Ice Jam Prediction

Ice Jam PredictionIce Jam Prediction

Ice Engineering Research DivisionIce Engineering Research Division

US Army Cold Regions Research and Engineering LaboratoryUS Army Cold Regions Research and Engineering Laboratory

Presented by Kate WhiteFor Hydromet 00-2

Thursday, 9 March 2000

Why try to predict ice jams?Why try to predict ice jams?

• Improve emergency responseImprove emergency response

• Increase flood fighting effectivenessIncrease flood fighting effectiveness

• HindcastHindcast– Synthesize a historical record of jams Synthesize a historical record of jams

for design or flood warning purposesfor design or flood warning purposes

Ideal ice jam prediction modelIdeal ice jam prediction model

• Will provide a quantitative probability of Will provide a quantitative probability of ice jam occurrence or flooding with ice jam occurrence or flooding with enough lead time to institute mitigation enough lead time to institute mitigation measuresmeasures

• Type I error (i.e., a jam occurs when it Type I error (i.e., a jam occurs when it was not predicted) rate smallwas not predicted) rate small

• Type II error (“cry wolf”) rate smallType II error (“cry wolf”) rate small

• Variables easily and accurately Variables easily and accurately measured or forecastmeasured or forecast

Ice Jam Prediction ModelIce Jam Prediction Model

The ideal:The ideal: “generalized, site-transferable “generalized, site-transferable methods” which address all issues. methods” which address all issues.

The reality:The reality: A collection of site-specific methods, A collection of site-specific methods, each of which addresses only those issues each of which addresses only those issues deemed important at that specific site. A wide deemed important at that specific site. A wide variety of forecasting techniques are used. variety of forecasting techniques are used.

Stable Ice Cover Formation

Mechanical Breakup: (ice pieces generated)

Transport Capacity Exceeded

Transport Capacitynot Exceeded

Jam No JamNo JamNo Jam

Thermal Meltout (ice melts in place)

Ice Run

No Ice Cover or

Insufficient Ice to Form Jam

Thermal Decay

Ice Jam Prediction IssuesIce Jam Prediction Issues

• Stable ice cover formationStable ice cover formation

• Ice cover growth, strength, and Ice cover growth, strength, and decay decay

• Mechanical vs. thermal breakupMechanical vs. thermal breakup

• Ice transportIce transport

• Ice jam formationIce jam formation

• Flood levels and rate of rise Flood levels and rate of rise

Important VariablesImportant Variables• Ice Cover FormationIce Cover Formation

• Strength and DecayStrength and Decay

• BreakupBreakup

– Air temperatureAir temperature– AFDDAFDD– DischargeDischarge– Other met dataOther met data

– Air TempAir Temp– TDD /SunlightTDD /Sunlight– Ice thicknessIce thickness– Snow coverSnow cover

– Freezeup stageFreezeup stage– Rate of change in Q and/or Rate of change in Q and/or

stagestage– Antecedent meteorological Antecedent meteorological

conditionsconditions

Important VariablesImportant Variables• Ice TransportIce Transport

• Ice Jam FormationIce Jam Formation

• Flood LevelsFlood Levels

– DischargeDischarge

– Floe size and strengthFloe size and strength

– River plan formRiver plan form

– River geometry and slopeRiver geometry and slope

– DischargeDischarge

– Ice volumeIce volume

– Granular ice strength Granular ice strength parametersparameters

– Downstream stagesDownstream stages

– DischargeDischarge

– Ice volumeIce volume

Modeling ConsiderationsModeling Considerations

EasierEasier HarderHarder

Many Many Historical Ice JamsHistorical Ice Jams FewFew

Consistent Consistent Winter WeatherWinter Weather VariableVariable

One Spring PeakOne Spring Peak HydrologyHydrology Many peaksMany peaks

Long Long Duration of Ice CoverDuration of Ice Cover Short Short

Comprehensive Comprehensive Data AvailableData Available Little Little

Moderate Moderate Ice Jam SeverityIce Jam Severity Extreme Extreme

Frequent problems withFrequent problems withice-related dataice-related data

• Short or interrupted period of recordShort or interrupted period of record

• Low frequency ice jamsLow frequency ice jams

• Perception stagePerception stage

• MisclassificationsMisclassifications

• Reliability of measurementsReliability of measurements

• Error in discharge estimationError in discharge estimation

• Interrelationships among variables Interrelationships among variables understood to varying degreesunderstood to varying degrees

