Climate scenariosApplications 102: Selection of models and practical applications

1
Climate scenarios/Applications 102 Selection of
models and practical applications
Levi Brekke (Reclamation, Technical Service
Center)
OCCRI Workshop on Scenarios of Future Climate,
Portland, OR, October 28-29, 2009
2
Preview

• Example Reclamation CVP OCAP 2008
• Goal represent climate change uncertainty in an
  ESA consultation
• Example Extensions
• SJRRP 2009 add central CC scenario
• RMJOC 2010 leveraging UW CIG HB 2860 data
• Moving beyond Small Scenario Sets

3
Planning Example Reclamation CVP OCAP 2008
4
Reclamation CVP OCAP 2008

• Project Team
• Reclamation TSC
• Reclamation Mid-Pacific Region
• CA DWR
• Report available at
• http//www.usbr.gov/mp/cvo/ocap_page.html

5
Central Valley Project (Reclamation, Mid-Pacific
Region)
State Water Project (CA Department of Water
Resources (DWR))
6
CVP OCAP 2008Context

• ESA consultation on several listed/threatened
  species
• Delta smelt, multiple salmon fisheries
• Process
• Water Agencies provide Fisheries Agencies a
  Biological Assessment (BA) on the effects of
  long-term CVP operations
• through 2030
• Water and Fisheries Agencies consult
• Fisheries Agencies issue Biological Opinions
• Key Issues
• Geographically overlapping water systems
  (fed/state/local)
• Upstream vs. downstream fisheries management
• Sensitivity of BA to future assumptions on
  climate and sea level?

7
Framework for using Climate Change Information in
Planning
1) Survey Global Climate Projections that have
been spatially downscaled for the study region
3) Define supplies, demands, and/or operating
constraints in terms of climate info from 2.b.
Natural Systems Response
Social Systems Response
2.a) Decide whether/how to cull the information.
4) Assess operations and dependent resource
responses characterize uncertainties
2.b) Decide how retained information will be used.
8
parts used in OCAP BA
1) Survey Global Climate Projections that have
been spatially downscaled for the study region
3) Define supplies in terms of climate info from
2.b (i.e. inflows, supply forecasts, year-type
classifications)
Natural RUNOFF Response
2.a) Decide whether/how to cull the information.
4) Assess operations and dependent resource
responses characterize uncertainties
2.b) Select bracketing Climate Changes
9
1) Information Survey

• Global Climate Projections at PCMDI
• World Climate Research Programmes Coupled Model
  Intercomparison Project, phase 3 (WCRP CMIP3)
• http//www-pcmdi.llnl.gov/ipcc/about_ipcc.php
• 100 projections from multiple models, emission
  paths, initial conditions (runs), served IPCC
  (2007)
• Regional Climate Projections
• Prioritize data resource with many downscaled
  projections
• Statistically Downscaled WCRP CMIP3 Projections
• http//gdo-dcp.ucllnl.org/downscaled_cmip3_project
  ions/
• 112 projections, 12km res, monthly, 1950-2099,
  contiguous U.S.
• Sea Level Rise (SLR)
• IPCC (2007) CA CAT 2008 tools from CA DWR

10
2.a) Do we cull the Info? How?

• Focusing on regional climate projections
• Should we regard all available downscaled
  projections as equally plausible, or
• Should we regard some of these projections as
  more credible, and focus on them?
• If the latter, how would we rate credibility?

11
Guidance for culling is unclear and effect may
be minor anyway
Focusing on CA, Brekke et al. (2008) considered
historical simulations from 17 GCMs, and found
similar skill when enough metrics were
considered. Focusing globally, Gleckler et al.
2008 and Reichler et al. 2008 found similar
results.
Focusing on CA, projection distributions didnt
change much when the GCM-skill assessment (Brekke
et al. 2008) was used to reduce the set of 17
GCMs to a better set of 9 GCMs.
Santer et al. PNAS 2009 results from a global
water vapor detection and attribution (DA) study
were largely insensitive to skill-based model
weighting. Pierce et al. PNAS 2009 results
from western U.S. DA study were more sensitive
to ensemble size than skill-based model weighting.
12
2.b) How do we use the retained climate
information?

