incorporating large-scale climate information in water resources decision making

Incorporating Large-Scale Climate Information in Water Resources

Decision Making

Balaji RajagopalanBalaji Rajagopalan

Dept. of Civil, Env. And Arch. Dept. of Civil, Env. And Arch. Engg. And CIRESEngg. And CIRES

Katrina Grantz, Edith Zagona Katrina Grantz, Edith Zagona

(CADSWES)(CADSWES)

Martyn Clark (CIRES)Martyn Clark (CIRES)

A Water Resources Management Perspective

Time

Horizon

Inter-decadal

Hours Weather

ClimateDecision Analysis: Risk + Values

Data: Historical, Paleo, Scale, Models

• Facility Planning

– Reservoir, Treatment Plant Size

• Policy + Regulatory Framework

– Flood Frequency, Water Rights, 7Q10 flow

• Operational Analysis

– Reservoir Operation, Flood/Drought Preparation

• Emergency Management

– Flood Warning, Drought Response

INDEPENDENCE

DONNERMARTIS

STAMPEDE

BOCA

PROSSER

TRUCKEERIVER

CARSONRIVER

CARSONLAKE

Truckee

CarsonCity

Tahoe City

Nixon

Fernley

DerbyDam

Fallon

WINNEMUCCALAKE (dry)

LAHONTAN

PYRAMID LAKE

NewlandsProject

Stillwater NWR

Reno/Sparks

NE

VA

DA

CA

LIF

OR

NIA

LAKE TAHOE

Study Area

TRUCKEE CANAL

Farad

Ft Churchill

NEVADA

CALIFORNIA

Carson

Truckee

Motivation

• US Bureau of Reclamation (USBR) searching for an US Bureau of Reclamation (USBR) searching for an improved forecasting model for the Truckee and improved forecasting model for the Truckee and Carson Rivers (accurate and with long-lead time)Carson Rivers (accurate and with long-lead time)

Truckee Canal

• Forecasts determine reservoir Forecasts determine reservoir releases and diversions releases and diversions

• Protection of Protection of listed specieslisted species

Cui-ui

Lahontan Cutthroat Trout

Outline of Approach

ClimateDiagnostics

ClimateDiagnostics

ForecastingModel

ForecastingModel

DecisionSupport System


• Forecasting ModelForecasting ModelNonparametric stochastic model Nonparametric stochastic model conditioned on climate indices and conditioned on climate indices and snow water equivalentsnow water equivalent

• Climate DiagnosticsClimate DiagnosticsTo identify relevant predictors To identify relevant predictors to spring runoff in the basinsto spring runoff in the basins

• Decision Support SystemDecision Support SystemCouple forecast with DSS to Couple forecast with DSS to demonstrate utility of forecastdemonstrate utility of forecast

Data Used

•1949-2003 monthly data sets:1949-2003 monthly data sets:

•Natural Streamflow (Farad & Ft. Churchill gaging stations)

•Snow Water Equivalent (SWE)- basin average

•Large-Scale Climate Variables

500mb Geopotential Height Sea Surface Temperature

Carson Spring Flow

Winter Climate Correlations

Fall Climate Correlations

500mb Geopotential Height Sea Surface Temperature

Carson Spring Flow

Physical Mechanism

L

• Winds rotate Winds rotate counter-counter-clockwise clockwise around area around area of low of low pressure pressure bringing bringing warm, moist warm, moist air to air to mountains in mountains in Western USWestern US

Climate Indices• Use areas of highest correlation to develop Use areas of highest correlation to develop

indices to be used as predictors in the indices to be used as predictors in the forecasting model forecasting model

• Area averages of geopotential height and SST Area averages of geopotential height and SST

500 mb Geopotential Height Sea Surface Temperature

Outline of Approach

ClimateDiagnostics

ClimateDiagnostics

ForecastingModel

ForecastingModel



Forecasting ModelForecasting ModelNonparametric stochastic model Nonparametric stochastic model conditioned on climate indices and conditioned on climate indices and SWESWE

• Climate DiagnosticsClimate DiagnosticsTo identify relevant predictors to To identify relevant predictors to spring runoff in the basinsspring runoff in the basins

