Best Practices on the Application of Climate Information for

Water Resources Management

M.N. Ward1, U. Lall1,2, C. Brown1, H.-H. Kwon2

1International Research Institute for Climate and Society (IRI), The Earth Institute at Columbia University, New York, USA

2Department of Earth and Environmental Engineering, Columbia University, New York, USA

Open seminar on the applications of climate information in various socio-economic sectors. Tokyo, Tuesday 20th February 2007

Managing Water Resource Systems• Balance Water Supply and Demand, avoid flood

• Historical rules for resource allocation

• How much, and when should these rules be modified based on new climate technologies

• How do we assess and communicate potential impacts of action & inaction ?

• Background risks for sustainable strategies and infrastructure development

Health

Human Activity

Energy

Climate

Water

Agriculture

New City

Irrigated Farms

Irrigated Farms

Dam 1Dam 2

Dam 3

Electric Grid Well Field

Muddy River

1. Monitoring and Short-term (several days) projections

2. Seasonal Prediction (next 3-6 months)

3. Merging knowledge on natural multidecadal (e.g. 10-30 years) and global change for water resources management

Management options at different timescales of the available information

Section 1

Monitoring and Short-term projections

Flood prediction and management(including Mozambique case study)

Conception of FEWS Flood Model

FEWS Flood Risk Monitoring System Flow Diagram

Preprocessing

MAP

MAE

Basin

Linkage

Routing Parameters

Soil Parameters

Flood Inundation Mapping

Landsat 7 SPOT

Output / Decision Support System

Data

RFE

PET

Soil

LU/LC

DEM

QPF

Stream Flow Model

Water Balance

Lumped Routing

Dist. Routing

Updating

Case Study: surface hydrology in Sri Lanka

Potential for enhanced monitoring and prediction of weather-driven component of surface hydrology

New opportunity: Reanalysis weather data

ECMWF reanalysis weather data drives stream flow simulation for Mahaweli gauge location, Sri Lanka

NASA, Mahaweli River Authority, IRI1979 1994

Black = observedRed = simulated

Flow

Time

(Reanalysis rainfall is bias corrected)

Eds, Hellmuth et al., 2007. Mozambique case study by Lucio et al.

Recent climate-related natural disasters in Mozambique

***********

(Lucio et al., 2007)

Limpopo basin includes Zimbabwe, Botswana andS. Africa

1. Seasonal forecast recognized increased risk of flooding through the rainy season due to presence of La Nina and other climate aspects (but no methods yet to quantify increased risks)

2. November – National disaster committee meets frequently and produces National Contingency Plan

Mozambique floods, Jan-Feb 2000Aspects of good practice that were already in place

1. Flood risk analysis for vulnerable areas(see section 3 of lecture)

2. Hydromet monitoring system enhanced

3. Linking monitoring/forecast information to trigger response

4. Consider news media, and communication

Improvements in practice after 2000 Mozambique flood

Section 2

Bringing Seasonal Prediction Technology into Water Resources Management

Especially in tropical regions, capability exists now to forecast climate patterns 3-6 months into the future

Forecasting Reservoir Inflow for

Reservoir Operations

Reservoir Operation Model

Reservoir Inflows

NEED AS A PDF OR ENSEMBLE

Forecasting Water Supply and Demand

General Circulation Model

“Downscaling”Regional Climate Model

Or Statistical Model

Hydrologic Model

Crop Model

Reservoir Operation Model

Economic Model

Regional Climate Predictors

Statistical Model

Reservoir Inflows

NEED AS A PDF OR ENSEMBLE

Possible ProceduresSeasonal GCM Rainfall Forecast Statistical or Dynamical Downscaling

Daily Weather Sequences

Crop Irrigation Demand

“Climate Predictors”

Empirical StatisticalModel

Reservoir Inflow

CurrentReservoir Volume

Probability that Demand > Supply

Revise Crop Choice or Planted Area based on

Expected Net Value or other criteria

Toolavailable

Exploring the management of Angat Dam, Philippines using seasonal inflow forecasts(Most value in such low storage to inflow ratio settings)

Rainfall-Runoff (Oct-Feb) Relation

y = 0.8331x + 27.464R2 = 0.79

0

50

100

150

200

250

300

350

400

0 50 100 150 200 250 300 350 400 450Rainfall (mm/month)

Stre

amflo

w (M

cm/m

onth

)

Reliable Seasonal Climate Forecasts are possible in many tropical locations

Skill of Oct-Dec rainfall Predictions from a GCM

Software tool to translate GCM seasonal forecasts into a target variable

Freely available from IRI website

From General Circulation Model (GCM)to Reservoir Inflow Forecast

The GCM gives a large-scale climate forecast

Then apply a statistical transformation to predict reservoir inflow

050

100150200250300350400

1968 1978 1988 1998Year

Stre

amflo

w (M

CM

)

