Levi Brekke (Reclamation, Technical Service Center)
Co-Investigators: J. Anderson (DWR), E. Maurer (Santa Clara University), and M. Dettinger (USGS, Scripps Institution of Oceanography)
28 February 2007CWEMF Annual Meeting
Pacific Grove, CA
Exploring the use of Risk Analysis to study the effects of climate change on CVP and SWP operations
Acknowledgements• Reclamation
– R&D Office, Tech Service Center, and Mid-Pacific Region
• DWR – Bay-Delta Office (Modeling Support), Flood Management
• USACE– Sacramento District, ERDC-CRREL
• Climate Research Groups– Scripps Institute of Oceanography (Mike Dettinger)– Santa Clara University (Edwin Maurer)– Lawrence Livermore National Laboratory – Program for
Coupled Model Diagnosis and Intercomparison (PCMDI)
Context
• Reclamation is exploring options in how to use future climate information in planning.
• This is research on potential methods.
The findings and conclusions of this presentation have not been formally disseminated by Reclamation and should not be construed to represent any agency determination or policy.
Outline
• Analysis Overview:– Choose scenarios and assess impacts (runoff, operations)– Assess climate projection uncertainty, scenario probabilities– Combine scenarios, impacts and probabilities into risk– Explore strategies to manage risk
• Questions Today1. How do climate projection distributions depend on
apparent climate model skill?2. How do relative climate scenario probabilities depend on
projected variable (e.g., temperature, precip., or both)?3. How does operations risk depend on the basis for deriving
relative probabilities for climate scenarios?
Question #1:How do climate projection
distributions depend on apparent climate model skill?
Methods• Premise: Quality of 20th Century Simulation indicates
credibility of 21st Century Projection• Approach:
– Survey climate simulations, 20th to 21st Century• 17 models, {20c3m +SRES A2 or B1}• annual mean T and P during base & 3 future periods
– Evaluate the models’ 20th Century simulation skill• Get simulated and reference climate variables relevant to Nor. CA• Compute statistical metrics on the monthly values, 1950-1999• Compute metric differences between models and reference• Translate differences into “distances” and then weights
– Construct 21st Century climate change pdfs• pdf(T), pdf(P), pdf(T,P) • with and without climate model weighting
Climate Model Weights: sensitivity to variables & metrics
Weighted: based on different basis variables and metrics
pdf (Temperature), 3 futures: sensitivity to model weights
Weighted: based on “All Variables and Metrics”
pdf (T,P), 1 future: unweighted & weighted
Question #2: How do relative climate
scenario probabilities depend on projected variable?
Methods• Consider 3 variable-specific pdfs, with/without weighting
– pdf(T), pdf(T | “all vars & metrics” climate model weight)– pdf(P), pdf(P | “all vars & metrics” climate model weight)– pdf(T,P), pdf(T,P | “all vars & metrics” climate model weight)
• Choose scenarios of interest, locate their projected climate change values within the pdfs– E.g., 75 used to fit the pdfs; 22 of those 75 scenarios were
assessed for impacts (discussed later); focus on the 22…
• Scenario probability = ?– ? point probability density in the pdf– ? integrated probability within the scenario’s neighborhood with
the pdf, after dividing the pdf accordingly
Relative Scenario Probabilities (1 future, 6 pdfs, 2 use methods)
Question #3:
How does operations risk depend on the basis for
deriving climate scenario probabilities?
Impacts Assessment Methods(similar to DWR 2006)
• Choose Climate Scenarios (22) and get GCM output– Downscaled and bias-corrected relative to observed variability
• Simulate Headwater Runoff for base and 2 futures– NWS CNRFC models, base period 1963-1992– futures consistent with projected climate (2011-40, 2041-70)
• Simulate Operations for base and 2 futures – Compute performance metrics on output, by scenario– Compute changes in future from base, by scenario
• Updated, Dec 2006
• Construct Distributions of Metric Changes (Impacts)– Resample the distributions proportionately to scenario
probabilities
Runoff Impact:CVP North, April-July Inflow
Operations Impact: CVP Delta Exports
Operations Impact: SWP Delta Exports
Operations Impact:Lake Shasta Carryover Storage
Questions Revisited
1. How do climate projection distributions depend on apparent climate model skill?
– Some effect on local aspects of distribution; – aggregately, not much effect
2. How do relative climate scenario probabilities depend on projected variable?
– Significantly, also on how the pdf is used to get probabilities
3. How does operations risk depend on the basis for deriving climate scenario probabilities?
– Some effect on local aspects of distribution; – aggregately, not much effect
Next Steps
• Documentation– Project Report expected Summer 2007– Brekke, L.D., M.D. Dettinger, E.P. Maurer, M. Anderson, 2006. “Significance
of Model Credibility in Projection Distributions for Regional Hydroclimatological Impacts of Climate Change”, submitted to Climatic Change, In Review
– Other articles planned…
• Additional Impacts and Risk Analyses – Delta WQ/Levels, Stream Temps, Power
• Risk Management Studies– Flood Control Rules– Conjunctive Use– Others?