top of page
Post-fire Debris Flow Initiation Model
We collected post-fire data on rainfall and hydrogeomorphic outcomes from fourteen burned regions in Arizona and New Mexico. This dataset includes 163 observations from 200 watersheds in the first year after the fire, totaling 3,144 records for specific events. Rainfall intensities were averaged over 10, 15, 30, and 60 minutes. Following Staley et al. (2017), we derived features using the product of rainfall with terrain characteristics (RT), fire metrics (RF), and soil properties (RS).
We developed models using logistic regression (LR), linear discriminant analysis (LDA), random forest (RF), and XGBoost (XGB), resulting in 3,200 two-feature and 16,000 three-feature models. We chose two two-feature LR models, SWA and SWB, for their interpretability and user-friendliness, offering comparable or better performance than complex models.
The SWA model incorporates rainfall, terrain, and fire metrics based on 15-minute averaged rainfall intensities, mean slope, and the fraction of watershed area burned at different severity levels. In contrast, the SWB model uses the same rainfall and slope metrics but incorporates the mean watershed differenced normalized burn ratio (dNBR) divided by 1000.
We calculated rainfall intensity–duration (ID) thresholds for a 15-minute duration using the SWA and SWB models. For comparison, we also used the M1 model (Staley et al., 2017), which is widely utilized for assessing postfire debris-flow risks. The study included the 651 South watershed from the 2017 Pinal Fire in Arizona and Watershed D from the 2020 Tadpole Fire in New Mexico, while Watershed CON1 was withheld from the training/testing dataset and used only for calculating intensity–duration (ID) thresholds.
At 651 South, thresholds were 64 and 51 mm h⁻¹; at Tadpole’s Watershed D, 60 and 58 mm h⁻¹; and at CON1, 51 and 59 mm h⁻¹ for models SWA and SWB, respectively. In all three watersheds, the M1 model gave approximately 21 mm h⁻¹ and many more false positives. In particular, on the withheld Contreras Fire, watershed CON1, SWA/SWB had 2 false positives each, while the M1 model gave 12 false positives. Our models generate spatial debris-flow likelihood maps and 15-minute intensity–duration thresholds for small watersheds, supporting postfire hazard assessment in the western U.S., and they outperform M1 while acknowledging regional differences in rainfall and sediment transport.
Pre-fire dNBR Prediction Model
Post-fire debris-flow initiation models can be paired with burn-severity models that predict dNBR (differenced Normalized Burn Ratio), linking fire behavior to downstream hazards (Staley et al., 2018; Wells et al., 2023). Understanding drivers of high-severity fire and building predictive models is critical for protecting people and ecosystems (Parks et al., 2018; Klimas et al., 2025). Remote sensing improves burn-severity prediction and supports both pre-fire risk assessment and post-fire evaluation (Klimas et al., 2025; Wells et al., 2023). As climate change shifts fire regimes, these tools enable adaptive management, including risk reduction near the wildland–urban interface (Collins et al., 2020). We aim to use pre-fire burn-severity models to evaluate post-fire debris-flow (PFDF) likelihood, connecting burn-severity forecasts with downstream hazard assessments. However, burn-severity forecasts are rarely integrated directly with post-fire debris-flow likelihood. Our objective is to produce prefire burn-severity predictions for direct integration into PFDF models.
References
Collins, L., McCarthy, G., Mellor, A., Newell, G., and Smith, L. (2020). Training data requirements for fire severity mapping using Landsat imagery and random forest. Remote Sensing of Environment, 245:111839.
Gorr, A. N., McGuire, L. A., Beers, R., and Hoch, O. J. (2023). Triggering conditions, runout, and downstream impacts of debris flows following the 2021 Flag Fire, Arizona, USA. Natural Hazards.
Klimas, K. B., Yocom, L. L., Murphy, B. P., David, S. R., Belmont, P., Lutz, J. A., DeRose, R. J., and Wall, S. A. (2025). A machine learning model to predict wildfire burn severity for pre-fire risk assessments, Utah, USA. Fire Ecology, 21(1):8.
Liu, T., McGuire, L. A., Youberg, A. M., Gorr, A., and Rengers, F. K. (2023). Guidance for parameterizing post-fire hydrologic models with in situ infiltration measurements. Earth Surface Processes and Landforms, 48(12):2368–2386.
McGuire, L. A., Rengers, F. K., Youberg, A. M., Ganesh, I., Gorr, A. N., Hoch, O., Johnson, J. C., Lamom, P., Prescott, A. B., Zanetell, J., and Abramson, N. S. (2020). Post-wildfire debris-flow monitoring data, 2019 Woodbury Fire, Superstition Mountains, Arizona, USA. U.S. Geological Survey data release, https://doi.org/10.5066/P9URVNXI.
McGuire, L. A., Youberg, A. M., Rengers, F. K., Abramson, N. S., Ganesh, I., Gorr, A. N., Hoch, O., Johnson, J. C., Lamom, P., Prescott, A. B., Zanetell, J., and Fenerty, B. (2021).Extreme Precipitation Across Adjacent Burned and Unburned Watersheds Reveals Impacts of Low Severity Wildfire on Debris-Flow Processes. Journal of Geophysical Research: Earth Surface, 126(4).
McGuire, L. A., Rengers, F. K., Youberg, A. M., Gorr, A. N., Hoch, O. J., Beers, R., and Porter, R. (2024). Characteristics of debris-flow-prone watersheds and debris-flow-triggering rainstorms following the Tadpole Fire, New Mexico, USA. Natural Hazards and Earth System Sciences, 24(4):1357–1379.
Parks, S. A., Holsinger, L. M., Panunto, M. H., Jolly, W. M., Dobrowski, S. Z., and Dillon, G. K. (2018). High-severity fire: evaluating its key drivers and mapping its probability across western us forests. Environmental research letters, 13(4):044037.
Raymond, C., McGuire, L., Youberg, A., Staley, D., and Kean, J. (2020). Thresholds for post-wildfire debris flows: Insights from the Pinal Fire, Arizona, USA. Earth Surface Processes and Landforms, 45(6).
Staley, D. M., Negri, J. A., Kean, J. W., Laber, J. L., Tillery, A. C., and Youberg, A. M. (2017). Prediction of spatially explicit rainfall intensity–duration thresholds for post-firedebris-flow generation in the western United States. Geomorphology, 278:149–162.
Staley, D. M., Tillery, A. C., Kean, J. W., McGuire, L. A., Pauling, H. E., Rengers, F. K., Smith, J. B., Staley, D. M., Tillery, A. C., Kean, J. W., McGuire, L. A., Pauling, H. E., Rengers, F. K., and Smith, J. B. (2018). Estimating post-fire debris-flow hazards prior to wildfire using a statistical analysis of historical distributions of fire severity from remote sensing data. International Journal of Wildland Fire, 27(9):595–608.
Wells, W. G. (1987). The effects of fire on the generation of debris flows in southern California. Reviews in Engineering Geology, 7:105–114.
Youberg, A. M. (2015). Geodatabase of Post-Wildfire Study Basins: As sessing the predictive strengths of post-wildfire debris-flow models in Arizona, and defining rainfall intensity-duration thresholds for initiation of post-fire debris flow.: Digital Information Series DI-44: Tucson, AZ, Arizona Geological Survey.
bottom of page