Reclamation Research:Water Management Under Uncertain Hydrologic

Conditions

Levi Brekke, Chief, Research and Development

Slides acknowledgment:

Ken Nowak, Water Availability Research Coordinator

Bureau of Reclamation The mission of the Bureau of Reclamation is to manage, develop, and protect

water and related resources in an environmentally and economically sound

manner in the interest of the American public.

• Manage water in 17 western states

• Operate 337 reservoirs

• Largest wholesaler of water in the

country; provides 1 out of 5

Western farmers with irrigation

water for 10 million farmland acres

• Second largest hydropower

producer; 53 hydroelectric

powerplants providing 14000+ MW

capacity; generate enough

electricity to power 3.5M U.S.

homes

Reclamation R&D

• Address Technical Challenges in:

– Water infrastructure

– Power and energy

– Environmental issues for water mngt.

– Water operations and planning

– Developing water supplies

• Customers

– Reclamation facility and resource

managers, customers, stakeholders,

private sector, public

• Programs

– Science & Technology

– Desalination & Water Purification

www.usbr.gov/research

R&D Objectives

• Increased Water Deliveries

• More Hydropower

Generation

• Reduced Maintenance

Costs

• Improved Worker Safety

• Longer Infrastructure Life

• Improved Decision Making

• More Effective

Environmental Compliance

8

Oct 2018 Presidential Memorandum

• Promoting the Reliable Supply and Delivery of Water in

the West (October 19, 2018)

• Reclamation R&D contributing to…

– Sec. 3: Improve Forecasts of Water Availability

• facilitate greater use of forecast-based management

• improve the information and modeling capabilities related to water

availability and water infrastructure projects.

– Sec. 4: Improve Use of Technology to Increase Water Reliability

• promote expanded use of technology for improving the accuracy and

reliability of water and power deliveries

• invest in programs that promote and encourage innovation, research

and development of technology that improve water management.

• invest in technology to enable broader deployment of desalination

and recycled water technologies

R&D Tools

Prize Competitions

Technology Transfer

AgreementsResearch

R&D Highlights

• Better Prediction

– Prize Competition: Sub-Seasonal Climate Forecast Rodeo

• Handling Uncertainty:

– Research: Decision Making under Deep Uncertainty for

water management

• Other Research Efforts

What are Prize Competitions?

… “citizen solver”, crowd sourcing, a way to reach a

broader group of thinkers to address difficult topics.

Forecast Rodeo

• Goal: Improve S2S

Prediction, reduce water

management uncertainty

• Objectives– Advance Science

– Raise S2S Awareness

– Provide Evaluation Platform

A year long, real-time sub-seasonal forecasting competition

Forecast Rodeo Structure • Forecasts

– Domain: 17 Western States

– Resolution: a 1°x1° grid

– Variables: Temperature and Precipitation

– Outlooks: Weeks 3&4 and 5&6

– Issue Frequency: 2 Weeks

– Rodeo Period: 13 Months

• Benchmarks– CFSv2 (32 member ensemble)

– Damped persistence

• Final Submission– Code

– Documentation

– Hind-Cast*

• Scoring– Spatial Anomaly Correlation

– Drought.gov Leader Board

Forecast Rodeo Structure

• Prizes– 4 Forecast Categories

– Real-time Issues • 1st: $100,000

• 2nd: $50,000

• 3rd: $25,000

– Hindcast Issues• $25,000 prize available per

category

• Schedule– Announced: Dec 12, 2016

– Pre Season: Mar 21, 2017

– Regular Season: Apr 18, 2017

– Final Submission: May 3, 2018

– Winners Announced: March, 7

2019

Forecast Rodeo – Design,

Refereeing and Communication

• Design participants from…

– NOAA, USGS, USACE, USDA, BOR, CA DWR

• Thanks to NOAA for providing multiple services

– NOAA CPC provided obs datasets for forecast skill evaluation

and also the CFSv2 as a benchmark (post-processed for bias

and regressed to competition template)

– NOAA ESRL provided damped persistence benchmark, set up

the forecast skill assessment apparatus, and produced the

forecast visualization products and stats that are served at the

NIDIS leaderboard.

– NIDIS – set up web portal for competition information, including

Leader Board, articles, etc.

• Features– Rankings by forecast category

– Time series plots of forecast scores for top teams and benchmark forecasts

– Individual team pages• Visualize each forecast with corresponding observations and benchmark

forecasts

• Time series plots of forecast scores

• Tabular forecast scores

– Competition “news feed”

Forecast Rodeo Leader Board (NIDIS)Search ‘Forecast Rodeo’

Final Real-Time Standings

Winning Solvers

Category1st Place

($100,000)

2nd Pace

($50,000)

3rd Place

($25,000)

Hind-Cast

($25,000)

Weeks 3&4

TemperatureStillLearning - - StillLearning

Weeks 5&6

Temperature- - - StillLearning

Weeks 3&4

PrecipitationSalient Lupoa13 - Salient

Weeks 5&6

PrecipitationSalient StillLearning Lupoa13 Salient

Solution Methods Overview

• StillLearning (T34, P56)

– Primary approach: backwards stepwise local linear regression

with feature selection shared across grid points using a diverse

pool of predictor datasets (meteorological observations, sub-

seasonal forecasts from NMME members, and climate indices)

– Secondary method (checked first, but often not used): analog

forecast technique when periods from previous years were

sufficiently similar to the conditions preceding issuing a forecast

• Salient (P34, P56)

– Artificial intelligence model trained on sea surface temperatures

• Lupoa13 (P34, P56)

– Analog periods from the historical record identified based on

global pressure anomalies

Two Solvers withdrew to commercialize

solutions, foregoing prize-eligibility

PRXWX

bgzimmerman

Rodeo Next Steps

• Facilitate engagement between

winners and Federal forcasters

• Document lessons learned

• Rodeo 2.0 - teaming with NASA

Center of Excellence for

Collaborative Innovation (CoECI)

and the NASA Tournament Lab

• Partners Welcome!

