Tasks projection

					Draft, 4 June 2009                       For Review Only, Do Not Cite or Distribute




Climate and Hydrology Datasets for use in the RMJOC
Agencies’ Longer-Term Planning Studies




External Review Draft

4 June 2009




Prepared by:

(Reclamation) Levi Brekke, Leslie Stillwater
(Bonneville Power Administration) Nancy Stephan
(U.S. Army Corps of Engineers) Randy Wortman, Seshagirir Vaddey




                                  1
Draft, 4 June 2009                                    For Review Only, Do Not Cite or Distribute



                             Executive Summary
Acting in coordination under the Reservoir Management Joint Operating Committee
(RMJOC), the Bonneville Power Administration (BPA), U.S. Army Corps of Engineers
(USACE), and U.S. Bureau of Reclamation (Reclamation) (i.e. RMJOC agencies) are
initiating a collaborative work plan to adopt a climate change and hydrology dataset for
their longer-term planning activities in the Columbia-Snake River Basin (CSRB). In
addition to these data, these ―operating agencies‖ will also work together to adopt a set of
methods for incorporating these data into such longer-term planning activities. The
purpose of adopting such data and methods is to promote consistent incorporation of
regional climate projection information in the agencies’ planning efforts, and to promote
efficient development of these data and methods by pooling agency resources.

Incorporation of climate change information into the longer-term reservoir systems
planning generally involves four basic steps:

    1. Survey available climate projection information over the region.

    2. Decide what portion of this information should be related to agency planning, and
       how the retained information should be integrated into planning scenarios.

    3. Adjust traditional planning assumptions to be consistent with the new climate
       context (i.e. assumptions on water supplies & hydrology, water demands,
       reservoir operations or regulation constraints).

    4. Assess reservoir operations or regulations given these adjusted assumptions, along
       with associated uncertainties.

Recognizing that there are multitudes of data and methods that might be applied to this
planning outline, the RMJOC agencies have scoped this work plan to feature data and
methods that are representative of current research and planning efforts being conducted
in the Pacific Northwest. This document outlines work plan deliverables, tasks, staffing
requirements, costs and schedule. Work plan content was framed by four key scoping
considerations

       Leverage the ongoing University of Washington Climate Impacts Group (UW
        CIG) effort to develop regional climate and hydrologic dataset for use in longer-
        term water resources planning within the Pacific Northwest region.

       Utilize two of the information types being developed by UW CIG: information
        that reflects step-change in climate and hydrology from historical to future
        periods (i.e. Hybrid scenario data) and information that reflects time-developing
        climate and hydrology conditions, continuously through historical and future
        periods (i.e. Transient Climate Projections data). The purpose of considering both
        types is to gain understanding on which type is more appropriate for a given type
        of longer-term planning effort conducted by RMJOC agencies.


                                             2
Draft, 4 June 2009                                      For Review Only, Do Not Cite or Distribute




       For this initial collaboration, utilize limited selections of both information types
        (Hybrid scenarios and Transient Climate Projections data) from the larger
        ensemble of scenarios and projections being issued by UW CIG.

       Verify the data received from UW CIG through independent hydrologic modeling
        performed by RMJOC agencies staff, and thereby enable RMJOC Agencies’
        technical staff to develop a firmer understanding on UW CIG data development
        procedures and information limitations.

The work plan is scoped to produce seven deliverables, which are outlined under the
three task areas listed below. The first two task areas lead to generation of the first three
deliverables, where the tasks primarily involve adopting and reviewing UW CIG data.
Task 3 leads to development of the last four deliverables, and will be implemented by
RMJOC agencies staff and include input from interested stakeholders during scoping and
review of preliminary results.

       Task 1 - Climate Projections Survey and Selection
           (1) monthly climate change scenarios (UW CIG’s Hybrid data type; a
               maximum of 10 scenarios) and time-developing climate projections (UW
               CIG’s Transient data type; a maximum of 10 projections).

       Task 2 - Hydrologic Data Selection and Verification
           (2) daily weather inputs for hydrologic modeling (both data types)
           (3) hydrologic modeling results (natural streamflow, snowpack).

       Task 3 - Operations Analyses Preparation and Demonstration to Reveal
        Implications of Hybrid- or Transient-Style Approach
           (4) adjusted streamflows for reservoir systems modeling
           (5) adjusted seasonal runoff volume forecasts for reservoir systems modeling
           (6) adjusted reservoir storage targets for flood control and adjusted variable
           energy content inputs, consistent with adjusted inflows and seasonal runoff
           volume forecasts
           (7) demonstration study on reservoir systems analysis using inputs associated
           with either Hybrid- or Transient-style approach

A project implementation team has been developed to include four team-member levels
from each agency: sponsor, liaison to programs/planning, technical coordination, and
technical implementation. Details on staffing requirements, costs and schedule by
agency are described herein. Total work plan costs are estimated to be $500,000, with
agency-specific shares estimated to be $122,000 for BPA, $201,000 for USACE, and
$177,000 for Reclamation (based on maximum scenario and projection counts listed
above). Work plan implementation would begin in October 2009 and be completed in
July 2010.




                                              3
Draft, 4 June 2009                                                                                 For Review Only, Do Not Cite or Distribute



                                             TABLE OF CONTENTS

I. INTRODUCTION..................................................................................................................................... 6
    MOTIVATION .............................................................................................................................................. 6
    RELATING THESE DATA TO LONGER-TERM DECISION MAKING ................................................................. 8
    SCOPING CONSIDERATIONS ........................................................................................................................ 9
    RMJOC AGENCIES IMPLEMENTATION TEAM ............................................................................................11
    EXTERNAL COLLABORATION.....................................................................................................................12
II. DELIVERABLES ...................................................................................................................................13
    DELIVERABLE #1 – MONTHLY REGIONAL CLIMATE CHANGE AND CLIMATE PROJECTION DATA FROM UW
    CIG (ADOPTED) ........................................................................................................................................13
    DELIVERABLE #2 – DAILY WEATHER INPUTS FOR HYDROLOGIC MODELING (ADOPTED) .........................13
    DELIVERABLE #3 – SIMULATED WATER BALANCE AND NATURAL STREAMFLOW (ADOPTED) .................14
    DELIVERABLE #4 – STREAMFLOW INPUTS FOR OPERATIONS MODELING (ADOPTED AND DEVELOPED) ....14
    DELIVERABLE #5 – SEASONAL RUNOFF VOLUME FORECASTS FOR OPERATIONS MODELING (DEVELOPED)
    ..................................................................................................................................................................15
    DELIVERABLE #6 – STORAGE TARGETS FOR FLOOD CONTROL & VARIABLE ENERGY CONTENT INPUTS
    CONSISTENT WITH STREAMFLOW AND SEASONAL RUNOFF VOLUME FORECAST INPUTS...........................16
    DELIVERABLE #7 – DEMONSTRATION STUDY ON OPERATIONS ANALYSIS USING EITHER HYBRID-
    SCENARIO OR TRANSIENT CLIMATE PROJECTION INFORMATION ...............................................................16

III. TASKS ...................................................................................................................................................18
    TASK 1 - CLIMATE PROJECTIONS SURVEY AND SELECTION .......................................................................18
      Task 1.1 - Review of Regional Climate Projection Information available from UW CIG ...................18
      Task 1.2 - Select Subset of UW CIG Regional Climate Projection Information (Deliverable #1) .......19
      Task 1.3 - Documentation and Internal Review ...................................................................................21
    TASK 2 - HYDROLOGIC DATA SELECTION AND VERIFICATION ..................................................................22
      Task 2.1 – Obtain and Review Hydrologic Model ...............................................................................22
      Task 2.2 – Obtain and Review Daily Weather Inputs (Deliverable #2) ...............................................22
      Task 2.3 - Obtain and Review Simulated Water Balance and Natural Streamflow (Deliverable #3) .24
      Task 2.4 - Independently Verify Deliverables #1, #2, and #3 ..............................................................25
      Task 2.5 - Internal Review, Revised Documentation............................................................................26
    TASK 3 - OPERATIONS ANALYSES PREPARATION AND DEMONSTRATION TO REVEAL IMPLICATIONS OF
    HYBRID- OR TRANSIENT-STYLE APPROACH ..............................................................................................27
      Task 3.1 – Prepare Adjusted Streamflow Inputs (Deliverable #4) .......................................................28
      Task 3.2 - Prepare Adjusted Seasonal Runoff Volume Forecasts (Deliverable #5) .............................31
      Task 3.3 - Prepare Adjusted Flood Control Storage-Targets and Variable Energy Content Curves
      (Deliverable #6) ...................................................................................................................................33
      Task 3.4 – Demonstration Analyses using Hybrid-scenario and Transient Projection Inputs
      (Deliverable #7) ...................................................................................................................................34
      Task 3.5 - Peer Review, Revisions, Finalize Documentation ...............................................................35
IV. COSTS AND SCHEDULE ..................................................................................................................37
V. LIMITATIONS .......................................................................................................................................43
VI. REFERENCES .....................................................................................................................................45
APPENDIX A. DETAILS ON SCOPING CONSIDERATIONS ...........................................................47
    FACTOR #1 – LEVERAGING UW CIG DATA DEVELOPMENT ......................................................................47
    FACTOR #2 – CHOOSING WHICH UW CIG DATA TYPES TO USE ................................................................47
    FACTOR #3 – CHOOSING HOW MUCH UW CIG DATA TO USE ...................................................................50
    FACTOR #4 – VERIFYING UW CIG DATA ..................................................................................................51




                                                                                  4
Draft, 4 June 2009                                                                           For Review Only, Do Not Cite or Distribute


APPENDIX B. A BRIEF COMPARISON OF DOWNSCALING METHODS FOR CLIMATE
CHANGE STUDIES (UW CIG) ................................................................................................................52



FIGURE 1 - CLIMATE-RELATED ASSUMPTIONS IN LONGER-TERM OPERATIONS PLANNING. ............................ 7
FIGURE 2. FRAMEWORK FOR RELATING CLIMATE PROJECTION INFORMATION TO LONGER-TERM OPERATIONS
     PLANNING. ............................................................................................................................................. 8
FIGURE 3 - SUMMARY OF WORK PLAN COSTS AND SCHEDULE (DETAILS ON FIGURE 4 THROUGH ERROR!
     REFERENCE SOURCE NOT FOUND.). .......................................................................................................38
FIGURE 4 - WORK PLAN COSTS AND SCHEDULE DETAILS: RECLAMATION. ...................................................39
FIGURE 5 - WORK PLAN COSTS AND SCHEDULE DETAILS: BONNEVILLE POWER ADMINISTRATION. ............40
FIGURE 6 - WORK PLAN COSTS AND SCHEDULE DETAILS: USACE NORTHWESTERN DIVISION....................41
FIGURE 7 - WORK PLAN SCHEDULE. ..............................................................................................................42




                                                                             5
Draft, 4 June 2009                                    For Review Only, Do Not Cite or Distribute



I. Introduction
The Bonneville Power Administration (BPA), U.S. Army Corps of Engineers (USACE),
and U.S. Bureau of Reclamation (Reclamation) are initiating a collaborative work plan to
adopt a climate change and hydrology dataset for their longer-term planning activities in
the Columbia-Snake River Basin (CSRB). In addition to these data, the agencies will
also work together to adopt a set of methods for incorporating these data into such
longer-term planning activities. The purpose of adopting such data and methods is to
promote consistent incorporation of regional climate projection information in the
agencies’ planning efforts, and to promote efficient development of these data and
methods by pooling agency resources.

This document presents a draft work plan outlining deliverables, tasks, staffing
requirements, costs and schedule. Work plan development has been collaboratively
supported by BPA, USACE and Reclamation through participation in the River
Management Joint Operating Committee (RMJOC).

This document is being distributed to internal and external parties that may be interested
in this future effort, hopefully to inspire discussions on broader collaboration
opportunities during work plan implementation. Interested parties include project-
specific planning groups internal to each RMJOC agency, CSRB parties serving as
stakeholders in these planning efforts, and external parties who have experience with
relating regional climate projection information to water resources planning.

Motivation

RMJOC agencies recognize the need to move toward incorporating climate projection
information into their longer-term planning. Each agency regularly evaluates
management or regional proposals that involve operational and/or infrastructure actions
that would apply during some future period. Studying the benefits and effects of these
proposals requires making future assumptions about possible water supplies, demands
and operational constraints that would affect system operations under these proposals.
As illustrated on Figure 1, each of these assumptions has an assumed climate context.
Traditionally this climate context has been provided by data from historical records.

In this document, the proposals of interest are those that have planning periods distant
enough in the future to be relevant on a ―climate change time scale‖ (i.e. ―longer-term‖
proposals having look-aheads of multiple decades and longer (IPCC 2007)). Several
upcoming studies or planning processes involving RMJOC agencies might be classified
as having ―climate change relevant‖ planning periods. Notable studies include the
Columbia River Treaty 2014/2024 Review, USACE Stage Damage Analysis Project,
BPA’s capital investment scheduling and budgeting process, and Reclamation’s suite of
potential storage studies in the Boise, Yakima, Umatilla, and Columbia Basins.




                                             6
Draft, 4 June 2009                                               For Review Only, Do Not Cite or Distribute




       Paleoclimate Data:                   Instrumental             Climate Projections:
         reconstructed                   Records: observed            modeled weather
         weather and/or                  weather (T and P)               and runoff
             runoff                        and runoff (Q)


                         statistical modeling            watershed simulation


                     Q                              T, P and Q                    T, P, and/or Q


       Supply Variability                Demand Variability                Operating
                                                                           Constraints




                                            Operations and
                                             Uncertainties


Figure 1 - Climate-related Assumptions in Longer-Term Operations Planning.

Given the prospective need for incorporating climate change information into longer-term
evaluations, the RMJOC agencies have agreed that they would each benefit from the use
of a common Pacific Northwest (PNW) climate change hydrologic dataset, and from
collaboration on data and usage methods development. More specifically:

    1. Collaboration would reflect consensus agreement among the RMJOC agencies
       about which climate projection and hydrology data are suitable for their
       respective long-term planning studies, and how the hydrology data should be
       developed.

    2. Collaboration would reflect progress among the RMJOC agencies in adopting and
       demonstrating consensus methods for using these climate and hydrology data to
       adjust planning assumptions on supplies, demands, and constraints. (Specifically,
       this work plan focuses on usage methods for adjusting supply-related
       assumptions. Subsequent work efforts are expected to focus on demand and other
       operational constraint assumptions.)

    3. Collaboration would promote efficient use of each agency’s limited resources that
       might be available for such an effort.

The relevance of the second consideration is illustrated on Figure 2. The process of
incorporating climate projection information into a longer-term planning process raises
several fundamental questions, many of which are explored in this work plan.



