Seasonal to InterannualHydroclimate Forecasts

Portland, Oregon USA | 28-31 July 2013

AGU Chapman Conference on Seasonal to InterannualHydroclimate Forecasts

Portland, Oregon28-31 July 2013

2

Note: Attendees at the Chapman conference may be photographed by AGU for archival and marketing purposes. No photography will be permitted during scientific sessions.

Sankar ArumugamNorth Carolina State Universitysankar_arumugam@ncsu.edu

Balaji RajagopalanUniversity of Coloradobalajir@colorado.edu

Andrew WoodNOAA/NWSColorado Basin River Forecast CenterAndy.wood@noaa.gov

John SchaakeConsultantjcschaake@comcast.net

Program CommitteeCHAIR: Hamid MoradkhaniPortland State Universityhamidm@cccs.pdx.edu

Lifeng LuoMichigan State Universitylluo@msu.edu

Andy RobertsonColumbia Universityawr@iri.columbia.edu

Martyn ClarkNCARmclark@ucar.edu

Ximing CaiUniversity of Illinoisxmcai@illinois.edu

Tom PaganoBureau of Metorologyclimatron@gmail.com

Katrina GrantzUS Bureau of Reclamationkgrants@uc.usbr.gov

Alan RobersonAmerican Water Works Associationaroberson@awwa.org

Levi BrekkeUS Bureau of ReclamationLBREKKE@do.usbr.gov

David RaffU.S. Army Corps of Engineersdavid.raff@usace.army.mil

Dan SheerHydrologicsdsheer@hydrologics.net

Ryan BoylesNorth Carolina State Universityryan_boyles@ncsu.edu

Ashish SharmaUniversity of New South Walesa.sharma@unsw.edu.au

Eric PytlakBonneville Power Administrationespytlak@bpa.gov

Karl KansbergU.S. Army Corps of Engineerskarlis.kanbergs@usace.army.mil

Peter BrooksU.S. Army Corps of Engineerspeter.f.brooks@usace.army.mil

Conveners

3

AGU Chapman Conference on Seasonal to InterannualHydroclimate Forecasts

Meeting At A Glance Sunday, July 2812:30 p.m. – 5:30 p.m. Optional Field Trip to Bonneville Dam6:00 p.m. – 7:30 p.m. Welcome Reception

Monday, July 298:20 a.m. – 10:00 a.m. Session – Climate to Streamflow Forecasting at Seasonal to Interannual

Time Scales I10:00 a.m. – 10:30 a.m. Coffee Break10:30 a.m. – 12:30 p.m. Session - Climate to Streamflow Forecasting at Seasonal to Interannual

Time Scales II12:30 p.m. – 1:30 p.m. Group Lunch1:30 p.m. – 3:30 p.m. Poster Session - Climate to Streamflow Forecasting at Seasonal to

Interannual Time Scales 3:00 p.m. – 3:30 p.m. Coffee Break3:30 p.m. – 4:10 p.m. Climate to Streamflow Forecasting at Seasonal to Interannual

Time Scales III6:30 p.m. – 8:30 p.m. Banquet Dinner and Program

Tuesday, July 308:20 a.m. – 10:00 a.m. Session – Streamflow Forecasting at Seasonal to Interannual Time Scales10:00 a.m. – 10:30 a.m. Coffee Break10:30 a.m. – 12:30 p.m. Session – Risk Management Tools and Probabilistic Frameworks for Water

Management I12:30 p.m. – 1:30 p.m. Group Lunch1:30 p.m. – 3:30 p.m. Poster Session – Climate Forecasts Applications and Risk Management

Framework 3:00 p.m. – 3:30 p.m. Coffee Break3:30 p.m. – 4:10 p.m. Session – Risk Management Tools and Probabilistic Frameworks for Water

Management II

Wednesday, July 318:20 a.m. – 10:00 a.m. Session – Water Supply and Quality Management Applications of Climate

and Streamflow Forecasts10:00 a.m. – 10:30 a.m. Coffee Break10:30 a.m. – 11:50 a.m. Session – Moving Forward: What is Required to Make Better use of

Medium-range Weather and Climate Information in Water ResourcesManagement? I

12:00 p.m. – 1:00 p.m. Group Lunch1:00 p.m. – 2:00 p.m. Session – Moving Forward: What is Required to Make Better Use of

Medium-range Weather and Climate Information in Water ResourcesManagement? II

2:00 p.m. Adjourn

Poster Presentation GuidelinesPoster sessions are active on Monday, 29 July and Tuesday, 30 July, 1:30 p.m. – 3:30 p.m.

The poster dimensions are 4 feet high by 8 feet wide. All boards will be marked with the appro-priate poster number and category.

Those presenting their posters on Monday are asked to set up their posters on Sunday, 28 Julybetween the hours of 4:00 p.m.-6:00 p.m. Posters must be removed by 7:00 p.m., on Monday, 29July.

Those presenting their posters on Tuesday are asked to set up their posters on Tuesday, 30 Julybetween the hours of 8:00 a.m.-12:00 p.m. Posters must be removed by 12:00 p.m. on Wednes-day, 31 July.

4

5

SUNDAY, 28 JULY

12:30 p.m. – 5:30 p.m. Optional Field Trip - Bonneville Dam

6:00 p.m. – 7:30 p.m. Welcome Reception

MONDAY, 29 JULY

Climate to Streamflow Forecasting at Seasonal toInterannual Time ScalesPresiding: Andrew W. Robertson, Hamid MoradkhaniColumbia Falls

8:20 a.m. – 8:40 a.m. Lisa Goddard | Seasonal-to-Interannual Forecasts for Use inDecision Systems

8:40 a.m. – 9:00 a.m. Lifeng Luo | Seasonal streamflow prediction with VIC and ClimateForecast System version 2

9:00 a.m. – 9:20 a.m. Joshua K. Roundy | The optimal time and space scale fordownscaling the CFSv2 forecast for seasonal hydrologic predictions

9:20 a.m. – 9:40 a.m. Carly Tozer | Identification of Drivers of Rainfall Variability forInforming Seasonal Forecasting Schemes

9:40 a.m. – 10:00 a.m. Xing Yuan | The role of climate models in global seasonalhydrologic forecasting

10:00 a.m. – 10:30 a.m. AM Coffee Break (Monday)

Climate to Streamflow Forecasting at Seasonal toInterannual Time Scales IIPresiding: Ashish Sharma, Lifeng LuoColumbia Falls

10:30 a.m. – 10:50 a.m. David Gutzler | Hydroclimatic Forecasting in Southwestern U.S.River Basins

10:50 a.m. – 11:10 a.m. Chetan Pandit | Introducing EHP in India

11:10 a.m. – 11:30 a.m. Julien Boé | Theoretical and practical predictability of decadalvariations in river discharges in France and implications for watermanagement

SCIENTIFIC PROGRAM

11:30 a.m. – 11:50 a.m. Mohammed Bari | Operational streamflow forecasting at differenttime scales – hours to seasons to decades

11:50 a.m. – 12:10 p.m. Mohammad Reza Najafi | Evaluating Multi-Modeling Techniqueswith Varying Complexities for Seasonal Hydrologic Forecasts

12:10 p.m. – 12:30 p.m. Andrew J. Newman | Impacts of Hydrologic Model Uncertainty onthe Skill of Seasonal Streamflow Forecasts

12:30 p.m. – 1:30 p.m. Group Lunch (Monday)

1:30 p.m. – 3:30 p.m. Climate to Streamflow Forecasting at Seasonal toInterannual Time Scales PostersPoster Hall

M-1 Li-Chuan Chen | Seasonal Runoff Forecasts Based on the ClimateForecast System Version 2

M-2 Takeshi Doi | Why was the prediction of the 2012 positive IndianOcean Dipole Mode difficult?

M-3 Thomas Mosier | Downscaling Global Climate Data: Productionand Optimization of 30 Arc-Second Surfaces

M-4 Shraddhanand Shukla | Forecasting East Africa Spring rainfall atSeasonal Scale using a Hybrid Approach

M-5 Julie Vano | Mapping the diversity of hydrologic responses toseasonal climate in the Pacific Northwest

M-6 Mengqian Lu | Space-time structure of tropical moisture exportsand their precursors associated with high precipitation inducedfloods

M-7 Shahrbanou Madadgar | A Probabilistic Framework for Predictingthe Spatial Variation of Future Droughts

M-8 Mohammad Reza Najafi | The Impact of Climate Change onRunoff Extremes Based on Regional Climate Models andHierarchical Bayesian Modeling

M-9 Andrew M. Chiodi | An OLR perspective on La Nina and El Ninoprecipitation impacts over North America

M-10 Moty Cohen | Depth-Area-Duration analysis for an EasternMediterranean basin

M-11 Tanvir H. Bhuiyan | Ensemble Methods for Seasonal StreamflowPrediction

M-12 Lianne Daugherty | Probabilistic Operational Forecasting in theSan Juan Basin using Stochastic Weather Generator based SeasonalEnsemble Streamflow Forecasts

6

M-13 Caleb M. DeChant | Understanding the Effects of Initial Conditionand Model Structural Uncertainty in Seasonal HydrologicalForecasts with Data Assimilation and Bayesian Model Averaging

M-14 Ben Livneh | From catchments to regional scales: Hydrologicimpacts of land cover disturbances in the Upper Colorado RiverBasin

M-15 Katherine Lownsbery | Development of Seasonal Forecasting ofUpper Niger River Streamflow

M-16 Golnazalsadat Mirfenderesgi | A Comparative Assessment ofParticle Filtering-Markov Chain Monte Carlo and ParticleSmoothing Methods for Seasonal Hydrologic Forecasting andUncertainty Quantification

M-17 Sarah Whateley | Seasonal Hydrologic Forecasting in theNortheastern United States

M-18 Hongxiang Yan | A Regional Bayesian Hierarchical Model for FloodFrequency Analysis in Oregon

M-19 Ying Zhang | Development of a rainfall type prediction model forNYC

M-20 Nina Caraway | Advancing Ensemble Streamflow Prediction withStochastic Meteorological Forcings for Hydrologic Modeling

M-21 Rajarshi Das Bhowmik | Multivariate Downscaling of DecadalClimate Change Projections Over the Sunbelt

M-22 Seung Beom Seo | Near-term Climate Change Impacts on SurfaceWater and Groundwater Interactions Over the Sunbelt

3:00 p.m. – 3:30 p.m. PM Coffee Break (Monday)

Climate to Streamflow Forecasting at Seasonal toInterannual Time Scales IIIPresiding: David Raff, Edith ZagonaColumbia Falls

3:30 p.m. – 3:50 p.m. Ashish Sharma | THE DO’S AND DON’T’S OF COMBININGMODEL PREDICTIONS FOR SEASONAL HYDROLOGICFORECASTING

3:50 p.m. – 4:10 p.m. Andrew W. Wood | The potential value of NCEP climate forecastsfor hydrologic prediction in the Pacific Northwest

4:20 p.m. – 5:00 p.m. Climate to Streamflow Forecasting a Seasonal toInterannual Time Scales III

7

Dinner Program Introductions - Sankar ArumugamWelcome - Hamid MoradkhaniPresiding: Sankar ArumugamWillamette Ball Room

6:30 p.m. – 7:30 p.m. Upmanu Lall | The Columbia Global Flood Initiative — fromClimate Causal Modeling to Supply Chain Risk Management

TUESDAY, 30 JULY

Streamflow Forecasting at Seasonal to Interannual TimeScalesPresiding: Martyn ClarkColumbia Falls

8:20 a.m. – 8:40 a.m. Jan Danhelka | IMPLEMENTING THE EXTENDEDHYDROLOGICAL PREDICTION: A PAST AND FUTURE WMOACTIVITIES

8:40 a.m. – 9:00 a.m. Christian Zammit | Effect of initial conditions of a catchment onseasonal streamflow prediction using ensemble streamslowprediction technique (ESP) for 2 river basins on New Zealand’sSouth Island

9:00 a.m. – 9:20 a.m. Daehyok Shin | Seasonal Hydrologic Forecasting for WaterResources Management

9:20 a.m. – 9:40 a.m. Bart Nijssen | Relative Contributions of the Sources ofUncertainties in Seasonal Hydrologic Prediction Globally

10:00 a.m. – 10:30 a.m. AM Coffee Break (Tuesday)

Risk Management Tools and Probabilistic Frameworks forWater Management IPresiding: Ximing Cai, Levi BrekkeColumbia Falls

10:30 a.m. – 10:50 a.m. Carlos H. Lima | SEASONAL STREAMFLOW FORECASTS FORTHE BRAZILIAN HYDROPOWER NETWORK: PRESERVING THESPATIO-TEMPORAL VARIABILITY IN STATISTICAL MODELS

10:50 a.m. – 11:10 a.m. Tushar Sinha | Experimental Inflow and Storage Forecasts Portalfor Major Reservoirs in the Southeastern US

11:10 a.m. – 11:30 a.m. Mengqian Lu | Multi-time scale Climate Informed StochasticHybrid Simulation-Optimization Model (McISH model) for Multi-Purpose Reservoir System

8

11:30 a.m. – 11:50 a.m. Andrew W. Robertson | Combining seasonal climate forecasts withstochastic simulation of interannual-to-interdecadal streamflowvariability for reservoir Optimization over NW India

11:50 a.m. – 12:10 p.m. Leon Basdekas | STOCHASTIC STREAMFLOW GENERATIONFOR INTER-ANNUAL MUNICIPAL WATER SUPPLY PLANNING

12:10 p.m. – 12:30 p.m. Nathalie Voisin | Improved Medium Range to Seasonal StreamflowForecasts Through Simultaneous Assimilation of Snow andStreamflow Observations: Evaluation Over the Feather River Basin,CA

12:30 p.m. – 1:30 p.m. Group Lunch (Tuesday)

1:30 p.m. – 3:30 p.m. Climate Forecasts Applications and Risk ManagementFramework PostersPoster Hall

T-1 Indrani Pal | Detecting the Shift in Timing of SeasonalHydrological Cycle in India and Understanding the DynamicalAssociations

T-2 Kara DiFrancesco | Inclusion of climate change projections intoflood frequency analysis to assess the robustness of proposedmanagement actions – application to the American River,California, USA

T-3 Leon Basdekas | Using Multi-objective Optimization forInterdecadal Drought Planning

T-4 Yeonjoo Kim | A robust decision making framework for integratedwatershed management under climate change: a case study in aKorean urban watershed

T-5 Maria C. Mateus | Vulnerability of water resources with changingland use, and climate in the Santiam River Basin, Oregon

T-6 James L. McCreight | Global Positioning System (GPS) observationsof hydrologic variables for water management: current and futureapplications

T-7 Augustina Odame | Water-Saving Technology Adoptions underEcological and Economic Uncertainty

T-8 Sunal Ojha | Role of snow cover and its Hydrological impact onHimalayan River basins

T-9 Adriana D. Piemonti | ANALYSIS OF STAKEHOLDERATTITUDES ON OPTIMIZED WATERSHED MANAGEMENTPLANS

T-10 Mehmet U. Taner | Use of seasonal forecasting for improvingreservoir operations in the Niger River Basin

9

T-11 Logan Callihan | A Robust Decision Making Technique for WaterManagement Under Uncertainty Due to Climate Variability

T-12 Ashish Sharma | AN UPPER LIMIT TO SEASONALSTREAMFLOW PREDICTABILITY?

T-13 Amirhossein Mazrooei | Inter-model Comparison of SeasonalStreamflow and Soil Moisture Forecasts for the US Sunbelt

T-14 Maryam Pournasiri Poshtiri | Understanding the Droughts in theMajor River Basins of the U.S. and their Dynamical Connections:Implications for Water Resources Management

T-15 Majid Shafiee-Jood | Assess the value of seasonal forecast tomitigate farmers’ losses from drought

T-16 Pablo A. Mendoza | A multisite ensemble seasonal streamflowforecasting framework for semi-arid Andean basins

T-17 Juan J. Nieto | Establishment of a regional Hydrological Outlookfor the West Coast of South America

T-18 Zhaohui Lin | Seasonal Hydrological Ensemble Prediction SystemOver Huaihe River Basin and Its Preliminary Verification

3:00 p.m. – 3:30 p.m. PM Coffee Break (Tuesday)

Risk Management Tools and Probabilistic Frameworks forWater Management IIColumbia Falls

3:30 p.m. – 3:50 p.m. Edith Zagona | Tools and Techniques for Complex WaterManagement Models on Interannual to Multidecadal Time Scales

3:50 p.m. – 4:10 p.m. Upmanu Lall | Innovations in Climate Informed Water Supply andDemand Forecasts and Water System Management

4:20 p.m. – 5:20 p.m. Climate Forecasts Applications and Risk ManagementFramework Panel

10

WEDNESDAY, 31 JULY

Water Supply and Quality Management Applications ofClimate and Streamflow ForecastsPresiding: Andrew W. RobertsonColumbia Falls

8:20 a.m. – 8:40 a.m. Laila B. Parker | Adaptively managing variable environmental flowreleases from a water supply reservoir into First Herring Brook,Scituate, MA, in response to unexpected climactic conditions

8:40 a.m. – 9:00 a.m. Tirusew Asefa | Challenge and opportunities in usingHydroclimate Predictions to drive weekly to seasonal scaledecisions: A water supply utility’s perspective

9:00 a.m. – 9:20 a.m. Benjamin L. Harding | A Probabilistic Seasonal Forecasting Systemfor Water Utilities

9:20 a.m. – 9:40 a.m. Daniel P. Sheer | Using Forecast Information to Improve WaterManagement

10:00 a.m. – 10:30 a.m. Coffee Break (Wednesday)

Moving Forward: What is Required to Make Better use ofMedium-range Weather and Climate Information in WaterResources Management?Presiding: Hamid Moradkhani, Lisa GoddardColumbia Falls

10:30 a.m. – 10:50 a.m. Alan Roberson | Improving Municipal Water Demand Forecasting

10:50 a.m. – 11:10 a.m. Erik Pytlak | Water Supply Forecast Techniques and Challenges onthe Columbia River

11:10 a.m. – 11:30 a.m. Kevin Werner | Forecasting the Life Blood of America’s Southwest:Challenges for the Next Decade

11:30 a.m. – 11:50 a.m. Paul Trimble | Application of Seasonal and Multi-seasonal ClimateOutlooks for Water Management in South Florida

1:00 p.m. – 2:30 p.m. Group Lunch (Wednesday)

1:00 p.m. – 2:00 p.m. Moving Forward: What is Required to Make Better Use ofMedium-range Weather and Climate Information in WaterResources Management?

11

Asefa, TirusewChallenge and opportunities in using HydroclimatePredictions to drive weekly to seasonal scaledecisions: A water supply utility’s perspectiveAsefa, Tirusew1; Adams, Alison1

1. Tampa Bay Water, Clearwater, FL, USA

Tampa Bay Water is the largest wholesale water providerin the Southeast U.S. and with its Member Governments(Hillsborough, Pasco and Pinellas Counties and Cities ofNew Port Richey, Tampa, and St. Petersburg) serves over 2.3million water customers. It provides potable water using aunique blend of ground water, surface water and seawatersources. Over the past decade, average annual delivery hasranged between 222 mgd to 262 mgd. In the last three years,45% to 65% of supply came from groundwater sources, 35%to 45% from surface water sources and 1% to 9% fromdesalinated seawater. As an “on demand” water provider, theutility forecasts water supply availability and expected waterdemands from weekly to seasonal scales. It has developed asuite of decision support tools, several of which areprobabilistic, to account for uncertainties in input, such asrainfall, and the process being modeled. Its short-termmodels use forecast product such as Weather Prediction’sCenter’s (WPC) Quantitative Precipitations Forecast (QPF),whereas seasonal models make use of established large-scaleteleconnections and ENSO ensemble outlooks. The short-term operational objective, given the seasonal outlook andsurface water availability, is optimizing well field operationsto meet projected demand. Seasonal to annual objectives,based on planning and management of different supplysources, include cost-efficiency and environmentallysustainability goals. This presentation highlights the use ofhydroclimate predictions to manage a complex water supplysystem and identifies the space-time gaps, accuracy, anduncertainty of forecast products as they propagate throughvarious decision support tools.

