Transcript JPL

Making satellite data available for global and
regional climate model evaluation
June 2010
Peter Lean, Duane Waliser, Jinwon Kim, Joao Teixeira
Jet Propulsion Laboratory
California Institute of Technology
Copyright 2010 California Institute of Technology. Government Sponsorship Acknowledged.
Jet Propulsion Laboratory
California Institute of Technology
Motivation
• Evaluation of strengths/weaknesses of climate models
is essential to understand confidence in model
projections and impact assessments
• Satellites provide a wealth of data that can be used,
along with reanalysis and other observations, to
evaluate model performance
• Challenge to make remote sensing datasets more
accessible to the modeling community
Jet Propulsion Laboratory
California Institute of Technology
Observations in the IPCC process
Observations used to examine
trends and impacts & evaluate and
improve regional and global models.
Perform global model projections
Downscale
resultschanges
to regional
and characterize
scales and end-user quantities
Observations
used to
score/weight
model
projections.
Impacts on decision-makers,
public opinion/action & policy.
Jet Propulsion Laboratory
California Institute of Technology
What observations are available?
Satellite:
• AIRS (profiles of temperature, moisture)
• TRMM (tropical precipitation)
• QuickScat (surface winds over ocean)
• CloudSat (cloud water/ice)
• AMSU
+ many more
= lots of data
+Reanalysis
+In-situ observations
Jet Propulsion Laboratory
California Institute of Technology
Talk outline
Efforts at JPL/NASA to provide remote sensing
datasets to the modeling community:
• Providing observations for CMIP5
• A new tool for regional climate model evaluation
Theme: making satellite datasets more readily
available for use by modelers
Jet Propulsion Laboratory
California Institute of Technology
Part I:
Observations in CMIP5
JPL/NASA and PCMDI Collaboration
J. Teixeira, D. Waliser, D. Crichton, A. Braverman, S. Boland
Jet Propulsion Laboratory, California Institute of Technology,
Pasadena, CA
D. Williams, P. Gleckler, K. Taylor
Program on Climate Modeling Diagnostics and Intercomparison
(PCMDI), Livermore, CA
J. Potter
University of California, Davis, CA, and Goddard Space Flight Center,
Greenbelt, MD
Jet Propulsion Laboratory
California Institute of Technology
Observations for CMIP5 Simulations
MOTIVATION
How to bring as much observational
scrutiny as possible to the IPCC
process?
How to best utilize the wealth of NASA
Earth science information for the
IPCC process?
Jet Propulsion Laboratory
7
California Institute of Technology
Background: CMIP5 observations
•
Taylor et al (2008) have defined the protocol for the CMIP5
simulations that will be used for the next IPCC Assessment Report,
AR5.
•
The protocol defines the scope of simulations that will be undertaken
by the participating modeling groups.
•
For several of the prescribed retrospective simulations (e.g, decadal
hindcasts, AMIP and 20th Century coupled simulations) observational
data sets can be used to evaluate and diagnose the simulation
outputs.
•
However, to date, the pertinent observational data sets to perform
these particular evaluations have not been optimally identified and
coordinated to readily enable their use in the context of CMIP5.
Jet Propulsion Laboratory
8
California Institute of Technology
Objective: CMIP5 observations
To Provide the community of researchers that will access and evaluate the CMIP5
model results access to analogous sets of observational data.
• Analogous sets in terms of periods, variables, temporal/spatial frequency
•This activity will be carried out in close coordination with the corresponding
CMIP5 modeling entities and activities.
•It will directly engage the observational (e.g. mission and instrument) science
teams to facilitate production of the corresponding data sets.
Jet Propulsion Laboratory
9
California Institute of Technology
Observations for CMIP5 Simulations
NASA/JPL, PCMDI and ESG
• JPL and PCMDI have established a collaboration through the ESG to
share observations to support model-to-data comparison
– A prototype ESG node was established at JPL in 2009 demonstrating sharing
of AIRS data between JPL and the ESG
– Next, JPL and PCMDI plan to deploy a NASA/JPL “gateway” for access to
multiple NASA observational data sets in June 2010, pending NASA support.
• Provide access to a wealth of NASA observations through the ESG
– AIRS will be the first planned test data set that will be operationally available
when the gateway is released in June 2010
– PCMDI and JPL are planning enhancements to the ESG portal to improve
access to observations in conjunction with the models
– Additional observational data sets will be subsequently added through March
2011, pending NASA support.
Jet Propulsion Laboratory
California Institute of Technology
The Next-generation ESG with Observations
Jet Propulsion Laboratory
California Institute of Technology
Summary: CMIP5 observations
•
A collaborative effort between JPL/NASA and PCMDI is underway to
provide the community of researchers that will access and evaluate the
CMIP5 model results access to analogous sets of observational data.
•
A number of NASA satellite data sets have been identified that have model
equivalents. Thus far: AIRS, MLS, TES, QuikSCAT, CloudSat,
Topex/Poseidon, CERES, TRMM, AMSR-E.
•
Plans have been developed for converting the data into CF-compliant
format, documenting it for technical details for their use/application to IPCC
model assessment, and to make them available via ESG and links from
PCMDI model access web portal.
•
This activity is being carried out in coordination with the corresponding
CMIP5 modeling entities and activities (e.g. WGCM, PCMDI).