                                                     7
Draft, 4 June 2009                                     For Review Only, Do Not Cite or Distribute


       Should all available climate projections be regarded as suitable for planning
        purposes, or should only a portion of available projections be regarded as credible
        enough for planning while the others are discarded, or culled, from consideration?
        If yes on the latter, what rationale supports culling of projections?

       For the retained projections, should they be used to describe step-changes in
        climate based on their portrayal of historical period to future period conditions?
        Or should the time-developing nature of the projections be used for planning?

       Given the choice on how the retained information will be used, what steps follow
        on assessing natural and/or social systems responses that ultimately translate into
        planning assumptions for supplies, demands and constraints?




Figure 2 - Framework for relating Climate Projection Information to Longer-Term
Operations Planning.



Relating These Data to Longer-Term Decision Making

Some may question how these data are going to be used to serve longer-term planning
decisions by RMJOC agencies. It is presumed that such decisions and planning processes
are served by longer-term operations analyses, conducted with the use of operations
models maintained and applied by each agency for their respective systems in the CSRB.


                                             8
Draft, 4 June 2009                                    For Review Only, Do Not Cite or Distribute


For example, Reclamation maintains operations models for reservoir systems in the
Yakima, Deschutes, and Snake River basins. Proposals for longer-term changes to these
systems (either operationally, or through infrastructure) would be assessed using longer-
term simulations of such proposed operations (Figure 1).

Outputs of interest from such simulations may vary with the operational objective. For
studies focused on operational performance for water supply management objectives,
decisions might be based on a variety of storage, river flow, and water delivery metrics
(e.g., end-of-water year carryover volumes of stored water in reservoirs, mean annual
volumes for water deliveries, mean monthly volumes for streamflow in river reaches of
interest). For studies focused on operational performance for hydropower resource
management objectives, decisions might be based on generation, probability of reservoir
refill, system reliability, probability of meeting flow objectives, and load variation and
shape. For decisions focused on operational performance in providing flood risk
management, decisions might be based on the percent of time flood objectives are met,
while attempting to meet the objective to refill. Flood objectives could include
consideration of the frequency and duration of flooding, the risk of economic damages,
and the relative priority of a location.

Regardless of operating objective, the assessment of operational performance is
predicated on being able to translate the climate and hydrology data featured in this work
plan into adjusted hydrology- and supply-related inputs feeding the operations models
used to conduct such assessments. Further, it is predicated on understanding how the
various method options for conducting such data translation affect portrayal of operations
performance, and ultimately the decisions served by such analysis.


Scoping Considerations

As stated, the primary goal of this work plan is to develop (or adopt) an appropriate
climate and hydrology dataset and set of usage methods to permit RMJOC agencies to
consistently apply the hydrology- and supply-related planning assumptions associated
with climate change in their longer-term planning studies. These hydrology- and supply-
related assumptions include inflow characteristics, seasonal water supply forecasts, and
rule curve development including variable energy content curves and flood control rule
curves. Such assumptions are collectively featured in operations analyses conducted by
RMJOC agencies (i.e. BPA’s HydSim model of the Federal Columbia River Power
System (FCRPS), Reclamation’s ModSim applications for the Yakima Project, Deschutes
Project and Snake River above Brownlee, and USACE ResSim, HYSSR and AutoReg
applications for the Columbia River Basin).

Changes in water demand (e.g., irrigation) and existing storage reservation diagrams
(e.g., as in the Corps’ current Flood Control Operating Plan) that may be impacted by
climate change are not addressed in this work plan, but could be addressed in subsequent
efforts depending on the outcomes of this work plan. It is recognized that methods for
adjusting hydrology and surface water supply assumptions under climate change are


                                             9
Draft, 4 June 2009                                    For Review Only, Do Not Cite or Distribute


relatively well established compared to methods for adjusting demands and set operating
rules for flood risk management. Assuming success of this collaboration, RMJOC
agencies would be interested in follow-on efforts that engage stakeholders and focus on
developing methods for adjusting demand and operating rule curve assumptions.

In addition to the decision to focus on hydrology and surface water supply assumptions
under climate change, there were four additional scoping factors and decisions that
shaped this draft work plan. ―Appendix A. Details on Scoping Considerations‖ provides
additional background on each decision. Briefly,

       Factor #1 - Leveraging UW CIG Data Development: This work plan leverages
        the current data-development efforts of University of Washington Climate
        Impacts Group (UW CIG), which will produce regional climate and hydrologic
        information associated with current climate projections over the Columbia-Snake
        River Basin (CSRB). Those data will serve as starting points in this effort.

       Factor #2 - Choosing which UW CIG data types to use in the RMJOC dataset:
        Through their current work, UW CIG is generating several types of regional
        climate and hydrologic information. Referencing discussion in Appendix A.
        Details on Scoping Considerations, this work plan will explore how to use the
        following two types of UW CIG information: Hybrid-scenario ―climate change‖
        data generated for two future periods, and Transient Climate Projection data that
        feature climate and hydrology conditions that statistically evolve through the 21st
        century. These two information types represent fundamentally different
        portrayals of future climate conditions in longer-term operations analyses. The
        Hybrid-scenario ―climate change‖ data are useful for studies meant to reveal
        system operational sensitivity to incremental change in climate. The Transient
        Climate Projection data are useful for revealing time-developing climate and
        operational performance, which can be useful for adaptation planning (Brekke et
        al. 2009a). However, the latter involves more complex use of regional climate
        projection information and would seem to have limited applicability to planning
        at more local and sub-monthly scales (Elsner et al. 2009, Appendix A, Appendix
        B). In any case, the work plan features development of methods to utilize both
        information types, and demonstration of both methods sets to reveal which
        information type may be more appropriate for use by RMJOC agencies depending
        on which operational performance metrics are of greater interest in a given
        planning study

       Factor #3 - Deciding how much of the UW CIG dataset to use for chosen types:
        RMJOC agencies could use all of the available Hybrid-scenario and Transient
        Climate Projection data available from UW CIG. Doing so would increase
        computational burdens for related planning studies, and could also introduce
        challenges when interpreting results. In contrast, RMJOC agencies are interested
        in identifying a small, but representative set of information (e.g., box 2a).
        Preliminary focus is on identifying sets bracketing and central for the Hybrid-
        scenarios for early and mid-21st century periods assessed by UW CIG, and then to


                                            10
Draft, 4 June 2009                                    For Review Only, Do Not Cite or Distribute


        also obtain Transient Climate Projection data associated with these scenarios, as
        available.

       Factor #4 - Deciding how to verify data received from UW CIG: In addition to
        the downscaled climate projections dataset and hydrologic model used by UW
        CIG, RMJOC has access to another downscaled climate projections dataset
        developed using the same UW CIG methodology, but at a coarser spatial
        resolution over the CSRB. RMJOC also has access to a similar hydrologic model
        for the CSRB, but also developed at that same coarser spatial resolution (see
        ―Appendix A. Details on Scoping Considerations‖ and ―Task 2.4 - Independently
        Verify Deliverables #1, #2, and #3‖). Using the alternative dataset and model,
        this work plan features using them to verify the regional climate and hydrologic
        information received from UW CIG. The purpose of the verification task is for
        RMJOC technical staff to develop better understanding on how UW CIG data
        were developed, and to be better prepared for interpreting the limitations of using
        these data in RMJOC agencies’ planning activities.


RMJOC Agencies Implementation Team

The remainder of this document presents a project management plan outlining
deliverables, tasks, staffing requirements, costs, and schedule information. The work
plan has been scoped to include several internal team roles (Table 1). Sponsor-level
individuals from each agency would oversee efforts and address programming issues
(e.g., strategize on pooling resources in a complimentary and collaborative fashion,
crafting project outreach messages for interested internal/external parties). Liaisons
would represent interests of other programs, planning and policy service groups internal
to each agency, noting that development of these data and usage methods are meant to
benefit these other efforts and programs. Finally, a team of technical staff will be
assembled to coordinate and implement work plan tasks. Such tasks include facilitating
internal and external review processes for work plan deliverables.

While this document outlines staffing requirements and budget resources necessary to
complete work plan objectives, it does not identify the funding strategy that each
participating RMJOC agency will implement to secure these resources. During summer
2009, each agency will be having internal discussions about securing necessary staff time
and funding sources. It is expected that these issues will be resolved before the work
plan start date of 1 October 2009.




                                            11
Draft, 4 June 2009                                             For Review Only, Do Not Cite or Distribute


Table 1. Team Levels and Members from the RMJOC Agencies.

       Team Level                                         RMJOC Agency
                                    BPA                      Reclamation             USACE NWD1
    Sponsor                 Rick Pendergrass             Pat McGrane               Jim Barton
    Liaison to              Rick Pendergrass             Pat McGrane               Peter Brooks
    Programs,               Birgit Koehler
    Planning, Policy        Nancy Stephan

    Technical               Nancy Stephan                Levi Brekke               Randy Wortman
    Coordinator
    Technical               Available staff from         Leslie Stillwater,        vice-Modini,
    Implementation          Regional                     Tom Pruitt                Ron Malmgren
                            Coordination
                            (PGPL) and
                            Operations Planning
                            (PGPO)



External Collaboration

It is anticipated that an important aspect of work plan implementation will be interaction
with interested external parties. Part of this interaction involves inviting external parties
to share their reactions to proposed methods featured in this effort. RMJOC agencies are
particularly interested in hearing from external parties who have experience using climate
projection information in water resources planning. Another part of this interaction is
educational, where interested stakeholder groups would be kept apprised of work
progress.

External involvement, through methods scoping and progress briefings, has been
integrated into various parts of the draft work plan (see ―III. Tasks‖). RMJOC agencies
recognize that there may be other opportunities for external involvement. For that
reason, RMJOC agencies would like to invite parties reviewing this draft work plan to
suggest further opportunities for collaboration, perhaps in this effort or in subsequent
efforts focused on relating these climate and hydrology data to planning assumptions for
demands and other operational constraints.




1
 Although not identified in Table 1, the project team may also include USACE staff participating in the
Columbia River Treaty 2014/2024 Review process.


                                                    12
Draft, 4 June 2009                                     For Review Only, Do Not Cite or Distribute




II. Deliverables
The work plan is scoped to produce seven deliverables through activities outlined in three
task areas, which are previewed below and discussed in detail in ―III. Tasks.‖ The first
two task areas lead to generation of the first three deliverables of this effort (numbered in
the outline below these task areas), and feature activities that primarily involve adopting
and reviewing UW CIG data. The third task area leads to development of the last four
deliverables, which will be implemented by RMJOC agencies staff and include input
from interested stakeholders during scoping and review of preliminary results.

       Task 1 - Climate Projections Survey And Selection
           (1) monthly climate change scenarios (UW CIG’s Hybrid data type; a
               maximum of 10 scenarios) and time-developing climate projections (UW
               CIG’s Transient data type; a maximum of 10 projections).

       Task 2 - Hydrologic Data Selection And Verification
           (2) daily weather inputs for hydrologic modeling (both data types)
           (3) hydrologic modeling results (natural streamflow, snowpack).

       Task 3 - Operations Analyses Preparation And Demonstration To Reveal
        Implications Of Hybrid- Or Transient-Style Approach
           (4) adjusted streamflows for reservoir systems modeling
           (5) adjusted seasonal runoff volume forecasts for reservoir systems modeling
           (6) adjusted reservoir storage targets for flood control and adjusted variable
               energy content inputs, consistent with adjusted inflows and seasonal
               runoff volume forecasts
           (7) demonstration study on reservoir systems analysis using inputs associated
               with either Hybrid- or Transient-style approach


Deliverable #1 – Monthly Regional Climate Change and Climate
Projection Data from UW CIG (Adopted)

Resulting from steps 2a and 2b of Figure 2, this dataset includes the selected sets of
monthly regional climate change and climate projection information that will frame
hydrology- and supply-related planning assumptions in studies conducted by RMJOC
agencies involving climate change. As stated in ―Scoping Considerations‖, this dataset
will include selections from UW CIG’s broader data development efforts: Hybrid-
scenario ―climate change‖ data and Transient Climate Projection data.


Deliverable #2 – Daily Weather Inputs for Hydrologic Modeling
(Adopted)



                                             13
Draft, 4 June 2009                                   For Review Only, Do Not Cite or Distribute


In addressing step 3 of Figure 2, the monthly data from Deliverable #1 are translated into
hydrologic data, which then support planning assumptions for hydrology- and water
supply-related inputs to the RMJOC agencies’ longer-term operations models.
Deliverable #2 consists of the synthetic, daily weather data that are developed to be
consistent with the monthly climate change and climate projections data from Deliverable
#1, but at a space and time-step resolution consistent with the weather input requirements
of the hydrologic modeling featured in this translation effort. More specifically, UW
CIG has been analyzing CSRB hydrologic response using a daily time step hydrologic
model. As a result, Deliverable #2 includes UW CIG’s daily observed and synthetic
weather inputs for the hydrologic simulations conducted for each selected Hybrid-
scenario and Transient Climate Projection.

The daily observed weather data reflect historical weather station data in the CSRB.
These data are used directly for the ―base‖ hydrologic simulation of a given Hybrid-
scenario (see ―III. Tasks‖). These same daily observed weather inputs are also used in
the generation of ―synthetic weather‖ data for hydrologic simulation reflecting the
―climate change‖ part of a Hybrid-scenario, or hydrologic simulation associated with a
Transient Climate Projection. The common theme in both applications is that the daily
observed weather inputs provide realistic daily sequencing patterns that are scaled or
adjusted to reflect the monthly aspects of the data in Deliverable #1. The method of
adjustment differs according to whether Hybrid-scenario or Transient Climate Projection
information is being represented (Appendices A and B).