ONDJF-obsONDJF-pred

ρρ((QQpredpred,Q,Qobsobs):0.58):0.58

Translating large-scale forecast output from a GCM intoOct-Feb Reservoir inflow forecasts for reservoir management

Fig 4Angat Watershed

Angat Reservoir

Hydropower (200 MW)

Bustos Dam

Hydropower (Auxillary) – 48 MW)

Bulacan Irrigation (31000 ha)

Metro Manila (97%)

La Mesa Dam

Manila Bay

0

200

400

600

800

1000

1200

1400

1987 1989 1991 1993 1995 1997 1999 2001

Year

Hyd

ropo

wer

Gen

erat

ed (

in G

WH

)

0

50

100

150

200

250

300

350

400

Observed In

flowActualUpdated ForecastOctober ForecastObserved Inflow

Estimating Improved Hydropower Production using Seasonal ForecastsOutput from software illustrated in previous slide

Lall and Arumugam, 2006

Two Caveats for Changing PracticeBased on Seasonal Prediction

1)Technical: Care with downscaling the prediction signal

2)Societal: Participatory process and often need for policy change

High mountains can make downscaling information critical and complex

(Zubair et al.)

Seasonal forecasts vary across Sri Lanka

Modeling small scale seasonal rainfall anomalies across Java in El Nino Years

(Qian et al., 2007)

Sep-Nov

Dec-Feb

Brown = Below normalGreen = Above normal

#S

#S

#S#S

%U

%U%U

%U %U

%U %U

%U

%U

#

#

#

#

#

#

#

#

#

#

# #

#

#

#

#

#

#

##

#

#

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

# #

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

# #

#

#

##

#

#

#

#

##

#

#

#

#

#

#

##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

##

#

#

##

#

#

#

#

#

#

##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

##

##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

$T

$T

$T$T$T

$T$T

$T

$T$T

#

Bacia.shpAçudespol.shpRios.shpCanais.shpLink

#S NódePassagem

Demanda%U 0.3%U 0.3- 0.

N

EW

S

Fortaleza

Jaguaribe-Metropolitano Hidrosystem

Adoption of new allocationstrategies is complex process

Jan-Dec Water Macro-Allocation Plan --- Developed July-Oct 2002 Ensemble Forecast

0.0

5.0

10.0

15.0

1 3 5 7 9 11Month

Flow

Ensemble Forecast

FUNCEME/IRI

Feedback to revise offers

Water Committee

Water UsersIrrigation, Permanent

Water UsersIndustry, Canning

Water Agency

COGERH

Demand & Priority Scenario

Reservoirs Simulation &Optimization

Assess Feasible Allocation

Communicate

Propose Contracts:• Desired Reliability• Desired Price

Annual AllocationUser j gets Wj m3 water pj% reliability for price R$With specified monthly patternand priority for failure

When negotiations

conclude

ReviseRevise

Revise

Water allocation matters to many people

• For resource management strategies including infrastructure development

• For disaster risk management• cf Mozambique example

Section 3 Background Hydroclimatic Risk Information

Analyses to inform strategies for infrastructure

Knowledge of climate variability is a key factor

Here estimates of storage volume needed by country

Brown and Lall, 2006

Multi-decadal variability is nowrecognized as a natural partof the climate system

There is growing understanding ofits sources and statistical properties

Motivates finding best ways to incorporate statistics for long-termplanning

Luterbacher and Xoplaki, 2003

Expression in Regional Climate Fluctuations

Developing information to support South Florida Water Management District

Kwon and Lall, 2006

Models simulate low frequency statistical properties to guide management strategies

PowerSpectrum

Context of Global Change

Climate/Environment and Socioeconomic

Availability of water for agriculture in the coming decades depends not only on changing climate, but also on population, economic development, and technology

Linking Regional Water Supplies and Water Demands in a changing world

(C. Rosenzweig, NASA GISS& Columbia University)

Luterbacher and Xoplaki, 2003

Expression in Regional Climate Fluctuations

0

50

100

150

200

250

1910 1920 1930 1940 1950 1960 1970 1980 1990

Demanda AtualDemanda 2030AfluenciaMédiaQuantil25%

Annual Oros Reservoir Inflow in m3/s

Projected & Existing Demand

Water Resources Setting – Study in NE Brazil

How to extract the inflow statistics expected for the next 30 years?

Insurance as a natural tool to better manage climate and

hydroclimatic risk

Weather / hydrology index insuranceExample for Peru, flood precipitation proxy (y-axis)x-axis is Nino index, introducing predictability to the insurance problem

Khalil et al, 2007

Insurance could be a natural partner for innovative water resources management based on

probabilistic climate information

1. Monitoring and Short-term (several days) projections

2. Seasonal Prediction (next 3-6 months)

3. Merging knowledge on natural multidecadal (e.g. 10-30 years) and global change for water resources management

Management options at different timescales of the available information