Frozen Ground Frozen Ground ModelModel

Cold Regions Workshop 2004Cold Regions Workshop 2004Brian ConnellyBrian Connelly

North Central River Forecast CenterNorth Central River Forecast CenterChanhassen, MNChanhassen, MN

The Frozen Ground The Frozen Ground ScenarioScenario

Soils are wet as Soils are wet as freeze-up beginsfreeze-up begins

There is little snow There is little snow cover to insulate the cover to insulate the soilsoil

Hard freeze occurs Hard freeze occurs making the soil less making the soil less permeable to permeable to subsequent snowmelt subsequent snowmelt eventsevents

Somewhat rare, but Somewhat rare, but responsible for big responsible for big floodsfloods

Mankato, Minnesota, 1965

Modeling Frozen Ground Modeling Frozen Ground RunoffRunoff

The frozen ground model reduces the percolation and interflow rates.

From Center for Hydrometeorology and Remote Sensing, University of California, Irvine

How is index calculated…How is index calculated…

If Tair < 0°C then ΔFI = -C ∙ ( Tair2 + FI1

2)1/2 - C ∙ FI1 + Hc

If Tair > 0°C then ΔFI = C ∙ Tair + Cthaw ∙ P + Hc

where C = Cground ∙ ( 1 – Asnow ) + Cground ∙ Asnow ∙ ( 1 – Csnow )W

In other words…• Frost Index = FI ≤ 0°C• When Tair < 0°C then the FI decreases ΔFI α (Tair – FI)• When Tair > 0°C then the FI increases ΔFI α Tair

• Snow retards the change in the FI because it insulates the soil.• As water enters the soil it increases the FI

Percolation & Interflow Percolation & Interflow ReductionReduction

0.0

0.2

0.4

0.6

0.8

1.0

0 0.2 0.4 0.6 0.8 1

Soil Moisture (0.0=wet; 1.0=dry)

Re

du

ctio

n in

Pe

rco

latio

n &

Inte

rflo

wFI = -5 C FI = -10 C FI = -15 C FI = -25 C

For example…For example…

Frost Index adjusted to -25 CFrozen Ground model off

Baldhill Cr, North Baldhill Cr, North DakotaDakota

Frost Index = -50 C

Frozen Ground model off

Forest River, North Forest River, North DakotaDakota

Frozen Ground Model on

Ensemble Streamflow PredictionEnsemble Streamflow Prediction

500

1000

1500

2000

2500

3000

3500

Pe

ak

Dis

ch

arg

e (c

fs)

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1Chance of Exceeding Discharges

-6%-19%-14%-10%-10%

-10%-5%

-18%

19%

Control

FGIX = -15 C

Baldhill Creek near Dazey, NDForecast Period: February - May

0

1000

2000

3000

4000

5000

6000

Pe

ak

Dis

ch

arg

e (c

fs)

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1Chance of Exceeding Discharges

-22%3%

51%60%55%

83%105%

92%

106%

Control

FGIX = -25 C

Baldhill Creek near Dazey, NDForecast Period: February - May

How well does it work?How well does it work? On large events it is quite sensitive to On large events it is quite sensitive to

snowmelt volume.snowmelt volume. Since the Frost Index depends on air Since the Frost Index depends on air

temperatures and simulated SWE, QC of temperatures and simulated SWE, QC of these variables is these variables is extremelyextremely important. important.

Sometimes it works well, sometimes not, Sometimes it works well, sometimes not, but it almost always requires runtime but it almost always requires runtime modifications.modifications.

Bottom line: We need a frozen ground Bottom line: We need a frozen ground model to increase runoff ratios in the model to increase runoff ratios in the spring and this model does that.spring and this model does that.

Final Thoughts…Final Thoughts… The Frozen Ground Model allows us to The Frozen Ground Model allows us to

increase runoff ratios in the spring.increase runoff ratios in the spring. QC of temperatures and SWE is very QC of temperatures and SWE is very

important.important. The model is very sensitive to snowmelt The model is very sensitive to snowmelt

volume; Use model output and ESP results volume; Use model output and ESP results with care.with care.

Forecaster can use run-time modifications Forecaster can use run-time modifications to affect results.to affect results.

It It cannotcannot simulate ice layers at the snow/soil simulate ice layers at the snow/soil interface (interface (Variable Impervious Area modVariable Impervious Area mod?).?).