Deliverable #3 – Simulated Water Balance and Natural
Streamflow (Adopted)

This dataset includes the outputs from UW CIG’s hydrologic simulation modeling in the
CSRB under the selected regional climate change and climate projection information.
The hydrologic model simulates water balance through time at regularly distributed, or
gridded, locations throughout the CSRB. Thus, the first type of hydrologic simulation
output is a spatially-distributed (gridded) water balance (i.e. natural runoff,
evapotranspiration, snow water equivalent, soil moisture). The second type of output is
developed after simulation, as gridded natural runoff is aggregated, or ―routed,‖ into
location-specific streamflow at river locations of interest. The temporal and spatial
coverage attributes of these data outputs are described in ―III. Tasks.‖


Deliverable #4 – Streamflow Inputs for Operations Modeling
(Adopted and Developed)

This dataset will likely be a mix of adopted data from UW CIG’s effort, and data
developed by the RMJOC agencies implementation team as an extension from
Deliverable #3. The deliverable will feature both daily and monthly streamflow data,
with monthly data being aggregated from the daily data. The general purpose is to
develop streamflow inputs to each agency’s reservoir operations or regulation model that


                                           14
Draft, 4 June 2009                                    For Review Only, Do Not Cite or Distribute


reflect natural runoff patterns under climate change, plus other flow impairment factors
(e.g., scenario consumptive water use leading to flow depletions, natural lake effects on
routing). The degree to which Dataset #3 will need to be massaged or post-processed to
reflect the latter varies with the needs of the various RMJOC agencies’ models. Many of
the longer-term operations and planning models use system inflows that are typically
referred to as ―Modified Flows‖, which represent historical unregulated flows that have
been modified to bring all years to a common level of depletion (e.g., irrigation diversion
practices and return flow efficiency). One form of the deliverables in dataset #3 from
UW CIG is expected to be monthly or 14-period (April and August are split months)
streamflows reflective of modified flows at all points used by models such as BPA’s
HYDSIM model and USACE’s HYSSR model.

Other models will require natural inflows and require further processing to bring the
flows at some or all inflow points to the proper form. One such example is the
incorporation of historic natural lakes into the timing and routing of runoff. Depending
on how this was accounted for in UW CIG inflow development will determine the need
to post-process Deliverable #3 for certain operations and planning models.

Any further post-processing will need to be developed for each selected Hybrid-scenario
and Transient Climate Projection. The effort will feature method development, with
input invited from external parties.


Deliverable #5 – Seasonal Runoff Volume Forecasts for
Operations Modeling (Developed)

In the context of RMJOC agencies’ real operations planning, seasonal runoff volume
forecasts drive monthly reservoir operations and related decisions (e.g., annual water
allocations to system water users). Likewise, when operational or system changes are
proposed, and when such proposals must be simulated to assess impacts on operational
performance, it is necessary to also simulate forecast-informed operations and decision-
making. The simplest assumption in this endeavor is to assume ―perfect knowledge‖ of
seasonal runoff volumes during operations simulation. A more realistic assumption is to
simulate decisions in response to imperfect forecasts of seasonal runoff volume, where
forecast uncertainty is similar to that from real-world forecast providers.

Our current capability in forecasting seasonal runoff volumes is based on the
understanding of historical relations between runoff volume during a given forecast
season, snowpack at the time of forecast issue, and accumulated water year precipitation
prior to the time of forecast issue (ENSO indices are also incorporated in several forecast
procedures). This relation has a climate context that could evolve with climate change,
which is meant to be reflected in Deliverable #5.

The list of influential seasonal runoff volume forecasts varies by sub-system in the
CSRB, which means that each RMJOC agency may feature different forecast periods in
their respective operations models. Also, for a given operations model, the list of


                                            15
Draft, 4 June 2009                                              For Review Only, Do Not Cite or Distribute


forecast variables varies by location, month of issue, and forecast season. For each
model, forecast data must be developed for each of the selected Hybrid-scenario and
Transient Climate Projections. Lastly, the current forecast seasons are tied to historical
climatology. There may be need to generate seasonal forecast volumes for seasons that
differ from traditional seasons depending on how climate change impacts runoff patterns
and predictability (e.g., warming causing earlier spring runoff, and motivating interest in
forecasts during March-June instead of April-July). The effort will feature method
development, with input invited from external parties.

Deliverable #6 – Storage Targets for Flood Control & Variable
Energy Content Inputs Consistent with Streamflow and Seasonal
Runoff Volume Forecast Inputs

The deliverable is a set of operating rule curves consistent with underlying climate and
hydrology assumptions that are used to guide the simulated operation of system
reservoirs for the various uses and needs. Two such input sets will need to be developed:
the flood control rule curves that reflect changed hydrology and (as an initial
consideration) application of current storage reservation diagrams and refill methodology
from USACE2; and, variable energy content curves (VECCs) that define the reservoir
contents necessary to refill the reservoir by July 31.

Deliverable #7 – Demonstration Study on Operations Analysis
Using Either Hybrid-scenario or Transient Climate Projection
Information

This deliverable is a comparative study of operations analyses conducted by each
RMJOC agency using both types of regional climate information: Hybrid ―climate
change‖ scenarios and Transient Climate Projections. The intent of the study is to show
how results on operational performance metrics revealed by such analyses vary with
information type, how they might be communicated to decision-makers, and what that
means for decision-making and planning response. Results should provide RMJOC
agencies a better understanding on the implications of choosing either perspective choice
for framing various types of RMJOC longer-term planning analyses addressing issues
2
  Note that this deliverable is predicated on use of existing storage reservation diagrams (SRD). New SRD
are not scoped to be developed in this effort. To explain, consider that in longer-term reservoir regulation
analyses, storage targets for flood control must be simulated. When USACE conducts such analyses, these
targets are set based on a seasonal volume forecast and the existing SRD for the draft portion.
Traditionally, USACE has completed this effort has a preliminary step before regulation simulation, with
targets are entered into the simulation as inputs. In this work plan, it is expected that climate change will
affect seasonal volume forecasts, and thus affect refill targets. What’s not known at this point is whether
there will be a significant change in refill targets under future climate conditions (Hybrid-scenario or
Transient Climate Projection). If changes are significant, then questions will be raised about whether
existing SRD remain appropriate for simulation in that future context. Pending the significance of such
target changes, it is possible that there will be interest in redeveloping the SRD for future climate
conditions, and then using those SRD with future seasonal volume forecasts to modify flood control targets.
This is not a trivial task and is not included in the cost estimate of this work plan. USACE estimates that
this could involve two experienced people working full time for a year to build a new set of SRD.


                                                     16
Draft, 4 June 2009                                For Review Only, Do Not Cite or Distribute


from water and power resources management to operating systems for flood risk
management.




                                         17
Draft, 4 June 2009                                         For Review Only, Do Not Cite or Distribute




III. Tasks
This section provides an outline of work plan tasks, as previewed in the introduction of
―II. Deliverables.‖ For each task, discussion is provided on considerations leading to
tasking choices, RMJOC Technical Team activities, and expected outcomes including
work plan deliverables (II. Deliverables).


Task 1 - Climate Projections Survey and Selection

The objectives under Task 1 are to survey, select, and gather regional climate projection
information that will frame the development of work plan deliverables (Section II).
Decisions for each step would be documented. Reclamation would serve as lead
coordinating agency on this task.


Task 1.1 - Review of Regional Climate Projection Information
Available from UW CIG

The purpose of this task is to survey and prepare summary descriptions of available
regional climate projection information. Without reliance on previous work by UW CIG,
this would ordinarily be a time-consuming and resource-intensive task. However, as
discussed in Scoping Considerations, it was decided that this review effort would be
framed by the survey and collection of climate projection’s featured in the UW CIG data
development efforts. Their preliminary work has considered 40 climate projections from
the more than 100 available at the WCRP CMIP33 multi-model data archive4. (However,
UW CIG has recently indicated that they may reduce the count of projections in their
ultimate data deliverable.5)

To summarize, RMJOC Technical Team activities include:

       Meet with CIG and interested external collaborators to discuss CIG's collection of
        climate projections, and CIG’s rationale for focusing on these projections among
        those available in the CMIP3 dataset. Also, discuss CIG’s evaluation of these
        projections over the Pacific Northwest (Mote and Salathé 2009).

       Summarize discussions.

Outcome will be a documented survey of information.

3
  World Climate Research Programme Coupled Model Intercomparison Project Phase 3 (Meehl et al. 2007)
4
  http://www-pcmdi.llnl.gov/ipcc/about_ipcc.php
5
  Communication with Alan Hamlet, UW CIG, 5/14/2009: CIG is leaning towards serving CSRB datasets
that are derived from 20 of the 40 climate projections, representing 10 global climate models each
simulating two future greenhouse gas emissions scenarios.


                                                18
Draft, 4 June 2009                                    For Review Only, Do Not Cite or Distribute




Task 1.2 - Select Subset of UW CIG Regional Climate Projection
Information (Deliverable #1)

Two factors discussed under Scoping Considerations influence the framing of this task:
(#2) Choosing which UW CIG data types to use in the RMJOC dataset, and (#3)
Deciding how much of the UW CIG dataset to use for chosen types. On factor #2, UW
CIG is featuring different methodologies for translating climate projection information
into CSRB hydrologic response. The different methodologies give rise to the different
data types discussed earlier (e.g., Hybrid-scenario, Transient Climate Projection, and also
other methods discussed in Appendix B). Scoping decisions were made to focus on the
Hybrid-scenarios and Transient Climate Projections data.

Hybrid-scenarios represent incremental climate (or ―step-change‖ in climate) and thus
feature two hydrologic analyses: one for a base climate and another for a changed
climate. For their Hybrid-scenario work, UW CIG defines these ―climate change‖
scenarios as ―change in 30-year climate‖ computed within climate simulations spanning
historical to future periods. To elaborate, the ―historical climate‖ in these climate change
computations are from a given projections simulated historical period during 1971-2000
(i.e. a 30-year simulation period centered on 1985). The future period is then any of three
future 30-year periods centered on 2025, 2045, and 2085. Comparing climate between
historical and future periods within the projection then reveals a climate change
increment, which can then be used in subsequent hydrologic analysis ―with climate
change‖ (i.e. observed historical weather adjusted by this climate change increment).

Transient Climate Projections underlie diagnosis of Hybrid-scenarios. However, they can
also be used directly to infer hydrologic impacts. Whereas a Hybrid-scenario would
involve two hydrologic analyses (one with base, or observed historical, weather and
another with ―climate changed‖ weather), a Transient analysis would involve a single
hydrologic simulation where a transient climate projection (i.e. continuous time series of
precipitation and temperature from historical to future as simulated over the region) is
translated into a transient hydrologic projection. For comparison with Hybrid-scenario
analysis, the period of transient analysis would have to encapsulate the two periods
defining the Hybrid-scenario (e.g., if a Hybrid-scenario was based on 1971-2000 and
2031-2060 climates, the transient analysis would feature a single hydrologic simulation
represent 1971-2060 information from the transient climate projection).

For this work plan effort, the Hybrid-scenarios data would be regarded as the default
information type. The Transient Climate Projections data are considered here as the
alternative information type. The latter are potentially attractive for time-developing
adaptation planning, but potentially limited on its scale of application (possibly not
appropriate for local and sub-monthly evaluations (Elsner et al. 2009, Appendix B)).
Task 3 features demonstration operations analysis designed to inform which type may be
more appropriate for different longer-term planning situations.



                                            19
Draft, 4 June 2009                                    For Review Only, Do Not Cite or Distribute


On factor #3, a scoping decision was made to focus on a small, but representative, subset
of the UW CIG information. RMJOC agencies have held preliminary discussions on how
to select projections that might ―bracket‖ the spread of Hybrid information (e.g.,
Reclamation 2008a, CH2M-Hill 2008). A preliminary scoping assumption is that two of
the three future periods considered in UW CIG’s Hybrid-scenarios work would be
considered in this effort: the future 30-year periods centered on 2025 and 2045. Then,
for both future periods, a ―bracketing‖ set of Hybrid-scenarios would be selected along
with a ―central projection‖. In other words, five Hybrid-scenarios would be selected for
each future period, leading to selection of ten Hybrid-scenarios for consideration in this
work plan. For comparison purposes in Task 3, the Transient Climate Projections
underlying these Hybrid-scenarios would also be selected. Assuming no overlap of
climate projections providing Hybrid-scenarios for both periods, there would be a
maximum of ten Transient Climate Projections also considered in this work plan.

Also on factor #3, it is recognized that there may be interest in exploring whether the
selection of Hybrid-scenarios might be informed by an evaluation of relative credibility
among the climate models generating such information. Reclamation 2008a includes
discussion on this issue. In that study, a decision was rationalized to not cull any of the
projections surveyed. Briefly, the rationale was founded on previous studies for tailored
for the Northern California region, showing that:

       The contrast in skill among climate models decreases as more climate simulation
        performance metrics are considered (Reichler et al. 2008, Gleckler et al. 2008 and
        Brekke et al. 2008).

       Even though a climate model skill analysis might be performed prior to a
        projection selection process to justify culling ―lower skill‖ models from
        consideration, doing so does not necessarily guarantee an altered sense of climate
        projection uncertainty. This is because the assumed future scenario for
        greenhouse gas emissions (which defines future climate forcing) and projection
        initial-condition choices (which defines climate simulation starting point) both
        introduce substantial uncertainties. This was found to be the case for the Northern
        California region, where the spread of climate changes did not differ substantially
        when defined by climate projections produced by only ―higher skill‖ models
        versus a mix of ―higher and lower‖ skilled models (Brekke et al. 2008).

Such considerations will be revisited with UW CIG and with specific consideration to
projection considerations over the CSRB. Such discussion could offer new insight on
selection rationale.

To summarize, RMJOC Technical Team activities include:

       Adopt an approach designed to enable RMJOC agencies’ to utilize UW CIG’s
        Hybrid-scenario information to reflect an envelope of climate change possibilities
        and a centrally expected change. Preliminary approach would be similar to
        Reclamation (2008a), and involve:


                                             20
Draft, 4 June 2009                                            For Review Only, Do Not Cite or Distribute




             o selecting a bracketing set of Hybrid-scenarios among those analyzed by
               UW CIG during early- and mid-21st century periods (i.e. 4 to bracket from
               less warming to more warming and lesser to greater precipitation change)

             o selecting a central Hybrid-scenario for both period above (i.e. scenario that
               comes closest to median changes in temperature and precipitation evident
               in the UW CIG’s 40 climate projections)

       Obtain UW CIG’s Transient monthly climate projections associated with the
        sampled Hybrid-scenarios.

       Meet with CIG and collaborators to discuss selection approach and refinement
        opportunities.6

       Incorporate meeting reactions and finalize selection decisions

Outcome will be Deliverable #1 (Section II) constituting a set of 5 Hybrid ―climate
change‖ scenarios for two future periods (i.e. periods centered on 2025 and 2045), and
Transient Climate Projections that correspond to these Hybrid-scenarios. A master copy
of these data would be stored at BPA; each Agency would possess a resident copy for
work purposes.


Task 1.3 - Documentation and Internal Review

The purpose of this task is to summarize decisions made under Tasks 1.1 and 1.2, and
characterizing background information as necessary. RMJOC Technical Team activities
include:

       Prepare draft documentation on Task 1. 1 and 1.2 for internal review.

       Distribute for internal review.

       Gather and incorporate comments; prepare external review draft.

Outcome will be draft documentation for external peer review to be initiated in Task 3.4.




