1

Evaluation of the Areal Mean Basin Estimated Rainfall (AMBER) FlashFlood Algorithm at the Tulsa, OK and Sterling, VA NWSFOs

Ami ArthurWestern Intermountain Storms and Hydrometeorology TeamCooperative Institute for Mesoscale Meteorological Studies

National Severe Storms LaboratoryNational Oceanic and Atmospheric Administration

February 8, 2000

ce witheved

and

2

ACKNOWLEDGMENTS

We would like to thank the Sterling, VA and Tulsa, OK NWSFOs, the Middle AtlanticRiver Forecast Center, and the Arkansas-Red Basin River Forecast Center for their assistanthis evaluation. They provided excellent guidance in the selection of case studies and retripertinent archived data for each flash flood event. We would especially like to thank SteveZubrick and Melody Paschetag of the Sterling NWSFO, Eric Howieson of the Tulsa NWSFO,Jason Nolan of the Middle Atlantic River Forecast Center for all of their efforts.

......

.......

....

........12........15.......17......21........24

.....

........28

.......31

........34

.......38

.

........44

..

..

..

..

3

TABLE OF CONTENTS

Page

EXECUTIVE SUMMARY........................................................................................................5

Introduction......................................................................................................................................7

Objective..............................................................................................................................8Background.......................................................................................................................8Methodology....................................................................................................................9

Tulsa CWA Case Studies............................................................................................................11

Case Study #1 - January 4, 1998................................................................................Case Study #2 - September 14, 1998..........................................................................Case Study #3 - October 5-6, 1998.............................................................................Case Study #4 - April 25, 1999....................................................................................Case Study #5 - June 20, 1999...................................................................................

Sterling CWA Case Studies.......................................................................................................27

Case Study #1 - June 27, 1995...................................................................................Case Study #2 - October 20-21, 1995.........................................................................Case Study #3 - January 19, 1996..............................................................................Case Study #4 - July 1-2, 1997....................................................................................

Results and Discussion..................................................................................................................40

Conclusions and Recommendations......................................................................................

REFERENCES..............................................................................................................................47

APPENDIX ATulsa Case Study #1 - January 4, 1998DATA AND FIGURES................................................................................................................48

APPENDIX BTulsa Case Study #2 - September 14, 1998DATA AND FIGURES................................................................................................................62

APPENDIX CTulsa Case Study #3 - October 5-6, 1998DATA AND FIGURES................................................................................................................71

APPENDIX DTulsa Case Study #4 - April 25, 1999DATA AND FIGURES................................................................................................................87

..

..

..

..

..

4

APPENDIX ETulsa Case Study #5 - June 20, 1999DATA AND FIGURES..............................................................................................................102

APPENDIX FSterling Case Study #1 - June 27, 1995DATA AND FIGURES..............................................................................................................118

APPENDIX GSterling Case Study #2 - October 20-21, 1995DATA AND FIGURES..............................................................................................................134

APPENDIX HSterling Case Study #3 - January 19, 1996DATA AND FIGURES..............................................................................................................144

APPENDIX ISterling Case Study #4 - July 1-2, 1997DATA AND FIGURES..............................................................................................................153

rmsling,ccu- pur-in

tudiesist in

asedval-gage/townrma-

s ispre-

BRt

ow-ateflashter hasition,

no

ecase

nrempo-

was

with an

mber

5

EXECUTIVE SUMMARY

This project involves the evaluation of the Areal Mean Basin Estimated Rainfall(AMBER) flash flood algorithm, which was implemented as part of the National Severe StoLaboratorys (NSSL) Warning Decision Support System (WDSS) in the Tulsa, OK and SterVA National Weather Service Forecast Offices (NWSFOs) in 1998. The AMBER program amulates rainfall on the basin level to alert forecasters to the potential for flash flooding. Thepose of this study was to conduct an objective evaluation of the AMBER algorithm and obtaquantitative feedback on its performance. AMBER output was analyzed for several case sin the Tulsa and Sterling county warning areas (CWAs) to determine its utility as a tool to assflash flood warning decisions.

With guidance from the two NWSFOs, five case studies in the Tulsa CWA and four cstudies in the Sterling CWA were identified. Pertinent data was obtained, including archiveLevel II data from the KINX and KLWX radars, archived county flash flood guidance (FFG) ues from the Middle Atlantic and Arkansas-Red Basin River Forecast Centers, archived raindata from the Sterling NWSFO and the Arkansas-Red Basin River Forecast Center, and citypopulation data from the Geographic Names Information System (GNIS). Storm event infotion and flash flood reports were taken from theStorm Data and Unusual Weather Phenomenapublication.

For each case study, a description of the event and pre-existing hydrologic conditionprovided. The AMBER output is then analyzed, including a discussion of the accuracy of thecipitation estimates, the utility of using the average basin rainfall (ABR) rate and/or ratio of Aaccumulation to FFG value to determine flash flood potential, variation of results on differenbasin scales, and prediction errors.

Overall, the AMBER output provided good information about flash flood potential. Hever, it should be noted that AMBER output is only as good as the radar precipitation estiminput on which it is based. When a radar is overestimating or underestimating precipitation,flood guidance or other threshold guidance values become meaningless unless the forecasaccess to reliable ground truth and can adjust the precipitation estimates accordingly. In addin areas where there is beam blockage or incomplete radar coverage, AMBER is of little oruse.

In general, it is more beneficial to monitor ABR rate rather than the ratio or differencbetween ABR accumulation and FFG or other threshold guidance value. In each of the ninestudies, most or all of the reported flash floods occurred prior to the time ABR accumulatioexceeded FFG. However, in most cases the ABR rate reached a significant value well beforeported flash flooding. Therefore, it was concluded that ABR rate is the most important conent of AMBER to produce timely warnings.

Basin size was examined to determine the utility of including various basin scales. Itconcluded that basins delineated with a minimum drainage area threshold less than 10 mi2 wouldbetter capture the high rainfall rates and accumulations in comparison to basins delineatedminimum drainage area threshold of 10 mi2 or greater. While accumulation and rate informatioin larger basins is useful to forecasters, basins smaller than 10 mi2 are recommended for flashflood forecasting purposes.

Perhaps the most significant problem associated with the AMBER output was the nu

alseteereas

Sec-who

ofabilitymini--

ts

6

of flash flood false alarms. However, there are several factors which may result in these falarms. First, flash flood verification is a difficult if not impossible task. There is no guaranthat because a flash flood was not reported it did not occur. This is particularly true in rural awhere low population density increases the chance that a flash flood will not be observed. ond, this evaluation of the AMBER algorithm was strictly objective. Experienced forecastersare familiar with the terrain, hydrologic conditions, and flash flood prone basins in their areawould likely have eliminated many of the false alarms.

Ultimately, AMBERs operational success will depend on three factors: 1) the qualitythe basin delineation, 2) the accuracy of the precipitation estimates, and 3) the forecastersto interpret the output meaningfully. It is recommended that basins are delineated using a