Post-wildfire debris flows (PFDF) threaten life and property in western North America. They are triggered by short-duration, high-intensity rainfall. Following a wildfire, rainfall thresholds are developed that, if exceeded, indicate high likelihood of a PFDF. Existing weather forecast products allow forecasters to identify favorable atmospheric conditions for rainfall intensities that may exceed established thresholds at lead times needed for decision-making (e.g., & GE;24 h). However, at these lead times, considerable uncertainty exists regarding rainfall intensity and whether the high-intensity rainfall will intersect the burn area. The approach of messaging on potential hazards given favorable conditions is generally effective in avoiding unanticipated PFDF impacts, but may lead to "messaging fatigue" if favorable triggering conditions are forecast numerous times, yet no PFDF occurs (i.e., false alarm). Forecasters and emergency managers need additional tools that increase their confidence regarding occurrence of short-duration, high-intensity rainfall as well as tools that tie rainfall forecasts to potential PFDF outcomes. We present a concept for probabilistic tools that evaluate PFDF hazards by coupling a high-resolution (1-km), large (100-member) ensemble 24-h precipitation forecast at 5-min resolution with PFDF likelihood and volume models. The observed 15-min maximum rainfall intensities are captured within the ensemble spread, though in highest & SIM;10% of members. We visualize the model output in several ways to demonstrate most likely and most extreme outcomes and to characterize uncertainty. Our experiment highlights the benefits and limitations of this approach, and provides an initial step toward further developing situational awareness and impact-based decision-support tools for forecasting PFDF hazards.

Early warning of post-fire debris-flow occurrence during intense rainfall has traditionally relied upon a library of regionally specific empirical rainfall intensity–duration thresholds. Development of this library and the calculation of rainfall intensity-duration thresholds often require several years of monitoring local rainfall and hydrologic response to rainstorms, a time-consuming approach where results are often only applicable to the specific region where data were collected. Here, we present a new, fully predictive approach that utilizes rainfall, hydrologic response, and readily available geospatial data to predict rainfall intensity–duration thresholds for debris-flow generation in recently burned locations in the western United States. Unlike the traditional approach to defining regional thresholds from historical data, the proposed methodology permits the direct calculation of rainfall intensity–duration thresholds for areas where no such data exist. The thresholds calculated by this method are demonstrated to provide predictions that are of similar accuracy, and in some cases outperform, previously published regional intensity–duration thresholds. The method also provides improved predictions of debris-flow likelihood, which can be incorporated into existing approaches for post-fire debris-flow hazard assessment. Our results also provide guidance for the operational expansion of post-fire debris-flow early warning systems in areas where empirically defined regional rainfall intensity–duration thresholds do not currently exist.

First posted June 30, 2016 For additional information, contact: Center Director, Geologic Hazards Science CenterU.S. Geological SurveyBox 25046, MS 966Denver, CO 80225-0046http://geohazards.usgs.gov/ Wildfire can significantly alter the hydrologic response of a watershed to the extent that even modest rainstorms can generate dangerous flash floods and debris flows. To reduce public exposure to hazard, the U.S. Geological Survey produces post-fire debris-flow hazard assessments for select fires in the western United States. We use publicly available geospatial data describing basin morphology, burn severity, soil properties, and rainfall characteristics to estimate the statistical likelihood that debris flows will occur in response to a storm of a given rainfall intensity. Using an empirical database and refined geospatial analysis methods, we defined new equations for the prediction of debris-flow likelihood using logistic regression methods. We showed that the new logistic regression model outperformed previous models used to predict debris-flow likelihood.

During the North American Monsoon, low-to-midlevel moisture is transported in surges from the Gulf of California and Eastern Pacific Ocean into Mexico and the American Southwest. As rising levels of precipitable water interact with the mountainous terrain, severe thunderstorms can develop, resulting in flash floods that threaten life and property. The rapid evolution of these storms, coupled with the relative lack of upper-air and surface weather observations in the region, make them difficult to predict and monitor, and guidance from numerical weather prediction models can vary greatly under these conditions. Precipitable water vapor (PW) estimates derived from continuously operating ground-based GPS receivers have been available for some time from NOAA's GPS-Met program, but these observations have been of limited utility to operational forecasters in part due to poor spatial resolution. Under a NASA Advanced Information Systems Technology project, 37 real-time stations were added to NOAA's GPS-Met analysis providing 30-min PW estimates, reducing station spacing from approximately 150 km to 30 km in Southern California. An 18-22 July 2013 North American Monsoon event provided an opportunity to evaluate the utility of the additional upper-air moisture observations to enhance National Weather Service (NWS) forecaster situational awareness during the rapidly developing event. NWS forecasters used these additional data to detect rapid moisture increases at intervals between the available 1-6-h model updates and approximately twice-daily radiosonde observations, and these contributed tangibly to the issuance of timely flood watches and warnings in advance of flash floods, debris flows, and related road closures.

