THE IMPACTS OF CLIMATE CHANGETHE IMPACTS OF CLIMATE CHANGETHE IMPACTS OF CLIMATE CHANGE THE IMPACTS OF CLIMATE CHANGE AND VARIABILITY ON WATER AND VARIABILITY ON WATER

RESOURCES IN A SEMIRESOURCES IN A SEMI ARID REGIONARID REGIONRESOURCES IN A SEMIRESOURCES IN A SEMI--ARID REGION ARID REGION IN MEXICO: THE RIO YAQUIIN MEXICO: THE RIO YAQUI--BASIN.BASIN.

by:by:Andrea MunozAndrea Munoz--Hernandez.Hernandez.

Dr. Alex S. Mayer.Dr. Alex S. Mayer.

OBJECTIVESOBJECTIVES

GENERALDevelop an Integrated Hydrologic-Economic-Institutional

Water Model for the Yaqui River BasinWater Model for the Yaqui River Basin.

SPECIFICDetermine the impacts of climate change and variability on

precipitation and reservoir storage in the Yaqui Basinprecipitation and reservoir storage in the Yaqui Basin.

TASKSTASKS

• Develop a water balance model to determine storage in thereservoirs on a monthly basis.

•Create and calibrate a seasonal rainfall-runoff model.

•Incorporate climate change and climate variability into thewater balance model.

BACKGROUND

The Yaqui basin ischaracterized by semi aridcharacterized by semi aridconditions.

The basin consists of roughlyg y72,000 square kilometers.

One of the most important agricultural regions in Mexico is located within thebasin.

Water users include farmers, rural and urban municipalities, industries, andmines.

WATER BALANCE MODELWATER BALANCE MODEL

A node- link network is theconceptual basis for themodel.model.

This node-link network includesthe primary reservoirs within thep ybasin, river reaches, locationsof water demand and supply,and the Yaqui Valley.

WATER BALANCE MODELWATER BALANCE MODEL

A MATLAB code was developed in order to estimate the monthly storage

The model considers each surface water rights holder within the basin and takes

p y gof the main reservoirs.

within the basin and takes into account priorities in allocating the water.

The operating rules include the release of water downstream once the water

The main objective was to determine the storage in the reservoirs in

needs have been met.

j gOctober of every year, when cropping decisions are made.

WATER BALANCE MODEL

Di

WATER BALANCE MODEL

Direct

Precipitation

Direct

Evaporation

Reservoir0 9 Releases E t ti b d

ConveyanceFactor

Reservoir∆S

0.9 Releases Extractions based on water rightsRunoff

∆Si = ∆Si-1+ Runoff +Precipitation+ 0.9 Releases - Evaporation – Extractions

RUNOFF =YRUNOFF =Yij

CREATION AND CALIBRATION OF THE RAINFALL-RUNOFF MODEL

The watershed and sub-basinboundaries were delineated using GISb d DEMbased on DEMs.

The sub-basins were aggregated andclassified into an Upper basin, a Middlebasin, and a Lower basin each with asingle outflow point.

PRECIPITATION MAPS

Precipitation data was interpolated

PRECIPITATION MAPS

Precipitation data was interpolatedon a monthly basis over a 33-yeartime span using GIS.

The precipitation data was mergedinto three climatic seasons:

Spring season: February-May.Summer season: June-SeptemberWinter season: October-January

A static runoff coefficient map wasalso produced based on publisheddatadata.

PIXEL-BASED RUNOFF MODEL

Area of R ffPrecipitation(m)

Runoff coefficient*2500 m *

Area of the pixel

(m2)=

Runoff

(Mm3)

2500 m

PijGIS * CGIS * A= Xij

i=Sub basinJ=Season

CALIBRATION OF THE MODELCALIBRATION OF THE MODEL

PijGIS * CGIS * A= Xij

Y = β X +αYij = βij Xij +αij

Min = ∑ (Y Y* )2Min = ∑ (Yij-Y ij)2

i=Sub basinJ=Season

Alpha and Beta were found minimizing the sum of the square errors.

RESULTS: CREATION AND CALIBRATION OF THE MODELCALIBRATION OF THE MODEL

2500

3000

)Middle Basin

Winter Season Model (Yij)

C.N.A (Y*ij)2500

3000

)Middle Basin

Winter Season Model (Yij)

C.N.A (Y*ij)

1500

2000

2500

Run

off (

MC

M)

1500

2000

2500

Run

off (

MC

M)

0

500

1000

1970 1975 1980 1985 1990 1995 20000

500

1000

1970 1975 1980 1985 1990 1995 2000

The timing of the peaks matches reasonably well but the model tends

1970 1975 1980 1985 1990 1995 2000

Year1970 1975 1980 1985 1990 1995 2000

Year

The timing of the peaks matches reasonably well, but the model tends to under predict the runoff in the wettest year (1984).