• Options
• Reflect change in climate norms, use bracketing
  scenarios
• Reflect evolving climate projection, use many
  scenarios
• (e.g., Christensen Lettenmaier 2007, Maurer
  2007, CASCaDE)
• Approach depends on study purpose and goals
• E.g., focus on change in climate norms is useful
  for sensitivity analysis we retain our
  historical hydrologic variability, but scale it
  to show system response to change in climate
  norms
• E.g., focus on many evolving projections (i.e. a
  projection ensemble is useful for adaptation
  planning relative to a time-developing climate
  we replace historical hydrologic sequence with
  projection-derived sequence

13
Motives for Bracketing ApproachTwo Perceptions,
fair or not

• Keep it simple.
• Post-Step 4) Communicating Results
• Audience used to reviewing operations scenarios
  told in the sequence of the historical gage.
• Bracketing approach w/ period-change allows
  retention of this sequence understanding (e.g.,
  What happens in the drought of 1987-92 if climate
  changes?)
• Keep it manageable.
• Steps 3) and 4) Scenario-Handling capacity
• Each scenario feeds several analytical areas
  hydrology, Operations, Delta hydrodynamics, Water
  Temperatures,
• Considering pre/post processing simulation,
  capacity-limit is determined by most intensive
  step
• Large scenario-sets may require different
  procedures/tools

14
Selecting the Bracketing ScenariosMP OCAP
Choices

• Climate Periods 1971-2000, 2011-2040
• consultation horizon is through 2030
• Climate Metrics Period Mean-Annual Tair P
• Focus here is on assessing projections spread,
  key metrics
• Decision mean-annual, heavy influence on CVP
  performance
• Location Above Folsom
• Focus here is on assessing projections spread at
  relevant location
• Interested in hydrologic changes in all Sierra
  Nevada tributaries
• Decision Focused on central, representative
  tributary
• Change Range 10 to 90 -tile DTair, DP
• Range depends on risk attitude, regard for edge
  of spread
• Decision cast a wide net, represent breadth of
  changes

15
Implementation of Selection Factors
Step 1) Survey downscaled projections at the
Above Folsom location.
1a. From website, download 112 projected monthly
Tair P time series. 1b. Compute historical and
future period climate metrics for every
projection. 1c. Compute historical-to-future
period changes in climate metric for every
projection.
http//gdo-dcp.ucllnl.org/downscaled_cmip3_project
ions/
16
(No Transcript)
17
Step 3) Scatter plot DTair and DP, all
projections overlay -tile thresholds
18
Result chosen projections express changes that
span spread of changes
19
Extensions of CVP OCAP 2008 Scenario-Selection
Method
20
Related Studies

• Same approach, but with central selection added
• Reclamation SJRRP 2009 (draft)
• SJRRP San Joaquin River Restoration Program
  PEIS
• Front Range Climate Change Vulnerability Study
  2009 (draft)
• CWCB 2009 (draft) coordinated with study above
• Same approach, but different kind of scenario
• Change in period monthly variability rather than
  change in period monthly mean (UW CIG calls this
  Hybrid in their HB 2860 work)
• TX LCRA-SAWS 2008
• UW CIG 2009 HB 2860 (Hybrid Scenarios)
• RMJOC 2010
• Will inherit UW CIG Hybrid Scenarios
• Scoped to handle small-scenario set follow OCAP
  method?

21
SJRRP 2009 vs. CVP OCAP 2008Selected 5
projections rather than 4 based on spread Above
Millerton rather than Above Folsom
22
Related Studies

• Same approach, but with central selection added
• Reclamation SJRRP 2009 (draft)
• SJRRP San Joaquin River Restoration Program
  PEIS
• Front Range Climate Change Vulnerability Study
  2009 (draft)
• CWCB 2009 (draft) coordinated with study above
• Same approach, but different kind of scenario
• Change in period monthly variability rather than
  change in period monthly mean (UW CIG calls this
  Hybrid in their HB 2860 work)
• TX LCRA-SAWS 2008
• UW CIG 2009 HB 2860
• RMJOC 2010
• Will inherit UW CIG Hybrid Scenarios
• Scoped to handle small-scenario set follow OCAP
  method?