Global Positioning System (GPS) meteorology provides enhanced density, low-latency (30-min resolution), integrated precipitable water (IPW) estimates to NOAA NWS (National Oceanic and Atmospheric Adminis tration Nat ional Weather Service) Weather Forecast Offices (WFOs) to provide improved model and satellite data verification capability and more accurate forecasts of extreme weather such as flooding. An early activity of this project was to increase the number of stations contributing to the NOAA Earth System Research Laboratory (ESRL) GPS meteorology observing network in Southern California by about 27 stations. Following this, the Los Angeles/Oxnard and San Diego WFOs began using the enhanced GPS-based IPW measurements provided by ESRL in the 2012 and 2013 monsoon seasons. Forecasters found GPS IPW to be an effective tool in evaluating model performance, and in monitoring monsoon development between weather model runs for improved flood forecasting. GPS stations are multi-purpose, and routine processing for position solutions also yields estimates of tropospheric zenith delays, which can be converted into mm-accuracy PWV (precipitable water vapor) using in situ pressure and temperature measurements, the basis for GPS meteorology. NOAA ESRL has implemented this concept with a nationwide distribution of more than 300 GPSMet stations providing IPW estimates at sub-hourly resolution currently used in operational weather models in the U.S.

Rainfall intensity–duration (ID) thresholds are commonly used to predict the temporal occurrence of debris flows and shallow landslides. Typically, thresholds are subjectively defined as the upper limit of peak rainstorm intensities that do not produce debris flows and landslides, or as the lower limit of peak rainstorm intensities that initiate debris flows and landslides. In addition, peak rainstorm intensities are often used to define thresholds, as data regarding the precise timing of debris flows and associated rainfall intensities are usually not available, and rainfall characteristics are often estimated from distant gauging locations. Here, we attempt to improve the performance of existing threshold-based predictions of post-fire debris-flow occurrence by utilizing data on the precise timing of debris flows relative to rainfall intensity, and develop an objective method to define the threshold intensities. We objectively defined the thresholds by maximizing the number of correct predictions of debris flow occurrence while minimizing the rate of both Type I (false positive) and Type II (false negative) errors. We identified that (1) there were statistically significant differences between peak storm and triggering intensities, (2) the objectively defined threshold model presents a better balance between predictive success, false alarms and failed alarms than previous subjectively defined thresholds, (3) thresholds based on measurements of rainfall intensity over shorter duration (≤60 min) are better predictors of post-fire debris-flow initiation than longer duration thresholds, and (4) the objectively defined thresholds were exceeded prior to the recorded time of debris flow at frequencies similar to or better than subjective thresholds. Our findings highlight the need to better constrain the timing and processes of initiation of landslides and debris flows for future threshold studies. In addition, the methods used to define rainfall thresholds in this study represent a computationally simple means of deriving critical values for other studies of nonlinear phenomena characterized by thresholds.

Following wildfires, emergency-response and public-safety agencies can be faced with evacuation and resource-deployment decisions well in advance of coming winter storms and during storms themselves. Information critical to these decisions is provided for recently burned areas in the San Gabriel Mountains of southern California. A compilation of information on the hydrologic response to winter storms from recently burned areas in southern California steeplands is used to develop a system for classifying magnitudes of hydrologic response. The four-class system describes combinations of reported volumes of individual debris flows, consequences of debris flows and floods in an urban setting, and spatial extents of the hydrologic response. The range of rainfall conditions associated with different magnitude classes is defined by integrating local rainfall data with the response magnitude information. Magnitude I events can be expected when within-storm rainfall accumulations ( A ) of given durations ( D ) fall above the threshold A = 0.4 D 0.5 and below A = 0.5 D 0.6 for durations greater than 1 h. Magnitude II events will be generated in response to rainfall accumulations and durations between A = 0.4 D 0.5 and A = 0.9 D 0.5 for durations less than 1 h, and between A = 0.5 D 0.6 and A = 0.9 D 0.5 or durations greater than 1 h. Magnitude III events can be expected in response to rainfall conditions above the threshold A = 0.9 D 0.5 . Rainfall threshold-magnitude relations are linked with potential emergency-response actions as an emergency-response decision chart, which leads a user through steps to determine potential event magnitudes and identify possible evacuation and resource-deployment levels. Use of this information in planning and response decision-making process could result in increased safety for both the public and emergency responders.