Bari, MohammedOperational streamflow forecasting at differenttime scales – hours to seasons to decadesBari, Mohammed1; Tuteja, Narendra1; Enever, David1;Perkins, Jeff1; Jayasuriya, Dasarath (Jaya)1; Feikema, Paul1

1. Climate and Water Division, Bureau of Meteorology, WestPerth, WA, Australia

The Australian Government has assigned the Bureau ofMeteorology with national responsibilities for waterforecasting services through a legislative mandate under theWater Act 2007. These forecasting services complement andadd to other services for weather, climate and oceans. Theprolonged drought in recent decades has been followed bythree successive very wet years since 2010 in eastern

Australia. Demands for domestic, agricultural and industrialwater use have increased significantly over the past fewdecades. These factors have made improved water availabilityforecasts more important for water resources planning andoperations. Water availability forecast services at the Bureaurange from hours to days to seasons to several decades. Thehourly forecasting service includes extremes of rainfall andflood. This service provides real time operational forecasts ofriver heights and enables the public to take protective actionacross Australia (http://www.bom.gov.au/australia/flood/).The short-term (up to 7 days) streamflow forecast service isan extension of the flood forecasting service. This servicecombines hydrologic and weather forecasts and provideadvanced notice of high flow events, emergency storagemanagement and environmental water releases. The seasonalforecasting service is currently delivering 3-months totalstreamflow outlooks at 50 key water supply catchmentsacross eastern Australia (http://www.bom.gov.au/water/ssf).Many water sectors currently benefit from the seasonalforecasting service, including water allocations, croppingstrategies development, water markets planning, reservoiroperations, setting restrictions and environmental release.The long term forecasting service covers time spans of yearsto decades. This service is being developed to provide wateravailability forecasts with an emphasis on the next 10 to 30years and focus on regions considered more vulnerable toextremes in water availability. As a precursor of the longterm water forecasting service, a set of 221 HydrologicReference Stations has been identified for streamflowchange detection and attributions(http://www.bom.gov.au/water/hrs/). Using climateprojections from the Fifth Assessment Report of the IPCC,the Bureau plans to make a comprehensive nationalstatement on the long-term trends in water availability forall Hydrologic Reference Stations and major water sourcecatchments. These hydrological forecasting services are beingused for enhancing the existing drought service(http://www.bom.gov.au/climate/drought/), including themonitoring and prediction of meteorological, agriculturaland hydrological drought. Important elements of theoperational services include partnership development andactive and cooperative engagement of users andstakeholders. The Bureau is also leveraging complex researchand development collaboration – in particular with CSIROthrough the Water Information Research and DevelopmentAlliance (WIRADA). Other collaborators include Centre forAustralian Weather and Climate Research (CAWCR), eWaterCRC, and academic institutions.

12

ABSTRACTSlisted by name of presenter

Basdekas, LeonSTOCHASTIC STREAMFLOW GENERATION FORINTER-ANNUAL MUNICIPAL WATER SUPPLYPLANNINGBasdekas, Leon1; Stewart, Neil1, 2; Balaji, Rajagopalan2

1. Colorado Springs Utilities, Colorado Springs, CO, USA2. University of Colorado Boulder, Boulder, CO, USA

During periods of drought, near term (season to 24months) planning by municipal water supply providers isone of several tools used to assess and address risk.Generation of stochastic flows used to inform decisionmakers on potential near term risk is an important step inthis process. In this process, climate-based seasonalstreamflow forecast for the near term are combined withhistorical averages for the subsequent 18 months or so toproduce a streamflow sequence that is 24 months long.Thus, a produced sequence is used to drive water resourcessystems model for planning and management decisions. Themain drawback of this approach is that often historicalaverages are wetter than the near term seasonal forecasts,thus suggesting an optimistic future to the planners thereby,biasing the planning process. We propose two differentmethods to simulate streamflow sequences for the 2-yearperiod following the current seasonal forecast that bettercaptures potential variations. These methods are the K-nearest neighbor (KNN) and a variation of the K-nearestneighbor approach (KNN Plus) that uses the recenthydrologic conditions as the feature vector to selecthistorical neighbors and consequently the simulations. Forexample, if the recent hydrologic state is drier then thestochastic simulations will be biased towards drier flows.Thus, it can produce consecutive dry years such as thoseproduced in current climate can be simulated, which areimportance for risk based planning. By adding in a memorycomponent, this potential new climate regime can be bettercaptured even without a similar situation in the observedrecord. As part of continuing drought response measures,Colorado Springs Utilities (CSU) is interested in modelingthe variation in potential future flows, which is not capturedwell with typical methods. We apply these methods for CSUwater resources near term planning.

Basdekas, LeonUsing Multi-objective Optimization forInterdecadal Drought PlanningBasdekas, Leon1; Stewart, Neil1, 3; Triana, Enrique2

1. Colorado Springs Utilities, Colorado Springs, CO, USA2. MWH Americas, Ft. Collins, CO, USA3. University of Colorado Boulder, Boulder, CO, USA

Colorado Springs Utilities (CSU) is currently engaged inan Integrated Water Resource Plan (IWRP) to address thecomplex planning scenarios currently faced by CSU. Themodeling framework developed for the IWRP uses a flexibledata-centered Decision Support System (DSS) with aMODSIM-based modeling system to represent the operation

of the current CSU raw water system coupled with a state-of-the-art multi-objective optimization algorithm. The longterm planning effort and the ongoing drought in theWestern U.S. provided an opportunity to utilize analyticaltools not previously available for CSU drought planning. Wedescribe a new drought metric that captures duration andintensity. This drought metric was then used to filter over20,650 10 year time series down to 2328 potential droughttime series for system modeling. Those timeseries thatproduced system demand shortages were further analyzed inorder to select a stochastic ensemble of timeseries for multi-objective optimization. Of CSU’s 37 reservoirs or reservoiraccounts, nine high priority reservoirs were identified for useas preferred storage sites during drought periods. Asophisticated mutli-objective optimization approach wasused to identify individual and system storage pool volumes,identify drought response triggers and associated droughtpolicy responses. We utilize multiple system performancemetrics such as demand shortage, volume of water savedduring restrictions, and reservoir reliability as optimizationobjective functions to be either minimized or maximized.The goal of this multi-objective optimization is creating a setof non-dominated solutions for which the multiple objectivemetrics collectively could not be improved and for which thetrade-offs in satisfying the different objectives can bequantified. There is opportunity for adapting thismethodology to an interannual time frame.

Bhuiyan, Tanvir H.Ensemble Methods for Seasonal StreamflowPredictionBhuiyan, Tanvir H.1; French, Mark N.1

1. Civil & Environmental Engg., University of Louisville,Louisville, KY, USA

Hydrological extremes such as floods and droughts aretwo of the more hazardous natural disasters in context ofeconomic loss and human impact. The frequency and theseverity of these events are related to precipitation extremesand linked to climate variation. Prediction of floods anddroughts is challenging yet can provide benefits in terms ofmanaging irrigation water requirements, reservoiroperations, and planning for recovery. Accurate forecastsdepend on identification of relevant variables anduncertainties involved in prediction methodologies.Uncertainties in seasonal flow predictions can be consideredin ensemble forecast methods which combine a variety ofmodel predictions and allows the variability of estimation tobe considered. In this study, methods including time seriesand neural network (NN) are evaluated for seasonal flowforecasts ensembles. Forecasts based on persistentcharacteristics are included as a baseline performancecriterion. Initially, relationships between the seasonal flowvariations of the five selected rivers namely Congo, Yangtze,Rhine, Columbia, and Parana, and fluctuations of externalenvironmental variables such as El Niño, La Niña episodes,sun spot numbers, Pacific Decadal Oscillation, and NorthAtlantic Oscillation are quantified. River flows and external

13

variables (forcing variables) are incorporated in stochastictime series and NN methods to examine the ability toreproduce seasonal variability evident in historical flows.The forcing variables are expected to contribute to naturalstreamflow conditions. Model performance is evaluated forimprovement in prediction skills, while includingmultivariate approaches with river flow and externalvariables. Overall model performance indicates inclusion ofriver flows and forcing variables on average improve modelperformance. To further assess model efficiency forpredicting seasonal flow, both stochastic and NN modelforecasts are categorized into high, average and low flowlevels. Persistence model results are often comparable tomore complex multivariate approaches, indicating theinclusion of multiple variables does not always improvemodel performance. An ensemble approach mergingstochastic and the NN based forecasts with persistence aredeveloped to forecast seasonal river flow. Results indicateensembles forecasts are useful to improve seasonal flowsestimates. Variation of forecast magnitudes based on model-type underscores the usefulness of utilizing multi-modelensembles. The ensembles of the time series and the NNmodel predictions indicate seasonal flow variations are well-captured in the ensemble range. The ensemble approachappears promising for seasonal flow predictions in flowextremes associated with climate variation.

Boé, JulienTheoretical and practical predictability of decadalvariations in river discharges in France andimplications for water managementBoé, Julien1; Habets, Florence2, 3

1. CNRS/CERFACS URA1875, Toulouse, France2. Geosciences Department, MINES ParisTech,

Fontainebleau, France3. CNRS/UPMC, UMR 7619 Sisyphe, Paris, France

River discharges in France exhibit large decadalvariations, especially in spring. The existence of such low-frequency variations raises two important questions in thewater management context. (i) Are we able to evaluatecorrectly and take into account the uncertainties that areassociated with those decadal variations for watermanagement. (ii) Are those decadal variations theoreticallyand practically predictable? A positive answer to the secondquestion would open the door to the elaboration of moreefficient water management strategies at the inter-annualand decadal time-scales. If decadal variations are notpredictable, it is even more important to answer the firstquestion. For example, internal decadal variations instreamflows can either seriously aggravate or moderate theimpact of global warming on water resources during thenext decades, and the entire possible range of variations hasto be taken into account in the water management context.As a first step to answer those questions, it is necessary tounderstand the mechanisms responsible for the decadalvariations in river discharges in France. We show that thedecadal variations in streamflows are principally related to

the modulation of large-scale atmospheric circulation by theAtlantic Multidecadal Oscillation (AMO), the main mode ofdecadal climate variability in the North Atlantic region. Thisresult suggests that it is theoretically possible to predict apart of the the decadal variations in streamflows, thanks todecadal climate predictions. Decadal climate prediction is anew and experimental branch of climate modelling thatconsists in using observations to initialize the ocean incoupled climate simulations in the hope to predict theinternal low-frequency variations of the climate system. Wewill assess whether a state-of-the-art decadal predictionsystem (based on the CNRM-CM5 coupled climate modelfrom the Coupled Model Intercomparison Project Phase 5)shows any skill in the retrospective prediction ofprecipitation over France and/or of large-scale circulation,that could then be used with a statistical downscalingmethod to obtain the atmospheric forcing necessary tosimulate river discharges. As current decadal predictionsystems generally show more skill for oceanic variables, analternative approach will be tested, using the prediction ofthe AMO and the observed link between the AMO and riverdischarges. Finally, a purely statistical method, based on thelagged relationship between the AMO and river dischargeswill be investigated. If none of the approaches describedpreviously exhibit any skill, it is even more crucial to evaluateas well as possible the whole possible range of variationsassociated with internal low-frequency variability in riverdischarges during the next decades. The capacity of currentcoupled climate models to represent correctly the amplitudeof decadal variations in French hydroclimate will be assessedin order to determine whether the uncertainties they causein the projections for the next decades are correctly captured.

Callihan, LoganA Robust Decision Making Technique for WaterManagement Under Uncertainty Due to ClimateVariabilityCallihan, Logan1, 2; Zagona, Edith1, 2; Rajagopalan, Balaji1, 3

1. CEAE, University of Colorado, Boulder, CO, USA2. CADSWES, University of Colorado, Boulder, CO, USA3. CIRES, University of Colorado, Boulder, CO, USA

Robust decision making, fundamental to managingwater resources in light of deep uncertainties associated withclimate variability at inter-annual to multi-decadal timescales, is an analytical framework that detects when a systemis in or approaching a vulnerable state, and implementsstrategies to address the vulnerabilities that perform wellover a wide range of plausible future scenarios. Varioustechniques have been developed to identify vulnerableconditions and to select the options that are most favorablein given situations. To characterize strategies in terms oftheir potential success given various system states requiresextensive modeling of a wide range of conditions. Recentresearch(1) that increases our understanding of decadal scalevariability has the potential to improve decisions made inthis framework. This research develops a comprehensiverobust decision making framework that utilizes the power of

14

extensive modeling to develop relationships between climateindicators and future system performance to identify thetype and severity of vulnerable conditions, and evaluate a setof options and strategies in terms of benefits, costs,likelihood of mitigating vulnerable conditions, as well as theresiliency of system performance under a range of eventualconditions. The research utilizes the RiverSMART suite ofsoftware modeling and analysis tools developed underReclamation’s WaterSMART initiative and built around theRiverWare modeling environment. To provide a wide rangeof possible hydrologic futures, the framework generatesstochastic streamflow scenarios using a K-nearest neighbornonparametric method, resampling observed and paleoreconstructed hydrology; other data sets can be producedthat combine characteristics of the paleo, historic andclimate change projections using various Markov Chaintechniques(2). A case study is developed for the GunnisonBasin of the upper Colorado River. Various demandscenarios are projected and system performance indicatorsmeasure the ability of the system to meet water demands foragriculture, municipalities, environmental flows,hydropower and recreation. Options and strategies foraddressing vulnerabilities include such measures asconservation, reallocation, and adjustments to operationalpolicy. Projections that utilize teleconnections with decadalscale signals such as AMO are evaluated for improving theefficacy of the decisions. Results of extensive simulationsprovide both guidance for decision-making and evaluationof the effects of the decisions on performance and resiliencyof the systems. This methodology can also be readily adaptedto inter-annual and multi-decadal time scales. References: 1.Nowak, Kenneth C. (2011), Stochastic StreamflowSimulation at Inter-decadal Times Scales and Implicationsfor Water Resources Management in the Colorado RiverBasin, Civil, Environmental, and Architectural EngineeringPh.D. Dissertation, University of Colorado. 2. Prairie, J., K.Nowak, B. Rajagopalan, U. Lall, and T. Fulp, (2008), AStochastic Nonparametric Approach for StreamflowGeneration Combining Observational and PaleoReconstructed Data, Water Resources Research, 44.

Caraway, NinaAdvancing Ensemble Streamflow Prediction withStochastic Meteorological Forcings for HydrologicModelingCaraway, Nina1; Wood, Andrew2; Rajagopalan, Balaji3

1. University of Colorado, Boulder, Boulder, CO, USA2. Northwest River Forecast Center, Portland, OR, USA3. CIRES, University of Colorado, Boulder, Boulder, OR,

USA

River Forecast Centers of National Weather Service(NWS) produce seasonal streamflow forecasts via a methodcalled Ensemble Streamflow Prediction (ESP). NWS ESPforces the temperature index SNOW-17 and conceptualSacramento Soil Moisture Accounting model (SAC-SMA)models with historical weather sequences for the forecastingperiod, starting from models’ current watershed initial

conditions, to produce ensemble streamflow forecasts.Among several serious drawbacks of this method, two are: (i)the ensembles are limited to the length of historical record,limiting ensemble variability; and (ii) incorporating seasonalclimate forecasts (e.g., related to the El Nino SouthernOscillation) relies on adjustment or weighting of ESPstreamflow sequences. These drawbacks motivated theresearch presented here, which has two components: (i) amulti-site stochastic weather generator and (ii) generation ofensemble weather forecast inputs to the NWS models toproduce ensemble streamflow forecasts. We enhanced the K-nearest neighbor bootstrap based stochastic generatorinclude conditioning the weather forecasts on probabilisticseasonal climate forecast. This multi-site stochastic weathergenerator runs in R and the NWS models run within thenew Community Hydrologic Prediction System (CHPS), aforecasting sequence we label ‘WG ESP’. The WG ESPframework was applied to generate ensemble forecasts ofrunoff season (April-July) streamflow in the San Juan RiverBasin, one of the major tributaries of the Colorado River,USA, for the period 1981-2010. The hydrologic modelrequires daily weather sequences at 66 locations in the basin.Runoff season ensemble forecasts stepped throughNovember to April lead times for the period 1981-2010 andwere made from both WG ESP and traditional ESP. The WGESP approach provides a skillful and comprehensive varietyof flow ensembles compared to the ESP. Furthermore, itexhibited higher skill in predicting seasonal and monthlyflows at long lead times, particularly in wet years. Theflexible and robust framework augments the ESPframework, and provides a new linkage between climate andstreamflow prediction, two developments that will bevaluable for water resources management.

Chen, Li-ChuanSeasonal Runoff Forecasts Based on the ClimateForecast System Version 2Chen, Li-Chuan1; Shukla, Shraddhanand2

1. Earth System Science Interdisciplinary Center, Universityof Maryland, College Park, College Park, MD, USA

2. Department of Geography, University of California,Santa Barbara, Santa Barbara, CA, USA

In this study, we conduct an assessment of thehydrological drought predictability as derived by the use ofthe Climate Forecast System version 2 (CFSv2) forecasts. Wefound that direct runoff forecasts from CFSv2 not only havelarge biases but also limited skill to be used for operationaldrought prediction. Climatological analysis has revealed thatthe partition between the water balance components (e.g.,runoff and soil moisture) in Noah model, the only landsurface model employed in CFSv2, needs improvements. Wefurther investigate the usability of CFSv2 forecasts inhydrological drought prediction by evaluating the forecastskill of standardized cumulative runoff index simulated by amacroscale hydrologic model (the Variable InfiltrationCapacity model) forced with daily precipitation,temperature, and wind forecasts from CFSv2 (i.e.

15

hydroclimate forecasts denoted as CFSv2_VIC). The skill ofCFSv2_VIC forecasts is compared with respect to baselineforecasts generated using the Ensemble StreamflowPrediction (ESP) method (i.e. ESP_VIC), that derives its skillsolely from the knowledge of the initial hydrologicconditions. We observed that the relative skill of CFSv2_VIC(w.r.t ESP_VIC) varies throughout the year and with thelocation. In general, the skill of CFSv2_VIC forecasts isgreater than ESP_VIC over the areas where precipitation isdriven by large-scale circulation or dynamical forcing isactive, such as the Southwest monsoon region in July andthe eastern United States in the fall. The skill of CFSv2_VICforecasts is mostly increased from the baseline ESP_VICforecasts when and where precipitation forecasts are skillful.CFSv2-based hydroclimatic forecasts are a useful tool forreal-time hydrological drought prediction; however, theirskill is limited to specific seasons and regions.

Chiodi, Andrew M.An OLR perspective on La Nina and El Ninoprecipitation impacts over North AmericaChiodi, Andrew M.1; Harrison, Don E.2, 1

1. JISAO/ University of Washington, Seattle, WA, USA2. NOAA Pacific Marine Enviironmental Laboratory,

Seattle, WA, USA

El Niño-Southern Oscillation (ENSO) impacts onseasonal weather anomalies form the basis for skillfulstatistical seasonal weather prediction in the regions aroundthe globe where the statistical links between ENSO andseasonal weather anomalies are strong. A warm-ENSO (ElNiño) index based on outgoing longwave radiation (OLR)conditions in the tropical Pacific has recently been proposedand found to have a stronger statistical linkage to seasonalweather anomalies over the contiguous U.S. than thecommonly used ENSO indices, which rely on sea surfacetemperature (SST) or sea level pressure (SLP) conditions. Acomplimentary OLR-based cool-ENSO (La Niña) index isproposed and this pair of OLR-based ENSO indices isevaluated for their respective connections to interannualprecipitation and atmospheric circulation anomalies overNorth America using composite analysis. We find that overthe period for which satellite-based OLR observations areavailable, almost all of the useful (statistically significantand consistent from event to event) ENSO impacts onseasonal precipitation are due to the years distinguished bythe OLR-based ENSO indices (“OLR El Niño”; “OLR LaNiña”). On the other hand, composites based on other yearswith ENSO status based on the current NOAA definition(“non-OLR El Niño”; “non-OLR La Niña”) do not havenearly as robust or statistically significant anomaly patternsas the OLR ENSO events. To the extent this observedbehavior holds in the coming decades, OLR diagnostics canserve as indicators of times in which greater confidence canbe placed in seasonal forecasts than has perhaps beenpreviously realized. It should be noted also, however, thatsuch times occur only in a subset of the conditions nowcommonly identified as ENSO years.