• Lack of observationsLack of observations

Types of jam prediction modelsTypes of jam prediction models

• Probabilistic forecastingProbabilistic forecasting• Empirical thresholdEmpirical threshold• Empirical cluster-type analysisEmpirical cluster-type analysis• Multiple linear regressionMultiple linear regression• Logistic regressionLogistic regression• Discriminant function analysisDiscriminant function analysis• DeterministicDeterministic• OtherOther

Brief review of existing breakupBrief review of existing breakupjam prediction modelsjam prediction models

• Use a wide variety of variablesUse a wide variety of variables

• Some focus on small part of Some focus on small part of processprocess

• Results variableResults variable

• Few easily transported to other Few easily transported to other locationslocations

Threshold identificationThreshold identification

• Goal is to identify one or more Goal is to identify one or more variables for which a threshold variables for which a threshold exists, or where statistically exists, or where statistically significant differences are presentsignificant differences are present

Shuliakovskii 1963Shuliakovskii 1963500

400

300

200

100

0

Non Ice Jams Ice Jams

Fre

eze

-up

Sta

ge

(cm

)

wh 7/6/94 (i)

2000

1800

1600

1400

1200

1000

800

600

400

200

Jam

cat

No Jam Jam

0

+

SD 080

Jam

cat

No Jam Jam

0

*

0

+

70

60

50

40

30

20

10

0

SD 081

AFDD’s

Snow Depth

Box & Whisker PlotsBox & Whisker Plots

Missouri River at Williston, NDMissouri River at Williston, NDWuebben et al.

Variable Range Indicated PotentialLow High

AFDD (°F-days) 918/3300 <1700 >2600

Julian Day of AFDDmax 135/180+ <150 >165

Julian Day Qmax 145/175 <155 >170

JD of Qmax minus JD ofFDDmax

-13/+30 <-8 or >+10 >-5 or <+7

Qmax (K cfs) 17/124 <25 or >90 >30 or <70

Garrison Dam Stage (feet) 1798/1844 <1835 >1840

Total Snowfall (inches) 5/60 <20 >40

Early/Late Snowfall N.A. <5” after JD = 90 >10” after JD = 90 or >5”after JD = 120

Threshold ModelsThreshold Models• White & KayWhite & Kay (Platte R., North Bend, NE)(Platte R., North Bend, NE)

• AFDD > 400AFDD > 400• Q > 6000 cfs Q > 6000 cfs oror Q>=.39(JD) Q>=.39(JD)1.91.9

• whichever is largerwhichever is larger

• Identified 65% of the ice events +-7 daysIdentified 65% of the ice events +-7 days• 41% wrong date or Type II error41% wrong date or Type II error• One Type I errorOne Type I error

Threshold ModelsThreshold Models• Tuthill et. alTuthill et. al. . (Winooski R., Montpelier, VT)(Winooski R., Montpelier, VT)

1 Dec - 31 March1 Dec - 31 March Q > 1,800 cfsQ > 1,800 cfs No peaks > 1,000 cfs in previous 30 daysNo peaks > 1,000 cfs in previous 30 days Time to peak less than 3 daysTime to peak less than 3 days Identified 13 of 17 known historic ice jamsIdentified 13 of 17 known historic ice jams Also identified 22 “potential” jams (I.e., Also identified 22 “potential” jams (I.e.,

large Type II error)large Type II error)

Threshold ModelsThreshold Models• White & Daly White & Daly (Oil Creek, Oil City, PA)(Oil Creek, Oil City, PA)

1515AFDD > 120AFDD > 120 11Q > 1,000 cfsQ > 1,000 cfs

Predicted PredictedPredicted Predicted Jam No Jam___Jam No Jam___Actual Jam 19 2 21Actual Jam 19 2 21Actual No Jam 35 6 41Actual No Jam 35 6 41 54 8 6254 8 62