23
Texas Study - Hybrid Example

• Report
• Prepared by CH2M-Hill
• http//www.lcra.org/lswp/about/study/climatechange
  .html
• Section 7.5 describes application of Hybrid
  Methodology

24
Texas Study - Hybrid Example
1. Base climate variability 1950-1999 monthly
observed distributions, each downscaled grid
cell
25
Texas Study - Hybrid Example
2. Future climate variability 2066-2095
monthly distributions, each grid cell, from
chosen projection
26
Texas Study - Hybrid Example
3. Assess change in monthly distributions at
each quantile quantile map
27
Hybrid Downscaling Method

• Performed for each VIC grid cell

Bias Corrected Future Monthly CDF
Hist. Daily Timeseries
30 yr window
1916-2006
Projected Daily Timeseries
Historic Monthly CDF
Hist. Monthly Timeseries
1916-2006
1970-1999
1916-2006
Base Case
UW CIG HB 2860 effort slide from A.
Hamlet, 10/16/09
28
Related Studies

• Same approach, but with central selection added
• Reclamation SJRRP 2009 (draft)
• SJRRP San Joaquin River Restoration Program
  PEIS
• Front Range Climate Change Vulnerability Study
  2009 (draft)
• CWCB 2009 (draft) coordinated with study above
• Same approach, but different kind of scenario
• Change in period monthly variability rather than
  change in period monthly mean (UW CIG calls this
  Hybrid in their HB 2860 work)
• TX LCRA-SAWS 2008
• UW CIG 2009 HB 2860 (Hybrid Scenarios)
• RMJOC 2010
• Will make use of UW CIGs Hybrid and Transient
  Scenarios
• Scoped to handle small-scenario set follow OCAP
  method?

29
About the RMJOC Effort

• Motive
• consistent incorporation of climate projection
  information into RMJOC agencies longer-term
  planning studies
• BPA, USACE Northwestern Division, Reclamation PN
• Needs
• adopt common dataset (climate and hydrology)
• establish consensus methods for data use
• demonstrate use with RMJOC agencies planning
  models
• efficiently use limited resources through
  coordinated development, collaboration with
  interested stakeholders
• (e.g., CRITFC, NPCC, NOAA Fisheries, USFWS, USGS)
• Work Plan
• Scoped in 2009, implementation began Oct 2009.

30
RMJOC Scenarios Selection

• Preliminary Approach
• Step 1) Start with UW CIG HB 2860 climate
  scenarios
• Use both Hybrid and Transient from CIG
• Step 2a) No culling of UW CIG information
• Theyve already culled models (Mote and Salathe,
  2009)
• Step 2b) Small-scenario set, mainly for Keep it
  Simple reason
• Pilot effort will serve planning for next few
  years.
• Follow Reclamation SJRRP 2009
• Select UW CIG Hybrid Scenarios first
• Represent spread of CIG scenarios, 2020s and
  2040s
• Represent central tendency of each
  period-ensemble
• Select UW CIG Transient Projections by
  association
• Want underlying Global Climate Projections to be
  the same

31
Applying Reclamation 2009 to RMJOC Selection
Factors?

• Climate Periods Already determined by CIG
• 1971-2000, 2010-2039 and 2030-2059
• Climate Metrics ?
• Oct 16 stakeholder meeting, reviewed options
• Oct 24 distributed tool to interested parties,
  inviting feedback
• Location ?
• Same status as Factor (2.)
• Metric Change Thresholds of Interest ?
• For bracketing scenarios, same status as Factor
  (2.)

32
Moving beyond Limited-Scenario Set Applications
33
Limited Scenario Sets and Period-Change Approach

• Pros
• Easy way to explore system response
• Retains familiar historical variability patterns
• Cons
• Less ideal for adaptation planning, where climate
  change timing matters
• Diagnosing period Climate Change is not obvious
  (more of a problem for DP than for DT)
• Beware of the various CMIP3 initial conditions
• Cant ignore multi-decadal variability

34
Period-Change Application (1) metrics mean T
and P, (2) bracket spread, (3) relate to impacts
Vancouver
Data from http//gdo-dcp.ucllnl.org/downscaled_cm
ip3_projections/
35
Transient vs. Period-Change Applications Do
they tell the same impact story?
Transient Analysis
Period-Change Analysis
Invites envelop tracking (1) change in mean, and
(2) change in variability.
Focus is only on change in mean, and suggests
large uncertainty consistent with Transient view?
36
Use of hydrologic ensembles is becoming more
common practice
Lake Mead End-of-December storage under the
No-Action Alternative 90th, 50th, and 10th
percentile values (Reclamation 2007, Figure
4.2-2)
37
Projection Ensembles Approach