Cohen, MotyDepth-Area-Duration analysis for an EasternMediterranean basinCohen, Moty1

1. Soil and Water, Faculty of Agriculture, The HebrewUniversity of Jerusalem, Rehovot, Israel

The eastern Mediterranean climate is characterized(Goldreich, 2003) by winter rainfall ranging between 600mm (north) to 200 mm (south) isohyets. Several factorsdetermine the intra-annual variability of rainfall levels inthis region. The occurrence of Cyprus lows is the dominantfactor, and the Mediterranean oscillation index (MOI2) isalso a significant factor. Another key factor is an enhancedtendency of upper-level troughs to be oriented in asouthwest–northeast direction. Alpert et al. (2004) proposedthat rainfall may increase as a function of the number ofdays characterized by the Red Sea Trough, a system mainlyimpacting the rainfall potential in southern Israel. Beingsituated in the margins of these systems leads to highvariability and intermittency of rainfall. Knowledge ofrainfall depth and its temporal and spatial occurrence isessential for all hydrological research performed with thegoal of mitigating flood hazards. These include the physicalbasis of rainfall-runoff phenomena, as well as the design ofhydraulic structures. One approach is to assessmeteorological and synoptic parameters (Ziv et al. 2013),however from an engineering point of view, the use ofempirical tools is more practical. Traditionally, intensity-duration-frequency (IDF) curves are derived using pointvalues of the rain gauge measurements. This approach is notuseful for describing the spatial distribution of the rain, northe areal reduction resulting from the averaging of rainvalues over the whole area. To generate this type of analysis,depth-area-duration (DAD) analysis is used to correlaterainfall areal averaged depth with the size of the rainfallevent area. Analysis of a large number of storms of varyingduration enables one to assess the probable rainfall depthfor a given watershed size and response time. In addition tothe ability to accurately compute rain volumes and flooddischarge (critical for engineering/environmental design),this DAD analysis serves as a tool to characterize stormpatterns prevailing in a given area as a result of differentclimatologic regimes (Jolly et al. 2008). An extensive study ofthis type has been carried out in the US (Svensson and Jones,2010), but scant research of this sort has been done in Israel.This type of research is of particular interest since flooddischarge in areas such as the Yarkon watershed (800 km2)near Tel-Aviv, Israel may be catastrophic from anenvironmental, political, and economic standpoint as theyare very densely populated (with over 1.5 million residents).In this study, the 24-hr rainfall measured at 32 stations over59 years via gauge data was analyzed to derive fixed areaDAD curves. Thus, to characterize the spatial distribution ofrainfall events occurring over the Yarkon watershed Resultsof the analysis show that the maximal daily rainfall for theYarkon watershed is 172 mm. The areal reduction factor forthe entire watershed may reach 39% and 47% for a 2-year and

16

a 10-year return period, respectively. In a forthcoming study,such areal reduction functions based on point “groundtruth” measurements will be compared to similar functionsderived from radar measurements which may be a moreappropriate database with which to achieve this goal.

Danhelka, JanIMPLEMENTING THE EXTENDEDHYDROLOGICAL PREDICTION: A PAST ANDFUTURE WMO ACTIVITIESDanhelka, Jan1

1. Czech Hydrometeorological Institute, Praha, CzechRepublic

Extended hydrological prediction (EHP) has thepotential to bring a wide range of benefits in watermanagement and other fields. The successfulimplementation of EHP will require the bridging of severalgaps between various communities and groups as well asdevelopment of reliable operational procedures and toolsand training of hydrologists and water managers. WMOorganized workshop and trainings on EHP in the WesternCoast of South America 2010 and an Expert meeting on EHPin Melbourne in 2011. The recommendations from aMelbourne meeting include the compilation of case studiesand guidance on the needs and justification of EHP, itsscientific basis, the issue of service delivery mechanisms andformats, stakeholder consultation and finally the transferfrom research to operation. Based on the outcomes of theMelbourne meeting, the Commission for Hydrology (CHy)of WMO has included in its working plan (2013–2016) anaim to compile case studies and provide guidance on theapplication of EHP to water resources management. Theoverall aim is to increase the ability of National HydrologicalServices (NHSs) to provide and make use of EHP. The chainof activities within the field of EHP usually starts withclimate observations and/or outlooks, proceeds withhydrological modeling and ends at user’s interpretation orprocessing of the prediction output and finally, decisionmaking. From the institutional point of view these activitiespresent National Meteorological Services as climateinformation providers, the NHSs (hydrological modeling) asconduits for service delivery and water managers and othersas service users. The WMO has established the GlobalFramework for Climate Services (GFCS) as the frameworkfor the delivery of climate data, information and services.The water sector is one of the key focus areas of the GFCS. Itis assumed that hydrologists and water managers will be theusers of the GFCS and will provide user’s input and feedbackthrough the User Interface Platform (UIP). EHP couldbenefit from the feedback and guidance provided throughthe UIP. However the EHP is also an excellent example ofpossible product provided by the hydrological communitythrough the GFCS. Relevant climate drivers of stream flowand their variability differ among various regionssignificantly. Accordingly the optimal method of extendedhydrological prediction may differ from basin to basin.Approach used for reservoir inflow forecast does not include

climate variables of future period but rely on simplifiedempirical water balance method of observed snow and rain.Monthly EHP in the Czech Republic accounts for 9 differentcombinations of expected future temperature andprecipitation (below-normal, normal, and normal) based onECMWF monthly forecast evaluation to select 40 ensemblemembers from synthetic 1000 y daily weather time series andits spatial and temporal downscaling to produce flowsimulations. Australia approach includes basin specificstatistical models using various climate and oceanoscillation indices as predictors. This contribution willpresent a template for case studies compilation andproposed structure of guidance for NHSs on implementingEHP and using GFCS. The participants of the conferencewill be encouraged to contribute to the case studiescatalogue and guidance.

http://www.wmo.int/pages/prog/hwrp/chy/index.php

Das Bhowmik, RajarshiMultivariate Downscaling of Decadal ClimateChange Projections Over the SunbeltDas Bhowmik, Rajarshi1; Arumugam, Sankar1; Patskoski,Jason1

1. North Carolina State University, Raleigh, NC, USA

Statistical downscaling of precipitation and temperatureis commonly required to bring the large scale variablesavailable from GCMs to a finer grid-scale that can beingested watershed models. Most of the currently employedprocedures on statistical downscaling primarily consider aunivariate approach by developing a statistical relationshipbetween large-scale precipitation/temperature with the local-scale precipitation/temperature ignoring theinterdependency between the two variables. In this study, aBayesian K-nearest neighbor (K-NN) approach is proposedthat obtains the posterior distribution of precipitation andtemperature at the local scale by preserving spatio-temporalcorrelation structure among these two variables. The studyalso shows an upper bound on the correlation between thedownscaled variables that are obtained based on univariatedownscaling using simple and asynchronous regressiontechniques. The proposed Bayesian K-NN approach will alsobe compared with simple univariate downscaling andasynchronous regression in preserving the multivariatecorrelation structure at the local scale by downscalingclimate models inter-comparison project-45 (CMIP5)projections over the Sunbelt.

17

Daugherty, LianneProbabilistic Operational Forecasting in the SanJuan Basin using Stochastic Weather Generatorbased Seasonal Ensemble Streamflow ForecastsDaugherty, Lianne1, 2; Zagona, Edith1, 2; Rajagopalan, Balaji1,3; Grantz, Katrina4; Miller, Paul5; Werner, Kevin5

1. CEAE, University of Colorado, Boulder, CO, USA2. CADSWES, University of Colorado, Boulder, CO, USA3. CIRES, University of Colorado, Boulder, CO, USA4. Bureau of Reclamation, Salt Lake City, UT, USA5. Colorado Basin River Forecasting Center, Salt Lake City,

UT, USA

Projections of reservoir conditions and operations ofmajor projects in the Colorado River Basin have beengenerated each month for many years by the Bureau ofReclamation (Reclamation) using the 24-Month Study, amonthly timestep deterministic model that incorporatesstreamflow forecasts produced by the National WeatherService (NWS), Colorado Basin River Forecast Center(CBRFC). Using an Ensemble Streamflow Prediction (ESP)method and a physically based hydrologic model, theCBRFC produces an ensemble of streamflow forecasts bysampling historical weather sequences conditioned on 3-7month seasonal climate forecasts starting from the model’scurrent initial conditions. Subsequent months use statisticalaverage conditions. The 24-Month Study is used mainly toprovide only the most probable (based on the 50%exceedance) reservoir operations projection and requiresmanual monthly reservoir operations input.(1) The twomain drawbacks of this forecasting process are i) the ESPmethod is limited by the limited variability of the historicalmeteorological record; and ii) the deterministic nature of theoperations model does not provide a probabilistic range ofthe forecast. The goal of this research is to join the enhancedESP technique and a probabilistic version of the monthlyreservoir operations model to demonstrate improvedforecasting skill at long lead times along with quantifieduncertainty that can be used in decision-making. Caraway(2) developed an enhanced ESP method that uses a K-nearestneighbor bootstrap based stochastic weather generator (WG)combined with a probabilistic seasonal climate forecast. Thegenerated weather sequences are coupled with the SAC-SMAmodel within the NWS Community Hydrologic PredictionSystem (CHPS) to produce a weather-generated ensemblestreamflow forecast, referred to as WG based ESP whichshowed improved long lead skills compared to thetraditional ESP2. Reclamation is developing an ensemble-based operations model referred to as the Mid-TermOperations Model (MTOM), which is capable ofincorporating multiple hydrologic traces, simulatingreservoir operations, and thus producing probabilisticprojections of reservoir conditions. We apply the forecastingtechniques in the San Juan River Basin (SJRB) using aportion of the Colorado River MTOM. The springstreamflow ensembles from ESP, WG based ESP and thoseconditioned on seasonal climate forecasts, issued at sixdifferent lead times, are proposed to be incorporated into

yearly MTOM simulations from 2000-2010, using historicalvalues as initial conditions. Ensembles of reservoir levels andreleases at Navajo Reservoir will be obtained and their skillsevaluated against variables obtained using historicalstreamflow (i.e., baseline). Reference: 1.http://www.usbr.gov/uc/rm/crsp/wtrops.html 2. Caraway,Nina Marie, 2012, Stochastic Weather Generator BasedEnsemble Streamflow Forecasting, Masters thesis, Universityof Colorado at Boulder.

DeChant, Caleb M.Understanding the Effects of Initial Condition andModel Structural Uncertainty in SeasonalHydrological Forecasts with Data Assimilation andBayesian Model AveragingDeChant, Caleb M.1; Moradkhani, Hamid1

1. Portland State University, Portland, OR, USA

Uncertainties are an unfortunate but inevitable part ofany forecasting system. Within the context of seasonalhydrologic predictions, these uncertainties can be attributedto three causes: our imperfect characterization of initialconditions, an incomplete knowledge of future climate anderrors within computational models. In order to effectivelymanage these uncertainties, each of these three factors mustbe managed, providing a framework to reduce uncertaintyand accurately convey persistent predictive uncertainty. Todate, only partial accounting of uncertainty has beenperformed. Understandably, research and operationalforecast systems have emphasized climate uncertainties inseasonal predictions. Forecasted climate is arguably thedominant source of uncertainty in a hydrologic forecastingsystem, but the other sources of uncertainty are significant.Since these uncertainties are significant, a completerepresentation of forecast uncertainty requires directconsideration of both initial condition and model structuraluncertainties, in addition to climatic uncertainties. Recentresearch has examined the use of data assimilation tocharacterize initial condition uncertainty, but this has stillignored aspects of uncertainty related to model errors. Inorder to manage all three sources of uncertainty, this studyutilizes a combined data assimilation, to characterize initialcondition uncertainty, and model averaging system, toexamine model structural errors, to give a completedescription of seasonal hydrologic forecasting uncertainty.This presentation will highlight the need to advance thedescription of hydrologic uncertainty. An assessment of thereliability of current and proposed methods will be provided,along with an estimate of the relative uncertainty in thedifferent sources of uncertainty. Overall this assessment willquantify the importance of accounting for state and modeluncertainties in seasonal hydrologic forecasting.

18

DiFrancesco, KaraInclusion of climate change projections into floodfrequency analysis to assess the robustness ofproposed management actions – application to theAmerican River, California, USADiFrancesco, Kara1; Tullos, Desiree1

1. Biological and Ecological Engineering, Oregon StateUniversity, Corvallis, OR, USA

The scientific community currently lacks both reliableclimate projections at the temporal and spatial resolutionrequired for flood frequency analysis as well as methods toincorporate multiple, highly uncertain future scenarios intoflood frequency analysis. This study addresses the later issueby applying a new method using Bayesian statistics to assessthe potential changes in flood frequency and flood riskunder climate change. With this method, we calculateExpected Annual Damages over a range of plausible futureflood frequency curves derived from a Monte Carlo MarkovChain algorithm and use this information to determine theportion of future scenarios under which the system canmaintain damages below a threshold. We apply this methodto the American River Basin in California, USA to assess therobustness of proposed management approaches in the 2012Central Valley Flood Protection Plan. Each of the proposedapproaches meets the performance threshold under asmaller portion of projected future scenarios than undercurrent conditions and are vulnerable to greater than 70% ofthe plausible range of future conditions. While the futureprojections align with historic trends of increasing floodmagnitude and variability, due to the uncertainty associatedwith the currently available downscaled projections, theresults do not represent a predictive model of future floodconditions in the American Basin per say. Rather, themethod represents a general technique to incorporatemultiple sets of future projections into flood risk studies.

Doi, TakeshiWhy was the prediction of the 2012 positive IndianOcean Dipole Mode difficult?Doi, Takeshi1; Sasaki, Wataru1; Behera, Swadhin1;Masumoto, Yukio1; Yamagata, Toshio1

1. JAMSTEC, Yokohama, Japan

The Indian Ocean Dipole Mode (IOD) is now known tohave impacts on climate forecasts and water management ofthe world, in particular, Indian Ocean rim countries such asAustralia, India, and several east African countries. Theseasonal prediction system based on the SINTEX-F ocean-atmosphere coupled model has so far demonstrated goodperformance of prediction on the IOD. However, the systemhas failed to predict the 2012 positive IOD event 1-seasonahead for the first time since it became operational. We haveexplored the reason, and found that the ocean subsurfacetemperature initialization in April was the key. Theobservation showed warmer-than-normal temperature below50 m depth in the western equatorial Indian Ocean in April

2012. This sustained the warmer-than-normal sea surfacetemperature (SST) there in May. The active convectionassociated with this warm SST anomaly in the western polefavored a positive IOD condition, which evolved throughocean-atmosphere coupled feedback from June, to reach amature state in August. The SST-nudging initialization,however, could not capture the unique subsurfaceprecondition off East Africa and hence negatively influencedthe 2012 IOD prediction.

Goddard, LisaSeasonal-to-Interannual Forecasts for Use inDecision SystemsGoddard, Lisa1

1. International Research Institute for Climate & Society,Columbia University, Palisades, NY, USA

This talk will provide an overview of research anddevelopment efforts for seasonal climate forecasts at theInternational Research Institute for Climate & Society (IRI).The IRI has been producing probabilistic multi-modelforecasts since 1997, and the methodology behind thoseforecasts has evolved considerably over the past 15 years. Themethodology has always sought to emphasize pastperformance of the models, by region and season, and tominimize model-related errors. The goal is to extract themaximum useful information from multiple dynamicalmodels and construct reliable probability forecasts. I willdiscuss our current approach to correct and combine themodel output and to construct the probability densityfunctions (PDFs). The methodology involves a penalizedregression approach, which can account for sampling errorwhere skill is low and the historical record is short. Theavailability of full PDFs, rather than the more standardformats of tercile categories or a deterministic best-guess,allows us to disseminate the forecast in a much more flexibleformat. This enables forecast consumers to view forecastspertinent to their specific category or threshold. As anexample, results from a study of hydropower in Ethiopiashow that the use of reliable probabilistic forecasts keyed onuser-defined thresholds yield improved hydropowerreliability and economic benefit.

Gutzler, DavidHydroclimatic Forecasting in Southwestern U.S.River BasinsGutzler, David1

1. Dept of Earth & Planetary Sciences, Univ of New Mexico,Albuquerque, NM, USA

Major rivers in the southwestern U.S., such as theColorado and Rio Grande and their major tributaries, derivemuch of their annual flow from snowpack that falls at highelevation near their headwater catchments. The downstreamreaches of these rivers extend through the climatic domainof the North American monsoon system, which exhibits apronounced summer season precipitation maximum thatcan provide a significant flood pulse several months after

19

the spring snowmelt season. Streamflow forecasts on theserivers are largely based on observations of winter snowpack,with some additional predictability possible early in thewater year based on seasonal prediction skill for winterprecipitation. In recent decades, however, some (but not all)of the most severe winter precipitation and snowpackdeficits were followed by extremely wet monsoon seasons.Seasonal prediction skill for summer precipitation is stillproblematic, so water managers receive minimal guidance ofsummer hydroclimatic conditions beyond knowledge ofclimatology that would allow them to plan for severe deficitsor high flows associated with pronounced anomalies inmonsoonal precipitation. Projected climate changemagnifies the significance and current uncertaintiesinherent in summertime hydrologic forecasts. Streamflowsin southwestern rivers are projected to decline in associationwith warmer temperatures, due principally to decreasingsnowpack and increasing warm season evaporation.Although precipitation itself shows no significant long-termtrend, the ratio of (observed spring runoff)/(observed winterprecipitation) is decreasing as temperatures have risen inrecent decades. Summer precipitation projections representa major uncertainty in adapting global climate modelsimulations to streamflow projections, especially on lowflow tributaries in smaller drainage basins that can beespecially strongly affected in strong monsoon seasons. Thecurrent hydrologic situation in the Southwest unfortunatelyprovides an excellent example of the dilemma faced by watermanagers. Poor snowpack and low streamflow means thatdifficult surface water allocation decisions are now beingmade at the beginning of the growing season. More reliableguidance on summer precipitation, if it were available andtrusted by stakeholders and hydrologic forecasters, couldaffect these decisions.

Harding, Benjamin L.A Probabilistic Seasonal Forecasting System forWater UtilitiesHarding, Benjamin L.1; Gangopadhyay, Subhrendu2;Rajagopalan, Balaji3; Rodriguez, Alfredo4

1. AMEC Environment & Infrastructure, Boulder, CO, USA2. Bureau of Reclamation, Lakewood, CO, USA3. University of Colorado, Boulder, CO, USA4. Aurora Water, Aurora, CO, USA

Currently available climate and streamflow seasonalforecasts provide general guidance for water basinmanagement at a basin or sub-basin scale. However, to makeoperational planning and management decisions, managersof water-supply systems could greatly benefit from site-specific forecasts of relevant system variables (e.g.,end-of-water-year reservoir contents). Direct forecasting ofrelevant system variables requires the use of system-specificdata and methods that are not currently available in large-scale forecasts. These shortcomings motivated thedevelopment of a Probabilistic Seasonal Forecasting System(PSFS) for water utilities. The PSFS uses large scale climatedata, local hydrologic data and system data as the basis to

develop a probabilistic statement about the degree to whichconditions over a forecast period (i.e., through the end of anupcoming or current runoff season) will be similar toprevious years. The methodology for deriving theprobabilistic forecasts is based on the notion of similarity ofyears, which is derived using nearest neighbor algorithmsfrom non-parametric statistics. The PSFS produces acumulative distribution function (CDF) of system statevariables (e.g., reservoir contents). The PSFS can generate aforecast for streamflows or for any system variable that iscalculated by a system simulation model. The system hasbeen used to generate forecasts for water utilities along theFront Range of Colorado since water year 2008 (October 1,2007 through September 30, 2008) at five lead-times,October, and January through April. Cross validation offorecasts for the four utilities showed that the forecasts hadsubstantial skill at certain lead times when forecastconditions are in the lower (dry) and upper tercile (wet), butno skill in the middle tercile (average). A significantcomponent of forecast skill comes from antecedent reservoirstates. The skill exhibited by the PSFS, and other existingforecasts, is useful to water system managers, but issufficient only to condition operating decisions in thecontext of other information.