• Decision Support SystemDecision Support SystemCouple forecast with DSS to Couple forecast with DSS to demonstrate utility of forecastdemonstrate utility of forecast

The Ensemble Forecast Problem

•Ensemble Forecast/Stochastic Simulation Ensemble Forecast/Stochastic Simulation /Scenarios generation – all of them are /Scenarios generation – all of them are conditional probability density function conditional probability density function problemsproblems

•Estimate conditional Estimate conditional PDF PDF and simulate (Monte and simulate (Monte Carlo, or Bootstrap)Carlo, or Bootstrap)

•K-NN Approach is UsedK-NN Approach is Used

f yy y y

f y y y y

f y y y y dyt

t t t p

t t t t p

t t t t p t

1 2

1 2

1 2

, , . . . ,( , , , . . . , )

( , , , . . . , )

Model Validation & Skill Measure

• Cross-validation: drop one year from the model Cross-validation: drop one year from the model and forecast the “unknown” valueand forecast the “unknown” value

• Compare median of forecasted vs. observed Compare median of forecasted vs. observed (obtain “r” value)(obtain “r” value)

• Rank Probability Skill ScoreRank Probability Skill Score

• Likelihood Skill ScoreLikelihood Skill Score

ology)RPS(climat

st)RPS(foreca1RPSS

k

j

i

nn

i

nn dP

kdpRPS

1 111

1),(

N

N

tc

N

tij

ijP

PL

1

1

1,

,

Forecasting ResultsTruckee RPSS results

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

Nov 1st Dec 1st Jan 1st Feb 1st Mar 1st Apr 1st

Month

Med

ian

RP

SS

(al

l yea

rs)

GpH & SWE

SWE

Carson RPSS results

-0.2

0.0

0.2

0.4

0.6

0.8

1.0


Month

Med

ian

RP

SS

(al

l yea

rs)

GpH & SWE

SWE

Truckee Forecasted vs. Observed Correlation Coeff

0

0.2

0.4

0.6

0.8

1


Month

Co

rrel

atio

n C

oef

f

GpH & SWE

SWE

Carson Forecasted vs. Observed Correlation Coeff

0

0.2

0.4

0.6

0.8

1


Month

Co

rrel

atio

n C

oef

f.

GpH & SWE

SWE

Truckee Likelihood Results

0

0.5

1

1.5

2

2.5


Month

Med

ian

Lik

elih

oo

d (

all

year

s)

GpH & SWE

SWE

Carson Likelihood Results

0

0.5

1

1.5

2

2.5


Month

Med

ian

Lik

elih

oo

d (

all

year

s)

GpH & SWE

SWE

Outline of Approach

ClimateDiagnostics

ClimateDiagnostics

ForecastingModel

ForecastingModel



• Forecasting ModelForecasting ModelNonparametric stochastic model Nonparametric stochastic model conditioned on climate indices and conditioned on climate indices and SWESWE

• Climate DiagnosticsClimate DiagnosticsTo identify relevant predictors to To identify relevant predictors to spring runoff in the basinsspring runoff in the basins

Decision Support SystemDecision Support SystemCouple forecast with DSS to Couple forecast with DSS to demonstrate utility of forecastdemonstrate utility of forecast

Seasonal Decision Support System

•Method to test the utility of the Method to test the utility of the forecasts and the role they play in forecasts and the role they play in decision makingdecision making

•Model implements major policies in Model implements major policies in lower basin (Newlands Project lower basin (Newlands Project OCAP)OCAP)

•Seasonal time stepSeasonal time step

Seasonal Model Policies

• Use Carson water firstUse Carson water first

• Max canal diversions: 164 kafMax canal diversions: 164 kaf

• Storage targets on Lahontan Reservoir: 2/3 Storage targets on Lahontan Reservoir: 2/3 of historical April-July runoff volumeof historical April-July runoff volume

• No minimum fish flows (release from No minimum fish flows (release from upstream reservoir to combat low flows)upstream reservoir to combat low flows)

Decision Model Flowchart

Truckee Forecast

Truckee Avail for Diversion= Truckee Forecast

Carson Forecast

Is Truckee Forecast > Max Diversion ?