6
 It is possible that an entirely different approach may be chosen as an outcome of this meeting.
Developing the preliminary approach provides meeting participants a preliminary concept and rationale on
how RMJOC agencies are thinking that the body of CIG information might be related to planning activities.
That doesn’t preclude meeting discussion leading to adjusted aspects of projection selection.


                                                   21
Draft, 4 June 2009                                             For Review Only, Do Not Cite or Distribute




Task 2 - Hydrologic Data Selection and Verification

The products of Task 1 are selections for Hybrid-scenarios and Transient Climate
Projections from the UW CIG data set. Each of these has various types of regional
hydrologic data associated with them, as will be outlined in this section. The subtasks
under Task 2 involve obtaining, evaluating, and verifying these data, and also preparing
summary documentation suitable for RMJOC agencies’ planning studies.


Task 2.1 – Obtain and Review Hydrologic Model

The purpose of this task is for RMJOC technical staff to become oriented with the
hydrologic model used by UW CIG to develop regional hydrologic information. The
model is a CSRB application of the Variable Infiltration Capacity (VIC) model (Liang et
al. 1994), developed by UW CIG at 1/16º spatial resolution. From the more detailed
model documentation provided by UW CIG, this task involves preparing summary model
description for RMJOC agencies’ planning purposes. Discussion will address how the
VIC model–application in the CSRB compares with other hydrologic model options in
the region (e.g., Northwest River Forecast Center’s model, the National Weather Service
River Forecast System, which supports operational flood risk management and system
operations). It is understood that glacier dynamics are poorly represented in available
models. Documentation will highlight this issue and include status discussion on
regional research to address this issue7.

To summarize, RMJOC Technical Team activities include:

       Review and summarize the VIC model structure.

       Obtain, review and summarize the UW CIG's Columbia-Snake Basin VIC
        hydrologic model applied at 1/16º spatial resolution.

Outcomes will be a documented model review and a stored copy of the VIC-application
used to develop Dataset #3 in this study (stored with master dataset at BPA).


Task 2.2 – Obtain and Review Daily Weather Inputs (Deliverable #2)



7
  Glacial dynamics may play an important role in determining runoff characteristics from northern reaches
of the Columbia Basin. As possible, consideration will be given to research developments that might
permit better depiction of these dynamics in VIC. However, given the schedule of this work plan and the
status of UW CIG and BC-Hydro research on glacier dynamics for hydrologic modeling, it is not expected
that such research developments would be available in time to influence implementation of this work plan.



                                                   22
Draft, 4 June 2009                                              For Review Only, Do Not Cite or Distribute


The purpose of this task is to gather and summarize the weather data developed by UW
CIG to translate Hybrid-scenarios and Transient Climate Projections into regional
hydrologic information. These data include weather variables at temporal and spatial
resolutions required by the VIC application for the CSRB (i.e. daily and at 1/16º
resolution). Briefly, those data include:

         Hybrid-scenarios

              o Base Weather (Historically Observed8)
                    Variables: Temperature, Precipitation, winds
                    Temporal Coverage: WY1916-2006, Daily
                    Spatial Coverage: CSRB, 1/16º (~6km)
              o Climate Change Weather (Synthetic9)
                    Variables and Coverage same as Base

         Transient Climate Projections

              o Time-Developing Weather (Synthetic10)
                    Variables: Temperature, Precipitation, winds
                    Temporal Coverage: (roughly) 1950-209911, Daily
8
  VIC simulates surface water balance using input information on temperature, precipitation and winds.
The data for these variables are derived from historical station observations that have been mapped to the
VIC CSRB application’s 1/16º spatial grid. A common ―Base Weather‖ dataset provides the ―base‖
reference in each Hybrid ―climate change‖ scenario.
9
  For temperature and precipitation, these data are scaled versions of the Base Weather’s daily, gridded
historically observed values. The scaling reflects period changes in monthly temperature and precipitation
over the region, as discussed in Appendix A. Details on Scoping Considerations. Because these
temperature and precipitation sequences have not been observed, they are labeled here as synthetic. For
winds, these data is identical to Base Weather’s values.
10
  For temperature and precipitation, as with Hybrid scenario, these data are scaled versions of the Base
Weather’s daily, gridded historical observed values. However, the scaling is done parsimoniously such that
the monthly evolving nature of the Transient Climate Projection is preserved (see ―Appendix A. Details on
Scoping Considerations‖). Briefly, the scaling is done uniquely for each month time step in a given
projection. Progressing through a monthly climate projection, an historical observed month is randomly
sampled and associated with each projection month to provide a daily sequence pattern of precipitation and
temperature. These daily sequences are then scaled to match month-total precipitation and mean-
temperature of associated projection months (Wood et al. 2002). The result is a synthetic weather series
that matches monthly sequencing of the given climate projection. On a sub-monthly basis, the daily
sequencing patterns usually resemble those from the observed record, but scaled as necessary to match a
given projection month. However, the flexibility in sampling observed historical months can create
artifacts that are challenging to interpret for local, sub-monthly impacts assessments (Elsner et al. 2009,
Appendix B). More restrictive sampling causes the pool of eligible historical months to be reduced and
promotes repetitious daily storm patterns. Less restrictive rules lead to instances of associating observed
historical months with climatically contrasting projection months (e.g., sampling and observed historical
that is relatively dry and warm and then trying to scale it match the month precipitation and temperature of
a projection month that is relatively wet and cool). Particularly for precipitation, such associations can
produce daily events that significantly exceed the envelope of observed daily events.
11
     Based on communication with Alan Hamlet, UW CIG, 6/1/2009


                                                    23
Draft, 4 June 2009                                            For Review Only, Do Not Cite or Distribute


                         Spatial Coverage: CSRB, 1/16º (~6km)

These data will be accompanied by documentation from UW CIG. From the more
detailed documentation provided by UW CIG, this task involves preparing summary data
development descriptions for RMJOC agencies’ planning purposes.

To summarize, RMJOC Technical Team activities include:

        Obtain the observed and synthetic weather data from UW CIG for selected
         Hybrid-scenarios and Transient Climate Projections (i.e. Deliverable #2).

        Review and summarize UW CIG methods for data development and application
         in generating regional hydrologic information.

Outcomes will be a documented data and methods, and a stored copy of Deliverable #2 in
the RMJOC master dataset at BPA.


Task 2.3 - Obtain and Review Simulated Water Balance and Natural
Streamflow (Deliverable #3)

The purpose of this task is to gather and summarize the resultant regional hydrologic
information developed by UW CIG for the selected Hybrid-scenarios and Transient
Climate Projections, respectively. These data include gridded water balance variables
simulated by the CSRB VIC application (Task 2.1), as well as streamflow routed to the
various locations relevant to operations analyses conducted by RMJOC agencies12.
Briefly, those data include:

        Hybrid-scenarios
            o Base Hydrology (historical)13
                    Gridded water balance outputs from VIC (runoff,
                       evapotranspiration, snow water equivalent, soil moisture) having
                       temporal and spatial coverage same as Observed Weather Input.
                    Streamflow derived from the gridded natural runoff, routed to
                       various basin locations or inflow points.
            o Climate Change Hydrology
                    Gridded water balance outputs from VIC, similar to Base
                    Streamflow routed to various basin locations or inflow points,
                       similar to Base

        Transient Projections



12
   RMJOC agencies supplied UW CIG a list of desired routing locations consistent with reservoir system
inflow locations represented in RMJOC operations models.
13
   The VIC-simulated Base data are the same for each Hybrid ―climate change‖ scenario.


                                                   24
Draft, 4 June 2009                                         For Review Only, Do Not Cite or Distribute


            o Time-Developing Hydrology
                  Gridded water balance outputs and streamflow similar to Hybrid
                    Base and Hybrid Climate Change, but with temporal coverage
                    matching that of the climate projection.

The routed streamflow data will be used to adjust water supply assumptions for
operations modeling in Task 3 (i.e. regulated flows where required, seasonal runoff
volume forecasts, rule curve computations). Development of seasonal runoff volume
forecasts will rely on the use of both Deliverables #2 (precipitation) and Deliverable #3
(snow water equivalent and routed streamflow).

To summarize, RMJOC Technical Team activities include:

        Obtain the gridded daily water balance outputs from UW CIG’s VIC simulation
         of each selected Hybrid-Scenario and Transient Climate Projection.

        Obtain the streamflow routed to various basin locations from these same UW CIG
         VIC simulations.

        Review and summarize VIC simulation and routing methods, and resultant data.

Outcomes will be a documented data and a stored copy of Dataset #3 in the RMJOC
master dataset at BPA.


Task 2.4 - Independently Verify Deliverables #1, #2, and #3

Taking advantage of independent downscaled climate projection information over the
CSRB and making use of an 1/8º CSRB VIC application already developed, a scoping
decision was made for RMJOC agencies staff to conduct an independent verification on
the data obtained from UW CIG (see ―Scoping Considerations‖). A benefit of this
verification task is that RMJOC technical staff would develop a better understanding
about CIG’s data-development methods and what the resultant data represent. Both
benefits would better prepare RMJOC technical staff for Task 3 in this work plan, where
operations analyses are conducted, utilizing both Hybrid-scenario and Transient Climate
Projection information.

The data and model used in verification activities would include:

        downscaled monthly climate projections at 1/8º spatial resolution over the
         CSRB14.




14
  ―Statistically Downscaled WCRP CMIP3 Climate Projections‖ at http://gdo-
dcp.ucllnl.org/downscaled_cmip3_projections/


                                                 25
Draft, 4 June 2009                                               For Review Only, Do Not Cite or Distribute


        gridded, observed 1950-1999 daily weather data (Maurer et al. 2002, or M2002)
         developed at 1/8º spatial resolution, serving as the archive of historical daily
         sequences used in synthetic weather generation10, 15

        CSRB VIC application developed at 1/8º spatial resolution16.

RMJOC Technical Team activities include:

        Verify Deliverable #1: Focus on the selected Transient Climate Projections from
         UW CIG that coincide with 1/8º downscaled climate projections from the data
         resource mentioned above. For these projections and each source, obtain the
         gridded monthly temperature and precipitation time series over the CSRB, and
         assess for similarities and differences with that obtained from UW CIG.

        Verify Deliverable #2: Focus on the gridded, daily historical weather over the
         CSRB. For the period of common historical overlap (1950-1999), compare the
         1/16º daily data obtained from UW CIG with the 1/8º M2002 data.

        Verify Deliverable #3: Focusing on the Transient Climate Projections from
         ―Verify Dataset #1‖ and their associated Hybrid ―climate change‖ scenarios,
         redevelop the simulated, gridded water balance information over the CSRB in
         Deliverable #3, but do so by apply the 1/8º CSRB VIC model. Compare
         simulated water balance data from this verification effort with that obtained from
         UW CIG.

        Meet with CIG and collaborators to discuss results.

Outcomes will be a documented development of verification data and their comparison
with UW CIG data. Verification data will be stored at BPA with the master dataset.


Task 2.5 - Internal Review, Revised Documentation

RMJOC Technical Team activities:

        Prepare draft documentation on Task 2. 1 through 2.4 for internal review.

        Distribute for internal review


15
   The 1/8º M2002 daily weather data are similar to the 1/16º data used by UW CIG. However, the period
of historical coverage differs, as do some of the criteria used to determine which station weather histories
were represented when mapping station data to the spatial grid. Data are available at:
http://www.engr.scu.edu/~emaurer/data.shtml#Gridded_Obs
16
   Application is featured in the UW West-Wide Seasonal Hydrologic Forecasting System
(http://www.hydro.washington.edu/forecast/westwide/index.shtml) and was provided to Reclamation by
Dr. Andy Wood (UW).


                                                     26
Draft, 4 June 2009                                    For Review Only, Do Not Cite or Distribute


       Gather and incorporate comments; prepare external review draft.

Outcome will be draft documentation for external peer review to be initiated in Task 3.4.



Task 3 - Operations Analyses Preparation and Demonstration to
Reveal Implications of Hybrid- or Transient-style Approach

The majority of work plan effort is spent under Task 3. Prior tasks primarily involve
reviewing and verifying regional climate and hydrologic information obtained from UW
CIG. These data are now put to the task of supporting typical operations analyses
required in longer-term planning studies conducted by RMJOC agencies. Specifically,
methods are adopted and implemented for translating these climate and hydrology data
into various hydrology- and supply-related planning assumptions for reservoir operations
modeling conducted by RMJOC agencies.

The first part of Task 3 focuses on methods for adjusting inflows for depletions and
associated seasonal runoff volume forecasts (Tasks 3.1 and 3.2). The next task involves
establishing methods for developing a set of operating rule curves consistent with
underlying climate and hydrology assumptions that are used to guide the simulated
operation of system reservoirs for the various uses and needs (Task 3.3). Two such input
sets will need to be developed: the flood control rule curves that reflect changed
hydrology and (as an initial consideration) application of current storage reservation
diagrams and refill methodology from USACE; and, variable energy content curves
(VECCs) that define the reservoir contents necessary to refill the reservoir by July 31. In
addition, it may be necessary to compute other operational targets for ESA operations
that are dependent on underlying hydrology and forecasts. For Tasks 3.1-3.3, methods
will be developed to interface RMJOC agencies’ operations models with either the
Hybrid- or Transient-type data received from UW CIG

Demonstrated use of the Task 3.1-3.3 methods is then featured in Task 3.4. As already
stated, the purpose of these demonstration operations analyses is to produce information
on how choice of Hybrid- or Transient-style information affects depiction of operations
performance. Results from this style comparison are expected to produce scoping
guidance for RMJOC agencies on the appropriate style of climate change treatment in
various planning processes.

It is re-emphasized here that this work plan only goes so far as to demonstrate operations
analysis assuming climate change impacts on hydrology and water supplies. Other
operations analysis assumptions might be impacted by climate change (e.g., demands,
operational constraints), but adjustment of such assumptions is not scoped in this initial
work effort.




                                            27
Draft, 4 June 2009                                                For Review Only, Do Not Cite or Distribute


Task 3.1 – Prepare Adjusted Streamflow Inputs (Deliverable #4)

Separate methods will be adopted/developed for relating the UW CIG streamflow
information in a given Hybrid scenario or Transient Climate Projections to reservoir
inflow adjustments for each RMJOC agency’s operations models as needed17. Each of
these models are used to support RMJOC agencies’ longer-term planning studies, and can
be used to simulate system operations response to scenario hydrology (i.e. system
inflows, seasonal runoff volume forecasts), water demands, and operating constraints.
The models targeted in this work effort are described below.

          BPA
             o Model name: HydSim
             o Model description: 14-period (monthly except for April and August
               which are split) operations simulation of Federal Columbia River Power
               System (FCRPS)). Streamflow input is un-regulated, total inflow at
               designated points. Only regulated flows are used for the Snake River at
               Brownlee, and for the Yakima and Deschutes basins.