Kim, YeonjooA robust decision making framework for integratedwatershed management under climate change: acase study in a Korean urban watershedKim, Yeonjoo1; Chung, Eun-Sung2

1. Korea Environment Institute, Seoul, Republic of Korea2. Seoul National University of Science and Technology,

Seoul, Republic of Korea

An robust decision making framework has beendeveloped for integrated watershed management (IWM) torehabilitate the distorted cycles of water quantity and qualityin a changing climate. The procedure consists of nine stepsand includes several engineering, economic, and socialtechniques, which are multi-criteria decision making,sustainable development index, a continuous rainfall runoffsimulation model and stakeholder involvement. Based onthe driver-pressure-state-impact-response concept, the indexis used to assess spatial vulnerability with variousmanagement alternatives. A stakeholder participates in thequantification of preferences with regard to managementobjectives. This procedure was applied to the Korean urbanwatershed, which has suffered from streamflow depletionand water quality deterioration. As a result, robust strategiesare selected according the proposed framework. Thisresearch provides a useful IWM tool for incorporating suchquantitative and qualitative information into the evaluationof various policies with regard to water resource planningand management.

20

Lall, UpmanuThe Columbia Global Flood Initiative — fromClimate Causal Modeling to Supply Chain RiskManagementLall, Upmanu1

1. Earth & Environmental Eng, Columbia Univ, new york,NY, USA

This talk introduces the Columbia Global FloodInitiative with an invitation to the community to join in aneffort to bring hydroclimatic science and financial riskmanagement together to provide sustainable solutions tothe Earth’s most damaging natural hazard. The key buildingblocks for the initiative are: 1) the need to understand andpredict the causal mechanisms of extreme floods so that weare better prepared to adapt to climate variability; 2) thedevelopment of a Lagrangian framework for understandingthe role of atmospheric circulation mechanisms on moisturetransport and convergence associated with extreme,persistent flooding; 3) an understanding of the globalconcordance of flood and drought and its probabilisticrepresentation; 4) the application of this knowledge tocoordinated financial risk management and infrastructuredevelopment; and 5) the application to global and localsupply chain risk management in addition to local damageassessment. My thesis is that we need a transformation inthe thinking on these issues and we are close to where theseideas can be brought together for an exciting synergy ofhydrology, climate, and the risk management community

Lall, UpmanuInnovations in Climate Informed Water Supply andDemand Forecasts and Water System ManagementLall, Upmanu1

1. Earth & Environmental Eng, Columbia Univ, new york,NY, USA

Over the last two decades, many approaches toforecasting climate and streamflow have evolved. There hasbeen some corresponding effort to translate these forecastsinto application. Some new innovations for managementhave also emerged. I plan to discuss some ways by whichmarket mechanisms and financial securitization can bebrought into play to address water quantity and qualitygoals in a hierarchical framework from regional planning toreservoir operation to user opportunity and riskmanagement. The approach may offer benefits even whereforecasts are not skillful.

Lima, Carlos H.SEASONAL STREAMFLOW FORECASTS FOR THEBRAZILIAN HYDROPOWER NETWORK:PRESERVING THE SPATIO-TEMPORALVARIABILITY IN STATISTICAL MODELSLima, Carlos H.1, 2

1. Civil and Environmental Engineering, University ofBrasilia, Brasilia-DF, Brazil

2. Water Center, Columbia University, New York, NY, USA

Hydropower plants play a major role in energyproduction in Brazil, being responsible for almost 100 % ofthe electricity yield during wet periods. Nearly all plantsconsist of large storage and multipurpose water reservoirsthat are unequally distributed across the country, with mostof them located in the South and Southeast regions.Thousands of kilometers of transmission lines interconnectall major hydropower plants into four hubs of energyproduction (North, Northeast, South and Southeast) thatsupply the electricity demand across the country. Based onmulti-scale streamflow forecasts, the System NationalOperator (ONS) defines the optimal operational policy foreach hydropower reservoir in order to maximize the systemefficiency and avoid unnecessary energy losses due to waterspills and evaporation from the reservoirs. Most currentstreamflow forecast models employed by ONS are of theauto-regressive (AR) class of statistical models and are builtindividually for each site by considering, given the covariatesused in the model, that the streamflow series are mutuallyindependent. Although this assumption may hold for shortlead time forecasts, when the persistence of the streamflowseries and the use of past information may lead to aconditional independence across the series, it may not bevalid for larger lead times, where persistence explains little, ifany, of the variance of the future inflows. Since there isabout 50 streamflow series with a limited number of data tobe modeled, it becomes a challenge to derive a jointprobability distribution for the entire set of streamflow siteswhich accounts for all details of the inherent spatio-temporal variability and still keeps the uncertainty of theparameter estimates at a satisfactory level of confidence.Here in this work we explore the concept of dynamicBayesian networks and we factorize the joint probabilitydistribution of the entire set of the streamflow series into aset of conditionally independent distributions that are moretractable and robust in terms of parameter estimates.Particularly, we consider the spatial arrangement of thestreamflow sites in order to define the nodes of the Bayesiannetwork and the associated conditional independentdistributions. The most upstream sites are considered theparent nodes and climate information (climate indices fromthe Pacific and Atlantic Oceans) is used to factorize the jointdistribution of streamflow sites that are not in the sameriver network, i.e., sites that are not hydraulic connected. Theproposed model is tested on 54 monthly streamflow seriesfrom the major hydropower reservoirs in Brazil. A LASSOregression framework is used to shrink the modelparameters and avoid overparameterization. Preliminary

21

results shows an improvement of the proposed model overclassical periodic auto-regressive and principal componentanalysis (PCA) based models in predicting the streamflowseries up to 6 months lead time while preserving the maincorrelation structure across and within streamflow sites.

Lin, ZhaohuiSeasonal Hydrological Ensemble Prediction SystemOver Huaihe River Basin and Its PreliminaryVerificationLuo, Lifeng2; Tang, Wei1; Lin, Zhaohui1

1. Department of Geography, Michigan State University,East Lansing, MI, USA

2. Michigan State University, Lansing, MI, USA

The Seasonal Hydrological Ensemble Prediction System(SHEPS) over Huaihe River basin will be introduced firstly inthis paper, which is based on the Coupled Land surface-Hydrology Model System (CLHMS) developed at IAP, CAS,and the dynamical seasonal prediction system (CFSV2), isaimed to predict the monthly streamflow anomalies over thebasin. Besides the dynamical model system, themeteorological ensemble preprocessor system is anotherimportant component of the SHEPS, which includes theBayesian merging pre-processor, the historical-analogcriterion ensemble selection pre-processor and the temporalmatching pre-processor. Using the observation data overHuaihe river basin, it’s found that the preprocessorseffectively improve the probability distribution of thereforecast rainfall and temperature by CFSv2 (CFSRR) andthe forecast skill of summer precipitation and temperatureover the Huaihe River basin can be improved significantlywith lead time of 1-3 months. The seasonal prediction skillof summer streamflow over the Huaihe River basin will beevaluated through the 25-year hindcasts by SHEPS during1982 to 2006, and comparison with the results fromExtended Streamflow Prediction (ESP) will also beconducted. Preliminary verification results show that, theensemble mean forecast skill of streamflow by SHEPS isrelatively higher when lead time is one or two months, thePearson Correlation Coefficient (PMC) of the June-averagedstreamflow between the observation and forecast can reach0.81 in Bengbu station with lead time of one month,increased by 0.11 when compared with ESP hindcast. Withthe lead time of 2 months, the Pearson CorrelationCoefficient (PMC) between the observation and SHEPSforecast is 0.49, however, the PMC between the observationand EPS hindcast is -0.12. For the probabilistic forecasts forabove-normal, normal and below-normal events, thehindcast skill of SHEPS are also higher than that of ESPhindcast. The ranked probability score (RPS) for Junestreamflow forecasts by SHEPS hindcast is 0.19 in Wangjiabastation with lead time of one month, which is significantlylower than the value of 0.24 for ESP hindcast. Moreover, it’sfound that the skill of ensemble streamflow forecasts bySHEPS is remarkably higher for wet years than that for dryyears. Finally, further efforts on the improvement of seasonalhydrological prediction over China will be discussed. Key

words: Seasonal hydrological prediction, Land surface-hydrology model, Huaihe River basin, Downscaling, Biascorrection

Livneh, BenFrom catchments to regional scales: Hydrologicimpacts of land cover disturbances in the UpperColorado River BasinLivneh, Ben1; Deems, Jeffery S.1, 2

1. CIRES, University of Colorado, Boulder, Boulder, CO,USA

2. National Snow and Ice Data Center, University ofColorado, Boulder, CO, USA

The U.S. Southwest and surrounding areas rely on theColorado River Basin to irrigate roughly 5.5 million acres offarmland, and to supply a population of 5.5 million peoplein 7 states. The majority of water originates as snowfall inthe headwaters region, which has experienced widespreadforest mortality due to bark beetle infestation across a rangeof forest types, elevation, and latitude. Further episodicdisturbance from deposition of dust from regional drylandsources on the mountain snowpack strongly alters the snowsurface albedo and influencing snowmelt runoff magnitudeand timing. In this study, we investigate the relative impactsof competing streamflow drivers through assessing systemsensitivities to individual and combined disturbances. Webegin at the catchment-scale by training the DistributedHydrology and Vegetation Model (DHSVM) over thebaseline historical period, and simulate hydrologicconditions over a set of 4 catchments within the headwatersregion that offer a gradient in bark beetle impacts, dust-deposition, elevation, and forest coverage. The observationaldata sets include meteorological forcings of precipitation,maximum and minimum temperature, aerial survey forestdisturbance data, time series maps of MODIS-derived leafarea index (LAI), as well as other ecological indices derivedfrom MODIS forest phenology products. In the second stageof the analysis we apply these parameterizations and resultsto larger areas of the headwaters region, using a meso-scalehydrologic model, the Variable Infiltration Capacity (VIC)Model. Experiments are aimed at quantifying the systemsensitivity and hydrologic impacts of changing LAI (fromforest disturbance) and reduced snowpack albedo (from dustdeposition and forest litter) on streamflow and hydrologicstates. Preliminary catchment-scale results suggest beetlekill-induced canopy loss leads to slightly greater snowaccumulation as a result of less snow interception andreduced sublimation canopy sublimation, which outweighincreases in sub-canopy ablation fluxes. Combined withreduced transpiration during the warm season, the increasedsoil moisture availability translates into an overall increase inwater yield (i.e. streamflow) on the order of 3 - 15%,depending on disturbance severity and extent. Dust-on-snowexerts a primary control on the timing and rate of melt, withearlier and more rapid melt rates associated with moreextreme dust deposition. It is anticipated that the finalresults will lead to a clearer understanding of system

22

components and will better inform mitigation strategies andplanning efforts.

Lownsbery, KatherineDevelopment of Seasonal Forecasting of UpperNiger River StreamflowLownsbery, Katherine1; Taner, Mehmet Umit1; Brown, Casey1

1. University of Massachusetts - Amherst, Amherst, MA,USA

The Upper Niger River basin is characterized by a sub-Sahelian to Sahelian climate, with distinct wet and dryseasons. Selingue Dam, the largest dam in the region locatedon a tributary to the Niger River, is operated for threepurposes: hydroelectricity generation, large scale irrigationand sufficient flood waters to the Niger Inner Delta.Operations decisions to maximize the value of dam releasescould be more optimal when informed by seasonalforecasting of wet season (July – September) streamflow.West African Sahelian rainfall during the wet season isgenerally predicted with low skill by general circulationmodels (GCMs); however, using model output statistics(MOS) with regional wind flow has been shown to improvepredictions. It has also been shown that depending on thesign of rainfall anomalies in the Sahel and Guinea Coastregions, Sahelian rainfall is negatively correlated to El-NinoSouthern Oscillation sea surface temperatures and theSouthern Oscillation Index. This study extends seasonalforecasting of rainfall to investigate the predictability ofUpper Niger River streamflow for use in dam operationsdecision-making processes. Specifically, the skill of seasonalforecasting from GCM outputs, MOS corrections of GCMoutputs and direct statistical prediction from climate indicesis evaluated.

Lu, MengqianSpace-time structure of tropical moisture exportsand their precursors associated with highprecipitation induced floodsLu, Mengqian1; Lall, Upmanu1; Nakamura, Jennifer2

1. Earth & Environmental Engineering, Columbiauniversity, New York, NY, USA

2. Lamont Doherty Earth Observatory of ColumbiaUniversity, Palisades, NY, USA

The recent extreme floods in the United States (1993,2011), China (1998), United Kingdom (2000, 2003), Pakistan(2010), Europe (1995, 2010) and Thailand (2011) highlightthe importance of understanding the hydrometeorologicalprocesses responsible for these extreme floods events and theassociated temporal and spatial characteristics of sequencesof the associated precipitation events. The intensity ofextreme precipitation is projected to increase under globalwarming in many parts of the world. However, thesearguments are driven largely by considerations of themoisture holding capacity as a function of temperature, asindicated by the Clausius-Clapeyron equation. One needs toalso consider the attendant atmospheric circulation and

moisture transport dynamics that lead to persistent extremeprecipitation and subsequent flooding as evidenced in therecent major floods cited earlier. Although the climatemechanisms governing precipitation vary by location,extreme precipitation events in the mid-latitudes aretypically associated with anomalous atmospheric moisturefrom warmer tropical or subtropical oceanic areas.Atmospheric Rivers (AR) are associated with direct polewardtransport of tropical moisture along the AR bands from theTropics all the way to the extratropics. For meridionaltransport at middle latitudes, ARs account for a substantialpart of the moisture transport. The AR concept indicated adirection to track the moisture from warmer oceanic sourceto the heavy precipitated regions. Stationary Rossby wavesaccount for a substantial fraction of summertime monthlymean surface temperature and precipitation variability overa number of regions of the Northern Hemisphere middlelatitudes. Tropical moisture exports (TMEs) to the NorthernHemispheric extratropics are an important feature of thegeneral circulation of the atmosphere and link tropicalmoisture sources with extratropical precipitation andoccasionally with explosive cyclogenesis. Here, a case studythat classifies N. Hemisphere tropical moisture export(TME) tracks using a clustering method, and relates thespatial flood incidence across the regions to the persistenceand spatial structure of these tracks. The origins andpathways of moist and warm tropical air masses, which arethe fuel for the heavy precipitation events and rapidcyclogenesis in the extratropics, are examined with watervapor content along the tracks considered be a massattribute for a mass moments analysis of centroid andvariance. The clustered TME tracks, thus, help identifydifferent moisture sources and climate dynamics, whichgovern and drive the movement of water vapor to thefrequent-flooded areas, together with an atmosphericcirculation pattern that leads to persistent multi-dayconvergence and precipitation in those regions. The genesislocation of TME can be linked to the seasonality and seasurface temperature; while the variance or the secondmoments of the tracks recognize the water vapor contentsand most importantly monitor the lease of the moisture,which may directly induces intensive precipitation, that arepotentially attributable for flooding events followed.

Lu, MengqianMulti-time scale Climate Informed StochasticHybrid Simulation-Optimization Model (McISHmodel) for Multi-Purpose Reservoir SystemLu, Mengqian1; Lall, Upmanu1

1. Earth & Environmental Engineering, Columbiauniversity, New York, NY, USA

In order to mitigate the impacts of climate change,proactive management strategies to operate reservoirs anddams are needed. A multi-time scale climate informedstochastic model is developed to optimize the operations fora multi-purpose single reservoir by simulating decadal,interannual, seasonal and sub-seasonal variability. We apply

23

the model to a setting motivated by the largest multi-purpose dam in N. India, the Bhakhra reservoir on the SutlejRiver, a tributary of the Indus. This leads to a focus ontiming and amplitude of the flows for the monsoon periods.The flow simulations are constrained by multiple sources ofhistorical data and GCM future projections, that are beingdeveloped through a NSF funded project titled “DecadalPrediction and Stochastic Simulation of Hydroclimate OverMonsoon Asia”. The model presented is a multilevel,nonlinear programming model that aims to optimize thereservoir operating policy on a decadal horizon and theoperation strategy on an updated annual basis. The model ishierarchical, in terms of having a structure that twooptimization models designated for different time scales arenested. The two optimization models have similarmathematical formulations with some modifications tomeet the constraints within that time frame. The first levelof the model is designated to provide optimization solutionfor policy makers to determine contracted annual releases todifferent uses with a prescribed reliability; the second level isa within-the-period (e.g., year) operation optimizationscheme that allocates the contracted annual releases on asubperiod (e.g. monthly) basis, with additional benefit forextra release and penalty for failure. The model maximizesthe net benefit of irrigation, hydropower generation andflood control in each of the periods. The model design thusfacilitates the consistent application of weather and climateforecasts to improve operations of reservoir systems. Thedecadal flow simulations are re-initialized every year withupdated climate projections to improve the reliability of theoperation rules for the next year, within which the seasonaloperation strategies are nested. The multi-level structure canbe repeated for monthly operation with weekly subperiods totake advantage of evolving weather forecasts and seasonalclimate forecasts. As a result of the hierarchical structure,sub-seasonal even weather time scale updates andadjustment can be achieved. Given an ensemble of thesescenarios, the McISH reservoir simulation-optimizationmodel is able to derive the desired reservoir storage levels asa function of calendar date, and the associated releasepatterns. The multi-time scale approach allows adaptivemanagement of water supplies acknowledging the changingrisks, meeting both the objectives over the decade inexpected value and controlling the near term and planningperiod risk through probabilistic reliability constraints. Forthe applications presented, the target season is the monsoonseason from June to September. The model also includes amonthly flood volume forecast model, based on a Copuladensity fit to the monthly flow and the flood volume flow.This is used to guide dynamic allocation of the flood controlvolume given the forecasts.

Luo, LifengSeasonal streamflow prediction with VIC andClimate Forecast System version 2Luo, Lifeng1; Wood, Andrew2

1. Department of Geography, Michigan State University,East Lansing, MI, USA

2. NOAA/NWS Northwest River Forecast Center, Portland,OR, USA

Decisions regarding water resource management,agricultural practice, and energy allocation often requireinformation about future climate conditions weeks tomonths in advance. Skillful and reliable subseasonal andseason climate prediction have the potential to significantlyfacilitate and benefit the decision making process. Forstreamflow predictions, the extended streamflow prediction(ESP) method heavily relies on the initial hydrologicalconditions such as soil moisture and snow water equivalentin the basin. During the forecast period, historical records ofmeteorological conditions are used as possible futureoutcomes to produce a multi-member ensemble streamflowprediction. Earlier studies have shown that potentialimprovement in the seasonal streamflow forecast skills canbe achieved by incorporating precipitation and temperatureforecast from dynamic climate models. In this study, we usethe precipitation and temperature forecasts from the NCEPClimate Forecast System version 2 (CFSv2) and the VIChydrological model to quantify the usefulness of the climateforecast information in seasonal streamflow forecasting. Thecentral piece of the work is the bias correction anddownscaling method that continuously updates theposterior distribution of future conditions based on dailyCFSv2 forecast. For a given day, the posterior distributionsof temporally aggregated temperature and precipitation arederived within a Bayesian framework. This new approach cantake advantage of the operational configuration of CFSv2and combine many of its forecast runs in a consistent way tomaximize the skill of temperature and precipitation forecast,and thereby improve the streamflow forecast. Thispresentation discusses the application of the system to theColorado River Basin. Streamflow forecast skill will beevaluated in both realtime and retrospective contexts, with afocus on several key drainage basins that are important forwater management.

http://drought.geo.msu.edu/research/forecast/

Madadgar, ShahrbanouA Probabilistic Framework for Predicting theSpatial Variation of Future DroughtsMadadgar, Shahrbanou1; Moradkhani, Hamid1

1. Civil and Environmental Eng, Portland State University,Portland, OR, USA

Spatial variation of future droughts across theGunnison river Basin in CO, USA is studied using a recentlydeveloped probabilistic forecast model. The StandardizedRunoff Index (SRI) is employed to analyze the droughtstatus across the spatial extent of the basin. To apply SRI in

24

drought forecasting, the Precipitation Runoff ModelingSystem (PRMS) is used to estimate the runoff generated inthe spatial units of the basin. Then, the statistical forecastmethod, whose main core is the copula functions, modelsthe joint behavior between the correlated variables ofaccumulated runoff over the forecast and predicting periods.Given the drought status of the predicting period, theprobability of different drought conditions in the forecastperiod is evaluated. Moreover, runoff variation over thebasin with the particular chance of occurrence is obtained.The forecast model also provides the uncertainty bound offuture runoff produced at each spatial unit across the basin.Our results indicate that the statistical method developed inthis study is a useful procedure in presenting theprobabilistic forecasting of droughts given the spatio-temporal characteristics of droughts in the past.