15 A

FD

D (

°F)

1

Q(c

fs)

10 20 3010 2010 20 3010 20 303020

Nov '82 Dec Jan '83 Feb Mar

5000

4000

3000

2000

1000

0 0

100

200

300

400

500

15 A

FD

D (

°F)

1 Q

(cfs

)

15 AFDD (°F)

1 Q(cfs)

K. White KW 48

Predicted no jam

15

1

10 20 3010 2010 20 3010 20 30

Dec '65 Jan '66 Feb Mar

5000

4000

3000

2000

1000

0 0

100

200

300

400

500

15

1 1 Q

15 AFDD

K. White KW 58

Predicted jam

Discriminant Function AnalysisDiscriminant Function Analysis• White & Daly White & Daly (Oil Creek, Oil City, PA)(Oil Creek, Oil City, PA)

LogARQ: Log Allegheny R DischargeLogARQ: Log Allegheny R Discharge LogLog11QOC: Log 1 Day Oil Creek QOC: Log 1 Day Oil Creek Q Q 22AFDD: 2 Day AFDD: 2 Day AFDD AFDD

Predicted PredictedPredicted Predicted Jam No Jam___Jam No Jam___Actual Jam 11 6 17Actual Jam 11 6 17Actual No Jam 1 33 34Actual No Jam 1 33 34 12 39 5112 39 51

Deterministic ModelsDeterministic Models• Stable ice cover formationStable ice cover formation• Ice cover strength and Ice cover strength and

decay ratedecay rate• Mechanical breakupMechanical breakup

• Ice transportIce transport• Ice jam formationIce jam formation

• Flood levelsFlood levels

• 1-D hydraulic models with 1-D hydraulic models with thermal & icethermal & ice

• solar penetration modelssolar penetration models

• Onset of breakupOnset of breakup

• Hydraulic models with ice: free Hydraulic models with ice: free drift;discrete parcel; discrete drift;discrete parcel; discrete elementelement

• Hydraulic model: unsteady vs. Hydraulic model: unsteady vs. steadysteady

Deterministic ModelsDeterministic Models

• Uncertainties enter due toUncertainties enter due to– parametersparameters– input data input data – structural problemsstructural problems

• Difficult to deal with uncertainty in Difficult to deal with uncertainty in any direct manner any direct manner

• Development lagging due to lack of Development lagging due to lack of compete analytical modelcompete analytical model

SummarySummary

• Ice jam formation is a very site specific Ice jam formation is a very site specific phenomenaphenomena

• There are a number of issues inhibiting There are a number of issues inhibiting development of ice jam prediction development of ice jam prediction modelsmodels

• The ideal of “generalized, site-The ideal of “generalized, site-transferable methods” remains a goaltransferable methods” remains a goal

• A number of techniques have been used A number of techniques have been used for forecasting jamsfor forecasting jams

• Threshold models have been successfully Threshold models have been successfully used in some casesused in some cases

• Deterministic models can achieve the goal Deterministic models can achieve the goal of a generalized method, but a complete of a generalized method, but a complete analytical model is required and the analytical model is required and the problems of uncertainty must be addressedproblems of uncertainty must be addressed

Partnerships between agencies Partnerships between agencies will increase effectiveness of will increase effectiveness of

response to ice jams response to ice jams

Examples: Examples:

• NWS inclusion of ice jams in Hydromet training NWS inclusion of ice jams in Hydromet training http://www.nws.noaa.gov/er/nerfc/riverice/Training00a/http://www.nws.noaa.gov/er/nerfc/riverice/Training00a/

• Corps-NWS partnerships (e.g., St. Paul District, NCRFC)Corps-NWS partnerships (e.g., St. Paul District, NCRFC)

• CRREL partnerships (e.g., MARFC, NERFC, NWS CRREL partnerships (e.g., MARFC, NERFC, NWS Glasgow MT)Glasgow MT)

http://www.nws.noaa.gov/er/nerfc/riverice/Part1/ http://www.nws.noaa.gov/er/nerfc/riverice/Part2/