National Weather Service Flash Flood Modeling and Warning Services Seann Reed, Middle Atlantic River Forecast Center Peter Ahnert, Middle Atlantic River Forecast Center August 23, 2012 USACE Flood Risk Management and Silver Jackets Workshop Harrisburg, PA
The Flash Flood Problem Intense rainfall, ice jam, or dam failure Rapid rise Danger!
What we do Collect data Analyze data and forecast Issue timely watches and warnings Ongoing Challenges: TIMELINESS, LOCATION SPECIFICITY, UNGAUGED STREAMS, VERIFICATION
Topics Observed and forecast rainfall calculations Hydrologic modeling Verification Communicating warnings, outreach The path forward, emerging technologies
RFC WFO Roles in Flash Flood Services RFCs prepare Flash Flood Guidance and Site Specific Models RFCs WFOs use flash flood monitoring tools along with rainfall observations, forecasts and monitoring tools to issue flash flood watches and warnings when appropriate WFOs
Observed and Forecast Rainfall
Observed and Forecast Rainfall Rain gauge data Radar data Satellite data Multi-sensor rainfall products QPF guidance
Rain Gauge Data NWS cooperative observers (daily) CoCoRAHS (daily) Telemetered equipment (hourly, sub-hourly) E.g. AFWS/IFLOWS (regional) Automated Data Collection Platforms (DCPs) from multiple state and federal agencies HADS national processing, acquisition and distribution of DCP data Stations with hourly or sub-hourly reporting are most useful for flash flood modeling 8
Automated Flood Warning Systems (AFWS) Using Integrated Flood Observing and Warning System (IFLOWS) Technology Joint effort between NWS and states 1663 rain gauges 12 states 5 minute radiotelemetry reporting http://afws.erh.noaa.gov/afws/national.php
Real time data acquisition, processing and distribution. Data Collection Platforms (DCPs) operated by more than 100 collaborators 14,700 data collection locations
Satellite and Radar Rainfall Estimates Satellite 24-hour precipitation Ending 1200 UTC, 5 Jan 2007 More complete spatial coverage than gauge or radar estimates Radar 24-hour precipitation Ending 1200 UTC, 5 Jan 2007 Generally higher accuracy than satellite estimates
Hourly Quantitative Precipitation Estimate (QPE) from Multi-Sensor Precipitation Estimator (MPE) for RFCs Quality-Controlled Precipitation Gage Measurement GOES Satellite Estimate MPE Software Radar Estimate Climate patterns Forecaster Analysis Highest quality NWS precipitation Manual quality control Latency too long for very flashy events 1 hour, 4 km resolution grids for river forecasts
Characteristics of Rain Gauge and Radar Estimates MPE Gauge-only Rainfall (~2.3 in.) MPE Radar-only Rainfall (~1.0 in.) Bias-adjusted Radar Rainfall (~1.7 in.) MPE Multisensor Rainfall (~2.2 in.)