Mateus, Maria C.Vulnerability of water resources with changing landuse, and climate in the Santiam River Basin,OregonMateus, Maria C.1; Tullos, Desiree1; Surfleet, Chris1

1. Biological Ecological Engineering, Oregon StateUniversity, Corvallis, OR, USA

The distribution of water supply and demand are likelyto vary across river basins and across water users in thefuture as the climate and land use change, though someareas are certain to be more sensitive than others to thechanges. This paper analyzes the influence of climate changeand land use change on the future availability of waterresources across two sub-basins with differenthydrogeological characteristics within the Santiam RiverBasin (SRB), in Oregon. The objectives of this study are to: 1)Investigate how patterns in hydrologic responses vary withsubbasin characteristics that contribute to hydrologicsensitivity to climate and land use change; 2) Identifysubbasins likely to experience vulnerability to water scarcityin the future; and 3) Explore how hydrologic sensitivityrelates to water scarcity vulnerability in the basin. Resultsdemonstrate that water demand exerts the strongestinfluence on basin’s vulnerability to water scarcity regardlessof basin characteristics. Results also demonstrate that basinscharacterized by higher permeability, with greatergroundwater recharge and storage (North Santiam Basin),are less sensitive to climate and land use changes whencompared to basins characterized by a mixed groundwaterand surface-water system (South Santiam Basin). There isthe need for water managers to take into account hydrologicvariability and basin characteristics when allocating anddistributing water to different users within the basin.

Mazrooei, AmirhosseinInter-model Comparison of Seasonal Streamflowand Soil Moisture Forecasts for the US SunbeltSinha, Tushar1; Mazrooei, Amirhossein1; Kumar, Sujay2;Peters-Lidard, Christa D.2; Arumugam,Sankarasubramanian1

1. Department of Civil, Construction and EnvironmentalEngineering, North Carolina State University, Raleigh,NC, USA

2. Hydrological Sciences, Goddard Space Flight Center,Greenbelt, MD, USA

Seasonal streamflow and soil moisture forecasts,contingent on climate forecasts, could provide valuableinformation to improving water resources management suchas optimizing hydroelectric power generation andagricultural operations. However, the uncertainty in climateforecasts poses a serious challenge in utilizing streamflowand soil moisture forecasts information in real timeoperations. Several studies have utilized Land SurfaceModels (LSMs) to develop seasonal streamflow and soilmoisture forecasts using observed or climatological forcings,but only fewer studies have used monthly updated climateforecasts. Furthermore, very limited studies focused onrainfall-runoff regimes, where precipitation forecasts play animportant role than initial hydrologic conditions (IHCs), asopposed to snowmelt-driven regimes. Therefore, ourobjectives are: 1) To develop seasonal streamflow and soilmoisture forecasts for the rainfall-runoff mechanismdominated region - the US Sunbelt, by implementingmultiple LSMs with climate forecasts from ECHAM4.5General Circulation Model (GCM), and 2) Compare andevaluate performance of different LSMs under differentclimatic (humid and semi-arid) and hydrologic (flood anddrought) conditions. In this study, we will implement widelyused LSMs such Noah, Catchment, Variable InfiltrationCapacity (VIC) and CLM, available through NASA’s LandInformation System (LIS) framework, with ECHAM4.5climate forecasts. Updated IHCs will be estimated, prior tothe forecasting period, by forcing each of the LSMs withNorth American Data Assimilation phase – 2 (NLDAS-2)forcings. We will statistically downscale monthly updatedECHAM4.5 precipitation forecasts (up to 3 months leadtime) to 0.25° scale using Canonical Correlation Analysis(CCA) and then temporally disaggregate to a daily time stepusing K-Nearest-Neighbor (KNN) approach. Hourlyclimatology for all other forcing variables will be estimatedfrom NLDAS-2 dataset over the 1981-2010 period. Then allthe LSMs will be implemented with monthly updatedprecipitation forecasts and hourly climatological forcingsduring the 3-month forecasting period to developretrospective seasonal streamflow and soil moisture forecastover the period 1993-2012. Finally, the performance ofdifferent land surface models in forecasting streamflow andsoil moisture under different climatic and hydrologicconditions will be analyzed over the US Sunbelt.

25

McCreight, James L.Global Positioning System (GPS) observations ofhydrologic variables for water management:current and future applicationsMcCreight, James L.1; Small, Eric2; Larson, Kristine M.1;Braun, John3

1. Aerospace Engineering, University of Colorado, Boulder,CO, USA

2. Geology, University of Colorado, Boulder, CO, USA3. University Corporation for Atmospheric Research,

Boulder, CO, USA

The indirect, or reflected, (L-band) microwave GPSsignal is sensitive to water in the environment. We have usedthe signal to noise ratio of the direct and reflected GPSsignals (at an antenna) to measure 3 variables of acuteinterest to water supply forecasting. 1) Snow depthobservations are critical to forecasting runoff volume insubsequent months. 2) Volumetric soil moisture directlyaffects recharge rates and runoff at seasonal timescales. 3)Vegetation water content observations inform the timing oftranspiration. (Unlike green-up measures, such as NDVI,GPS measures when plants increase their water content.)Estimates of these variables have been validated against insitu observations and found to be accurate. Improvement ofmethods is on-going. GPS provides high-frequency data atan intermediate spatial scale (between point and satellite):daily observations have spatial footprints of roughly 1000square meters. We have produced estimates of thesehydrologic variables using the Plate Boundary Observatory(PBO - originally intended to measure tectonic activity) inthe Western United States and our data are freely availableonline (http://xenon.colorado.edu/portal/). GPS antennasdesigned for positioning intentionally suppress the “noise”of reflected signals and are extremely expensive. We aredeveloping inexpensive GPS receivers for hydrologicpurposes to expressly measure reflected signals. Futurehydrologic and agricultural applications will be discussed.

http://xenon.colorado.edu/portal/

Mendoza, Pablo A.A multisite ensemble seasonal streamflowforecasting framework for semi-arid Andean basinsMendoza, Pablo A.1, 2; Rajagopalan, Balaji1, 2; Clark, Martyn3;McPhee, James4, 5

1. Department of Civil, Environmental and ArchitecturalEngineering, University of Colorado at Boulder, Boulder,CO, USA

2. Cooperative Institute for Research in EnvironmentalSciences, University of Colorado at Boulder, Boulder, CO,USA

3. Research Applications Laboratory, National Center forAtmospheric Research, Boulder, CO, USA

4. Departamento de Ingeniería Civil, Universidad de Chile,Santiago de Chile, Chile

5. Advanced Mining Technology Center, Universidad deChile, Santiago de Chile, Chile

In Chile, the rivers with headwaters in the AndesCordillera are the main source of water for humanconsumption, irrigation, industry, mining and energygeneration, and there is a strong need for streamflowforecasts to help optimize scarce water resources. In thispaper, we develop a multisite ensemble streamflowforecasting method to predict average streamflow forspring/summer in 10 basins located in the semi-arid Andeanregion between 30° and 34° S, where most of the surfacerunoff comes from the water accumulated during winter assnowpack and glaciers, which melts during thespring/summer seasons. The framework tested here is basedon the following steps: (i) Principal Component Analysis(PCA) over average Oct-Mar flows to find the mostsignificant patterns in the data (i.e., leading PrincipalComponents or PCs), (ii) climate diagnostics to findappropriate predictors for the dominant streamflow modes,(iii) selection of the best models using the Generalized CrossValidation (GCV) score, (iv) streamflow prediction in cross-validation mode by using all models and then backtransforming to flow space, and (v) final ensemble forecastgeneration by resampling results from the models based onweights computed using their associated GCV scores. Weperform probabilistic verification by computing the RankProbability Skill Score (RPSS) for each year using threecategories obtained from the 33% and 66% flows observed ateach location. As the first PC in spring/summer flowsexplains 88 % of the total variance, model development isbased on testing several combinations of predictors for thispattern. Our results demonstrate the advantage of applyinga framework like this in areas where meteorological data isquite sparse, with skill scores between 0.2 and 0.7. Futuretests will include forecast evaluation with increasing leadingtimes and the exploration of temporal disaggregationtechniques.

26

Mirfenderesgi, GolnazalsadatA Comparative Assessment of Particle Filtering-Markov Chain Monte Carlo and ParticleSmoothing Methods for Seasonal HydrologicForecasting and Uncertainty QuantificationMirfenderesgi, Golnazalsadat1; DeChant, Celeb1;Moradkhani, Hamid1

1. Civil and Environmental Engineering, Portland StateUniversity, Portland, OR, USA

The need for effective assimilation of useful hydrologicdata into the forecast model is becoming increasinglyemphasized. Application of ensemble data assimilationmethods to hydrologic forecasting provides a framework forimproving the accuracy of initial states and parameters, andsimultaneously estimating their respective uncertainties.Among all the assimilation methods, approaches likeparticle filtering (PF) and particle smoothing (PS) arebecoming popular for assimilation of a wide range ofhydrologic variables, while algorithmic developments aremaking these techniques increasingly effective for hydrologicapplications. This study provides a comparison of therecently developed PF-Markov chain Monte Carlo (MCMC)and PS methods for state-parameter estimation, within thecoupled Sacramento Soil Moisture Accounting (SAC-SMA)and Snow-17 models for seasonal streamflow forecasting.The experiment is performed over several watersheds withinthe upper Colorado RiverBasin. In this study, both asynthetic and a real experiment were performed to study andcompare the effectiveness and efficiency of the two dataassimilation techniques.

Mosier, ThomasDownscaling Global Climate Data: Production andOptimization of 30 Arc-Second SurfacesMosier, Thomas1; Hill, David1; Sharp, Kendra1

1. Oregon State University, Corvallis, OR, USA

This study compares variations of static downscalingmethods for applications with spatially distributedhydrologic models. The most physically representativedownscaling procedure is then implemented using globallyavailable data as inputs, resulting in 30 arc-second resolutionsurfaces for monthly precipitation and mean temperaturewith global land coverage. Both the Delta and BiasCorrection Spatial Disaggregation downscaling methods areimplemented; however, the Delta downscaling methodappears to produce more physically representative resultsand therefore analysis focuses on optimization of the Deltamethod. The primary step within the Delta method whichdiffers between implementations is the anomalyinterpolation step. Bilinear, cubic spline, and piecewise cubicHermite interpolating polynomials (PCHIP) are examinedfor the step of interpolating the anomaly field. The resultinggrids are then compared to Global Historical ClimatologyNetwork station records to assess the grids’ accuracy, whichshows that use of PCHIP anomaly interpolation with the

Delta method produces the most physically representativedownscaled surfaces. This downscaling procedure isimplemented using the 30 arc-second WorldClimclimatologies and 0.5 degree time-series grids by Willmott &Matsuura as inputs. The Delta downscaled grids arecompared to the corresponding Parameter-elevationRegressions on Independent Slopes Model (PRISM) datausing Oregon, USA as a test region. In the case ofprecipitation, the Delta grids have a root mean square error(RMSE) of 15.6 mm compared to PRISM’s RMSE of 12.9mm. In the case of mean temperature, though, the Deltadata performs slightly better than PRISM, with an RMSE of0.8 deg C compared to PRISM’s RMSE of 1.4 deg C. Astrength of the Delta downscaled dataset discussed herein isthat it is freely available for all global land surfaces at aspatial resolution of 30 arc-seconds. Analysis for additionaltest regions distributed globally indicate that the Deltadownscaled grids have a relatively consistent level ofaccuracy across regions.

Najafi, Mohammad RezaEvaluating Multi-Modeling Techniques withVarying Complexities for Seasonal HydrologicForecastsNajafi, Mohammad Reza1; Moradkhani, Hamid1; Wood,Andy2; Garen, David3

1. Portland State University, Portland, OR, USA2. NOAA/NWS Northwest River Forecast Center, Portland,

OR, USA3. USDA- Natural Resources Conservation Service,

Portland, OR, USA

Seasonal water supply forecasts (WSFs) are required bydecision makers for managing water deliveries, determiningindustrial and agricultural water allocation, and operatingreservoirs. Several operational agencies issue seasonal watersupply outlook forecasts of naturalized or unimpaired flow.This is commonly done using statistical and ensemblestreamflow prediction (ESP) methods. Recent work hasshown the potential for WSF improvement throughimproved forecasting of seasonal atmospheric forcings aswell as from the combination of a suite of predictionapproaches, including statistical and dynamical (i.e., physicalor conceptual models), with benefits to streamflow forecastskill at both short- and long- lead times. We present resultsfrom a project to incorporate several multi-modelcombination approaches that optimally merge forecastsfrom different sources, including ensemble hydrologicforecasts. The complexities of the applied techniques varyfrom simple averaging to Bayesian model averageapproaches. Given that a multi-model contains informationfrom all participating models, including the less skillfulones, the increase in forecast accuracy is assessed incomparison with the available best and worst prediction.The uncertainty related to each multi-model method is alsoestimated. The study is performed over four different riverbasins in the western US. The results indicate that multi-modeling methods can produce forecasts with a greater skill

27

than any one of the individual forecasts, but, of course, theskill of the combined forecast is limited by the skill of theindividual forecasts.

Najafi, Mohammad RezaThe Impact of Climate Change on Runoff ExtremesBased on Regional Climate Models andHierarchical Bayesian ModelingNajafi, Mohammad Reza1; Moradkhani, Hamid1

1. Portland State University, Portland, OR, USA

Climate change would impact the spatiotemporalvariability of hydrologic extremes especially in regions withtopographical variations. We investigate the runoff extremesover the Pacific Northwest (PNW) using a spatialhierarchical Bayesian approach. The changes in extremes inthe future period (2041-2070) are compared to the historicalperiod (1971-2000). Different regional climate modelprojections are analyzed and the seasonal variations ofrunoff extremes are studied. Climate scenarios are providedby the North American Regional Climate ChangeAssessment Program (NARCCAP) including nine regionalclimate model (RCM) simulations. Hydrologic modeling isperformed after downscaling the precipitation, maximumand minimum temperature and wind speed to 1/8th degreeresolution using the quantile-mapping approach. VariableInfiltration Capacity (VIC), a distributed hydrologic model isused to provide daily runoff estimates (mm) for each cell.Spatial hierarchical Bayesian model was then applied on thecell-wise extreme runoff (mm) for both time periods and forall seasons. The estimated spatial changes in extreme runoffsover the future period vary depending on the RCM drivingthe hydrologic model. The hierarchical Bayesian modelcharacterizes the spatial variations in the marginaldistributions of the General Extreme Value (GEV)parameters and the corresponding 100-yr return levelrunoffs. Results show an increase in the 100-yr return levelrunoffs for most regions in particular over the high elevationareas during winter. However, reduction of extreme events inseveral regions is projected during summer.

Newman, Andrew J.Impacts of Hydrologic Model Uncertainty on theSkill of Seasonal Streamflow ForecastsNewman, Andrew J.1; Sampson, Kevin1; Hopson, Tom1;Clark, Martyn1

1. Research Applications Laboratory, National Center forAtmospheric Research, Boulder, CO, USA

The National Weather Service (NWS) operationalapproach to streamflow forecasting involves (i) running ahydrologic model up to the start of the forecast period toestimate basin initial conditions; and (ii) running the modelinto the future with probabilistic weather forecasts andclimate information. Forecast skill depends on both theaccuracy of the basin initial condition estimates and theirimpact on the basin response, the accuracy of the weatherand climate forecasts, and the hydrologic model. A key

shortcoming of this approach is that it only accounts foruncertainty in climate forecasts, and neglects uncertainty inthe estimates of the basin initial conditions and uncertaintyin the hydrologic model (e.g., Schaake et al., 2007). Thepurpose of this study is to examine the spatial variability inthe performance of the operational NWS SNOW-17/SAC-SMA hydrologic modeling system in different hydroclimateregions (e.g., regions with/without substantial snow storage;regions with varying degrees of climate predictability), andat different forecast initialization times throughout the year(e.g., forecasts initialized on October 1st versus April 1st). Toaccomplish this, the SNOW-17/SAC modeling system isconfigured and calibrated for the 629 United StatesGeological Survey’s Hydro-Climatic Data Network 2009(HCDN-2009, Lins 2012) conterminous U.S. basin subset(gages having at least 10 years of flow data since 1990). Thecalibration is based on existing retrospective forcing datasets(e.g. Daymet, Thornton et al. 2012) and is performed usingthe shuffled complex evolution optimization strategy (Duanet al. 1993). Relationships between model parameters, skilland basin characteristics (climate, topography, vegetation,soils), and station density are examined and presented.

Nieto, Juan J.Establishment of a regional Hydrological Outlookfor the West Coast of South AmericaNieto, Juan J.1

1. Climate Services, CIIFEN, Guayaquil, Ecuador

In response to the need for an operational hydrologicalseasonal forecast, it was established an exploratory andconsultative work plan with experts from NMHSs of theregion, IRI, CIIFEN in a coordinating role, and thecontribution of the Hydrology and Water ResourcesDepartment of the World Meteorological Organization.After the evaluation process of the most appropriateapproach, considering the circumstances of the region thatallow to take advantage of the seasonal climate forecastmechanism, which remains operationally since 2003, wasdecided to establish a common methodology indemonstration basins in Bolivia (Achacachi), Chile (Maule),Ecuador (Paute), Peru (Coata) and Venezuela (Caroní). Themethodology consist in generate areal precipitation andevapotranspiration in the whole basin by the inclusion oftime series data from representative meteorological stationslocated within or nearby the basin. With the use of ahydrological model that receives as input, quarterly arealprecipitation, evapotranspiration, and flow data, a schemethat correlates flows and seasonal accumulated rainfall isconstructed. From seasonal climate forecasts operationallyperformed by the six countries, expected accumulatedprecipitation is included to the hydrological model toestimate expected values of quarterly flow for the basin.With the use of the method of analogue years, the quarterlyflow forecast is disaggregated into monthly expected flowvalues. The process is now in a demonstration phase in somecountries of Western South America, to achieve andconsolidate a regional mechanism that contributes to a

28

seasonal operational decision making and planning forvarious sectors and water resource users.