• Pros
• Avoids challenges of Climate Change diagnosis
• Supports master planning for CC adaptation
• schedule of interventions to maintain system
  reliability
• indication when implementation should begin for a
  given intervention to ensure that its online when
  needed
• Cons
• Information is more complex
• Multiple hydrologic sequences, different from
  historical
• Transient hydrology, not stationary
• May require learning phase for planning parties

38
Questions?

• Levi Brekke
• Reclamation, Technical Service Center
• lbrekke_at_do.usbr.gov

39
Extras

• Reclamation CVP OCAP 2008 Steps 3) 4)
  Scenario-Specific Analysis

40
3) 4) Scenario-Specific Analysis

• Sea Level Rise Scenario
• 1 foot rise with 10 increase in tidal energy
• consistent with IPCC 2007 and CA CAT 2008
  information
• rise increment constrained by available Delta
  model tools
• CVP/SWP Operations Studies (CalSim II)
• 9.0 historical regional climate, current sea
  level
• 9.1 historical regional climate, sea level rise
• 9.2-9.5 four selected future climates sea
  level rise

41
3) 4) Scenario-Specific Analysis Methods
Tools
42
3) 4) Scenario-Specific Analysis - Starting
Framework
I. Choose Climate Context
Instrumental Records observed weather (T and P)
and runoff (Q)
II. Relate to Planning Assumptions
Demand Variability
Operating Constraints
Supply Variability
III. Conduct Planning Evaluations
System Analysis, Evaluate Study
Questions (related to Resource Management
Objectives)
Adapted from USGS Circular 1331 (Brekke et al.
2009)
43
CVP OCAP 2008 Scenario-Specific AnalysisAdd
Regional Climate Projections
I. Choose Climate Context
Instrumental Records observed weather (T and P)
and runoff (Q)
Global Climate Projections Representing
various GCMs, forcings ? bias-correction,
spatial downscaling
watershed simulation
Regional T and P
II. Relate to Planning Assumptions
Runoff
Demand Variability
Operating Constraints
Supply Variability
III. Conduct Planning Evaluations
Future Operations Portrayal for OCAP BA (flows,
storage, deliveries, etc)
44
CVP OCAP 2008 Scenario-Specific AnalysisAdd
Global Climate, Sea Level Projections
I. Choose Climate Context
Instrumental Records observed weather (T and P)
and runoff (Q)
Global Climate Projections Representing
various GCMs, forcings ? bias-correction,
spatial downscaling
Single Sea Level Rise Scenario -- one foot rise
with 10 increase in tidal energy (meant to
reflect 2030 possibility at CA SF Bay/Delta,
consistent with Rahmstorf 2007, CalFed ISB
2007) -- number of rise scenarios limited by
scenario-handling capacity (analysis
communication) -- which rise increment also
constrained by available Delta modeling tools
already developed by CA DWR
watershed simulation
Regional T and P
II. Relate to Planning Assumptions
Global T Sea Level Rise
Runoff
Demand Variability
Operating Constraints
Supply Variability
Delta Flow-Salinity Relationship
Constraint on Upstream Operations
III. Conduct Planning Evaluations
Future Operations Portrayal for OCAP BA (flows,
storage, deliveries, etc)
45
CVP OCAP 2008 Scenario-Specific AnalysisAssess
Operations, Water Temperatures
I. Choose Climate Context
Instrumental Records observed weather (T and P)
and runoff (Q)
Global Climate Projections Representing
various GCMs, forcings ? bias-correction,
spatial downscaling
watershed simulation
Regional T and P
II. Relate to Planning Assumptions
Global T Sea Level Rise
Runoff
Demand Variability
Operating Constraints
Supply Variability
Delta Flow-Salinity Relationship
Constraint on Upstream Operations
III. Conduct Planning Evaluations
Reservoir Operations
Regional T
Future Operations Portrayal for OCAP BA (flows,
storage, deliveries, etc)
Stream Water Temperature analyses