High Resolution Precipitation Estimator (HPE); High Resolution Precipitation Nowcaster (HPN) for WFOs HPE HPN Rainfall accumulations at <5 minute intervals, 1km spatial resolution, < 1 hour latency (better than MPE but no manual QC) Uses radar-rainfall estimates from multiple radars Bias correction with recent MPE gauge/radar information Extrapolation forecasts using lag correlation pattern matching 15 minute temporal resolution, 4km spatial resolution, 1-2 hour forecast Production Produced in AWIPS environment at each WFO Lead time = rainfall forecast period + basin response time
MPE Precipitation (mm) 23Z April 21st to 00Z April 22 nd 2009 HPE Precipitation (mm) 23Z April 21st to 00Z April 22 nd 2009 Observations Hour 1 HPN Forecast (mm) 23Z April 21st to 00Z April 22 nd 2009 Hour 2 HPN Forecast (mm) 23Z April 21st to 00Z April 22 nd 2009 Forecasts 1 hour forecast 2 hour forecast
Probabilistic Precipitation Guidance from NCEP s Available every 6 hours out 3 days Good for situational awareness Can be used with Flash Flood Guidance (discussed next) No explicit spatial temporal patterns cannot be used for high resolution hydrologic modeling Probability of rainfall exceeding 0.25 in a given 6 hour period
Hydrologic Modeling for Flash Floods (Current Practice)
How much rain is needed for Flash Flooding? Rainfall needed dependent on: Antecedent conditions Basin characteristics Threshold Runoff Depth of runoff required to exceed flood flow in a channel (over a certain time and drainage area) Flash Flood Guidance Depth of rain required to exceed flood flow in a channel (over a certain time and drainage area)
Flash Flood Guidance (FFG) Curves 5 Threshold Runoff Cumulative Runoff (in) 4 3 2 SMI = 0.5 1.5 2.0 2.5 3.5 3.0 1 0 0 1 2 3 4 5 Flash Flood Cumulative Guidance Rainfall (FFG) (in) (in) In this basin, FFG ranges from 2 to more than 4 inches depending on antecedent moisture
Flash Flood Guidance (FFG) Products Headwater FFG Gridded FFG Issued by RFCs for different rainfall durations Updated 1-4 times daily
FFG Pros and Cons Pros Easy to use product used by WFOs and other outside organizations Can use with probabilistic precipitation forecasts Used internationally w/ satellite data Central America / Caribbean (Hydrologic Research Center) Cons Methodology inconsistencies exist among neighboring RFCs West rainfall rate more important than soil moisture Gridded FFG does not include routing Difficult to verify 21
Flash Flood Monitoring and Prediction (FFMP) software allows Forecasters to see current radar derived rainfall rates, rainfall totals, and compares data to Flash Flood Guidance. This software helps forecasters identify areas of possible flooding sooner. Add image from Pete J and text.
Hydrologic Modeling for Flash Floods, Moving Forward Distributed hydrologic modeling combining physical, conceptual, statistical Leveraging Community Hydrologic Prediction System (CHPS) technology Interagency collaboration
Lumped Versus Distributed Models Distributed models are well-suited for flash flood prediction and monitoring, offering high-resolution streamflow at outlet and interior points with ability to route flow Lumped Distributed 1. Rainfall and soil properties averaged over basin 2. Single rainfall/runoff model computation for entire basin or subbasin 3. Prediction/verification only at outlet point 1. Rainfall, soil properties vary by grid cell 2. Rainfall/runoff model applied separately to each grid cell 3. Prediction/verification at any grid cell
From Soil Properties to SAC-SMA Model Parameters (Koren et al., 2000; Koren et al., 2001; Anderson et al., 2005) Hydrologic Soil Group UZTWM for SAC Model Surface Texture UZFWM / UZTWM 25 4. Parameterization
Distributed Hydrologic Model Threshold Frequency Approach (DHM-TF) 4 Times on 1/4/1998 14 UTC 15 UTC Frequencies are derived from routed flows on a 2 km network. 16 UTC 17 UTC 26 6. Statistical-distributed
Community Hydrologic Prediction System (CHPS) http://www.nws.noaa.gov/ohd/hrl/chps/index.html New operational hydrologic modeling software architecture Easier to leverage models and data from other groups in operations Client server architecture Configurable/flexible GUIs may allow WFO to run flash Client-server architecture FEWS Models flood hydrology models maintained on the RFC CHPS systems. CHPS Other Models FEWS FC NWS Models USACE Models
Recent Interagency Collaboration on Dam Break Forecasting NWS GeoSMPDBK GIS Pre-processor for SMPDBK Produces SMPDBK input file FEMA GeoDam- BREACH Geospatial Dam Break, Rapid EAP, Consequences, and Hazards Includes GeoSMPDBK functionality and much more... inundation maps + velocity maps + time to peak maps EAPs other
NWS Use of Unsteady HEC-RAS Models 3200 mi 2 at outlet/coastal boundary 2183 mi 2 upstream Current HEC-RAS model applied to mainstem only Experimental HEC-RAS applications to smaller basins (~ 100 km2 ) will allow flood forecast mapping at flash flood scales
Communicating warnings, outreach
Communicating the Warning Message NO BENEFIT unless the end user gets the warning in time and understands it so they can act! End users Communication Mechanisms Homeowners Business owners Drivers Police/emergency responders EMT Transportation Department Local Emergency Official recreational boaters etc....