Nijssen, BartRelative Contributions of the Sources ofUncertainties in Seasonal Hydrologic PredictionGloballyNijssen, Bart1; Shukla, Shraddhanand2, 1; Sheffield, Justin3;Wood, Eric F.3; Lettenmaier, Dennis P.1

1. Civil and Environmental Engineering, University ofWashington, Seattle, WA, USA

2. Geography, University of California, Santa Barbara, CA,USA

3. Civil and Environmental Engineering, PrincetonUniversity, Princeton, NJ, USA

Skillful seasonal hydrologic prediction is crucial tomitigating the impacts of droughts and floods. Sources ofuncertainties in hydrologic prediction at seasonal lead times(1–6 months) are the initial hydrologic conditions (IHCs –primarily soil moisture and snow) and seasonal climateforecast skill (FS). In this presentation we report on therelative contributions of these two sources of uncertainty toseasonal hydrologic prediction skill globally as a function oftime of year, as well as on the development of a global systemfor estimating IHCs in near real-time. We conducted twomodel-based forecast experiments using the VariableInfiltration Capacity (VIC). In the first experiment,Ensemble Streamflow Prediction, we combined a single setof IHCs with ensembles of atmospheric forcings (based onobserved forcings during the period 1961–2007) to generatehydrologic forecasts for the target forecast period. In thesecond experiment, Reverse-ESP (rESP), we combined anensemble of IHCs with a single time series of atmosphericforcings (observed atmospheric forcings for the targetforecast period) to generate forecasts. By construct, the ESPexperiment derives its skill from the knowledge of the IHCsonly, whereas the rESP experiment derives its skill solelyfrom the observed forcings. We compared cumulative runoff(CR), soil moisture (SM) and snow water equivalent (SWE)forecasts obtained from each experiment with a controlsimulation of the VIC model forced with observedatmospheric forcings over the reforecast period (1961-2007)and estimated the ratio of Root Mean Square Error (RMSE)of both experiments for each forecast initialization date andlead time. The contributions of IHCs are greater than thecontribution of FS to forecast skill over the Northern(Southern) Hemisphere during the forecast period startingin October and January (April and July). Over snow-dominated regions in the high latitudes of the NorthernHemisphere, the IHCs dominate the CR forecast skill for upto 6 months lead-time during the forecast period starting inApril. Based on our findings we argue that despite thecurrent low levels of seasonal climate forecast skill (mainlyfor precipitation), better estimates of the IHCs could lead toimprovement in the current level of seasonal hydrologicforecast skill over many regions of the globe at least duringsome parts of the year. To produce IHCs globally in near

real-time, we describe the implementation of a multi-modeldrought monitoring system, which provides near real-timeestimates of soil moisture conditions for the global landareas between 50S and 50N with a latency of about one day.The global system we describe uses satellite-basedprecipitation as well as temperature estimates based onglobal weather model analysis fields to track the evolution ofsoil moisture in near real-time at a spatial resolution of 0.5degree using multiple land surface models.

http://www.hydro.washington.edu

Odame, AugustinaWater-Saving Technology Adoptions underEcological and Economic UncertaintyOdame, Augustina1; Sims, Charles1

1. Applied Economics Department, Utah State University,Logan, UT, USA

This study employs the real options approach toevaluate the decision to invest in water-saving technology inthe face of economic and ecological uncertainty in theWasatch Range Metropolitan Area. An agent, who employswater in production, must determine whether and when toinvest in water-saving technology in order to maximize hiswater-use efficiency and subsequent water savings gains. Theagent’s decision to invest in water-saving technology andthus make a switch from his current (inefficient) technologydepends on the trade-off between the expected value of theinvestment, and the cost of investment. The net-benefit ofinvesting in water-saving infrastructure is estimated by themarket-value of any water savings which may accrue to theinvestor minus the cost of investment. This depends on theprice of water which is determined by its availability relativeto demand. Thus the evolution of investment benefits (V)depends on the evolution of water-price (P) and water-supply(W). The availability, and hence the price of water isuncertain due to changing hydro-climatic conditions in theregion. Both water price and water supply are believed toevolve according to a Geometric Brownian Motion, astochastic process employed in explaining uncertainty-ladenevolutions over time. The two processes interact via a sharedWeiner process associated with stochastic changes in water-supply. Technology adoption is considered largelyirreversible due to the prohibitive costs of reversal and thelimited use and resale value given the specialized nature ofthe technology. Considering the uncertainty of future watersupplies in the region and the largely irreversible nature oftechnology adoption, the option to delay investment inwater-saving technologies may be valuable. By waiting toinvest, an investor can observe whether water-prices increaseor decrease before committing to the substantial sunkinvestment cost. This tends to delay investment longer thansuggested by traditional cost-benefit analysis. Carey andZilberman (2002) apply real-option theory to a farm’sdecision to adopt new irrigation technology, and is thepremise spurring this more general study. This study extendsCarey and Zilberman (2002) in four ways: First, it generalizestheir farm-specific model to allow consideration of

29

infrastructure investments by other agents such as canalcompany operators. Second, it explicitly considers ecologicaluncertainty. Using hydrological stream-flow data to estimatetrends and volatility in water-supplies that may compromisethe ability of water-users to perceive predictable water-supplies, the study investigates how ecological uncertaintyfrom uncertain future water-supplies impacts the potentialbenefits from investing in water-saving infrastructure andsubsequently, the decision to invest in these infrastructures.Third, it allows for impulse-control where the agent mayincrementally upgrade technology, investing where he canmake the greatest efficiency-gains first, with possible futureexpansion. Finally, it considers the impact of policy-inducedjumps in investment costs to verify the hypothesis that whilepolicies like subsidies aim to hasten technology adoption,uncertainty about the timing of such policy may delayinvestment by increasing the value of waiting to invest.

Ojha, SunalRole of snow cover and its Hydrological impact onHimalayan River basinsOjha, Sunal1

1. Nepal Electricity Authority, Ministry of Energy,Kathmandu, Nepal

Satellite remote sensing is an effective tool formonitoring snow covered area. However, complex terrainand heterogeneous land cover and the presence of clouds,impose challenges to snow cover mapping. This researchanalyzes snow cover and glaciers with a perspective ofclimate change in Himalayan Regions using remote sensingtechniques. The remote sensing snow cover data fromModerate Resolution Imaging Spectroradiometer (MODIS)satellite from 2000 to 2010 have been used to analyze someclimate change indicators. In particular, the variability in themaximum snow extent with elevations, its temporalvariability (8-day, monthly, seasonal and annual), itsvariation trend and its relation with temperature have beenanalyzed. The snow products used in this study are themaximum snow extent and fractional snow covers, whichcome in 8-day temporal and 500m and 0.05 degree spatialresolutions respectively. The results showed a tremendouspotential of the MODIS snow product for studying thespatial and temporal variability of snow as well as the studyof climate change impact in large and inaccessible regionslike the Himalayas. The snow area extent (SAE) (%) timeseries exhibits similar patterns during seven hydrologicalyears, even though there are some deviations in theaccumulation and melt periods. The analysis showedrelatively well inverse relation between the daily meantemperature and SAE during the melting period. Someimportant trends of snow fall are also observed. Inparticular, the decreasing trend in January and increasingtrend in late winter and early spring may be interpreted as asignal of a possible seasonal shift. However, it requires moreyears of data to verify this conclusion. Significant coverageof lake ice was found in lower elevation zone which is due to

flat terrain in this zone. Key Words: Climate change,Himalayas, MODIS, remote sensing, snow, lake ice.

Pal, IndraniDetecting the Shift in Timing of SeasonalHydrological Cycle in India and Understanding theDynamical AssociationsPal, Indrani1; Dimri, A. P.2

1. Department of Civil Engineering, University of Colorado,Denver, Denver, CO, USA

2. School of Environmental Sciences, Jawarharlal NehruUniversity, New Delhi, India

Determining when, where, and how much moisture isavailable for hydrological cycle is important for waterresources management, planning and development. Indiareceives almost two-thirds of its annual precipitation duringthe Indian summer monsoon (ISM) (Das, 1986), with someexpected spatial variability, whereas east coastal andsoutheastern states receive precipitation due to northeastmonsoon (NEM) (Rao, 1999). Reported changes inmagnitude, frequency, and timing of monsoon precipitationin India due to various reasons (for example, climate change,land use changes, aerosol, etc), and increase in post-monsoon precipitation and reduction in June-Julyprecipitation (Krishnakumar et al., 2009) especially inKerala, the state known as the “Gateway of ISM”, imperativequestion comes that “is there a shift-in hydrological cycleand its timings?” and if yes, “what is the magnitude of thosechanges and how it can be explained?”. As India’sprecipitation pattern is highly seasonal, it is expected thatthe maximum moisture influx will be concentrated duringthe monsoon season over most region. Therefore, thisresearch tries to detect the shift in timing and thecorresponding amount of moisture delivered. To understandsuch shifts, we also carry out an in depth analysis of thedynamical connections. The initial results showed that therehave indeed been significant ‘shifts’ in moisture influxtiming towards the end of the monsoon season (traditionallydefined as June-July-August-September) and that the shiftcould be explained by the large-scale intra-seasonal dynamicsrelated to large-scale climatic factors. This study alsodemonstrates that it is important to consider shift in timingof moisture availability for water resources management.References: Das PK. 1986. Monsoons. Geneva, Switzerland:World Meteorological Organization, Fifth WMO lecture,WMO No. 613. Krishnakumar KN, Prasada Rao GSLHV andGopakumar CS. 2009. Rainfall trends in twentieth centuryover Kerala, India. Atmospheric Environment 43 : 1940-1944. Rao GN. 1999. Variations of the SO relationship withsummer and winter monsoon rainfall over India: 1872–1993.Journal of Climate 12 : 3486–3495.

30

Pandit, ChetanIntroducing EHP in IndiaPandit, Chetan1

1. Consultant and Visiting Professor, Pune, India

This paper attempts to derive lessons for future success,from analysis of past failures. 335,160 SqKm of area in Indiais prone to floods and some 80,000 SqKm is annual affectedby floods. The annual flood damages are 575 mn $, affecting41 million people. The Central Water Commission (CWC)conducts hydrologic observations at 945 sites and providesFlood Forecasts (FF) at 175 sites, with a lead time of 12 to 24hours. The India Meteorological Department (IMD) isproviding QPF. CWC is set to double the number ofhydrologic observation stations, using telemetry. India isready to start EHP, yet previous attempts to introduce EHPhave failed. CWC continues with vintage practice ofcomputing the forecast at each of these 175 locations. CWCcan not provide hydrologic modeling experts at 175locations; and the divisional engineer is required to spendmore time in administrative work, leaving forecastcomputing to junior staff with no training even in basics ofhydrology. As a result, forecasting was, and continues to be,carried out by primitive methods such as gauge to gaugecorrelation. Following actions are necessary to introduceEHP in India. - Local forecast computation must end.Transmit all hydro-met data in real time to a central forecastcomputing facility, manned by trained experts, capable ofusing advanced models. - With catchments ranging fromarid zones to humid tropics to snow mountains to rainforests, FF softwares with a single option for rainfall-runoffmodeling just won’t do. This is a major obstacle in usingconceptual models. Develop FF packages with severaloptions for rainfall-runoff model, so that with all otherthings (data I/O, channel routing, user interface) remainingthe same, a rainfall-runoff model can be selected suitable foreach catchment. Even better, user should be able to tweakthe rainfall-runoff model at source code level. - A majorobstacle is user’s question “is EHP more accurate?”Unfortunately, it isn’t. The QPF based forecast with leadtime of 96 hours will necessarily be less accurate than the G-to-G correlation forecast with 12 to 24 hours lead time. Theonly way to convince the user is, develop reservoir operationmodels and do a “shadow reservoir operation” in real timeusing the EHP, to compare ‘with EHP’ and ‘without EHP’operations, and demonstrate that despite less accuracy, morelead time results in better performance in terms of powergeneration, water supplies, flood moderation. This by itselfis a major task. But unavoidable. EHP also makes a changefrom discrete forecasting of only the peak flow rate duringextreme events, to forecasting the continuous hydrograph;and from a deterministic forecast to probabilistic forecast.Considerable user education is necessary to demonstrate theutility of all this. The author is presently working as a‘mentor’, trying to assemble a group comprisinghydrologists, river engineers, and software writers, togenerate a forecasting package with above namedspecifications; also develop reservoir simulation package to

demonstrate the utility of EHP; bring together the mainactors – the CWC, the IMD, and the State Government(users of forecast); arrange the resources to do all this; andconvince at least one user to try it out. With some resources,and some luck, the aim is to try out the first EHP in Indiaduring the monsoon of 2015.

Parker, Laila B.Adaptively managing variable environmental flowreleases from a water supply reservoir into FirstHerring Brook, Scituate, MA, in response tounexpected climactic conditionsParker, Laila B.1; Grady, Sara2

1. Division of Ecological Restoration, MA Dept. of Fish &Game,, Boston, MA, USA

2. North and South Rivers Watershed Association, Norwell,MA, USA

In the fall of 2011, the Town of Scituate (MA) beganimplementing an environmental flow release program fromits water supply reservoir system into First Herring Brook.Scituate Water Division staff had worked with non-governmental organization and state and federal agencypartners for years to come up with this approach, whichinvolved bio-period flow targets designed to ensure adequatewater supply while improving conditions for aquatic life.Water Division staff and partners found themselvesconfronting the feasibility of implementing the plan soonerthan expected, as streamflows around the Northeast wereconsiderably below normal for much of that first year,particularly during the two key periods for herringmigration. By early April 2012, the season when river herringbegin to migrate upstream, discharge at a local gage wasabout 20-40% that usually seen at that time of year. TheWater Division found that they were having trouble meetingtheir flow targets while ensuring adequate water supply.After much consultation, Water Division and watershedassociation staff agreed to lower the flow target in mid-April.Dam operators were able to meet that revised target regularlyas measured at the fish ladder itself; although stage datacollected downstream suggested less success in meeting thetarget. The benefits of these efforts were demonstrated laterthat month, when volunteer fish counters saw river herringswimming into the lower reservoir for the first time indecades. By fall streamflows were still low and partners cameup with an approach to use a notched weir board thatallowed for releases that met the targets for both fish ladderwater depths and downstream flows. Although partners felt2012 was a relative success, they are nervous as to how thiseffort will fare as the local climate changes. In thispresentation, we will describe how the environmental flowrelease program came to be, and how partners adaptivelymanaged this program through a very dry inaugural year.Looking forward, we will also discuss the challenging ofusing multiple targets (flow over fish ladder, stage atdownstream staff gage) and how local data collected by thetown and the watershed association are being used toimprove those targets. We will seek input from Chapman

31

participants on how to better use regional forecasting toolsto plan ahead for allowing downstream flows and adequatesupply reservoir levels.

www.rifls.org

Piemonti, Adriana D.ANALYSIS OF STAKEHOLDER ATTITUDES ONOPTIMIZED WATERSHED MANAGEMENT PLANSPiemonti, Adriana D.1; Babbar-Sebens, Meghna2; Luzar, E. J.3

1. Civil and Construction Engineering, Oregon StateUniversity, Corvallis, OR, USA

2. Civil and Construccion Engineering, Oregon StateUniversity, Corvallis, OR, USA

3. School of Public and Environmental Affairs, IndianaUniversity-Purdue University Indianapolis, Indianapolis,IN, USA

The lack of a sustainable planning for land resources hasled to the alteration of hydrological cycles creating a series ofthreatening and challenging problems (such as climatechange, water scarcity, floodings, etc.) in the modernsocieties. It has been suggested that a distributed network ofecological systems (conservation practices) may provide asolution for such problems, restoring some of the benefitsthat nature uses to maintain a natural balance. But whereand how this distributed network should be placed is one ofthe most important challenges current land planners shouldconsider conservation practices have been widelystudiedfrom a multi-objective optimization perspective inorder to explore the most optimal alternative that will meetphysical, ecological or/and economic requirements toprovide the highest benefits from an implemented systemplan in a particular community. However, directparticipation of stakeholders is not considered during thedesign plan, delaying the adoption and implementation ofpractices, and avoiding the so important insightful learningof the different processes that will help to solve a particularproblem. In this research, we focus our attention towards theinfluence that a certain attitude may bring to a proposedoptimal design. An evaluation of socio-economic studies hasbeen performed to estimate possible outcomes of preferredsets of conservation practices depending on differentstakeholder’s attitudes. Using a weight approach system weseparate a set of 4 (four) significant land operatorpreferences and evaluate the effects that these preferenceshave over a plan design that includes a complete set ofconservation practices. . To perform the optimization, weused a NSGA algorithm assuming uniform distribution ofland tenure type. Our results showed that, for example, whenlandowner decisions are driven by economic profits, then themodified alternatives experienced a decrease in nitratereduction by 2-50%, and decrease in peak flow reductions by11-98 %, and decrease in sediment reduction by 20-77%.Other decision drivers such as flood control, erosion controland fertilizer lost were also evaluated. Interactive surveys andtests have been proposed in order to corroborate thehypothesis built and obtain a more accurate evaluation ofthe preferences from different land operators.

Pournasiri Poshtiri, MaryamUnderstanding the Droughts in the Major RiverBasins of the U.S. and their DynamicalConnections: Implications for Water ResourcesManagementPournasiri Poshtiri, Maryam1; Pal, Indrani1

1. Civil Engineering, University of Colorado Denver, Denver,CO, USA

Drought is one of the most expensive natural disastersworldwide and the United States is not an exception. Duringthe past 2000 years, U.S. has experienced widespread andprolonged droughts with devastating negative impacts onthe society and environment. According to the IPCC, theoccurrence of more severe, warmer and drier conditions isexpected to continue in the 21st century within the U.S.,which we might already be experiencing at these earlierdecades. Despite much progress in seasonal, annual, andlonger-term (e.g. decadal and multi-decadal to some extent)hydro-climate predictions, we need further understandingon how climate and region specific water sustainability arelinked at different spatio-temporal scales. Rivers are one ofthe important and key water resources for many parts of theU.S., which sustains vast portion of total freshwater needs inthis country with notable spatial and temporal variations.This research aims to study the characteristics of differenttypes of droughts (agricultural, hydrological, andmeteorological) occurring in the Major River basins withinthe U.S. and affecting the water sustainability, and also tounderstand their spatio-temporal links with large-scaleclimatic patterns, which ultimately helps determining theoptimal combinations of predictors. These findings will beuseful to not only have a basic dynamic and thermodynamicunderstanding of the linkages between different category ofdroughts in the major river basins in the U.S. but will alsocontribute enhancing the predictive capacity of the same.

Pytlak, ErikWater Supply Forecast Techniques and Challengeson the Columbia RiverPytlak, Erik1; van der Zweep, Rick1; Hall, Stephen2; Day,Ted3; Dittmer, Kyle4; Gobena, Adam5; Tama, Rashawn6

1. Bonneville Power Administration, Portland, OR, USA2. U.S. Army Corp of Engineers, Walla Walla, WA, USA3. U.S. Bureau of Reclamation, Boise, ID, USA4. Columbia River Inter-Tribal Fish Commission, Portland,

OR, USA5. BC Hydro, Burnaby, BC, Canada6. UDSA/Natural Resources Conservation Service, Portland,

OR, USA

Water supply forecasting in the Columbia River Basin iscrucial for supporting the many commercial, treaty-trustand cultural uses of this river system, and protecting theregion’s infrastructure, economy, and ecology. The ColumbiaRiver Forecast Group (CRFG) was established in 2008 as aninteragency (i.e., Federal, State, Tribal) effort to collectively

32

identify ways to improve long-range streamflow forecastingand water supply forecast accuracy in the Columbia RiverBasin. Since that time, the participating agencies have sharedtheir statistical and Ensemble Streamflow Predictiontechniques and verification results, evaluated thecomparative strengths and weaknesses of each technique,and in a few cases proposed and implemented changes in theforecasting techniques. The Group has also identifiedlocations where better and more frequent observational data,particularly in the wilderness areas of southeast BritishColumbia, may help improve situational awareness, forecastresponsiveness, and forecast reliability. Meanwhile, theCRFG has continued to support the existing regionalobservational network, despite growing budgetary pressuresand a loss of human observing capabilities. In thispresentation, the CRFG will share the forecast techniquesthe member agencies currently use for their long range andwater supply forecasts, their verification and validity, andwhich situations and locations the Group sees thetechniques providing more or less value for planningpurposes. The presentation will also identify some of thechallenges going forward as hydrologic model skill evolves,and new challenges develop in the region, including agrowing population, increasing ecological pressure, andclimate change. The CRFG hopes to share our forecastchallenges with the hydrologic research community and helpguide discussions on the current and futurehydrometeorological research opportunities, which in turncould be used to improve regional decision-making.

Roberson, AlanImproving Municipal Water Demand ForecastingRoberson, Alan1

1. American Water Works Association, Washington, DC,USA

Water utility managers face increasing risk in matchingwater supply and demand due to increasing uncertainties inpredicting future demand due to several factors, includingthe potential impacts of climate change. Climate changepresents many challenges to the water sector. The impacts ofclimate change on water supply include changes in thequantity and intensity of precipitation events, changes insnowpack and resultant runoff events, lowering of aquiferlevels, and seawater infiltration into coastal aquifers. Assuch, existing research on potential impacts to waterresources has focused on water supply and less is knownabout climate change’s potential impacts on water demand.Utility managers need accurate tools to both forecastchanges in water demand, and be able to respondaccordingly. Accurate water demand forecasting is critical, aswater demand equals water sales equals revenues for watersystems. This presentation will summarize the results from arecently completed research project - Decreasing RiskThought Improved Municipal Water Demand Forecasting. Astakeholder process was used to develop a gaps analysis andresearch plan to address the impacts of climate change onwater demand. This research started with an evaluation ofthe state of the science for municipal water demand

forecasting and then identified research gaps. Thestakeholder group developed several specificrecommendations for water utilities for reducing riskthrough improved demand forecasting, and developedstrategies for demand forecasting that would appropriateaddress several factors such as the penetration into themarket place of low-flow plumbing fixtures, Xeriscapelandscaping, installation of automatic sprinklers, smaller-scale lifestyles, among others. These recommendations fellinto six general categories that need further investigation bywater utilities to determine the appropriate means ofimplementation at their system: 1. Collecting the necessarydata; 2. Analyzing water use and related data ; 3. Evaluatingpotential changes in demand; 4. Evaluating potentialchanges in demographics; 5. Understanding andincorporating uncertainty; and 6. Planning for and copingwith drought This presentation will go into detail on each ofthe above recommendations. This research should increaseutility managers’ better understanding of how to plan forand adapt to the potential impacts on water demandforecasting from climate change.