A portable truck-mounted C-band Doppler weather radar was deployed to observe rainfall over the Station Fire burn area near Los Angeles, California, during the winter of 2009/10 to assist with debris-flow warning decisions. The deployments were a component of a joint NOAA-U.S. Geological Survey (USGS) research effort to improve definition of the rainfall conditions that trigger debris flows from steep topography within recent wildfire burn areas. A procedure was implemented to blend various dual-polarized estimators of precipitation (for radar observations taken below the freezing level) using threshold values for differential reflectivity and specific differential phase shift that improves the accuracy of the rainfall estimates over a specific burn area sited with terrestrial tipping-bucket rain gauges. The portable radar outperformed local Weather Surveillance Radar-1988 Doppler (WSR-88D) National Weather Service network radars in detecting rainfall capable of initiating post-fire runoff-generated debris flows. The network radars underestimated hourly precipitation totals by about 50%. Consistent with intensity duration threshold curves determined from past debris-flow events in burned areas in Southern California, the portable radar-derived rainfall rates exceeded the empirical thresholds over a wider range of storm durations with a higher spatial resolution than local National Weather Service operational radars. Moreover, the truck-mounted C-band radar dual-polarimetric-derived estimates of rainfall intensity provided a better guide to the expected severity of debris-flow events, based on criteria derived from previous events using rain gauge data, than traditional radar-derived rainfall approaches using reflectivity rainfall relationships for either the portable or operational network WSR-88D radars. Part of the reason for the improvement was due to siting the radar closer to the burn zone than the WSR-88Ds, but use of the dual-polarimetric variables improved the rainfall estimation by similar to 12% over the use of traditional Z-R relationships.

Following wildfires, emergency-response and public-safety agencies are faced often with making evacuation decisions and deploying resources both well in advance of each coming winter storm and during storms themselves. Information critical to this process is provided for recently burned areas in the San Gabriel Mountains of southern California. The National Weather Service (NWS) issues Quantitative Precipitation Forecasts (QPFs) for the San Gabriel Mountains twice a day, at approximately 4 a.m. and 4 p.m., along with unscheduled updates when conditions change. QPFs provide estimates of rainfall totals in 3-hour increments for the first 12-hour period and in 6-hour increments for the second 12-hour period. Estimates of one-hour rainfall intensities can be provided in the forecast narrative, along with probable peak intensities and timing, although with less confidence than rainfall totals. A compilation of information on the hydrologic response to winter storms from recently burned areas in southern California steeplands was used to develop a system for classifying the magnitude of the postfire hydrologic response. The four-class system is based on a combination of the reported volume of individual debris flows, the consequences of these events in an urban setting, and the spatial extent of the response to the triggering storm. Threshold rainfall conditions associated with debris flow and floods of different magnitude classes are defined by integrating local rainfall data with debris-flow and flood magnitude information. The within-storm rainfall accumulations (A) and durations (D) above which magnitude I events are expected are defined by A=0.3D0.6. The function A=0.5D0.6 defines the within-storm rainfall accumulations and durations above which a magnitude III event will occur in response to a regional-scale storm, and a magnitude II event will occur if the storm affects only a few drainage basins. The function A=1.0D0.5defines the rainfall conditions above which magnitude III events can be expected. Rainfall trigger-magnitude relations are linked with potential emergency-response actions in the form of an emergency-response decision chart. The chart leads a user through steps to determine potential event magnitudes, and identify possible evacuation and resource-deployment levels as a function of either individual storm forecasts or measured precipitation during storms. The ability to use this information in the planning and response decision-making process may result in significant financial savings and increased safety for both the public and emergency responders.