Flood Safety Education Best Practices Check for the latest forecast Get Insurance http://www.weather.gov/floodsafety/ index.shtml
Flood Safety Awareness - Outreach
Verification
Flash Flood Verification Data (all options have limitations) NWS Storm Database reports from local observers, officials, police, etc. Gauge-based verification Severe Hazards Analysis and Verification Experiment (SHAVE) Gourley, Jonathan J., Jessica M. Erlingis, Yang Hong, Ernest B. Wells, 2012: Evaluation of Tools Used for Monitoring and Forecasting Flash Floods in the United States. Weather and Forecasting, 27, 158 173. 35
The use of FFMP has improved the quality and timeliness of warnings A study conducted in 2010 compared national verification statistics Study looked at data from 1997-2000 (pre-ffmp) to those from 2006-2009 (post FFMP installation). Years of Study Probability of Detection (POD) % of warnings with > 0 min lead time Average lead time for warnings False Alarm Ratio (FAR) 1997-2000 0.85 66% 47 min 0.42 2006-2009 0.91 79% 64 min 0.56
Example Verification for Distributed Hydrologic Modeling Threshold Frequency Technique Binghamton WFO, October 1 st, 2010 Flood Event Excellent spatial agreement among areas of 2+ year return periods and local storm reports Isolated areas with > 100 year return periods # s = USGS Gauge Return Period = Local Storm Reports = NWS Storm Reports
The Path Forward Regional pilot projects - build a little, test a little, field a little MARFC flash flood improvement pilot planned Larger nationally coordinated projects IWRSS, National Water Center
Getting Ready for Weather Ready Nation Flash Flood Pilot DHM-TF pilot studies are underway in coordination with NWS Weather Forecast Offices (WFOs) and River Forecast Centers (RFCs) Binghamton and Pittsburgh WFO domains on WFO servers Baltimore/Washington WFO domain on OHD server (transitioning to WFO) Pittsburgh, Binghamton, and Balt/Wash WFO Domains 57,500 km 2 Binghamton 89,000 km 2 Pittsburgh 60,000 km 2 Balt/Wash
NEW! National Distributed Hydrologic Models Experimental 08/11/2012 @12z Surface Soil Saturation: 4 km Grid New inputs for decision support National gridded flash flood guidance
National gauge-based verification: A future possibility? USGS stations with Area < 100 mi 2 [2781 stations] Many gauges exist for small streams but minor, moderate, and major flood levels are not defined Average Recurrence Interval Associated w/flooding # of historical peaks available at each point since 1986 Could make better use of small gauges, national distributed hydrologic models, and frequency thresholds for systematic verification
Summary Keys to life saving flash flood forecasts Timeliness, lead time Model accuracy and resolution Warning and response Education and outreach NWS Improvements since 1980s have led to improved precipitation measurements, QPF, hydrologic modeling, forecaster analysis tools, warning dissemination methods, and education and outreach ( BUT DIFFICULT TO QUANTIFY IMPROVEMENT)
What s Next? New technologies and modeling techniques to leverage in next 5 10 years High resolution distributed models CHPS Dual pol radar Combine multisensor-extrapolative and higher-resolution meteorological outputs to develop more precise and accurate 0-6 hour precipitation forecasts Innovative verification Hydraulic models models for smaller streams public private partnerships Uncertainty estimates via ensembles IWRSS establishes formal program at the national level to facilitate interagency collaboration
Thank You! Contact Information: Seann Reed Middle Atlantic River Forecast Center 328 Innovation Blvd. (STE 330) State College PA, 16803 Email: seann.reed@noaa.gov