Flash Flood Modeling Using the DHM-TF Approach Current Status and Future Plans

1. Flash Flood Modeling Using the DHM-TF ApproachCurrent Status and Future Plans Brian Cosgrove, Seann Reed, Ziya Zhang, Victor Koren, and Mike Smith NOAA/NWS/OHD

2. Distributed Hydrologic Model-Threshold Frequency (DHM-TF) Approach • DHM-TF system: Combine a distributed hydrologic model with a threshold frequency post-processor to improve flash flood forecasts at ungauged locations • DHM produces gridded flow forecasts, from which gridded frequency forecasts are derived using historical simulations • Historical simulations are conducted using DHM and same type of forcing data used in forecast simulations to maintain consistency • Threshold frequency grids are derived from local information and compared with frequency forecasts for flash flood determination • Sacramento with kinematic routing used in OHD’s implementation, but any hydrologic model can be used

3. DHM-TF Motivation • DHM-TF is able to predict flash flooding at ungauged locations • System provides inherent bias correction to model flow predictions (concept that the relative rank of events—and therefore frequency—can be well-simulated even if the flow is biased) • Superior to standard flash flood guidance • Computed at 2km to 4km resolution versus > 260km2 basin scale • Can be updated every 15 minutes versus 1 to 6 hours • Produces verifiable small basin flow estimates • DHM-TF has the capability of providing flash flood forecasts two hours into the future, thus improving upon current NWS flash flood warning lead times

4. DHM-TF History • 2003: Threshold frequency concept advocated at DOH workshop • 2003: Concept expanded and initial development of software • 2004: Statistical flash flood modeling paper presented at AMS • 2006: DHM/QPF flash flood prediction paper presented at AMS • 2006-2007: DHM-TF simulations conducted over Oklahoma-domain, work is published in Journal of Hydrology • 2007: Discussions with Sterling WFO concerning use of DHM-TF • 2007: Maryland case developed to support Sterling WFO transfer • Now: Brian Cosgrove takes over DHM-TF work, Sterling WFO activities commencing soon, will work with MARFC and OHRFC as well. Discussions with Pittsburgh WFO underway.

5. HPN DHM-TF Overview Forecast Historical Analysis

6. Real-time DHM-TF Prototype Precipitation Observations Precipitation Forecasts MARFC MPE (4km, 1hr, high confidence) HPE (1km, 15min, lower confidence) HPN (4km, 15min) No Precipitation Model States Saved Switch Time T-16 hours T-4 hours Present Time T+2 hours T+6 hours HL-RDHM Run 2 (forecast run) HL-RDHM Run 1 (state update) *Cycle automatically repeated every hour in current setup *Can be set to update forecast run every 15 minutes, using more HPE and less HPN data

7. DHM-TF Maryland Pilot Project • Maryland area well-suited for implementation of prototype DHM-TF • HPE/HPN prototype currently running over this region • Includes Baltimore, a well-studied, high profile, flash flood prone area • Local basins, MARFC, and Sterling WFO are accessible for site visits • Large number of stream gauges are available in Baltimore County for validation • Mix of urban and suburban areas Current: 4 km model with 28 outlets defined Planned: 2 km model with 12 outlets defined

8. Case Study: Tropical Storm Hanna • Tropical Storm Hanna brought heavy rain to the Maryland region September 6th-7th, 2008 • Eight flash flood warnings were issued by the Sterling WFO for areas within Maryland and Virginia • Several local roads and large parkways were shut down due to flash flooding and several rescues and evacuations were necessary Precipitation (inches) associated with Tropical Storm Hanna from 12Z September 6th, 2008 through 12Z September 7th, 2008

9. Tropical Storm Hanna Maximum return period values from DHM-TF simulated during the period of time from 16Z on September 6th through 02Z on September 7th 2008. Spotter-confirmed flash flood events are indicated by the Blue wave symbols Depiction of the areas covered by the three Tropical Storm Hanna flash flood warnings (outlined in Red) issued by the Sterling WFO that lie within the Maryland DHM-TF case study domain. Figure also depicts maximum return period values from DHM-TF covering a similar time period.

10. Return Period (Years) Sample DHM-TF Output (TS Tammy, 10/2005) Grass GIS Images Return Period (Years) Prototype Google Earth Images

11. Additional Maryland DHM-TF Case Study • Multiple rain events, May 9th-12th, 2008 • Several flash flood warnings issued by Sterling WFO for Washington DC and surrounding counties on May 9th • Tornado warnings for Prince Georges and Charles Counties in Maryland

12. Maryland Case Study: May 12-13 2008 Sacramento UZFWC (fraction) 12Z May 12, 2008 Sacramento UZFWC (fraction) 00Z May 12, 2008 = Forecast Gauge Locations Maximum Discharge (m3/s) May 12-13, 2008 Average Recurrence Interval (Years) May 12-13, 2008 Frequencies are derived from routed flows, demonstrating capability to forecast floods in ungauged locations downstream of rainfall.

13. Assessing the Inherent DHM-TF Bias Correction (Are flow ranks and frequencies correct even if flow magnitudes are biased?) Dutch Mills Basin Flow Frequency Distributions Compute peak flow adjustments at gauged validation points using observed and modeled flow frequency distributions 2 1 Dutch Mills Basin 3 • Supports concept that frequency is well-simulated even if flows are biased • Potential exists for implementing DHM-TF approach even if model is not well calibrated Adjusted Raw simulation

14. Bias Correction of Precipitation Monocacy at Jug Bridge (2116 km2) Cumulative Bias, Monocacy River at Jug Bridge (2100 km2) • Bias detected in MARFC MPE archives prior to 2004 • Bias corrected precipitation needed to support unbiased simulation statistics for a reasonable historical period (~10 years) • Analysis of Monocacy River flow shows reduction in cumulative bias and improved consistency when bias corrected precipitation is used • Consistent bias can be removed through calibration or through DHM-TF approach

15. Bias Correction of Precipitation Monthly RFC MPE Precipitation 03/97 (mm) Monthly PRISM Precipitation 3/97 (mm) RFC Hourly MPE Precipitation 03/01/97 12z (mm) Adjusted RFC Hourly MPE Precipitation 03/01/97 12z (mm) Monthly Bias (ratio) Yu Zhang

16. Observations Original Original Re-analysis Reanalysis Impact of Precipitation Bias Correction Improved precipitation input leads to improved simulation of stream flow Monocacy River Flow: April 12 through June 11, 2003

17. Prototype Tools Available from the DHM-TF Project • Several new routines were needed for DHM-TF approach • monthlySum: generates gridded monthly total precipitation • xmrgAdjust: bias corrects hourly precipitation data with PRISM • AnnPeaks: produces annual grids of maximum peak discharges • FreqParams: reads annual peak files, computes frequency distribution at each grid cell and outputs distribution parameters • getMaxRet: computes maximum discharge and return period during forecast period • GRASS and GrADS scripts for visualizing input and output. Google Earth and AWIPS being investigated. • Cron scripts for real-time prototype • Full set of DHM-TF components available through LAD

18. Future Work • Continue pilot real-time DHM-TF runs at OHD • Increase run resolution to 2km (1/2 HRAP) over Maryland domain • Add Snow-17 into prototype DHM-TF system • Refine graphical tools and scripts needed for output visualization at RFCs and WFOs • Coordinate with Sterling and Pittsburgh WFOs, and MARFC and OHRFC on use of experimental DHM-TF system • Work to increase portability of system (parameter and forcing derivation, expansion to additional RFCs)