Robertson, Andrew W.Combining seasonal climate forecasts withstochastic simulation of interannual-to-interdecadal streamflow variability for reservoirOptimization over NW IndiaRobertson, Andrew W.1; Cook, Edward3; Greene, Arthur1;Kondrashov, Dmitri4; Lall, Upmanu2; Lu, Mengqian2

1. IRI, Columbia University, New York, NY, USA2. Earth & Env Eng, Columbia University, New York, NY, USA3. LDEO, Columbia University, New York, NY, USA4. Atmos & Ocean Sci, UCLA, Los Angeles, CA, USA

Multi-year storage reservoirs must be managed in theface of weather and climate variability across timescalesranging from daily weather to interannual climate. Whileseasonal climate may contain a predictable componentassociated with the El Nino-Southern Oscillation (ENSO),longer timescales are not yet usefully predictable. It is thus ofinterest to develop strategies that combine togetherpredictive seasonal climate information with stochasticsimulation of interannual-to-decadal reservoir inflowvariability based on historical or proxy records. For theBhakra reservoir in northern India, we develop seasonalstreamflow forecasts based on GCM seasonal climateforecast models for the spring snow-melt season andsummer monsoon. Stochastic simulations of interannual-to-interdecadal variability are constructed using a nonlinearnoise-driven multi-level autoregressive model, trained onensembles of tree-ring based 500-yr reconstructions ofreservoir inflow, incorporating the reconstructionuncertainties. This stochastic model is shown to reproducethe main characteristics of the reconstructed flow seriesincluding its interdecadal spectral peaks. We investigate theextent of seasonally predictive information in the context oflonger timescale stochastic information for input into areservoir optimization model.

33

Roundy, Joshua K.The optimal time and space scale for downscalingthe CFSv2 forecast for seasonal hydrologicpredictionsRoundy, Joshua K.1; Yuan, Xing1; Wood, Eric1

1. Civil and Environmental Engineering, PrincetonUniversity, Princeton, NJ, USA

Extreme hydrologic events in the form of droughts andflood are a significant source of social and economic damagein many parts of the world. Having sufficient warning ofthese extreme events allows managers to prepare for andpotentially reduce the severity of their impacts to society. Ahydrologic forecast system can give seasonal predictions thatcan be used by mangers to make better decisions; howeverhydrologic predictions rely on the skill of the climate modelin order to predict the hydrologic state at seasonaltimescales. The skill of climate forecast models diminishes inthe first month due to chaotic nature of the climate system.Nevertheless, the gradual improvement of climate models inthe last decade has lead to the skill of these models nowexceeding that of statistical models in El Niño-SouthernOscillation (ENSO) prediction. The extent to which this skillpropagates to the key forcing variables that are used indriving seasonal hydrologic prediction is not wellunderstood, especially given that the forcing variables fromclimate models need to be downscaled in order to match thefiner resolution of the hydrologic model. For most seasonalhydrologic forecasts systems the downscaling is done at amonthly time step for convenience in operationalprocedures. The optimal time and spatial scale at which todownscale the climate model is unknown and most likelyvaries with space and time. In this work we explore theoptimal space and time events of climate models for the usein downscaling the key drivers of the hydrologic forecastsystem. This work analyzes NCEP’s Climate ForecastsSystem version 2 (CFSv2) for its optimal events fordownscaling the forcing variables to the NLDAS domainduring the hindcast period from 1982-2009. The optimalevents are determined by considering the correlation of dailyprecipitation and minimum and maximum temperaturebetween the CFSv2 forecasts and the NLDAS version 2dataset while changing the temporal and spatial scale ofaveraging of the forecast variables. The change in optimalevents due to initialization time and spatial location is alsoanalyzed and its implication to downscaling and producingbetter hydrologic forecasts is discussed.

Seo, Seung BeomNear-term Climate Change Impacts on SurfaceWater and Groundwater Interactions Over theSunbeltSeo, Seung Beom1

1. North Carolina State University, Raleigh, NC, USA

Recent Climate Change Projections from ClimateModels Inter-comparison Project-5 (CMIP5) provide

opportunities for incorporating climate change informationin planning and management. Most of the climate changestudies primarily focus on either surface water orgroundwater resources alone. In this study, we utilize one ofthe fully coupled surface water and groundwater models –Penn-State Integrated Hydrologic Model (PIHM) to analyzeimpacts of near-term climate change over selected basins inthe Sunbelt. The study will evaluate different approaches forquantifying the changes in surface water and groundwaterinteractions by effectively capturing the interactions betweenthem. We intend to evaluate the proposed methodologies byfirst validating using the hindcasts from CMIP5 and thencomparing them with the observations. The difference inperformance between a simple downscaling to an improvisedbivariate downscaling scheme that captures theinterdependency between precipitation and temperature willalso be quantified. The primary goal of this study is to assesssustainability of both surface water and groundwaterresources utilizing both the hindcasts and future near-termclimate change projections over four selected basins thatbelong different hydroclimatic regimes across the Sunbelt.Keywords: Hydrologic Model, Streamflow, Groundwater,Streamflow generation, Climate Change

Shafiee-Jood, MajidAssess the value of seasonal forecast to mitigatefarmers’ losses from droughtShafiee-Jood, Majid1; Cai, Ximing1; Chen, Ligang2; Liang,Xin-Zhong3; Kumar, Praveen1

1. Civil and Environmental Engineering, University ofIllinois at Urbana-Champaign, Urbana, IL, USA

2. Earth System Science Interdisciplinary Center, Universityof Maryland, College Park, MD, USA

3. Atmospheric and Oceanic Science, University ofMaryland, College Park, MD, USA

In recent years, drought has turned into a widespreadconcern in the United States and farmers are now facingwith adverse impacts of droughts more frequently. Thissituation is exacerbated by the U.S. Midwest drought of2012, one of the worst droughts in the history of thecountry. This concern has raised again the question of whatis available and what is needed from science and engineeringcommunities to help mitigate drought impacts, especially onfarmers. The 2012 Midwest drought hurt most the rainfedagriculture, for which short-term forecast is not useful sinceonce a crop is planted, what farmers can do is very limitedwhen a drought is predicted or actually occurs. It is the mid-term seasonal forecast, if it provides some meaningfulinformation, that can assist farmers’ decisions right beforethe crop season. For example, farmers in Midwest decide thecrop pattern (e.g., corn or soybean), contract with ethanolindustry, etc., which can considerably affect farmers’ incomewhen a serious drought occurs. Indeed farmers who made alarge contract to ethanol industry are among the groupwhich was most seriously hurt by the 2012 drought. On theother hand, seasonal climate forecast has low skill score,which is usually too unreliable to be used. Furthermore,

34

assuming the forecast includes some useful information,there lacks an approach to convey the information tofarmers. In this study seasonal forecast data are generatedusing the regional climate extension of Weather Researchand Forecasting (CWRF) model driven by multiple generalcirculation models (GCMs) and initial soil states as well asmultiple physics configuration. The driving GCMs includethe Climate Forecast System version 2 (CFSv2), the ECHAMversion 4.5 and the European Center for Medium-RangeWeather Forecasts (ECMWF) Reanalysis-Interim (ERI). Theplanning framework is based on a stochastic optimizationmodel coupled with a crop growth simulation model(SWAP). Utilizing the seasonal forecast data, theoptimization-simulation framework determines the optimaldecision for crop type (i.e. corn or soybean), farmers’contract and amount of fertilizer for a rainfed agriculturesystem. The model and forecast (hindcast in this case) will beapplied to the US Midwest 2012 drought. As the outcome,this study specifically attempts to address the followingissues: 1) reliability of utilizing seasonal forecast to mitigatefarmers’ loss from drought, 2) what useful information canbe obtained from the proposed framework for both farmers(information users) and modelers (information providers),and 3) assessment of effective forecast horizon (e.g., 2-, 3-, 4-month or an even longer period) for practical use of climateforecast for drought mitigation.

Sharma, AshishTHE DO’S AND DON’T’S OF COMBININGMODEL PREDICTIONS FOR SEASONALHYDROLOGIC FORECASTINGSharma, Ashish1; Khan, Mohammad Z.1; Mehrotra, Raj1;Chowdhury, Shahadat1; Amurugam, Sankar2

1. School Civil and Environmental Engineering, TheUniversity of New South Wales, Sydney, NSW, Australia

2. Department of Civil, Construction and EnvironmentalEngineering, North Carolina State University, Raleigh,NC, USA

The use of ensemble combinations for formulatingimproved climate forecasts is now well established. Thesemethods have recently been extended for seasonal hydrologicpredictions, with applications showing a range of situationswhere the advantages can be significant. This talk reviewssome of the approaches for combining model predictions,with a view of identifying when such combinations areespecially of use. The talk then proceeds to present a basisfor combining model predictions when inter-modelcorrelations at each grid location are significant, illustratingthe improvements that can be achieved if procedures forfactoring this dependence are utilised. The talk concludeswith a case study on forecasting one-month-ahead global(gridded) sea surface temperature anomalies (SSTA), thecommonly used basis for formulating seasonal hydrologicand climate forecasts the world over. The study includesforecasts from five dynamical and statistical models used byvarious modelling groups around the world, and proceeds todemonstrate the improvements that result when the

forecasts are combined without considering inter-modeldependence, and then again when inter-model dependence istaken into account. It then proceeds to speculate on thenature of improvements that will be possible when thesecombinatorial methods put to use by hydrologic andmeteorological organisations over the world.

Sharma, AshishAN UPPER LIMIT TO SEASONAL STREAMFLOWPREDICTABILITY?Sharma, Ashish1; Taormina, Riccardo2; Westra, Seth3

1. School of Civil and Environmental Engineering, TheUniversity of New South Wales, Sydney, NSW, Australia

2. Coastal and Hydraulic Engineering, Hong KongPolytechnic University, Hong Kong, Hong Kong

3. School of Civil, Environmental and Mining Engineering,The University of Adelaide, Adelaide, SA, Australia

In a recently published paper, Westra and Sharma (2010)concluded that “asymptotic predictability of globalprecipitation was 14.7% of its total variance”, with thispredictability reducing to just 2.1% using standard cross-validation techniques when a long dataset (1900-2007) wasconsidered. This finding posed serious questions to theviability of seasonal rainfall prediction, especially the part ofthe prediction that involves using climate information thatextends beyond its largest mode of low-frequency variability -the El Nino Southern Oscillation. This presentation buildson the abovementioned work, using the methodology toascertain whether an upper limit to predictability of globalseasonal streamflow exists. As with the above study, it usesthe global sea surface temperature anomaly (SSTA) field torepresent the low-frequency variability of the climate system,and an assumed Markov order one representationcharacterising the catchment storage mechanism for theflows. It goes on to assess the variability that is obtainedwith and without the Markov order one assumption, and asa function of various catchment related attributes. Theresults show that a similar upper limit can be formulated inthe context of streamflow, but that this limit variesconsiderably on a regional basis, often as a function ofrelevant catchment attributes. Reference: Westra, S., and A.Sharma (2010), An Upper Limit to Seasonal RainfallPredictability?, Journal of Climate, 23(12), 3332-3351,doi:10.1175/2010JCLI3212.1.

Sheer, Daniel P.Using Forecast Information to Improve WaterManagementSheer, Daniel P.1

1. Hydrologics Inc., Columbia, MD, USA

Forecast based operating rules have and have thepotential to greatly increase the benefits derived from themanagement of water resources in the United States andinternationally. This paper will present examples of howpoint forecasts, probabilistic forecasts, and ensembleforecasts can be incorporated in the rules used to guide the

35

operations of water management facilities. It will alsodescribe some of the forms of rules best suited to improvingmanagement with regard to specific uses, including watersupply, hydropower, navigation, recreation, fisheries, andenvironment and riparian habitat. Examples will be drawnfrom the author’s experience and include New York Citywater supply, Apalachicola/Chattahoochee/Flint (Atlantawater supply and Alabama/Florida/Georgia Water Wars),Roanoake Basin, and South Florida/Everglades. Displayswhich illustrate the benefits gained in both economic andnon-economic terms will be shown.

Shin, DaehyokSeasonal Hydrologic Forecasting for WaterResources ManagementBari, Mohammed1; Shin, Daehyok1; Tuteja, Narendra1;Feikema, Paul1; Zhou, Senlin1; Kabir, Aynul1; Kent, David1;Laugesen, Richard1; Lee, Bat1; Hawksworth, Clare1;MacDonald, Andrew1; Jayasuriya, Dasarath (Jaya)1

1. Climate and Water Division, Bureau of Meteorology, WestPerth, WA, Australia

A seasonal water forecasting service has been needed inAustralia for many years, and the Australian Government’srecent investment in water information is addressing thisneed. Skilful and reliable seasonal forecasts of streamflowsare highly valuable for providing water allocation outlooks,informing water markets, planning and managing water useand managing drought. The opportunity for the Bureau toexpand its seasonal forecasting service is a result of its newresponsibilities, which largely came about because of theimpacts of the prolonged drought. In early 2007, theAustralian Government announced ‘Water for the Future’, a$12.9 billion water investment program. This included $450million for the ‘Improving Water Information Program’administered by the Bureau and backed by theCommonwealth Water Act 2007 and key stakeholders. TheBureau launched a new Seasonal Streamflow Forecasting(SSF) service in December 2010. The service deliversprobabilistic forecasts of streamflow volumes for the nextthree months at a site or total inflows into major watersupply storages (http://www.bom.gov.au/water/ssf). Theseasonal water availability prediction service of Australiarelies on the development and integration of the statisticaland dynamic modelling approaches. A statistical predictionsystem is based on a Bayesian joint probability modellingapproach (BJP), and it provides reliable forecasts of ‘seasonal’streamflow over the next three months. In parallel, adynamic modelling approach has been developed wherebyoutputs from climate models are downscaled to a catchmentscale to provide inputs to hydrological models of variousdegrees of complexity. The key climate models are the Centrefor Australian Weather and Climate Research’s (CAWCR)Predictive Ocean Atmosphere Model for Australia (POAMA).All prediction systems are accompanied with verificationsystems that measure skill and reliability of the forecasts. Indelivering on these responsibilities, the Bureau relies onconsiderable research, particularly through CAWCR, the

Water Information Research and Development Alliance(WIRADA) between the Bureau of Meteorology and CSIRO,and the university sector.

Shukla, ShraddhanandForecasting East Africa Spring rainfall at SeasonalScale using a Hybrid ApproachShukla, Shraddhanand1; Funk, Christopher1; Husak, Geg1;McNally, Amy1; Hoell, Andrew1

1. University of California, Santa Barbara, Santa Baraba,CA, USA

East Africa ranks among one of the most food insecureregions in the world. This region experiences to two mainrainfall seasons: March-April-May (MAM, “Long Rains”,spring rainfall) and October-November-December (“ShortRains”). The agriculture in this region is mostly rainfed. Lowlevels of technological advances and poverty make thisregion particularly vulnerable to interannual variability ofrainfall during both seasons. Skillful rainfall predictions forthis region can help inform timely efforts to mitigate socio-economic losses. The relationship between “Short rains” andENSO conditions is well established, leading to reasonablelevels of predictability. “Long rains” predictability has provento be a challenge. In recent years, a few studies haveidentified a potential link between MAM rainfall and awarming trend in the western Pacific warm pool and easternIndian Ocean. While MAM decreases in rainfall and awarming trend have been attributed to the impact ofwarming sea surface temperatures (SSTs), this link couldpotentially provide greater levels of MAM rainfallpredictability. This presentation explores this relationshipusing a hybrid approach that combines the state-of-the-artClimate Forecast System version-2 (CFSv2) dynamicalforecasts with statistical downscaling methods to predictspring rainfall in East Africa. Specifically, we used CFSv2precipitation and SSTs forecasts over the Indian and PacificOceans as predictors of East Africa spring rainfall. Weimplemented two statistical approaches for this analysis: (1)Constructed analogue and (2) Matched filter regression.Both methods essentially perform a linear regressionbetween the CFSv2 forecasts and observed spring rainfallover hindcast years (i.e. the training period) and use thatrelationship to predict spring rainfall in real-time based onthe CFSv2 forecasts. We found that the hybrid approach wasable to identify ~75% of spring drought events between1996-2012 about a month in advance, and the correlationvalue between the Z-score of the predicted and observedMAM rainfall for that period was about 0.7. We also notedthat CFSv2 precipitation is a better predictor of the EastAfrica spring rainfall than the CFSv2 SST forecast. Weconclude with a brief exploration of this technique appliedto boreal spring and winter precipitation in western NorthAmerica.

36

Sinha, TusharExperimental Inflow and Storage Forecasts Portalfor Major Reservoirs in the Southeastern USSinha, Tushar1; Singh, Harminder1; Petersen, Thomas2;Boyles, Ryan3; Arumugam, Sankarasubramanian1

1. Department of Civil, Construction and EnvironmentalEngineering, North Carolina State University, Raleigh,NC, USA

2. Department of Civil and Environmental Engineering,Massachusetts Institute of Technology, Cambridge, MA,USA

3. State Climate Office of NC, North Carolina StateUniversity, Raleigh, NC, USA

Increase in water demand due to urbanization andpopulation growth has threatened the reliability of existingwater supply systems. In addition, increase in seasonal tointerannual variability in streamflow enhances thevulnerability of these systems. Given that most of thereservoirs in the eastern and southeastern US are designed aswithin-year storage systems, they intend to capture only theseasonal variability in streamflow, as opposed to reservoirsin the West, which typically carryover the deficit/surplusfrom year to year. Thus, relatively lower storage of reservoirsin the Southeast and increasing urban demand underscorethe importance of utilizing climate based information todevelop inflow and storage forecasts. Such forecasts canadaptively be used to update monthly to seasonal waterallocations to better manage existing reservoirs, particularlyduring prolonged droughts. As a part of recently fundedprojects by North Carolina (NC) Water Resources ResearchInstitute and NC Urban Water Consortium, we havedeveloped fully automated experimental monthly toseasonal inflow and storage forecasts for the two majorreservoirs, Falls Lake and Jordan Lake, located in the rapidlygrowing Triangle Area of NC (http://www.nc-climate.ncsu.edu/inflowforecast). These experimentalreservoir inflow forecasts are developed by downscaling theprecipitation forecasts available from ECHAM4.5 GeneralCirculation Model using Principal Components Regression(PCR) model. This prognostic information, which isavailable as shift in probabilities of streamflow, is convertedto experimental reservoir storage forecasts by combining theinflow forecasts with a reservoir simulation model. Theinflow forecasts portal is customized to project monthly toseasonal storage scenarios for any user prescribed releases.Both the inflow and storage forecasts can be generated forindividual years as well as for multiple years on aretrospective basis. These probabilistic forecasts can begenerated from 1990 to present on monthly or seasonalbasis for up to 3 months lead time using climatological orPCR models. Further, several skill evaluation metrics areestimated to compare the performance between utility ofprecipitation forecasts and climatological forecasts. Theseinflow and storage forecasts will be extended to majorreservoirs in the southeastern US. Finally, the existingexperimental inflow forecasts will be enhanced to an

operational streamflow and soil moisture forecasts usingmultiple semi-distributed models.

http://www.nc-climate.ncsu.edu/inflowforecast

Taner, Mehmet U.Use of seasonal forecasting for improving reservoiroperations in the Niger River BasinTaner, Mehmet U.1; Lownsbery, Katherine1; Brown, Casey1

1. Civil and Environmental Engineering, Umass Amherst,Amherst, MA, USA

Managing water resources in the Niger River Basin is achallenging task due to the high climate variability and theuncertainties in the future climate. The use of seasonalforecasting in an adaptive management framework is apromising option for improving reservoir operations,particularly for multi-objective planning. This study presentsa sampling stochastic dynamic program (SSDP) for derivingoptimal seasonal operation policies, and a simulation modelfor using the SSDP derived policies in the real-timeoperations. A Bayesian approach is used for employingseasonal forecasting to update the probabilistic streamflowdescriptions in the SSDP model and to update reservoirrelease decisions in the real-time simulation model. Theproposed framework is demonstrated in the Upper NigerBasin, which consists of Selingue Reservoir and a largeirrigation scheme, “Office Du Niger”. The systemperformance with proposed framework is compared againstthe existing operation policies in terms of reservoirreliability, irrigated acreages in rainy (June to October) anddry (February to May) seasons, and environmental flowreliability at the Niger Inner Delta, located downstream ofOffice Du Niger. The approach is considered as anadaptation strategy to climate change using plausibleprojections of future climate for the region.

