1. Improving Flash Food Prediction in Multiple Environments1
Using a Continuous Hydrologic Model in Support of Flash Flood Predictions
Patrick D. Broxton
Peter A. Troch, Michael Schaffner, Carl Unkrich, David Goodrich, Hoshin Gupta, Thorsten Wagener, Soni Yatheendradas

2. Motivation: Considerations for Modeling Extreme Streamflow Events
What is a catchment’s ability to absorb precipitation?
Precipitation
Runoff
Baseflow
Infiltration
Wet
Dry
Warm
Cool
Low Potential ET
Less Water in Storage
More Water in Storage
High Potential ET

3. Motivation: Considerations for Modeling Extreme Streamflow Events
What is the “true” precipitation input?
Rain Gauges
Radar
Satellite Observations
More Accurate
Less Accurate
Less Coverage
More Coverage
Large Scale
Small Scale
What about Snow?

4. SM-hsB Overview Soil Moisture – hillslope Bousinesq Model4
Soil Moisture – hillslope Bousinesq Model
Water and energy balance at the land surface
Land Surface Module
Incorporates Snow
Transmission Zone
Root Zone
hsB Aquifer
Deep Aquifer
Infiltration ET
Subsurface Module
Root zone water balance
Lateral transport of soil water
1) Keep track of the hydrologic state between flood model runs
2) Distributed so that it can account for spatial variability of terrain and
atmospheric forcing

5. 5
Study Sites

6. Study Sites – New York Watersheds6
Five watersheds in New York’s Catskill Mountains:
Humid catchments that are focus of current efforts
a) W. Branch Delaware River
(332 sq mi)
b) W. Branch Delaware River
(134 sq mi)
c) Platte Kill
(35 sq mi)
d) East Brook
(25 sq mi)
e) Town Brook
(14 sq mi)

7. Study Sites – Arizona Watersheds7
Three watersheds in southeastern Arizona:
Semi-arid catchments to compliment humid catchments
a) Sabino Canyon
(35.5 sq mi)
b) Rincon Creek
(44.8 sq mi)
c) Walnut Gulch
(57.7 sq mi)

8. Runoff Coefficient (Q/P) Runoff Coefficient (Q/P)
Hydrology of New York Watersheds 8
Month
Runoff Coefficient (Q/P)
0.4
0.8
1.6
1.2
Delaware River (Walton)
Delaware River (Delhi)
East Brook
Town Brook
Plate Kill
Jan
Dec
Nov
Oct
Sep
Aug
Jul
Jun
May
Apr
Mar
Feb
Longitude (degrees)
42.2
42.3
42.4
42.5
Latitude (degrees)
75.2
-75
-74.8
-74.6
1050
1100
1150
1200
New York Basins
Longitude (degrees)
31.8 32
32.2
32.4
Latitude (degrees)
-111
-110.6
-110.2
-109.8
PRISM – Average Yearly Precipitation (mm)
300
400
600
500
700
800
900
Month
0.2
0.6
0.4
0.8
Sabino Canyon
Rincon Creek
Walnut Gulch
Jan
Dec
Nov
Oct
Sep
Aug
Jul
Jun
May
Apr
Mar
Feb
Runoff Coefficient (Q/P)
Arizona Basins
Date

9. 9
Modeling with SM-hsB

10. Land Surface Module - Overview10
Atmospheric Inputs:
Shortwave Radiation
Longwave Radiation
Precipitation
Temperature
Pressure
Humidity
Wet Canopy Evaporation/
Snow Interception
Trees
Precipitation
Long Wave Radiation
Infiltration/Runoff
Shortwave
Radiation
Wintertime
Snowpack
Variable Canopy Cover
Near-Surface
Soil Layer
Stream
Fully distributed, runs on hourly timesteps (diurnal cycle is important)
Based on energy balance principles – similar to Utah Energy Balance Model

11. Land Surface Module - Calibration11
Can be run at a point:
e.g. Calibrate to a point measurement such as a snow pillow
...or over an area:
e.g. Calibrate over an area to remotely sensed data or to a data assimilation system
Photo courtesy Jim Porter at NYCDEP
Over a multi-year span, it is generally tuned to compare well with SNODAS, but for specific years, it can be refined using other measurements
SM-hsB SWE (mm)
SNODAS SWE (mm)
R2 = 0.81 20 40 60 80
100
120
140
SWE (mm)
1/1/2007
4/1/2007
1/1/2008
4/1/2008 40 80
120

