A Distributed Flash Flood Forecasting Rainfall-Runoff Model Applied to Watersheds in the Northeast United States Michael Schaffner NOAA National Weather Service WFO Binghamton, NY Carl Unkrich USDA ARS Tucson, AZ NROW November 6, 2008
Jan 02, 2016
A Distributed Flash Flood Forecasting Rainfall-Runoff Model Applied to Watersheds in the Northeast United States
Michael SchaffnerNOAA National Weather ServiceWFO Binghamton, NY
Carl UnkrichUSDAARS Tucson, AZ NROW
November 6, 2008
Flash Flood Monitoring and Prediction Program (FFMP)
• This allows the forecaster to pinpoint (i.e. provide site specific information) which basins are receiving the heaviest rainfall.
• Does not allow the forecaster in most cases to determine the timing and magnitude of an event.
Above image from: http://www.tucson.ars.ag.gov/kineros/
Plan View Schematic View
• Distributed Model composed of a cascade of planes and channel elements.• Takes into account spatial distribution of soil type, land use, topography, and
rainfall.• Kinematic wave equations for channel routing.• Model functions in real-time with radar interface.• Allows the user to run ensembles to investigate either multiple Z-R
relationships or variability in model parameters.
Outlet
Bush Kill near Arkville
Setting up KINEROS
• Delineate watershed boundary and interior sub-watersheds using a GIS application.
• Collect in the field stream manning roughness values.
• Collect channel width measurements from field work or using high resolution remote sensing.
Model Calibration
• Each event was calibrated for the magnitude and the timing of the peak flow.
• Calibration was accomplished by adjustment of multipliers that globally adjust individual model parameters.
• Saturated Hydrologic Conductivity (Ks) multiplier was utilized to calibrate the magnitude of the peak flow.
• Channel length multiplier was utilized to calibrate the timing of the peak flow.
Platte Kill near Dunraven
• 34 square miles.• Catskill mountains.• USGS stream gage. Outlet
Typical Hydrograph for Platte Kill
• Reasonable simulation of primary and secondary peak magnitude and timing.
• Later and more rapid rate of rise than seen in the gage record.
• More rapid rate of fall than seen in the gage record.
Platte Kill near Dunraven
0
500
1000
1500
2000
2500
3000
7/23/200812:00
7/23/200814:24
7/23/200816:48
7/23/200819:12
7/23/200821:36
7/24/200800:00
Date and Time
Discharge (cfs)
Gage
Simulation
Additional Platte Kill Hydrographs
Platte Kill near Dunraven
0
200
400
600
800
1000
1200
1400
1600
1800
6/26/2006
07:12
6/26/2006
09:36
6/26/2006
12:00
6/26/2006
14:24
6/26/2006
16:48
6/26/2006
19:12
6/26/2006
21:36Date and Time
Discharge (cfs)
Gage
Simulation
Platte Kill near Dunraven
0
100
200
300
400
500
600
700
800
3/7/2008
14:24
3/7/2008
16:48
3/7/2008
19:12
3/7/2008
21:36
3/8/2008
00:00
3/8/2008
02:24
3/8/2008
04:48
3/8/2008
07:12
3/8/2008
09:36
3/8/2008
12:00
3/8/2008
14:24
Date and Time
Discharge (cfs)
Gage
Simulation
Platte Kill near Dunraven
0
200
400
600
800
1000
1200
3/8/200812:00
3/8/200814:24
3/8/200816:48
3/8/200819:12
3/8/200821:36
3/9/200800:00
3/9/200802:24
3/9/200804:48
Date and Time
Discharge (cfs)
Gage
Simulation
Platte Kill near Dunraven
0
100
200
300
400
500
600
700
8/28/200619:12
8/29/200600:00
8/29/200604:48
8/29/200609:36
8/29/200614:24
8/29/200619:12
8/30/200600:00
Date and Time
Disc
harg
e (c
fs)
Gage
Simulation
Hydrographs point to the need for an additional model component
• Discussions with USDA and Peter Troch (UA) on the addition of his shallow-sub-surface lateral flow model (hsB) into KINEROS.
• We are actively seeking funding to incorporate this model into KINEROS.
Platte Kill Calibration for Operational Use
• Poor R-square when all events are taken into account.
• Decided to consider only those events where the basin average rainfall would have hydrological significance.
Platte Kill near Dunraven
R2 = 0.02
0.00
0.50
1.00
1.50
2.00
2.50
0 2 4 6 8 10 12
Ks Multiplier
Channel Length Multiplier
Platte Kill Calibration for Operational Use Continued
• Good trend in both channel length and Ks multipliers for events of 1 inch or greater basin average rainfall.
• Good R-square when only considering events such events.
Modeling Un-gauged Basins
• Modeling of un-gauged basins represents a significant challenge and operational need within the NWS.
• Watersheds in the Town of Colchester, Delaware County, NY were impacted by flash floods on June 19, 2007 and July 23, 2008.
• Time and magnitude of the peak flow and associated impacts were determined.
• Watersheds were calibrated to these flash flood events.
Spring Brook near Roscoe• 9 square miles• 4 lives lost on June 19,
2007• Multiple sections of
Route 206 impassible• Several homes
washed off their foundations
Spring Brook Hydrograph for June 19, 2007
• Flash flood warning issued at 7:01 PM.
• Major flash flooding and loss of life reported at 7:50 PM.
• KINEROS first began to forecast major flash flooding at 5:30 PM.
Berry Brook near Roscoe
• 4.5 square mile
Plane View
Indirect discharge measurement cross section for June 19, 2007 flash flood event located along Berry Brook Road
Berry Brook Hydrograph for June 19, 2007
• Flash flood warning issued at 7:01 PM.
• Major flash flooding and multiple wash- outs along Berry Brook Road were reported at 8:30 PM.
• KINEROS first began to forecast major flash flooding at 6:10 PM.
Future Work
• Test KINEROS on larger sized basins (200 – 250 square miles).
• Add ability for the forecaster to add QPF.• Have undergraduate intern work with KINEROS over
summer 2009.• Produce NWS technical publication on results.• Test KINEROS in areas outside of the Binghamton
forecast area.• Add shallow groundwater flow model component (funding
required)• Simple snow model (long term).
Questions / Comments?