          Reclamation:
              o Model name: ModSim applications for basins listed below
              o Model description: monthly operations simulations for tributary systems
                 in the Yakima basin, Deschutes basin, and Snake basin above Brownlee.

          USACE:
             o Daily Model names: HEC-ResSim and AutoReg/SSARR
             o Model description: daily operations simulation of FCRPS flood control
                operations; dependent on headwater flows, local inflows regulated flows
                from the Snake River at Brownlee, and from the Yakima and Deschutes
                basins.

                o Monthly Model: HYSSR
                o Model description: monthly (14 period) simulation of FCRPS
                  hydropower operations, streamflow input is un-regulated, total inflow at
                  designated points, dependent on pre-processed regulated flow from the
                  Snake River at Brownlee, and from the Yakima and Deschutes basins

Traditional development and application of these models has involved the assumption
that hydrology and basin runoff depletions from the historic record sufficiently portray
water supply variability that could be experienced during the future. The period of
historic record (i.e. range of water years) used by each model also varies:

          BPA’s HydSim:                                         1929-1999

          Reclamation’s ModSim Yakima:                          1926-1999

17
     As possible, this effort will be coordinated with the 2010 Modified Flows update.


                                                      28
Draft, 4 June 2009                                           For Review Only, Do Not Cite or Distribute




        Reclamation’s ModSim Deschutes:                    1929-1999

        Reclamation’s ModSim Snake:                        1928-2001

        USACE’s ResSim/AutoReg:                            1929-1999

        USACE’s HYSSR:                                     1929-1999

Generating streamflows that represent Hybrid-scenario data involves the following:

        For those RMJOC agencies that use monthly modified inflow points for their
         modeling, the Base estimates will be consistent with the most current available
         modified flows dataset based on historical hydrology and scenario consumptive
         use. For model applications that use natural flow as input, the historically created
         or measured natural flow will serve as the Base streamflow estimates.

        For each Hybrid-scenario, a scaled version of the Base natural or modified
         streamflow estimates will be developed, preserving the Base series’ sequencing
         aspects (e.g., occurrence of extreme years, timing of droughts), but also reflecting
         ratio change in natural streamflow statistics associated with the given Hybrid-
         scenario (i.e. climate change streamflow relative to base streamflow, Task 2.3).
         This adjustment may follow a variety of peer-reviewed approaches (e.g.,
         Anderson et al. 2008, Reclamation 2008a, Reclamation 2008b, CH2M-Hill 2008,
         Brekke et al. 2009b, or UW CIG’s recent assessment procedures in Vano et al.
         2009).18

        After developing the natural or modified streamflow time series that reflect the
         given Hybrid-scenario’s climate change assumption, RMJOC agencies’ methods
         for post-processing streamflow into the appropriate input to respective models
         will be applied. The post-processing methods may reflect requirements for
         natural lake regulation or basin water management activities. A preferred
         approach for handling the latter remains to be reconciled, and will be part of the
         effort in this task.

Generating reservoir inflows that represent Transient Climate Projections data involves
the following:

        Rather than starting with the historical monthly natural flow estimates as
         described for the Hybrid-scenario procedure, the starting point is instead a natural
         streamflow ―projection‖ (from historical to future) developed in association with a
         given time-developing Transient Climate Projection. This means that sequencing
18
  These inflow adjustments are assumed to be expedited by having UW CIG develop routed streamflow
information for Hybrid-scenarios and Transient climate projections at all ―inflow‖ locations used in
RMJOC operations models. Should that not be the case, assumptions will have to be introduced to
geographically extrapolate runoff at missing locations


                                                  29
Draft, 4 June 2009                                     For Review Only, Do Not Cite or Distribute


        (or frequency) aspects in the historical portion of this series do not align with
        historical experience. More specifically, drought and surplus periods occur in the
        historical portion of the transient natural streamflow projection, but not during
        years matching historical experience (due to limitations in climate projection
        development). However, they should exhibit historical period-statistical
        characteristics that match historically observed streamflow statistics given that
        climate projection bias-correction is implemented prior to developing transient
        natural runoff projections (Appendix B).

       Developing water supply sequences associated with transient natural runoff
        projections involves:
            o During a historical period of common overlap within the natural
                streamflow projection data from UW CIG and the RMJOC agencies’
                period of historical natural streamflow estimates, identify bias between
                these two data sources at each reservoir inflow locations of interest.
            o Adopt and apply a scheme to correct UW CIG simulated natural
                streamflow data to account for these biases.
            o Using bias-corrected natural streamflow projections, adopt scenario
                assumptions on basin depletions and natural lake effects (similar to
                assumptions featured in the Hybrid-scenario approach) to convert these
                natural streamflow data into modified flow data.18

RMJOC Technical Team activities:

       For each RMJOC operations model listed above, adopt/develop and apply
        methods to prepare streamflow inputs to reflect base and climate change
        hydrology for each of the selected Hybrid-scenarios (i.e. as identified in Task 1.2,
        ten scenarios, which are divided into five scenarios for the 30-year period
        centered on 2025 and five scenarios for the 30-year period centered on 2045).

       For each RMJOC operations model listed above, adopt/develop and apply
        methods to prepare streamflow projections consistent with streamflow projections
        associated with each of the selected Transient Climate Projections (i.e. as
        identified in Task 1.2, up to 10 projections depending on whether there’s any
        overlap between the five transient climate projections that underlie ―2025‖
        Hybrid-scenarios and the five transient climate projections that underlie ―2045‖
        Hybrid-scenarios).

       Meet with CIG and collaborators to discuss methods and implementations.

Outcomes will be a documented inflows data and development methods. A copy of
developed data and operations models will be stored at BPA with the master dataset.




                                             30
Draft, 4 June 2009                                              For Review Only, Do Not Cite or Distribute


Task 3.2 - Prepare Adjusted Seasonal Runoff Volume Forecasts
(Deliverable #5)

It is expected that a warming climate could reduce snow water equivalent (SWE)
conditions in the future. Such projected changes in the presence and amount of
snowpack leads to questions about what that means for the predictability of seasonal
runoff volume during the snowmelt season that traditionally occurs during spring-
summer19. This is relevant for contemporary operations and their depiction in longer-
term planning.

          Project operations (i.e. monthly operations on up to one-year look-ahead) is based
           on forecasts of spring-summer runoff volume, which are produced by forecast
           models designed to reflect historical relationships between autumn-winter
           precipitation, winter SWE, and spring-summer runoff (i.e. P-SWE-Q
           relationship).

          For certain planning studies, project operations simulated by RMJOC agencies’
           models (Task 3.1) have been based on use of perfect foresight of runoff volumes
           during volume forecast periods. For example, the seasonal (April-August)
           volume forecast in any given year produced in January, February, March, and
           April is the same as the volume of the April-August runoff volume during that
           given simulation year. For other studies, BPA and the Corps also have the ability
           to supply HydSim and HYSSR with simulated operations informed by imperfect
           seasonal runoff forecast information. In other words, a January, April-August
           volume forecast is different than February’s volume forecast, and March’s
           volume forecast, etc. Reclamation is currently developing similar capability to
           simulated operations scheduling in its ModSim applications based on input
           imperfect seasonal runoff forecasts20.

Under climate change, warming could impact the P-SWE-Q relationship enough so that
the historical relationships underlying current forecasting models may no longer be
appropriate (Dettinger and Culberson 2008). It is reasonable to assume that water supply
forecast service providers (i.e. NRCS National Water and Climate Center and NWS River
Forecast Centers) would recalibrate their water supply forecast models in a changing
climate to reflect more recent developments in the P-SWE-Q relationship.

This task involves developing seasonal runoff volume forecasts consistent with the
regional hydrologic information in each selected Hybrid-scenario and Transient Climate
Projection. Methods from ongoing Reclamation research are expected to be applicable
for this effort19. Briefly, these methods involve applying NRCS procedures for
developing statistical water supply forecast models within a basin hydroclimate defined
by spring-summer runoff, antecedent precipitation, and snowpack at the time of forecast

19
     ―Erosion of Water Supply Predictability under Climate Change?‖, information at:
https://www.usbr.gov/research/Propcweb/reviewer/print_research_question.cfm?fy=2009&proposalid=2726
20
     Contacts: John Roche and Mary Mellema, PN region.


                                                     31
Draft, 4 June 2009                                              For Review Only, Do Not Cite or Distribute


issue; and, associated with either a ―climate change‖ portion of a Hybrid-scenario or a
Transient Climate Projection.

        For a Hybrid-scenario, both ―climate change‖ and ―base‖ hydroclimate datasets
         would be available (i.e. VIC-simulation of snow water equivalent (SWE) and
         runoff based on VIC simulation inputs for precipitation). Parallel sets of seasonal
         runoff volume forecast data would be generated for both ―climate change‖ and
         ―base‖ conditions for the list of forecast locations, months of issue, and forecast
         seasons. As described under Task 2, for ―base‖ the VIC simulation would be
         forced by historical observed precipitation and produce runoff and SWE that’s
         close to historical observations. For ―climate change‖, the VIC simulation would
         be forced by a sequence of precipitation that’s correlated with historical observed,
         but adjusted to reflect change in period climate (i.e. the climate change increment
         associated with the Hybrid Scenario). These sets would be respectively used to
         generate season runoff volume forecasts that would then be respectively used in
         the operations analyses featuring ―climate change‖ and ―base‖ modified flows
         discussed in Task 3.1.21

        For Transient Climate Projections, a projection of forecasts would be generated
         for the list of forecast locations, months of issue, and forecast periods. Given that
         the underlying hydroclimate is a projection (i.e. time-developing, statistically non-
         stationary), it is expected that the forecast model used to develop these data will
         have to also reflect time-developments in underlying statistics. In fact, this
         mimics model-maintenance practices by both NRCS and some NWS RFCs,
         where statistical water supply forecast models are periodically redeveloped to
         reflect observed developments in the statistical relationships between
         precipitation, snow water equivalent and runoff. One choice with climate change
         is whether to assume that each cycle of model-maintenance features model re-
         development using an expanding complete period-of-record, or using a moving
         fixed-duration period of recent-record. The ongoing research effort19 is exploring
         that issue and is expected to inform application in Task 3.2.

As said, seasonal runoff volume forecasts have to be developed for the various operations
models considered in this effort (Task 3.1). For example, the HydSim model features
operations informed by input seasonal runoff forecast data at 10 locations. Depending on
location, there could be as many as 4 forecast seasons driving operations logic. For a

21
  It is recognized that this would mean developing seasonal runoff volume forecasts for the ―base‖ case
that are less accurate or have greater uncertainty relative to the actual forecasts served historically by
NRCS, NWS NWRFC and USACE. The rationale for proceeding with the former is that it provides a
consistent forecasting framework that can be applied using either observed historical or simulated basin
hydroclimate conditions (i.e. a statistical framework that reflects the relation between spring runoff and
antecedent precipitation and snowpack at the time of forecast issue). The degree to which ―base‖ case
forecasts are less accurate or more uncertain than historical actual forecasts will be evaluated as part of
Task 3.2 activities. Depending on results from this evaluation, consideration will be given for what these
results imply for use of these simulated ―base‖ forecasts for various types of longer-term planning studies
conducted by RMJOC (e.g., monthly time-step simulations informing power resources planning versus
daily time-step simulations informing flood control analyses).


                                                     32
Draft, 4 June 2009                                     For Review Only, Do Not Cite or Distribute


given forecast season, there may also be multiple issuances of a forecast as the season
progresses and updated information is available.

RMJOC Technical Team activities include:

       For each RMJOC operations model listed above, adopt/develop and apply
        methods tailored for Hybrid-scenario application, involving the development of
        ―base‖ and ―climate change‖ sets of seasonal runoff volume forecasts at model-
        dependent lists of forecast location, month of issue, and forecast period.

       For each RMJOC operations model listed above, adopt/develop and apply
        methods tailored for Transient Climate Projection application, involving the
        development of forecast projections at model-dependent lists of forecast location,
        month of issue, and forecast period.

       Meet with CIG and collaborators to discuss methods and implementations.

Outcome will be a documented dataset of water supply forecasts consistent with Datasets
#1 and #3 (i.e. Dataset #4). A copy of developed data and operations models will be
stored at BPA with the master dataset.


Task 3.3 - Prepare Adjusted Flood Control Storage-Targets and
Variable Energy Content Curves (Deliverable #6)

As mentioned, two hydrology-driven operations rule curve inputs are addressed in this
work plan: the flood control drawdown and refill rule curves developed by the USACE;
and, variable energy content curves (VECCs) that define the reservoir contents necessary
to refill the reservoir by July 31. In addition, any volume dependent ESA operations or
targets will also need to be computed.

Discussion here is focused on the flood control curves and VECCs. The flood control
curves specify end-of-month maximum reservoir contents and are driven by SRD2,
seasonal runoff volume forecasts, and assumptions about runoff shape. The VECCs are
used in BPA’s HydSim model and USACE’s HYSSR model, and specifies variable refill
curves, and is based on volume forecasts and shape of runoff. (Task 3.3).

RMJOC Technical Team activities include:

       The VECCs are monthly lower reservoir limits for each reservoir to assure a 95
        percent probability of refill by 31 July. The ―variable‖ comes from using the
        January – July volume forecast working backwards from June to January. The
        process incorporates using the forecast errors of the procedure to represent the
        probability of .95 of refilling by July 31st. The procedure also uses the historical
        percent of the forecast volume that comes off each month and the project



                                             33
Draft, 4 June 2009                                    For Review Only, Do Not Cite or Distribute


        discharge requirements or minimum outflow to set the end-of-month VECC
        elevations.

       On computing flood control rule curves for the operations and regulation models
        in this study, several levels of detail could be considered. The level of detail
        scoped for this work plan involves the Corps using existing SRD2 and seasonal
        volume forecasts from Task 3.2 to compute flood control rule curves; assumptions
        about runoff shape may be influenced based on hydrologic information gathered
        in Task 2. This is a simple procedure, can be accomplished quickly. However,
        this does not consider that climate change impacts on hydrology could motivate
        adjustments to SRD at some time in the future2. Thus, a more resource-intensive
        and time-consuming effort might involve first examining hydrologic conditions
        associated with the Hybrid-scenarios and Transient climate projections scoped in
        this work plan, as a result of climate change, determine whether new SRD are
        warranted, develop new SRD, and regulate each year individually to meet the
        desired flood objectives. It is expected that results generated in the demonstration
        analyses under Task 3.4 should provide preliminary illustration of how existing
        SRD serve flood control objectives under climate change conditions featured in
        this work plan. Such preliminary illustration could help scope any subsequent,
        more intensive efforts to evaluate SRD adjustments.