RESULTS: CREATION ANDRESULTS: CREATION AND CALIBRATION OF THE MODEL

This table shows all the parameters fitted with the linear model. In most cases, the R2 values indicate reasonable fits.

AT ACLIMATE CHANGE

• Two Global Climate Models were used : the UK’s Hadley Center model(HADCM3) and the Canadian Center for Climate Modeling and Analysis(CGCM) for the period 2011 2100(CGCM) for the period 2011-2100.

• A regional model was used for comparison for the period 2011-2040:Providing Regional Climates for Impact Studies (PRECIS)Providing Regional Climates for Impact Studies (PRECIS).

•Two SRES (Special Report on Emissions Scenarios) scenarios wereused: A2 (high emission) and B1(low emission).used ( g e ss o ) a d ( o e ss o )

BIAS CORRECTION ANDBIAS CORRECTION AND DOWN-SCALING

•The bias-correction and downscaling approach developed by Wood et al (2002)was used to downscale the GCMswas used to downscale the GCMs.

S ADOWNSCALING

For PRECIS, monthly percentage changes based on the 1961-1990 record were calculated. These percentage changes werethen applied to our baseline period (1970-2000), assuming thatthi d ld b t d i th f tthis record would be repeated in the future.

CHANGE ON PRECIPITATION

1HADCM3

Scenario A2 MeanStdev

1HADCM3

Scenario A2 MeanStdev 1 5

2 CGCMScenario A2 Mean

Stdev1 5

2 CGCMScenario A2 Mean

Stdev

0.2

0.4

0.6

0.8

lta

Stdev

0.2

0.4

0.6

0.8

lta

Stdev

0.5

1

1.5

Del

ta

Stdev

0.5

1

1.5

Del

ta

Stdev

-0.6

-0.4

-0.2

00 10 20 30 40 50 60 70 80 90

Del

-0.6

-0.4

-0.2

00 10 20 30 40 50 60 70 80 90

Del

1

-0.5

00 10 20 30 40 50 60 70 80 90

D1

-0.5

00 10 20 30 40 50 60 70 80 90

D-0.8

Year-0.8

Year

The CGCM model predicts a drier future than the HADCM3 model

-1Year

-1Year

The CGCM model predicts a drier future than the HADCM3 model

RESULTS:RESULTS:CLIMATE CHANGE

6000

CGCMA2 Scenario Mean

std. dev6000

CGCMA2 Scenario Mean

std. dev

6000

CM

)

HADCM3 A2 scenario

Mean

Std. dev

6000

CM

)

HADCM3 A2 scenario

Mean

Std. dev

3000

4000

5000

age

(MC

M)

3000

4000

5000

age

(MC

M)

3000

4000

5000

Stor

age

(MC

3000

4000

5000

Stor

age

(MC

0

1000

2000

0 10 20 30 40 50 60 70 80 90St

ora

Year

0

1000

2000

0 10 20 30 40 50 60 70 80 90St

ora

Year

0

1000

2000

0 10 20 30 40 50 60 70 80 900

1000

2000

0 10 20 30 40 50 60 70 80 90YearYearYearYear

The forecasted precipitation is reflected in the total storage for the period 2011 2100The forecasted precipitation is reflected in the total storage for the period 2011-2100.

RESULTS:CLIMATE CHANGE

•This table shows the percentage of times the reservoir storage falls below the users needs compared to the historical period.

•The storage varies depending on the climate model and the scenario used.

CLIMATE VARIABILITY

To assess the effects of year-to-year correlations in precipitation a l i it ti d t t (104 ) d (Ni h l t llonger precipitation data set (104 years) was used (Nicholas et al, 2007).

Winter Seasonl

Winter Seasonl

5

6(ln)

5

6(ln)

3

4

[ln (P

)] (m

m)

Data MA63

4

[ln (P

)] (m

m)

Data MA6

1

2

1900 1920 1940 1960 1980 2000

mean std dev

1

2

1900 1920 1940 1960 1980 2000

mean std dev

1900 1920 1940 1960 1980 2000Year

1900 1920 1940 1960 1980 2000Year

CLIMATE VARIABILITY

An auto-regressive model approach (AR) was fitted to the th d l t f d i it ti d d t tsmoothed, log-transformed precipitation and used to generate, on

a monthly basis, precipitation for a period of thirty years using the following approach:

( )∑ − +−+=p

tjtjt yy εμφμ

Where μ is the mean of the data p is the order of the model φj

( )∑=j

tjtjt1

Where μ is the mean of the data, p is the order of the model, φj

are fitted parameters and εt is a uncorrelated normal random variable.

