 0.7(B) 1/2 + 0.2(T) 1/2 + 0.3 V = volume (m 3 ) A = area of basin w/ slopes >30% (km 2 ) B = area of basin burned at moderate and high severity (km 2 ) T = total storm rainfall (mm)