12. Land Surface Module - Simulation12
Preliminary results for Snow Season in W. Branch Delaware River Watershed
All precipitation inputs are derived from the MPE
January 15,2010
100 mm
50 mm
SWE (mm)
12/1/2009
1/1/2010
2/1/2010
3/1/2010
Date 40 80
120
160
200
4/1/2010
January 25,2010
February 15,2010
February 28,2010

13. Subsurface Module - Overview13
Root Zone Water Balance / Baseflow ET
Infiltration
Runoff
Root Zone
Streamflow Routing
Transmission Zone
hsB Aq. Baseflow
hsB Aquifer
Deep Aquifer
Deep Aq. Baseflow
Semi distributed, runs on daily or hourly timesteps

14. Streamflow/Baseflow/Runoff (mm)
Subsurface Module - Calibration 14
Calibration procedure relies on a baseflow separation
Portions of the model are reconstructed from the steamflow signatures (hydrology backwards)
Deep Aquifer
HSB Aquifer
Streamflow/Baseflow/Runoff (mm)
Log(Streamflow-mm)
Baseflow
Streamflow
Runoff
1/1/2005
4/2/2006
7/3/2007 5 10 15
10/1/2008
12/31/2009 20 30 40 50 60 70
Effective Time (days)
Date 35 1 2 -1 25
Calibration procedure based on that developed by Gustavo Carrillo and Peter Troch at the University of Arizona

15. Normalized Streamflow Generation log(Streamflow – mm/day)
Subsurface Module - Simulation 15
Simulation for Delaware River (Walton) using MPE as input
Model
Data
1/1/2005
4/2/2006
7/3/2007
10/1/2008
12/31/2009 20 40 60
Streamflow (mm/day) 5 10 15
Baseflow (mm/day)
Normalized Streamflow Generation
Normalized Water Year Precipitation
0.2
0.4
0.6
0.8 1
Data
Model
log(Streamflow – mm/day) 20 40 60 80
100
10-1
101
Probability of Exeedance
Data
Model
Catchment
NSE Baseflow
NSE Streamflow
Delaware River (Walton)
0.61
0.34
Delaware River (Delhi)
0.62
0.10
East Brook
0.58
0.48
Town Brook
0.65
0.41
Platte Kill
Catchment
NSE Baseflow
NSE Streamflow
Sabino Canyon
0.10
0.41
Rincon Creek
-3.34
-0.96
Walnut Gulch
No Baseflow

16. Benefits of Modeling With SM-hsB16
Yields many useful modeled quantities for flood forecasting
Baseflow
Model
Data
Modeled Soil Moisture
Modeled Transpiration
1/1/2005
4/2/2006
7/3/2007
10/1/2008
12/31/2009
BF (mm/day)
SM (%)
Transp. (mm/day) 20 10 30 40 5
Storage (mm)
hsB Aquifer Storage
Discharge (mm) 15
Initial Conditions
Modeled soil moisture
Modeled water storage
Potential and actual evapotranspiration
Aquifer depth
Precipitaiton Estimates
Modeled SWE
Snow and Snowmelt
Storage-discharge relationships that can be inverted to estimate precipitation from streamflow

17. Summary 17
hsB-SM has been implemented in all NY watersheds, most AZ watersheds
Snow module reproduces wintertime snowpacks; subsurface module works well in the W. Branch Delaware River Basin
Model yields useful information such as snowmelt rates, estimates of catchment “wetness”, and can be useful for estimating rainfall/snowmelt from streamflow response
Although it has not yet been coupled with a flash flood model (KINEROS2), statistical combinations of rainfall and e.g. soil moisture suggest that there is information to be gained from using model data
Correlation with Flood Size - Top 10 Events
Total Precip
Soil Moisture
Combined
Deleware River (Walton)
0.80
0.43
0.88
Deleware River (Delhi)
0.65
0.59
0.82
East Brook
0.02
0.00
0.06
Town Brook
0.68
0.01
Platte Kill
0.48
0.05
0.72
AVERAGE
0.52
0.22
0.66

18. 18 18
Acknowlegements
Funding comes from a COMET grant (UCAR Award S )
Special thanks to Mike Schaffner, Peter Troch, Gustavo Carrillo, Jim Porter, Glenn Horton, and others

19. 19 19
Questions