Tozer, CarlyIdentification of Drivers of Rainfall Variability forInforming Seasonal Forecasting SchemesTozer, Carly1; Verdon-Kidd, Danielle C.1; Kiem, Anthony S.1

1. Environmental & Life Sciences, University of Newcastle,Callaghan, NSW, Australia

A key challenge in the development of skilful seasonalrainfall forecasts is the identification of the atmospheric andoceanic processes that drive the rainfall variability. Seasonalrainfall forecasts for South Australia (SA) currently have lowpredictive skill and we hypothesise that this is because thekey drivers of SA’s rainfall variability have yet to be fullyidentified and therefore are not adequately represented inthe forecast models. Previously, much of the focus inAustralia has been on determining the causes of seasonaland annual rainfall variability in eastern and WesternAustralia, with little research conducted on SA’s rainfallvariability. Therefore the aim of this this study was toidentify relationships between a host of potential climatedrivers (such as the El Niño Southern Oscillation, IndianOcean Dipole, Southern Annular Mode etc.), and seasonal

37

rainfall across South Australia. A threshold method is usedthat accounts for the inherently non-linear nature of thelinks between large scale climate phenomena andhydroclimatic variability. This will provide a set of metricsagainst which dynamical forecasting schemes can be testedand may also lead to the development of new statisticalforecasting schemes.

Trimble, PaulApplication of Seasonal and Multi-seasonalClimate Outlooks for Water Management in SouthFloridaTrimble, Paul1

1. Hydrologic & Environmental Systems Modeling, WestPalm Beach, FL, USA

Incorporation of seasonal to multi-seasonal climateoutlooks for improved hydrologic forecasting is importantfor operational planning in the management of large-scalewater resources systems. The South Florida WaterManagement District (SFWMD), regional agency responsiblefor management of water resources in South Florida, haseffectively employed the seasonal and multi-seasonal climateoutlooks for operational planning. Since the agency hasembarked on a major infrastructure improvement project tofacilitate Everglades Restoration while meeting the watersupply for agriculture and increasing urban population, theissue of climate change and decadal to multi-decadal climatevariability in water resources planning has come to theforefront as a major consideration in future investments.This presentation provides an application of the research todevelop operational rules for Lake Okeechobee, the liquidheart of South Florida, as well as the real-timeimplementation of rules based on climate outlooks. Thecurrent research to address the questions of climate changeand the decadal to multi-decadal climate variability ininfrastructure planning is also discussed. The South FloridaWater Management Model (SFWMM), a distributivehydrologic model applied to the SFWMD hydrologic system,is one of several decision tools in operational planning. Themodel is reinitialized to current conditions on the same dayfor every year in the simulation period. Model results arepresented as stage and discharge time series of percentiletraces for Lake Okeechobee and other impoundments in thesystem. Conditional position analysis is obtained when agiven climatic shift (dry or wet) is incorporated into theanalysis. Discussion of the possible application of thenational Climate Forecast System Model to drive theSFWMM will be considered.

Vano, JulieMapping the diversity of hydrologic responses toseasonal climate in the Pacific NorthwestVano, Julie1; Lettenmaier, Dennis1

1. University of Washington, Seattle, WA, USA

Temperature and precipitation changes will lead tofundamental changes in the seasonal distribution of

streamflow and consequently for water management in thePacific Northwest. Basin-specific implications of thesechanges are, however, complicated by the wide range ofprojections of future temperature and precipitation fromglobal climate models, the spatial resolution of which iscoarse from a hydrologic perspective, and therefore do noteasily translate to local water management applications. Tobetter understand local impacts of regional climate changesat basin scales, we conducted experiments to determinebasin-scale hydrologic sensitivities (to imposed annual andseasonal temperature and precipitation change) of seasonaland annual streamflow. We did so using the VariableInfiltration Capacity (VIC) land-surface hydrologic modelapplied at one-sixteenth degree latitude and longitudespatial resolution, over the Pacific Northwest, a scalesufficient to support analyses at the hydrologic unit codeeight (HUC-8) basin level, a scale at which there are ~200basins within the region. Our results allowed us todetermine basin-specific transfer functions to estimatefuture changes in long-term mean seasonal responses thatcan provide viable first-order estimates of the likely range ofstreamflow changes without performing detailed modelsimulations. These estimates of long-term changes allow forprojections of streamflow changes and mapping of thoseportions of the river basins and portions thereof within theregion that are most sensitive to temperature andprecipitation changes. The approach we suggest alsoprovides a basis for exploring sources of uncertainties infuture hydrologic change across the region.

Voisin, NathalieImproved Medium Range to Seasonal StreamflowForecasts Through Simultaneous Assimilation ofSnow and Streamflow Observations: EvaluationOver the Feather River Basin, CAVoisin, Nathalie1; Nijssen, B.2; Wigmosta, M.1; Lettenmaier,D. P.2

1. Hydrology Group, PNNL, Seattle, WA, USA2. University of Washington, Seattle, WA, USA

The Pacific Northwest National Laboratory and theUniversity of Washington are collaborating on thedevelopment of a streamflow forecasting system that is fullyintegrated into the DOE Water Use Optimization Toolset(WUOT). which can be used to inform reservoir and poweroperations and guide decisions that affect environmentalperformance decision making. WUOT adresses the uniqueforecast requirements of DOE and the hydropower industryin the face of increasing competition for water. The toolsettakes into consideration traditional objectives such as floodcontrol and water supply, as well as emerging needs forrenewable energy integration and environmentalrequirements. The streamflow forecast system uses the semidistributed Variable Infiltration Capacity (VIC) model,medium-range ensemble forecasts from the Global ForecastSystem (GFS), as well as ensemble seasonal climate forecastsfrom CFS v2 and climatological resampling. Forecastedhydrologic variables include gridded fields of runoff, snow

38

water equivalent, soil moisture and unregulated ensembleflow and side-flow forecasts at points of interest. The systemalso generates 14-day hourly meteorological forecasts,including precipitation, air temperature, air and vaporpressures, relative humidity, solar and longwave radiations,and wind speed. Forecasts are enhanced throughassimilation of two sources of information: i) streamflowobservations are assimilated to improve modeled soilmoisture states, which leads to improved forecasts at shortto medium range time scales, ii) snow water equivalentobservations to improve snow pack storage, which results inreduced errors at medium to longer time scales. Thechallenge in combining the two methods is that it may resultin an inconsistency at intermediate time scales withpotentially a deterioration in forecast skill. The system isevaluated at multiple time scales with multiplecombinations of data assimilation approaches over theFeather River, CA over the 1990-2012 period. We concludewith a discussion of the best-performing implementation ofthe forecast system over this rain-snow transition basin.

Werner, KevinForecasting the Life Blood of America’s Southwest:Challenges for the Next DecadeWerner, Kevin1

1. CBRFC, NWS, Salt Lake City, UT, USA

Demand on fresh water resources in major river basinsin the Southwestern United States have reached parity withsupply in recent years. This has placed tremendous pressureon water managers and the forecast agencies that supportthem. NOAA’s Colorado Basin River Forecast Center(CBRFC) provides forecasts from timescales of hours out toa year or more in the future to support water managers inthe Colorado Basin but also in the Great Basin of Utah.CBRFC relies on a hydrologic modeling and forecast systemcomposed of hydrologic and snow models, extensive dataquality control, ingestion of weather and climate forecasts,as well as the considerable forecaster expertise at the CBRFC.CBRFC strives to continually improve its forecast processesand methods through both through internal developmentalprojects and in collaboration with the greater sciencecommunity. This talk describes major areas ripe forimproving CBRFC forecasts including linking to andimproving climate prediction, improving snow modeling,leveraging data assimilation techniques to improve dataquality control, and connecting to stakeholder decisionsupport systems. This talk describes the current state andnear term opportunities for water supply forecasting insupport of water management in the Colorado River Basin.In particular, stakeholders have worked with the US Bureauof Reclamation to design more sophisticated managementmodels and techniques in order to optimize operations. Onemajor aspect has been the development of a probabilisticoperations model (the Mid Term Operations Model orMTOM) intended to provide stakeholders with probabilisticprojections of management activities to support risk baseddecision making. The MTOM is intended to supplement the

long standing 24 month study run by USBR. The MTOMpresents an opportunity for the NOAA Colorado Basin RiverForecast Center (CBRFC) to apply its existing ensembleforecast technique. It also presents significant opportunitiesto test NOAA’s next generation streamflow ensembleforecast system, the Hydrologic Ensemble Forecast System(HEFS) and other ensemble forecast techniques that leverageclimate and weather forecasts. Forecast skill in weather andclimate forecasts has steadily improved over the last severaldecades as its underlying science and technology hasadvanced. While these improvements have yet to besystematically leveraged into improved water supply forecastsfor the Colorado Basin, HEFS and other techniques underexploration by CBRFC and its partners presentopportunities to advance the state of water supplyforecasting.

Whateley, SarahSeasonal Hydrologic Forecasting in theNortheastern United StatesWhateley, Sarah1; Brown, Casey1

1. Department of Civil and Environmental Engineering,University of Massachusetts Amherst, Amherst, MA, USA

In many parts of the United States, seasonal hydrologicforecasts are critical to water resource management.Numerous studies have demonstrated significant skill inpredicting regional hydrology and river flows, namely in thewestern and southeastern US, using both initial hydrologicconditions (IHC) (i.e. soil moisture and snow water-equivalent measures) and climate indices that characterizeaspects of the geophysical system. However, efforts toproduce skillful seasonal hydrologic forecasts in theNortheast US have been limited. As such, long-rangestreamflow forecasts are often not incorporated intoseasonal operating policies of reservoir systems in this regionas they are in the West. Increasing concern for drought andflood management in the water resources community undera changing hydrologic regime emphasizes the need forimproved understanding of the influence of IHC and large-scale atmospheric-oceanic circulation patterns on seasonalhydrologic forecast skill. This work will illustrate our effortstoward creating a hydrologic forecast in the Northeastbeneficial to water management. Current efforts focus onpredicting spring peak and summer low streamflows inmajor river basins throughout the Northeast US using soilmoisture and snow water-equivalent anomalies andatmospheric-ocean circulation patterns related to the NorthAtlantic. Further exploration of the relationships betweenhydroclimatology (i.e. the influence of the climate system ontime and space variations of the hydrologic cycle),anomalous antecedent hydrologic conditions, andhydrologic extremes in the region is needed. Thispresentation will present the current state of understandingof the variability and predictability of the hydroclimatesystem in the Northeast US, evaluate forecast skill, andassess its applicability to water resource management.

39

Wood, Andrew W.The potential value of NCEP climate forecasts forhydrologic prediction in the Pacific NorthwestWood, Andrew W.1; Schaake, John2

1. NOAA/NWS Northwest River Forecast Center, Portland,OR, USA

2. Consultant, Annapolis, MD, USA

Seasonal climate forecasts have been producedoperationally by NCEP for over a decade. NCEP has recentlyupgraded to version 2 of the Climate Forecast System(CFSv2), and also participates in the new National Multi-Model Ensemble (NMME) prediction capability whichcombines ensemble forecasts from a multiple modelingcenters. Operational streamflow predictions at theNOAA/NWS Northwest River Forecast Center (NWRFC)that support water management in the Pacific Northwest(PNW) do not currently make use of climate forecastinformation, but are slated to do so in the near future via thenew NWS Hydrologic Ensemble Forecast Service (HEFS),which uses NCEP Global Ensemble Forecast System (GEFS)short to medium range predictions and CFSv2 climatepredictions. The weather and climate predictions aredownscaled by a statistical technique to provide calibratedinputs to catchment level streamflow prediction models.This presentation summarizes skill assessments at the coarseand catchment scales for these weather to climate predictiondatasets, and discusses their potential to improve seasonaloperational streamflow prediction.

Yan, HongxiangA Regional Bayesian Hierarchical Model for FloodFrequency Analysis in OregonYan, Hongxiang1; Moradkhani, Hamid1

1. Portland State University, Portland, OR, USA

In this study, we propose a regional Bayesianhierarchical model for flood frequency analysis. TheBayesian hierarchical method is an alternative to thetraditional regional flood frequency analysis. Instead ofrelying on the delineation of implicit homogeneous regions,the Bayesian model describes the spatial variability anddependence in its inner structure. Our model has two levels.The first level of the hierarchical model introduces theindependent fitted distributions (identified by L-moments)to the maximum streamflow data. The second level of thehierarchical model presents the spatial variability of theparameters by considering different covariates (watershedsize, elevation, etc.). The performance of the Bayesian modelis assessed in a case study over the Pacific Northwest region.The uncertainty of different quantiles can be quantifiedfrom the posterior distributions using Markov Chain MonteCarlo algorithm. Temporal changes for different quantilesare also examined using a 10 year moving window method.The calculated shifts in flood risk can facilitate future waterresource management.

Yuan, XingThe role of climate models in global seasonalhydrologic forecastingYuan, Xing1; Wood, Eric F.1

1. Department of Civil and Environmental Engineering,Princeton University, Princeton, NJ, USA

Using NCEP’s Climate Forecast System version 2(CFSv2), we recently found that climate models can providebetter seasonal hydrologic forecasts over CONUS than theEnsemble Streamflow Prediction (ESP) approach throughappropriate downscaling procedures, but significantimprovements are heavily dependent on the variables,seasons and regions. Now we are extending suchinvestigation from CONUS region to global land areas andfrom a single CFSv2 model to multiple climate forecastmodels. Based on the joint probability distribution betweenobservations and model hindcasts, we bias correct themonthly precipitation and temperature forecasts from nineclimate forecast models (COLA, GFDL, IRI-AC, IRI-DC,GMAO, CFSv1, CFSv2, CanCM3, CanCM4), downscale themto a daily time scale, and use them to drive the VariableInfiltration Capacity (VIC) land surface hydrologic model toproduce a set of seasonal hydrologic hindcasts globally atone degree resolution. The hindcasts are initiated on the 1stof each calendar month during 1982-2009 and run out to 6months with 20 ensemble members. For comparison, aparallel run using the ESP forecast method is also carriedout. Analysis is being conducted for the predicted soilmoisture and runoff over global major river basins, and istargeted to explore the global baseline hydrologicpredictability from ESP, and to see whether there are addedvalues from climate models in specific regimes.

http://hydrology.princeton.edu/~xingy/

Zagona, EdithTools and Techniques for Complex WaterManagement Models on Interannual toMultidecadal Time ScalesZagona, Edith1, 2; Rajagopalan, Balaji2, 3; Jerla, Carly4; Prairie,James4

1. CADSWES, Univ of Colorado, Boulder, CO, USA2. Civil Engineering, CU, Boulder, CO, USA3. CIRES, CU, Boulder, CO, USA4. Bureau of Reclamation, Boulder City, NV, USA

Managing water for an uncertain future - whether floodforecasting over the next hours or days, or planning forunknown climate change impacts over many decades –benefits from the ability to quickly and efficiently simulatemany possible realizations, organize and analyze out putdata, and securely archive decision-producing results. A newsuite of modeling and analysis tools developed at theUniversity of Colorado Center for Advanced DecisionSupport for Water and Environmental Systems facilitatesthis type of activity, from generating stochastic hydrology tostatistical analysis of outputs. The tools, developed under

40

Reclamation’s WaterSMART initiative, were designed tofacilitate Basin Studies - comprehensive water studies toexplore options for meeting projected imbalances in watersupply and demand under future climate change andvariability. Possible flow scenarios are produced by theHydrology Simulator, which uses non-parametric K-nearestneighbor resampling techniques to generate ensembles ofhydrologic traces based on historical data, optionallyconditioned on long paleo reconstructed data using variousMarkov Chain techniques. Resampling can also beconditioned on climate change projections from e.g.,downscaled GCM projections to capture increasedvariability. The simulations produced are ensembles ofhydrologic inputs to the RiverWare operations/infrastucturedecision modeling software. Alternative demand scenarioscan be produced with the Demand Input Tool (DIT), anExcel-based tool that allows modifying future demands bygroups such as states; sectors, e.g., agriculture, municipal,energy; and hydrologic basins. The demands can be scaled atfuture dates or changes ramped over specified time periods.Resulting data is imported directly into the decision model.Different Model Files can represent infrastructurealternatives and different Policy Sets represent alternativeoperating policies, including a means to identifyunacceptable vulnerabilities that trigger dynamicallyexecuting changes in operations or other options. The over-arching Study Manager provides a graphical tool to createcombinations of future supply scenarios, demand scenarios,infrastructure and operating policy alternatives; eachscenario is executed as an ensemble of RiverWare runs,driven by the hydrologic supply. The Study Manager sets upand manages multiple executions on multi-core hardware.The sizeable output files are typically direct model outputs,or post-processed indicators of performance based on modeloutputs. Post processing statistical analysis of the outputsare possible using the Graphical Policy Analysis Tool orother statistical packages. Several Basin Studies undertakenhave used RiverWare to evaluate future scenarios. TheColorado River Basin Study, the most complex and extensiveto date, has taken advantage of these tools and techniques togenerate supply scenarios, produce alternative demandscenarios and to set up and execute the many combinationsof supplies, demands, policies, and infrastructurealternatives. The tools and techniques will be described withexample applications.

Zammit, ChristianEffect of initial conditions of a catchment onseasonal streamflow prediction using ensemblestreamslow prediction technique (ESP) for 2 riverbasins on New Zealand’s South IslandZammit, Christian1; Singh, Shailesh1; Hreinsson, Einar1;Woods, Ross1, 4; Hamlet, Alan3; Clark, Martyn2

1. NIWA, Riccarton, New Zealand2. National Centre for Atmospheric Research Earth & Sun

systems Laboratory, Boulder, CO, USA3. Dept Civil and Environmental Engineering and Earth

Sciences, University of Notre Dame, Notre Dame, IN,USA

4. Department of Civil Engineering, Queen’sSchoolofEngineering-University of Bristol, Bristol, UnitedKingdom

Increased access to water is a key pillar of the NewZealand government plan for economic growths. Variableclimatic conditions coupled with market drivers andincreased demand on water resource result in criticaldecision made by water managers based on climate andstreamflow forecast. Because many of these decisions haveserious economic implications, accurate forecast of climateand streamflow are of paramount importance (eg irrigatedagriculture and electricity generation). New Zealandcurrently does not have a centralized, comprehensive, andstate-of-the-art system in place for providing operationalseasonal to interannual streamflow forecasts to guide waterresources management decisions. As a pilot effort, weimplement and evaluate an experimental ensemblestreamflow forecasting system for two major river basins inon New Zeland’s South island (Waitaki and Rangitata Riverbasins ) using a hydrologic simulation model (TopNet) andthe familiar ensemble streamflow prediction (ESP) paradigmfor estimating forecast uncertainty. To provide acomprehensive database for evaluation of the forecastingsystem, first a set of retrospective model states simulated bythe hydrologic model on the first day of each month werearchived from 1972-2009. Then, using the hydrologicsimulation model, each of these historical model states waspaired with the retrospective temperature and precipitationtime series from each historical water year to create adatabase of retrospective hindcasts. Using the resultingdatabase, the relative importance of initial state variables(such as soil moisture and snowpack) as fundamental driversof uncertainties in forecasts were evaluated for differentseasons and lead times. The analysis indicate that thesensitivity of flow forecast to initial condition uncertainty isdepend on the hydrological regime and season of forecast.However initial conditions do not have a large impact onseasonal flow uncertainties for snow dominated catchments.Further analysis indicates that this result is valid when thehindcast database is conditioned by ENSO classification. Asa result hydrological forecasts based on ESP technique,where present initial conditions with histological forcingdata are used may be plausible for New Zealand catchments.

41

Zhang, YingDevelopment of a rainfall type prediction model forNYCZhang, Ying1; Yu, Ziwen1; Block, Paul1

1. Drexel University, Philadelphia, PA, USA

Predicting precipitation in the Northeast US has directlyapplication to urban drainage, urban supply, runoff waterquality, recreational activities, etc. We develop a season-aheadprediction system for New York City focused on forecastingthe frequencies of precipitation type. Precipitation typesduring the June-August season are categorized by rain-forming mechanisms through a cluster analysis. The ensuingprediction system is statistically-based, using sea surfacetemperatures as predictors. After that, the precipitation typefrequencies are converted into precipitation depths throughbootstrapping. Preliminary results indicate skillful predictivecapabilities.

42