Outcome will be a set of flood control rule curves and VECC for each Hybrid-scenario
and each Transient Climate Projection and a documented set of methods for adjusting
these hydrology-drive inputs. A copy of developed planning inputs will be stored at BPA
with the master dataset.


Task 3.4 – Demonstration Analyses using Hybrid-scenario and
Transient Projection Inputs (Deliverable #7)

The purpose of this task is to produce an operations analyses using each RMJOC
agencies’ operations model, and featuring hydrologic-input adjustments (Tasks 3.1-3.3)
reflecting either Hybrid-scenario or Transient Climate Projections. The intent of the
study is to reveal differences in operations impacts relative to the incremental climate
change in a given Hybrid-scenario versus the time-developing climate in a given
Transient Climate Projection. This is relevant to how results and planning outcomes are
communicated to decision-makers. Results should provide RMJOC agencies a better
understanding on the implications of choosing either perspective choice for framing
RMJOC planning analyses.

RMJOC Technical Team activities:

       For each selected Hybrid-scenario and operations model, simulate operations
        using the associated streamflow, seasonal runoff volume forecast and Task 3.3
        input adjustments (as applicable). Each RMJOC agency would conduct analyses
        with their respective model tool. Reclamation analyses would necessarily


                                            34
Draft, 4 June 2009                                             For Review Only, Do Not Cite or Distribute


         preceded BPA and USACE analyses given that Reclamation analyses produce
         regulated flow inputs to BPA and USACE analyses.

        Do the same for each selected Transient Climate Projection. Sample simulated
         operations information during projection periods that coincide with the climate
         periods used to define ―climate change‖ in the Hybrid-scenarios.

        Evaluate and summarize changes in operations based on Hybrid-scenario analysis
         versus Transient projection analysis.

        Develop impressions on strengths and weaknesses for each style for given
         planning processes, hopefully to provide scoping guidance for future efforts.

Outcome will be a documented analyses and scoping guidance on the use of Hybrid-
scenarios versus Transient Climate Projections for various RMJOC planning efforts. A
copy of developed data and results will be stored at BPA with the master dataset.


Task 3.5 - Peer Review, Revisions, Finalize Documentation

This task involves facilitating internal and external review of methods development and
demonstration in Tasks 3.1 through 3.4.

RMJOC Technical Team activities:

        Prepare draft documentation on Tasks 3. 1 through 3.4 for internal review.

        Distribute for internal review.

        Gather and incorporate comments; prepare external review draft.

        Plan and organize one-day workshop to begin external peer-review and to orient
         the review panel on what was done and why, understanding that relying solely on
         project documentation to communicate such information may be insufficient22.

        Incorporate panel findings in final documentation.

Outcome will be final documentation of project datasets, usage methods and operations
analyses, as well as documented guidance on appropriate use of these data and methods
for RMJOC longer-term planning purposes.



22
  Reclamation has utilized external peer-review in this manner. An example is the public-release and
review of new San Joaquin hydrology and operations logic in the joint CVP-SWP operations model,
CalSim II. Peer review activities were initiated August 2005 with a kickoff workshop, occurred thereafter
during Fall 2005, and were finalized in early 2006 (http://www.cwemf.org/Pubs/index.htm).


                                                   35
Draft, 4 June 2009        For Review Only, Do Not Cite or Distribute




                     36
Draft, 4 June 2009                                              For Review Only, Do Not Cite or Distribute




IV. Costs and Schedule
Costs and schedule information are summarized on Figure 3 and described in detail by
agency on Figure 4 through Figure 6. The assumed schedule of tasks and sub-tasks are
illustrated on Figure 7. The four staffing levels indicated in Figure 4 through Figure 6,
are consistent with the staffing levels listed in Table 1

Several other assumptions had to be made to develop costs and schedule information:

        RMJOC managers would have necessary resources in place to support a work
         plan start date of 1 October 2009. This includes key staff listed in Table 1,
         financial resources necessary to accommodate costs outlined in Figure 3 through
         Figure 623, and the availability of data and models listed in III. Tasks.

        UW CIG development of Deliverables #1, #2, and #3 would be completed during
         summer 2009. These data and their documentation would be available for
         RMJOC use by 1 October 2009.

        It is most economical for a single agency to lead documentation of Tasks 1 and 2,
         and to conduct verification analyses on Deliverables #1, #2, and #3. The other
         two agencies would remain engaged, but primarily in a review and methods
         scoping capacity. Given Reclamation’s recent experience working with climate
         projection information and translating it into hydrologic response information, it
         seemed appropriate for Reclamation to lead these efforts. This leads to relatively
         greater staff time estimates for Reclamation on Tasks 1 and 2.

        Agency costs on Task 3 are fairly well balanced given that each agency will be
         responsible for implementing methods to adjust water supply assumptions in their
         respective planning models, and for conducting and documenting the
         demonstration operations analyses with their models. Some additional time is
         assumed for USACE staff on Task 3.1 because their efforts will focus on
         identifying an appropriate method for specify daily inflow time series to their
         reservoir regulation model, which is expected to involve more challenges than
         specifying the monthly inflow time series for BPA and Reclamation operations
         models.




23
  Internal financial resources are not identified in this work plan document. However, RMJOC agency
representatives have held discussions about financial source options. For the purposes of this public review
process, it is presumed that such resources will be available for this effort by October 1, 2009.


                                                    37
Draft, 4 June 2009                                       For Review Only, Do Not Cite or Distribute




Figure 3 - Summary of Work Plan Costs and Schedule (details on Figure 4 through Figure 7).




                                               38
Draft, 4 June 2009                                              For Review Only, Do Not Cite or Distribute




Figure 4 - Work Plan Costs and Schedule Details: Reclamation.
(Note: Tasks 3.2 and 3.3 are listed with a pooled schedule period, even though there are sequential aspects
to these tasks, with Task 3.3 ultimately completed after finishing Task 3.2.)



                                                    39
Draft, 4 June 2009                                              For Review Only, Do Not Cite or Distribute




Figure 5 - Work Plan Costs and Schedule Details: Bonneville Power Administration.
(Note: Tasks 3.2 and 3.3 are listed with a pooled schedule period, even though there are sequential aspects
to these tasks, with Task 3.3 ultimately completed after finishing Task 3.2.)



                                                    40
Draft, 4 June 2009                                              For Review Only, Do Not Cite or Distribute




Figure 6 - Work Plan Costs and Schedule Details: USACE Northwestern Division.
(Note: Tasks 3.2 and 3.3 are listed with a pooled schedule period, even though there are sequential aspects
to these tasks, with Task 3.3 ultimately completed after finishing Task 3.2.)



                                                    41
Draft, 4 June 2009                                              For Review Only, Do Not Cite or Distribute




Figure 7 - Work Plan Schedule.
(Note: Tasks 3.2 and 3.3 are listed with a pooled schedule period, even though there are sequential aspects
to these tasks, with Task 3.3 ultimately completed after finishing Task 3.2.)




                                                    42
Draft, 4 June 2009                                               For Review Only, Do Not Cite or Distribute



V. Limitations
This work plan only goes so far as to demonstrate operations analysis assuming climate
change impacts on hydrology and water supplies. Future RMJOC and stakeholder efforts
might consider introducing climate change impacts on other operations analysis
assumptions related to demands and system operating constraints. Some examples
follow:

        Establishing flood control rules consistent with climate projection information.
         The core of Task 3.3 work concerning the development of climate change based
         flood control rule curves is predicated on applying scenario hydrology with
         existing Storage Reservation Diagrams (SRDs) and refill procedures. A
         subsequent stage would be to permit scenario hydrology to motivate associated
         modifications to the current methodology and SRDs. It is recognized that there
         are many challenges to developing a rationale for such modification, and it is
         remains to be seen whether a rationale can be developed that would be mutually
         acceptable among parties participating in a regional flood risk evaluation.
         However, the choice to not consider the potential for flood control modifications
         under climate warming ignores the changing role of snowpack dynamics in
         modulating flood runoff events. In any event, it is assumed that USACE would
         help lead any effort to develop a methodology appropriate for CSRB planning
         purposes (and that such development would necessarily involve parties involved
         with Columbia River Treaty 2014/2024 Review). To that end, some time has
         been scoped in Task 3.3 to support initial exploration of such modification24
         (leading to additional USACE costs for Task 3.3). However, given uncertainty of
         what such a preliminary evaluation might produce, this work plan has been
         scoped to not assume that these initial exploration efforts in Task 3.3 will produce
         methods that can be featured in operations studies on the near term. For that
         reason, the operations analyses developed under Task 3.4 feature operations
         simulation under climate change as constrained by scenario hydrology and
         existing SRDs and flood control procedures.

        Power Demands: It is expected that climate change would have implications for
         regional and Western U.S. electricity demands, which will need to be addressed in
         any FCRPS or regional climate impact studies. It is assumed that BPA and the
         NW Power and Conservation Council would lead any effort to develop an
         appropriate methodology for adjusting such electricity demands.

        Environmental Water Demands: It is expected that climate change could have
         implications for ecosystem demands for water of appropriate quality at certain
         locations and times. Such impacts on ecosystem needs, combined with questions
         about how societal priorities in environmental management may evolve, could
24
  One general concept might be to design ―climate change‖ flood control rules (or projected changes in
rules over time) that preserve historical flood protection service (i.e. explicitly defined, or implicitly
represented in historical storage encroachments or flood-stage levels) given climate and hydrologic
projection information.


                                                     43
Draft, 4 June 2009                                  For Review Only, Do Not Cite or Distribute


        raise questions about future assumptions for environmental water demands in
        PNW reservoir management. Any rationale for making future assumption
        adjustments for environmental water needs may require stakeholder input in order
        to accurate portray future scenarios for use in RMJOC planning studies.




                                           44
Draft, 4 June 2009                                   For Review Only, Do Not Cite or Distribute



VI. References
Anderson, J., F. Chung, M. Anderson, L. Brekke, D. Easton, M. Ejeta, R. Peterson, and
        R. Snyder (2008) ―Progress on Incorporating Climate Change into Management
        of California’s Water Resources‖, Climatic Change, 89, Supplement 1, 91-108.
Brekke, L.D., M.D. Dettinger, E.P. Maurer, and M. Anderson (2008), Significance of
        Model Credibility in estimating Climate Projection Distributions for Regional
        Hydroclimatological Risk Assessments, Climatic Change, Published online 2-28-
        2008.
Brekke, L.D., Kiang, J.E., Olsen, J.R., Pulwarty, R.S., Raff, D.A., Turnipseed, D.P.,
        Webb, R.S., and White, K.D., (2009a), Climate change and water resources
        management—A federal perspective: U.S. Geological Survey Circular 1331, 65 p.
        (Also available online at http://pubs.usgs.gov/circ/1331/.)
Brekke, L. D., E. P. Maurer, J. D. Anderson, M. D. Dettinger, E. S. Townsley, A.
        Harrison, and T. Pruitt (2009b), Assessing reservoir operations risk under climate
        change, Water Resour. Res., 45, W04411, doi:10.1029/2008WR006941.
CH2M-Hill (2008) ―Climate Change Study – Report on Evaluation Methods and Climate
        Scenarios,‖ submitted to Lower Colorado River Authority and San Antonio Water
        System, WBS 3.3.3.1.CH2M.14.82676, 103pp.
Dettinger, M. D.; and S. Culberson (2008), Internalizing Climate Change—Scientific
        Resource Management and the Climate Change Challenges. San Francisco
        Estuary and Watershed Science. 6(2) (June), Article 5.
Elsner, M.M., L. Cuo, N. Voison, J.S. Deems, A.F. Hamlet, J.A. Vano, K.E.B.
        Mickelson, S.Y. Lee, and D.P. Lettenmaier (2009) ―‖, Chapter 3 in: The
        Washington Climate Change Impacts Assessment, 66pp.
Gleckler, P.J., K.E. Taylor, and C. Doutriaux (2008) ―Performance metrics for climate
        models,‖ Journal of Geophysical Research, 113(D06104),
        doi:10.1029/2007JD008972.
Hamlet, A.F., and D.P. Lettenmaier (2005) ―Production of temporally consistent gridded
        precipitation and temperature fields for the continental U.S.,‖ Journal of
        Hydrometeorology 6(3):330-336.
Liang, X, D. P. Lettenmaier, E.F. Wood, and S.J. Burges (1994), A simple hydrologically
        based model of land surface water and energy fluxes for general circulation
        models, Journal of Geophysical Research, 99(D7), 14415-14428.
Maurer E.P., A.W. Wood, J.D. Adam, D.P. Lettenmaier, and B. Nijssen (2002) A long-
        term hydrologically-based data set of land surface fluxes and states for the
        conterminous United States. J Climate 15(22):3237-3251.
Meehl, G.A., C. Covey, T. Delworth, M. Latif, B. McAvaney, J.F.B. Mitchell, R.J.
        Stouffer, and K.E. Taylor (2007) ―The WCRP CMIP3 Multimodel Dataset – A
        New Era in Climate Change Research,‖ Bulletin of the American Meteorological
        Society, 88(9), 1383-1394.
Mote, P.W. and E.P. Salathé (2009) ―Future climate in the Pacific Northwest,‖ Chapter 1
        in HB1303 Report, 35pp.
Reclamation (2007) Final EIS – Colorado River Interim Guidelines for Lower Basin
        Shortages and Coordinated Operations for Lake Powell and Lake Mead, Bureau



                                           45
Draft, 4 June 2009                                  For Review Only, Do Not Cite or Distribute


       of Reclamation, U.S. Department of the Interior
       (http://www.usbr.gov/lc/region/programs/strategies/FEIS/index.html).
Reclamation (2008a) ―Appendix R - Sensitivity of Future CVP/SWP Operations to
       Potential Climate Change and Associated Sea Level Rise‖ in CVP/SWP OCAP
       Biological Assessment, Bureau of Reclamation, U.S. Department of the Interior
       (http://www.usbr.gov/mp/cvo/ocapBA_2008.html).
Reclamation (2008b) ―The Effects of Climate Change on the Operation of Boise River
       Reservoirs,‖ Initial Assessment Report
       (http://www.usbr.gov/pn/programs/srao_misc/climatestudy/boiseclimatestudy.pdf)
Reichler, T., J. Kim (2008) How Well Do Coupled Models Simulate Today’s Climate?,
       Bulletin of the American Meteorological Society, 89(3), 303-311.
Vano, J.A., M. Scott, N. Voison, C.O. Stockle, A.F. Hamlet, K.F.B. Mickelson, M.M.
       Elsner, and D.P. Lettenmaier (2009) ―Climate change impacts on water
       management and irrigated agriculture in the Yakima River Basin, Washington,
       USA‖, Chapter 3.2 in: The Washington Climate Change Impacts Assessment,
       61pp.
Wood, A. W., E. P. Maurer, A. Kumar, and D. P. Lettenmaier (2002), Long-range
       experimental hydrologic forecasting for the eastern United States, J. Geophysical
       Research-Atmospheres, 107(D20), 4429, doi:10.1029/2001JD000659.
Wood, A. W., L.R. Leung, V. Sridhar, and D.P. Lettenmaier (2004) ―Hydrologic
       implications of dynamical and statistical approaches to downscaling climate
       model outputs,‖ Climatic Change, 15, 189-216.