Ensemble Forecasts

Truckee Canal Diversion= Avail for Diversion

Water Available for Fish = Truckee Fcst – Truckee Canal Diversion

Water Available for Irrigation= Carson Fcst + Truckee Canal Diversion

No Yes

Is Avail for Diversion> Diversion Request?

Truckee Avail for Diversion= Max Diversion

Is Carson Forecast> Lahontan Target ?

Diversion Requested = Target – Carson Forecast

Diversion Requested= 0.0 kaf

No Yes

No Yes

Truckee Canal Diversion= Diversion Requested

Repeat for each ensemble member

Decision Variables

• Lahontan Storage Lahontan Storage Available for IrrigationAvailable for Irrigation

• Truckee River Water Truckee River Water Available for FishAvailable for Fish

• Diversion through the Diversion through the Truckee CanalTruckee Canal

Decision Model Results

0 100 300 500

010

030

050

0

Perfect Forecast Irrigation Water (kaf)

Med

ian

of E

nsem

ble

Irrig

atio

n W

ater

(kaf

)

r=0.17

0 50 100 150

050

100

150

Perfect Forecast Diversion (kaf)

Med

ian

of E

nsem

ble

Div

ersi

on(k

af)

r=0.23

0 200 400 600

020

040

060

0

Perfect Forecast Fish Water (kaf)

Med

ian

of E

nsem

ble

Fis

h W

ater

(kaf

)

r=0.36

0 100 200 300 400 500

010

020

030

040

050

0


Med

ian

of E

nsem

ble

Irrig

atio

n W

ater

(kaf

)

r=0.54

0 50 100 150

050

100

150


Med

ian

of E

nsem

ble

Div

ersi

on(k

af)

r=0.38

0 200 400 600

020

040

060

0


Med

ian

of E

nsem

ble

Fis

h W

ater

(kaf

)

r=0.62

0 100 200 300 400 500

010

030

050

0


Med

ian

of E

nsem

ble

Irrig

atio

n W

ater

(kaf

)

r=0.79

0 50 100 150

050

100

150


Med

ian

of E

nsem

ble

Div

ersi

on(k

af)

r=0.78

0 200 400 600

020

040

060

0


Med

ian

of E

nsem

ble

Fis

h W

ater

(kaf

)

r=0.93

Canal Diversion Water for FishIrrigation Water

Dec 1st Forecast

Feb 1st Forecast

Apr 1st Forecast

Dry Year: 1994April 1st February 1st December 1st

Truckee ForecastTruckee Forecast

Carson ForecastCarson Forecast

Storage for IrrigationStorage for Irrigation

Canal DiversionCanal Diversion

Water for FishWater for Fish

Wet Year: 1993April 1st February 1st December 1st






Normal Year: 2003April 1st February 1st December 1st






Exceedance Probabilities 1994 (Dry Year) Apr 1st Feb 1st Dec 1st Historical

Irrigation Water mean value (kaf) 94 161 214 264264 kaf Irrigation Water exceedance probability 4% 14% 18% 50%Fish Flow mean value (kaf) 0 42 39 19960.5 kaf Fish Flow exceedance probability 0% 57% 58% 87%Canal Diversion mean value (kaf) 52 107 121 84

1993 (Wet Year) Apr 1st Feb 1st Dec 1st Historical


2003 (Normal Year) Apr 1st Feb 1st Dec 1st Historical


Summary & Conclusions

•Climate indicators improve forecasts Climate indicators improve forecasts and offer longer lead timeand offer longer lead time

•Water managers can utilize the Water managers can utilize the improved forecasts in operations and improved forecasts in operations and seasonal planningseasonal planning

Grantz et al. (2005) – submitted to BAMSGrantz et al. (2005) – submitted to BAMS

Grantz et al. (2005) – accepted in Water Grantz et al. (2005) – accepted in Water Resources Research.Resources Research.

Acknowledgements

• CIRES and the Innovative CIRES and the Innovative Research ProjectResearch Project

• Tom Scott of USBR Lahontan Tom Scott of USBR Lahontan Basin Area Office Basin Area Office

FundingFunding

incorporating large-scale climate information in water resources decision making

Documents

spring runoff

improved forecasting

relevant predictors

sweclimate diagnosticsto

reservoir releases

upstream reservoir

lahontan reservoir

decision makingmodel