RESULTS:CLIMATE VARIABILITYCLIMATE VARIABILITY

0.5000.500

0 300

0.350

0.400

0.450cy

Water rights

0 300

0.350

0.400

0.450cy

Water rights

0.150

0.200

0.250

0.300

Freq

uenc

0.150

0.200

0.250

0.300

Freq

uenc

0.000

0.050

0.100

0.150

0.000

0.050

0.100

0.150

2500 2750 3000 3250 3500 3750 4000 4250 4500 4750

Storage (MCM)2500 2750 3000 3250 3500 3750 4000 4250 4500 4750

Storage (MCM)

There is a small, but significant probability that the storage can fall below the water user’s needs.

CONCLUSIONS

• The results show that there is sufficient surface water tomeet users’ needs for a wide range of conditions(uncertainty, climate change, and climate variability), butthis is not always the case.

•The rainfall-runoff model produces acceptable results whencompared with historical data. The best and worst matchesare obtained in the middle and lower basin, respectively.

CONCLUSIONS

Th t ti t bt i d f th i ti f•The storage estimates obtained from the incorporation ofclimate change into the water model shows that the basincould suffer from water shortages during some yearsdepending on the climate model or the scenario used. Theuse of different GCMs or SRES scenarios that are moreoptimistic or pessimistic might produce different results.

•Future assessments of climate variability should considerseason to season correlation and different ways ofclassifying precipitation levels and correspondingprobabilities.

ACKNOWLEDGEMENTSACKNOWLEDGEMENTS

The authors wish to thank Dr. Cecilia Conde from the National AutonomousUniversity of Mexico (UNAM) for all the comments and guidance in therealization of this project.realization of this project.

Additionally, we also want to thank Dr. Jose Luis Minjares Moreno from theNational Commission of Water in Mexico and the Cuban Institute ofMeteorology for all the information and assistance provided.

We acknowledge the modeling groups, the Program for Climate ModelDiagnosis and Intercomparison (PCMDI) and the WCRP's Working Group onCoupled Modeling (WGCM) for their roles in making available the WCRPCMIP3 multi-model dataset.

REFERENCES

1 Mi j J L 2004 S i bl i f h Y i Ri M l i l

REFERENCES

1. Minjares J.L. 2004. Sustainable operation of the Yaqui River Multiple Reservoir System. Ph D. dissertation. New Mexico State University. Las Cruces, New Mexico.

2.Nicholas and Battisti (2007) Drought Recurrence and Seasonal ( ) gRainfall Prediction in the Rio Yaqui Basin, Mexico. In press, J. App. Meteor. Hydro.

3 PRECIS http://precis insmet cu/menu page htm3. PRECIS http://precis.insmet.cu/menu_page.htm

4. Wood, A. W., E. P. Maurer, A. Kumar, and D. P. Lettenmaier, 2002: Long range experimental hydrologic forecasting for the eastern g g p y g gU.S. J. Geophys. Res., 107, 4429, doi:10.1029/2001JD000659.

Thank You

UNCERTAINTY ANALYSIS: A T A A AMONTE CARLO APPROACH

Uncertainty in the rainfall-runoff model predictions were assessed using aMonte Carlo simulation approach, assuming that model errors arenormally distributed. Runoff was calculated using the relationship:

Y = α + β ± tn-2,1-α/2 SYXo

where α and β are best estimates, tn-2,1-α/2 is the t statistic, and SYXois the standard error of the estimate.

In the Monte Carlo simulations, 100 values of 1-α/2 were randomlygenerated from a uniform distribution.

RESULTS:RESULTS:MONTE CARLO

45005000

Total Storage

Best Estimate Water rights4500

5000

Total Storage

Best Estimate Water rights

200025003000350040004500

age

(MC

M)

Water rights

200025003000350040004500

age

(MC

M)

Water rights

0500

100015002000

Stor

a

0500

100015002000

Stor

a

Based on the results, there is enough surface water to satisfy current

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

Year1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

Year

water rights every year for the best estimates and, in some cases, forthe 10% confidence interval.