                                           46
Draft, 4 June 2009                                             For Review Only, Do Not Cite or Distribute




Appendix A. Details on Scoping Considerations
This appendix provides additional background information supplementing the discussion
under ―Scoping Considerations.‖

Factor #1 – Leveraging UW CIG Data Development

A considerable amount of climate projection and associated hydrologic data for the
CSRB will soon be made available by the University of Washington (UW) Climate
Impacts Group (CIG). These data are being developed by CIG through their effort
―Developing a Database of Regional Climate Change Scenarios‖
(http://www.cses.washington.edu/cig/res/rc/ccdb.shtml). The effort is expected to
produce a CSRB dataset representing information from 40 climate projections5. From
these projections, CIG is developing a dataset that includes regional temperature,
precipitation, natural water balance variables (runoff, snowpack, etc), and routed natural
runoff at locations requested by CIG stakeholders.

RMJOC agencies recognize the value of leveraging CIG’s data development efforts. It is
expected that the UW CIG dataset will be widely accepted for planning purposes by
CSRB water management agencies. Further, the developers at UW CIG had contact
RMJOC agencies to gather their input on desired output reporting from their hydrologic
modeling efforts, ideally to promote easier translation of their hydrologic information
into water supply assumptions used into longer-term planning efforts like those that may
be conducted by RMJOC agencies. Given these considerations, RMJOC agencies
decided to scope selection and adoption of UW CIG data for RMJOC purposes rather
than development of another dataset that would be redundant what that from UW CIG.

Factor #2 – Choosing which UW CIG Data Types to Use

As stated under Factor #1), the CIG effort is producing regional climate and hydrologic
information associated with 40 climate projections. CIG will be developing the regional
hydrologic information using three primary methods (see Approaches 1 through 3 in
Appendix B). The three methods are labeled for in this appendix as Delta, Transient, and
Hybrid. While only Transient and Hybrid methods are featured in work plan activities, it
is helpful to first understand both the Delta and Transient methods before discussing
CIG’s motivations for providing the Hybrid data. Briefly, these methods compare and
contrast as follows:

          Delta25 method: "step-change" hydrologic response to regional-average changes
           in period-mean monthly temperature (T) and (P) surveyed from raw GCM output.
               o This method is the simplest among these three to implement. It involves
                   defining historical and future periods to reveal step-change in climate and
                   hydrologic statistics specific to that pair of periods (e.g., 1971-2000
25
     In Appendix B, this is Approach 1 and labeled Delta Method Experiments.


                                                    47
Draft, 4 June 2009                                             For Review Only, Do Not Cite or Distribute


               climate to 2035-2064 climate). Change in hydrology for another future
               period would have to be defined using another Delta analysis specific to
               that future period. A challenge of this method is interpreting whether
               these differences represent ―climate change only‖ or a mix of ―climate
               change and climate variability‖ given that the multi-decadal variability
               present in the climate projections may differ depending on climate model
               used and on how projections were initialized (Brekke et al. 2009a).
             o On portrayal of sub-monthly weather (e.g., storms), this method produces
               storm intensity and occurrence patterns that are closest to historical
               observations.
             o On portrayal of monthly to decadal variability (e.g., drought and surplus
               periods), this method produces relative sequencing characteristics that are
               synchronized with historical sequencing, but with drought and surplus
               accumulations affected by step-changes in T and P.

        Transient26 method: time-developing hydrologic response to time-developing
         climate projection (i.e. time series, were period statistics for T and P evolve
         through time from historical to future). This method features the preliminary
         steps of Wood et al. (2002): (1) bias-correcting raw GCM output on a monthly
         and spatially at the GCM-scale to account for GCM tendencies to simulate
         historical climate that is too warm, cool, wet or dry; (2) spatially downscaling
         these monthly bias-corrected data to more local resolution (i.e. 1/16 degree); and,
         (3) temporally disaggregating these monthly bias-corrected and spatially
         downscaled (BCSD) data into daily BCSD T and P.
             o This method is considerably more complex than the Delta method in terms
                 of generating synthetic weather and in interpreting sequencing aspects
                 within the historical portion of the projections. However, it does not
                 involve the ―climate change‖ diagnosis challenges associated with the
                 Delta method. It instead involves a time-developing portrayal of
                 hydroclimate consistent with the climate projection considered. A
                 transient perspective may be more appropriate for adaptation planning,
                 where it’s necessary to understand the timing of impacts and necessary
                 adaptation actions (Brekke et al. 2009a). Challenges in the temporal
                 disaggregation method call into question the applicability of this method
                 for spatially localized or sub-monthly impacts assessments (Elsner et al.
                 2009)
             o On portrayal of sub-monthly weather, this method produces storm
                 intensity and occurrence patterns that can differ more substantially from
                 historical on a location and day- or month-specific basis.
             o On portrayal of monthly to decadal variability, this method produces
                 sequencing that matches monthly sequencing in the climate projections.
                 This means that projections determine possibilities for drought and surplus
                 spell and accumulation rather than historical records. There is some
                 influence of historical observation through the bias-correction step (1)

26
  In Appendix B, this is Approach 2 and labeled Bias Correction and Statistical Downscaling (BCSD).
That method label differently here to highlight its emphasis on the time-development of future climate.


                                                    48
Draft, 4 June 2009                                             For Review Only, Do Not Cite or Distribute


                   where the historical portion of climate projections are adjusted to be
                   statistically consistent with historical observations. This causes the
                   drought and surplus events in the historical portions of climate projections
                   to be somewhat similar to events observed historically. However,
                   differences remain because the sequencing aspects of the climate
                   projections are left unchanged. As a result, projections may exhibit
                   drought or surplus characteristics that are more or less persistent relative
                   to historical, depending on GCM tendencies for simulating monthly to
                   decadal variability.

          Hybrid27: "step-change" hydrologic response to spatially-distributed changes in
           period-distributions of monthly T and P surveyed from BCSD GCM output.
               o This method blends the ―step-change‖ view from Delta method and two of
                  the three projections-processing steps from the Transient method (i.e. bias-
                  correction and spatial downscaling). It also introduces the view that the
                  ―step-change‖ should reflect climate changes specific to relatively wetter,
                  drier, warmer, or cooler years (hence the comparison of period-
                  distributions rather than period-means) . The Delta method’s challenge
                  remains on interpreting these step-changes as ―climate change only.‖
                  However, the Transient method’s challenges of supporting sub-monthly
                  and localized impacts assessments are somewhat mitigated by this
                  approach.
               o On portrayal of sub-monthly weather (e.g., storms), this method produces
                  storm intensity and occurrence patterns that differ more from historical
                  compared to the Delta method, but are more similar to historical than the
                  Transient method, particularly on a month-specific and location basis.
               o On portrayal of monthly to decadal variability (e.g., drought and surplus
                  periods), this method produces relative sequencing characteristics that are
                  synchronized with historical sequencing, but with drought and surplus
                  accumulations affected by step-changes in T and P.

CIG provided information on tentative dataset completion plans (Alan Hamlet, personal
communication, 18 March 2009). For the scoping purposes of this work plan, it is
anticipated that CIG will make these data available during summer 2009. Tentatively, it
is understood that CIG’s Hybrid type would be the primary type featured in the dataset,
which would include climate and hydrology data specific to all 40 climate projections
analyzed for three future periods, leading to 120 Hybrid datasets that are projection- and
period-specific. The BCSC/Hybrid data would be complimented by limited
representation of the two other types. A few projections of the Transient type are
expected to be made available. The Delta type is expected to be represented for the same
three future periods featured in the Hybrid type, but with a change from that discussed
above. Rather than focus on projection-specific Delta analyses, CIG would only serve
projection-average, or ―composite‖, Delta analyses where the step-changes in period-
mean monthly T and P represent multi-projection, region-average changes.


27
     In Appendix B, this is Approach 3 and labeled Hybrid Bias Correction/Delta Method Downscaling.


                                                    49
Draft, 4 June 2009                                            For Review Only, Do Not Cite or Distribute


For this scoping effort, it was determined that each type has strengths and weaknesses in
portraying hydrologic response to projected climate. It was also determined that it is
difficult to make a choice on which data type is most appropriate for RMJOC purposes
without gaining experience on how to utilize these data in RMJOC planning models, and
subsequently evaluating the operations portrayals in these planning exercises. For that
reason, it was decided that this work plan should feature use of multiple CIG data types
to develop understanding and guidance on which types are appropriate for various types
of RMJOC longer-term planning (e.g., localized to system-wide conditions; operations
performance on sub-monthly or longer time scales). Further, it was decided that this
work plan should focus on use of Hybrid and Transient rather than all three data types.
The Transient method is more fundamentally different than the other two due to its time-
evolving portrayal or climate and hydrology compared to the step-change portrayal in the
other two. Given that Hybrid will be touted as CIG’s default ―step-change‖ method, a
decision was made to simplify efforts and only focus on one of the two step-change
methods.


Factor #3 – Choosing How much UW CIG Data to use

RMJOC agencies want to ensure that the spread of climate and hydrologic information
represented in the CIG dataset is represented in their agencies’ longer-term planning and
portrayal of climate change possibilities. In addition, there is the desire to do this with a
minimum amount of information from the CIG dataset. To be clear, the CIG effort is
expected to produce Hybrid datasets for each of their 40 climate projections5 multiplied
by 3 future periods labeled 2025, 2045, and 208528, and multiple Transient projections5.
On the Hybrid type, rather than assume use of all period- and projection-specific datasets
in longer-term planning, it was decided that a strategy would be explored and developed
in this work plan effort to rationalize how to cull down the CIG information for RMJOC
agencies’ planning purposes. Two guiding principles have been discussed:

         represent climate projection uncertainty in this information

         represent future periods that are relevant to RMJOC longer-term planning
          analyses

For scoping purposes, it was reasoned that CIG’s 2025 and 2045 periods would seem to
be more meaningful for RMJOC agencies’ longer-term planning activities and that this
first phase effort should focus on those two periods. It was also recognized that a
projection selection strategy might be featured similar to that of Reclamation (2008a) or
CH2M-Hill (2008) where ―bracketing‖ projections are identified for how they represent
the spread of step-change possibilities (e.g., representing the spread of step changes from
wetter to drier and less warming to more warming conditions),. Such bracketing

28
  The periods featured in the Hybrid analysis are expected to be 30-year windows centered around 1985,
2025, 2045, and 2085 (i.e. 1970-1999, 2010-2039, 2030-2059, and 2070-2099). The period centered on
1985 serves as the common ―historical climate‖ period from which step-changes in T and P are assessed in
any of the three future periods.


                                                   50
Draft, 4 June 2009                                             For Review Only, Do Not Cite or Distribute


projections might be complimented by selection of a fifth and ―central‖ step-change
projection. Regardless, the point is that a culling strategy would be explored in the work
plan. Whether such a culling strategy would also be supported by evaluation of relative
climate model skill remains to be seen, but will be discussed with UW CIG. It is
expected that there would be considerable stakeholder interest in establishing such a
strategy. Accordingly, the work plan calls for stakeholder interaction in the strategy
development and projection selection process. On the Transient data used in this work
plan, it was decided that projection selections from CIG’s Hybrid data would drive
selections from CIG’s Transient data.


Factor #4 – Verifying UW CIG Data

RMJOC agencies are interested in reviewing and understanding how CIG data were
developed. A basic understanding and review will be supported by the data-development
documentation provided by CIG. However, RMJOC agencies have the opportunity to
verify selected data through independent data-development. The opportunity stems from
the following:

          CIG’s Hybrid and Transient methods are being served by a CIG application of the
           BCSD method (Wood et al. 2002) at the 1/16º over the CSRB. This spatial
           resolution matches that of the hydrologic model being used to simulate hydrologic
           response for the entire CSRB (i.e. a 1/16º application of the Variable Infiltration
           Capacity (VIC) hydrologic model (Liang et al. 1994)).

          Independent of CIG’s efforts, the BCSD method has already been applied at the
           1/8º over this region and for most of the projections considered in CIG’s efforts.
           Data from the latter are freely available from the online ―Statistically Downscaled
           WCRP CMIP3 Climate Projections‖ (DCP Archive).29

          Consistent with DCP Archive resolution, a 1/8º application of the VIC model to
           the CSRB is also available and can be used for this effort, in conjunction with
           DCP Archive data, to verify Transient and Hybrid hydrologic data from UW CIG.

A benefit of doing more in-depth verification is better understanding on the strengths and
challenges of CIG’s data-development efforts. RMJOC technical staff would also
become more familiar with what the CIG data represent, thus helping them to provide
insight during the operations analysis portion of this work plan, where the final task is to
offer guidance on appropriate application of Hybrid versus Transient data types for
RMJOC agencies’ longer-term planning purposes.




29
     http://gdo-dcp.ucllnl.org/downscaled_cmip3_projections/


                                                     51
Draft, 4 June 2009                                     For Review Only, Do Not Cite or Distribute




Appendix B. A Brief Comparison of Downscaling
Methods for Climate Change Studies (UW CIG)
The following discussion is a draft narrative provided by UW CIG on March 18, 2009.
The discussion summarizes the various techniques that have been used by UW CIG in
their climate change impacts research within the CSRB. The following sections on Delta
Method Experiments, Bias Correction and Spatial Downscaling, and Hybrid Bias
Correction/Delta Method Downscaling collectively informed the discussion in ―Appendix
A. Details on Scoping Considerations‖ on ―Factor #2 – Choosing which UW CIG Data
Types to Use.‖ The remainder of this appendix is a reprint of the draft narrative provided
by UW CIG.


Prepared by Alan F. Hamlet (UW CIG)
3/17/2009


Introduction
Downscaling is a term used to describe the process of relating information or data at
relatively coarse spatial and temporal scales to desired products at finer spatial and
temporal scales. In the case of climate change experiments, the process is commonly
used to relate monthly simulations of temperature (T) and precipitation (P) data at
approximately 200km resolution produced by a global climate model (GCM) to finer-
scale information needed to drive a hydrologic model (e.g. daily data at 1/16th degree
resolution needed to drive the VIC model). This paper outlines several different
downscaling approaches, and discusses the strengths and weaknesses of each approach in
the context of different water resources planning applications.