R&D Highlights

• Better Prediction

– Prize Competition: Sub-Seasonal Climate Forecast Rodeo

• Handling Uncertainty:

– Research: Decision Making under Deep Uncertainty for

water management

• Other Research Efforts

Decision Making under Deep

Uncertainty (DMDU)

• Emerging field in academic water management in mid-2000s; first meeting of DMDU Society in 2013

• Deep uncertainty1: when parties do not know or cannot agree on most appropriate system model(s), probability distributions of key external conditions, or how to value performance measures

• Use analysis of system response and vulnerability to inform development of management alternatives.

• Goal is to find a robust solution – one that balances different types of performance in a wide range of potential futures

Courtesy of Rebecca Smith

1Lempert, R. J., D. G. Groves, S. W. Popper, and S. C. Bankes. (2006). “A General, Analytic Method for

Generating Robust Strategies and Narrative Scenarios.” Management Science 52 (4): 514–28

Reclamation DMDU Project

Building capacity for addressing uncertainty in

Reclamation's planning and decision making processes

• Partnership with Rand Corporation

– Leader in DMDU field

• Scope

– (1) Assess potential application and current uses of DMDU

methods within Reclamation

– (2) Develop training materials and case studies on using DMDU

– (3) Peer review publication on DMDU for water managers

– (4) Deploy developed resources for Reclamation and others to use

DMDU Example: Colorado River

Basin Study

• Long-term supply and

demand study

• Developed and evaluated

opportunities for resolving

imbalances

• Study conducted by

Reclamation and the Basin

States, in collaboration with

stakeholders throughout the

Basin

• Questions: Which

management alternatives make

sense to implement, and

when?

• Evaluate questions within

vulnerability assessment

framework

• Use metrics and vulnerabilities

to quantify impacts to Basin

resources

• Identify measurable hydrologic

or system conditions that are

likely precursors to threshold

violations

Vulnerability

Assessment

Indicator

Metrics

Vulnerability

Thresholds

Vulnerable

Conditions

Signposts

Courtesy of Rebecca Smith and Alan Butler

DMDU Example: Colorado River

Basin Study

R&D Highlights

• Better Prediction

– Prize Competition: Sub-Seasonal Climate Forecast Rodeo

• Handling Uncertainty:

– Research: Decision Making under Deep Uncertainty for

water management

• Other Research Efforts

Projected Hydroclimate

• Develop large ensemble with

multiple

– Emissions Scenarios

– Climate Models

– Downscaling Methods

– Hydrology Models

• Use quantitative evaluation

metrics to develop “storylines”

subsets for different

applications

• Assess impact of subset

choice on application

National Water Model Evaluation

Project Team

NOAA ESRL

NCAR

Reclamation R&D

Reclamation Policy

NOAA RFC

Advisory Panel

Reclamation Regional Staff

Reclamation Technical

Service Center Staff

NOAA RFC Staff

NOAA NWC Staff

Reflective

Evaluation

Quantitative

Evaluation

NWM Understanding

Future Development

Recommendations

• Retrospective

Simulation

• Operational

Forecasts

• Hydrologic Process

Representation

• Ensemble Model

Configuration

• Better Water Use

and Management

Representation

• Current Use/Utility

• Tools/Resources for

NWM Use

• Review of

Quantitative

Evaluation Findings

• Desired Features/

Functionality

Automated Streamflow Forecasting

• Automated “Over the

Loop” ensemble forecast

system

– What’s the potential of “over

the loop” vs. traditional “in

the loop” with lots of

manual mods?

– Goals: more forecast

locations, buy time for

forecasters for other duties

• Piloting in MP and UC

regions by fall 2019

• Utilizes SUMMA hydrology

model

Forecasting Research Project Partners

Seasonal and extended-range predictability of

atmospheric rivers and their associated precipitation

SIO

Can better representation of low-elevation snowpack

improve operational forecasts?

Reclamation TSC

and GP, NCAR

Improving seasonal runoff volume forecasting tools for

snow dominated basins (PyForecast)

Reclamation GP and

PN Regions

Development of short-range forecasts of weather-driven

channel losses and gains to support Reclamation water

management

Reclamation LC,

NCAR, NOAA ESRL

Improving the robustness of southwestern US water

supply forecasting

Reclamation UC,

NCAR

Detecting, Interpreting, and Modeling Hydrologic

Extremes to Support Flexible Water Management and

Planning

Reclamation UC,

NCAR

Sub-seasonal Heatwave Prediction SIO, University of

Colorado

Uncertainty Research

Project Partners

Improving seasonal runoff volume

forecasting tools for snow dominated

basins (PyForecast)

Reclamation GP and PN

Risk-based decision making in

reservoir operations

Reclamation GP and TSC, University of

Colorado – CADSWES, NCAR

Identifying Sources of Uncertainty in

Flood Frequency Analysis

Reclamation TSC, NCAR

Summary

• Reclamation R&D aims to improve water

management outcomes by advancing science and

engineering to meet the needs of water managers

• Uncertainty is inherent in water management

– Improved prediction can reduce uncertainty but is unlikely

to eliminate it

– Planning and decision processes specifically for contenting

with uncertainty have the potential to also be impactful

toward better water management outcomes

www.usbr.gov/research

Questions?