NOAA/NWS/Office of Hydrologic Development, Silver Spring, MarylandNOAA/OAR/National Severe Storms Laboratory, Norman, OklahomaUSGS, Golden, Colorado; COSTA ANDUSGS, Vancouver, WashingtonNOAA/NWS/WSFO Oxnard, Oxnard, CaliforniaCIRES/University of Colorado and NOAA Earth System Research Laboratory, Boulder, ColoradoNOAA/NWS/San Diego, San Diego, CaliforniaNOAA/NWS/Western Region Headquarters, Salt Lake City, UtahCORRESPONDING AUTHOR: Pedro Restrepo, National Weather Service W/OH, SSMC-2 Room 8176, 1325 East West Highway, Silver Spring, MD 20910, E-mail: pedro.restrepo@noaa.gov

Debris flows generated during rain storms on recently burned areas have destroyed lives and property throughout the Western U.S. Field evidence indicate that unlike landslide-triggered debris flows, these events have no identifiable initiation source and can occur with little or no antecedent moisture. Using rain gage and response data from five fires in Colorado and southern California, we document the rainfall conditions that have triggered post-fire debris flows and develop empirical rainfall intensity-duration thresholds for the occurrence of debris flows and floods following wildfires in these settings. This information can provide guidance for warning systems and planning for emergency response in similar settings.Debris flows were produced from 25 recently burned basins in Colorado in response to 13 short-duration, high-intensity convective storms. Debris flows were triggered after as little as six to 10 min of storm rainfall. About 80% of the storms that generated debris flows lasted less than 3 h, with most of the rain falling in less than 1 h. The storms triggering debris flows ranged in average intensity between 1.0 and 32.0 nun/h, and had recurrence intervals of two years or less. Threshold rainfall conditions for floods and debris flows sufficiently large to pose threats to life and property from recently burned areas in south-central, and southwestern, Colorado are defined by: I=6.5D(-0.7) and I=9.5D(-0.7), respectively, where I=rainfall intensity (in mm/h) and D=duration (in hours).Debris flows were generated from 68 recently burned areas in southern California in response to long-duration frontal storms. The flows occurred after as little as two hours, and up to 16 h, of low-intensity (2-10 mm/h) rainfall. The storms lasted between 5.5 and 33 h, with average intensities between 1.3 and 20.4 mm/h, and had recurrence intervals of two years or less. Threshold rainfall conditions for life-and property-threatening floods and debris flows during the first winter season following fires in Ventura County, and in the San Bernardino, San Gabriel and San Jacinto Mountains of southern California are defined by I=12.5D(-0.4), and I=7.2D(-0.4), respectively. A threshold defined for flood and debris-flow conditions following a year of vegetative recovery and sediment removal for the San Bernardino, San Gabriel and San Jacinto Mountains of I-14.0D(-0.5) is approximately 25 mm/h higher than that developed for the first year following fires.The thresholds defined here are significantly lower than most identified for unburned settings, perhaps because of the difference between extremely rapid, runoff-dominated processes acting in burned areas and longer-term, infiltration-dominated processes on unburned hillslopes. Crown Copyright (C) 2007 Published by Elsevier B.V. All rights reserved.

The effects of precipitable water vapor (PWV) retrievals from the Southern California Integrated GPS Network (SCIGN) on quantitative precipitation forecast (QPF) skill are examined over two flood-prone regions of Southern California: Santa Barbara (SB) and Ventura County (VC). Two sets of QPFs are made, one using the initial water vapor field from the NCEP 40-km Eta initial analysis, and another in which the initial Eta water vapor field is modified by incorporating the PWV data from the SCIGN receivers. Lateral boundary data for the QPFs, as well as the hydrostatic component of the GPS zenith delay data, are estimated from the Eta analysis. Case studies of a winter storm on 2 February during the 1997/98 El Nino, and storms leading up to the La Conchita, California, landslide on 10 January 2005, show notably improved QPFs for the first 3-6 h with the addition of GPS PWV data. For a total of 47 winter storm forecasts between February 1998 and January 2005 the average absolute QPF improvement is small; however, QPF improvements exceed 5 mm in several underpredicted rainfall events, with GPS data also improving most cases with overpredicted rainfall. The GPS improvements are most significant (above or near the 2 sigma level) when the low-level winds off the coast of Southern California are from the southern (SW to SE) quadrant. To extend the useful forecast skill enhancement beyond six hours, however, additional sources of water vapor data over broader areas of the adjacent Pacific Ocean are needed.

(1) NOAA National Weather Service, Silver Spring, MD, USA (Pedro.Restrepo@noaa.gov), (2) United States Geological Survey, Golden, CO, USA (cannon@usgs.gov), (3) NOAA National Weather Service, Oxnard, CA, USA (jayme.Laber@noaa.gov), (4) NOAA Oceanic and Atmospheric Research, Norman, OK, USA (David.P.Jorgensen@noaa.gov), (5) NOAA National Weather Service, Salt Lake City, UT, USA (Kevin.Werner@usgs.gov)