Limitations of GCMs
Downscaling approaches are generally designed to avoid the many limitations of GCMs
in simulating regional climate, while preserving the most important and defensible
signals that the models generate related to greenhouse forcing. Because GCMs do not
resolve the coastal mountains or smaller mountain ranges like the Cascades, east-west
temperature and precipitation gradients in the Pacific Northwest are not appropriately
resolved in the models, and attempts to use this data in its raw form will produce highly
erroneous results. North-south gradients are better resolved on the west coast of the U.S.,
because storm tracks in cool season are related to large-scale circulation that GCMs
simulate reasonably well (Salathé 2006) . The position of the dominant storm track,
however, can be strongly biased, and these biases can be different for different climate
models, which creates difficulties when attempting to interpret changes at relatively small
spatial scales (i.e. a particular river basin). Different climate models also show wide
variations in their ability to accurately reproduce the key features of regional climate, and
the time series behavior of different models also varies widely. Some models simulate a
reasonably accurate ENSO cycle, for example, whereas others simulate this important



                                             52
Draft, 4 June 2009                                   For Review Only, Do Not Cite or Distribute


driver relatively poorly. Some models may have too much interannual variability, others
too little, etc.

GCM simulations of precipitation are much more problematic than those for temperature,
and each GCM has a unique sequence of decadal scale precipitation variability that
makes comparisons between different GCM precipitation signals in a given future time
frame problematic. This is particularly true in the PNW, where systematic changes in
annual precipitation simulated by GCMs are relatively small, and decadal variability
remains an important driver of future impacts in any given decade (Mote and Salathé
2009).

Approach 1: Delta Method Experiments
In delta method experiments, regional scale (e.g. the entire Columbia basin), monthly
changes in T and P are extracted from a particular GCM simulation for some future time
horizon. A common approach is to regionally average the GCM simulations over 30-year
time horizons. In the case of the WA Climate Change Impacts Assessment (WACCIA),
for example, GCM data for 30-year windows centered around 1985, 2025, 2045, and
2085 were analyzed to estimate changes in T and P for future time periods (Mote and
Salathé 2009). These monthly changes are then applied to daily historic time series data
used as inputs to hydrologic models. So, for example, if the analysis of a particular GCM
shows that Januarys in the 30-year window centered on 2045 are 2 degrees C warmer
than those in the 30-year window centered on 1985, then all the Januaries in the historic
daily data are made 2 degrees C warmer in January over the entire domain.

Another application of the approach is to average the results from several climate models
together to produce what is sometimes called a ―composite‖ delta experiment. In this
approach the contribution from any one GCM is minimized, and the consensus is
emphasized. These approaches are most valuable when the central tendency of the
results is most important.

Although very simple, the delta approach remains very useful for sensitivity analyses and
water planning studies (Hamlet and Lettenmaier 1999; Snover et al. 2003). The
approach combines regional-scale changes from the GCM (thus avoiding most of the
spatial problems discussed above) with realistic spatial and temporal patterns of
variability from meteorological station observations. Delta method experiments also
provide many different realizations of natural variability with a constant change in
climate applied. This framework is well-suited to water resources planning studies,
whose primary objective is to ensure robust behavior of a water resources system over a
wide range of conditions. Delta method experiments also allow water managers and
planners to use their understanding of the historic record to inform the understanding of
climate change impacts (Snover et al. 2003). For example, critical drought sequences of
the past can be examined in the context of a warmer climate with reduced snowpack,
altered precipitation, etc.

Limitations of the delta method are primarily related to the fact that some kinds of
potentially useful information from the GCMs are ignored. Because only the monthly



                                           53
Draft, 4 June 2009                                    For Review Only, Do Not Cite or Distribute


mean climate is perturbed, other changes in variability (such as changes in sequencing,
variance, or extremes) are ignored. Potentially unique changes in different parts of the
region are also ignored.

Approach 2: Bias Correction and Statistical Downscaling (BCSD)
In this approach, more spatial and temporal information is extracted from the GCM
simulations in an attempt to overcome some of the limitations of the delta method
described above. As noted above, most GCMs display substantial bias in T and P, even
at a relatively coarse grid scale. Wood et al. (2002, 2004) avoided this difficulty by
statistically removing the monthly bias from the large scale GCM simulations using non-
parametric quantile-mapping techniques. In this procedure, the monthly quantile (e.g. the
99th percentile) in a GCM simulation of T or P is mapped to the equivalent monthly
quantile in a corresponding observed data set aggregated to the same spatial scale. Thus,
by construction, the cumulative distribution function (CDF) of the observations is
reproduced for the historic training period. This relationship between the GCM
simulations and the historic record is then assumed to remain constant for the future
scenarios (i.e. the same mapping is used to remove the bias in the future periods).

The bias corrected monthly GCM simulations are then statistically downscaled, first
spatially (by interpolating monthly anomalies to the finer scale grid) and then temporally
(monthly to daily) by randomly resampling from observed sequences from the historic
record (Wood et al. 2002; Wood et al. 2004). Salathé (2004) refined these methods by
incorporating analogue techniques in the temporal resampling procedure that account for
unique patterns in the GCM pressure or wind field. It is important to note that no daily
time scale information is incorporated directly from the GCM simulations, and these
characteristics of the downscaled products derive entirely from the temporal resampling
procedure.

The primary strength of the method is that potentially altered statistics simulated by the
GCM are incorporated in the downscaled product. The effects of altered monthly mean,
variance, and sequencing are explicitly accounted for in the downscaling procedure, and
these may differ substantially from the historic period. Such effects are of considerable
interest to planners. The method also produces a transient simulation, which may be
useful for trend analysis, exploration of decadal-scale variability, etc. Any future time
period may also be examined without making additional runs, which adds flexibility to
the products produced.

Ironically, the primary strength of the method is also its fundamental weakness.
Incorporating more information from the GCM is valuable if this information is of high
quality, and is a liability if the GCM simulations are of poor quality. Thus the approach
requires careful screening of the GCMs used to ensure that they closely reproduce
important statistics (such as sequencing and variability) over the area of interest. GCMs
that perform relatively poorly in reproducing historic climate should generally be avoided
when using this approach.




                                            54
Draft, 4 June 2009                                      For Review Only, Do Not Cite or Distribute


Most of the studies using this downscaling approach have been carried out at monthly
time scales in relatively large river basins (e.g. Payne et al. 2004), and for these
applications the approach has been shown to be reasonably robust. For studies at smaller
temporal and spatial scales, significant artifacts of the downscaling begin to appear, and
the results are less useful, particularly for GCMs that have marginal performance in
simulating the current climate in the region of interest. This downscaling method is
probably not a good choice for examining changes in hydrologic extremes such as Q100,
or 7Q10, because it inherits the many fundamental limitations of the GCMs affecting
these statistics, and the downscaled daily time history may not be very representative of
actual conditions.

Approach 3: Hybrid Bias Correction/Delta Method Downscaling
This approach begins with the bias correction and spatial downscaling procedure
described above for the BCSD method, but then takes a different approach for final
temporal/spatial downscaling. The approach uses the quantile mapping technique
(described above in Approach 2) to transform a month in the observed time series (at a
particular grid location) so that it reproduces the new statistics associated with a future
time period from the bias adjusted GCM data. To see how this works, consider a
particular January from the observed period for a particular grid location. If this January
was the 88th percentile in the observed period, this would be mapped to the 88th percentile
of the bias-corrected GCM months extracted from the future time period of interest (e.g. a
30-year window centered on the 2040s). The historic daily values within the month are
then rescaled so that the historic daily time series reproduces the future GCM monthly
value. This procedure is carried out over all grid cells for the entire observed time
history. The result is a gridded time series product that closely matches the daily time
series behavior and spatial extent of storms from the historic observations, but includes
potentially complex changes in the monthly probability distributions from the GCMs.
Thus more information from the GCMs is extracted (as for BCSD), but the realistic daily
sequencing and spatial variability from the observations is largely preserved (as for the
Delta Method).

Strengths of this method are that some of the best features of BCSD and Delta Method
approaches are combined in a single approach, while some of the primary limitations of
both techniques are avoided. This method is probably the most defensible of the
statistical downscaling methods for examining hydrologic extremes such as Q100 and
7Q10, because of the realistic daily time step realizations at the river basin scale. This
method is also probably preferred for examining impacts at relatively small spatial and
temporal scales (e.g. smaller basins at daily or weekly time scales), where artifacts of the
BCSD resampling method create the greatest difficulties. The method can be combined
with techniques from stochastic hydrology (or paleostreamflow reconstructions), to create
new (or more) time series realizations incorporating the future changes in climate.

The primary weakness of the method is that some aspects of potentially altered temporal
and spatial variability simulated by the GCMs are ignored. For example, the interarrival
time, spatial extent, and length of droughts will be essentially equal to that in the historic
record in the downscaled product, even though the intensity of droughts may change.



                                              55
Draft, 4 June 2009                                     For Review Only, Do Not Cite or Distribute




Approach 4: Dynamical Downscaling
In this approach a meso-scale climate model is driven at the outer boundary by GCM
simulations, and the internal climate dynamics are then simulated from first principals at
a higher spatial resolution (e.g. at 36km) (Salathé et al. 2009). Such approaches
explicitly resolve mountain ranges such as the Cascades, and are able to capture some the
unique climatic features of areas defined by complex terrain in the PNW (e.g. eastern vs.
western WA).

Strengths of this method include the ability to explicitly capture dynamic feedbacks
between different climatic effects. For example the differing rates of snowpack loss
across the PNW associated with regional warming creates unique changes in surface
temperatures in different areas of the domain because of positive feedback mechanisms
associated with the snow-albedo affect (Salathé et al. 2009). Such effects cannot be
captured by GCMs since the mountain snowpack in the Cascades is not simulated.
Similarly the models have the potential to resolve effects related to short-time-scale
precipitation and temperature extremes at relatively small spatial scales that statistically
downscaled GCMs cannot simulate.

A weakness of the approach is that meso-scale models inherit bias from the GCMs via
the large scale forcing at the outer boundary and then add some of their own. Thus bias
correction and downscaling is generally still required in most applications (Wood et al.
2004). Meso-scale climate models are computationally intensive to run, and only a few
realizations are available (e.g. two in the case of the WACCIA—Salathé et al. 2009). It
is also true that these models have only been exercised on a very limited basis to date in
comparison with GCMs, and it is not clear what potential problems will emerge when
these kinds of tools are applied more widely in applications research.

Implications for the Columbia Basin (“2860”) Climate Change Scenarios Project
A key objective of this project is to provide a comprehensive set of hydrologic products
using a common methodology at a finer spatial and temporal scale than has been
previously available in the PNW. In our survey of the user community, most of the
respondents cited daily time step simulations as the most useful product. Thus, in this
study, daily time step realizations in relatively small basins must be well supported by the
core study products. In addition one of the important products for the study is an
estimate of changes in Q100 and 7Q10 for each basin, which further highlights the need
for relatively accurate simulations of short-time-scale hydrologic variability.

As noted above, the BCSD approach is a useful one when applied at relatively large
spatial scales at monthly time steps, but requires careful screening of the GCMs to avoid
undesirable artifacts, and does not work well in small basins for daily time scales. As a
result we will probably produce these products for a relatively small number (~5) of
GCMs that perform the best over the historic record (Mote et al. 2009). This will also
allow considerable flexibility for those who want to examine different time horizons in
large-scale basin-wide studies.




                                             56
Draft, 4 June 2009                                   For Review Only, Do Not Cite or Distribute


The hybrid BC/Delta approach is arguably the best for use in relatively small basins or
where daily flow statistics are required. For the exploration of changes in hydrologic
extremes such as Q100 and 7Q10, this approach will probably produce the most
defensible baselines while incorporating considerably more information from the GCMs
in comparison with traditional delta approaches. The approach is also not very sensitive
to the specific limitations of particular GCMs in simulating time series behavior, which
permits a larger group of models to be used to examine the uncertainties in the climate
projections. Based on these desirable characteristics, I’m proposing this approach should
used to downscale the ten best climate models to form the core of the hydrologic products
for this project.

Composite delta method simulations may also be useful for those wishing to examine the
implications of the consensus from all the climate models in a single run. These can be
captured in six model runs (3 future time periods times two emissions scenarios), and are
probably well worth archiving as part of the study.




                                           57
Draft, 4 June 2009                                   For Review Only, Do Not Cite or Distribute


References
Hamlet, A.F., Lettenmaier, D.P., 1999: Effects of Climate Change on Hydrology and
Water Resources in the Columbia River Basin, J. of the American Water Resources
Association, 35 (6): 1597-1623

Mote, P.W., E.P. Salathe, 2009: Future climate in the Pacific Northwest, WACCIA (in
review)
http://cses.washington.edu/cig/res/ia/waccia.shtml

Payne, J.T., A.W. Wood, A.F. Hamlet, R.N. Palmer, and D.P. Lettenmaier, 2004:
Mitigating the effects of climate change on the water resources of the Columbia River
basin, Climatic Change, 62 (1-3): 233-256

Salathé, E.P., 2004: Methods for selecting and downscaling simulations of future global
climate with application to hydrologic modeling, International J. of Climatology, 25: 419-
436

Salathé, E.P. 2006. Influences of a shift in North Pacific storm tracks on western North
American precipitation under global warming. Geophysical Research Letters 33, L19820,
doi:10.1029/2006GL026882, 2006.

Salathé, E.P., P.W. Mote, and M.W. Wiley. 2007. Review of scenario selection and
downscaling methods for the assessment of climate change impacts on hydrology in the
United States Pacific Northwest. International Journal of Climatology 27(12): 1611-
1621, DOI: 10.1002/joc.1540.

Salathé, E.P., L.R. Leung, Y.Qian, and Y. Zhang, 2009: Regional climate model
projections for the State of Washington ,WACCIA (in review)
http://cses.washington.edu/cig/res/ia/waccia.shtml

Snover, A.K., Hamlet, A.F., Lettenmaier, D.P., 2003: Climate Change Scenarios for
Water Planning Studies, Bulletin of the Amer. Meteorological Soc., 84 (11): 1513-1518

Wood A.W., Maurer E.P., Kumar A. and Lettenmaier, D.P., 2002: Long range
experimental hydrologic forecasting for the eastern U.S. J. Geophys. Res., 107 (D20):
4429

Wood, A.W., Leung, L.R., Sridhar, V. and Lettenmaier, D.P., 2004: Hydrologic
implications of dynamical and statistical approaches to downscaling climate model
outputs, Climatic Change, 62 (1-3): 189-216




                                           58

				
DOCUMENT INFO