Jet Propulsion Laboratory
12
California Institute of Technology
Part II:
A new end-to-end tool for regional
climate model evaluation
Peter Lean, Duane Waliser Jinwon Kim, Dan Crichton, Chris Mattmann,
Tim Stough, Cameron Goodale, Andrew Hart
Jet Propulsion Lab / Caltech
Alex Hall
UCLA
Jet Propulsion Laboratory
California Institute of Technology
Motivation
• Typical regional climate model study
– spend lots of time downloading GCM forcing data
– spend lots of time running RCM downscaling
– not much time left available for performing model evaluation
– large effort required to download and use satellite datasets so
often not done.
• We want to make evaluating models using satellite data simpler and
faster to perform so that more modelers have time to do it
Jet Propulsion Laboratory
California Institute of Technology
A new regional climate model evaluation tool
• Goal:
• make the evaluation process for regional climate models simpler and
quicker
• Things that used to take weeks should take days.
• allow researchers to spend more time analysing the results and less
time worrying about file formats, data transfers and coding.
• Benefits:
• Improved understanding of model strengths/weaknesses allows
model developers to improve the models
• Improved understanding of uncertainties in predictions of specific
variables over specific regions for end-users
Jet Propulsion Laboratory
California Institute of Technology
System Features
•
Multiple observational datasets:
• Database will store wide variety of satellite and blended data for comparison with model
simulations
• “Add once, use many” philosophy
•
Evaluation metrics:
• Comprehensive model evaluation using well established statistical measures
• Using model variables directly related with regional climate including atmosphere, land, SWE,
ocean, and air quality
•
Portable:
• Location: system easily reconfigured to allow relocation to any location
• End user: different quantities can be evaluated dependent on end user
• Time periods
•
Expandable/Flexible:
• Designed to make it easy to add new datasets and metrics in future
Jet Propulsion Laboratory
California Institute of Technology
How this work relates to CMIP5 work
• CMIP5 work concentrating on sharing satellite data for model intercomparison
• Regional climate model work is providing an end-to-end tool for
model evaluation
– including calculation of statistical metrics and plotting
• More flexibility as less constraints imposed
• Leveraging experience and work already completed as part of
CMIP5 preparations
Jet Propulsion Laboratory
California Institute of Technology
System architecture:
ingest data files
Jet Propulsion Laboratory
California Institute of Technology
What observations are available?
Satellite:
• AIRS (profiles of temperature, moisture)
• TRMM (tropical precipitation)
• QuickScat (surface winds over ocean)
• CloudSat (cloud water/ice)
• AMSU
• IASI
+ many more
= lots of data
+Reanalysis
+In-situ observations
Jet Propulsion Laboratory
California Institute of Technology
Proposed Metrics
•
Bias
•
Mean Absolute Error
•
Root Mean Square Error
•
Pattern Correlation
•
Anomaly Correlation
•
Coefficient of Determination
•
Coefficient of Efficiency
•
PDF similarity score
Jet Propulsion Laboratory
California Institute of Technology
Evaluation of full distribution for each data point
Jet Propulsion Laboratory
California Institute of Technology
User experience mock-up:
Select Observation Dataset:
Select model data source:
AIRS level III gridded
ERA-Interim reanalysis
NCEP reanalysis
TRMM
CloudSat
QuickScat
Select Variable:
Surface Temperature
Specific humidity
Outgoing LW rad (TOA)
Cloud fraction (surface)
10m wind speed
Next >
Jet Propulsion Laboratory
California Institute of Technology
User experience mock-up:
Select Date Range:
Select output grid:
Use model grid
Use observational grid
Other regular grid (specify)
Select output grid:
Map
Time series
Select granule size:
Daily data
Pentad data
Monthly data
Seasonal (e.g. DJF, JJA)
Annual
Decadal
Select statistical metrics:
RMS error
Mean error
Mean absolute error
Anomaly correlation
PDF similarity score
Coefficient of efficiency
Pattern Correlation
Process
Jet Propulsion Laboratory
California Institute of Technology
User experience mock-up:
• Several minutes later…
[purely illustrative]
Jet Propulsion Laboratory
California Institute of Technology
Initial Application
• Joint Institute For Regional Earth
System Science and Engineering
regional model
– WRF coupled with ROMS over California
– Hindcasts already completed (Alex Hall,
UCLA)
IPCC AR4 Projections
• Focus on water resource end users
– Working with California Department of
Water Resources
– Sierra Nevada snowpack:
• Using new Snow Water Equivalent dataset
But how realistic?
Quantify Uncertainties.
Jet Propulsion Laboratory
California Institute of Technology
Progress and plans:
• Sample AIRS data file successfully ingested into
database.
• netCDF/GRIB interfaces under development
• Python processing code (in progress)
• Demonstration: September 2010
– End-to-end demonstration:
• Compare 1 dataset with 10 year model hindcast
e.g. AIRS level III water vapor v model run at UCLA
• Longer term aims:
– Use system to evaluate locally run RCM
IPCC AR5 regional downscaling evaluation
Jet Propulsion Laboratory
California Institute of Technology
Summary
• Remote sensing data provide a wealth of data for
evaluating regional climate models which are
currently under-exploited.
• JPL/NASA will provide satellite observations for
inter-comparison with CMIP5 model hindcasts
• A new tool to facilitate regional climate model
evaluation studies is being developed.
Jet Propulsion Laboratory
California Institute of Technology
Thanks for listening!
Any Questions?
Jet Propulsion Laboratory
California Institute of Technology
Progress to date
• System architecture design (complete)
• Server setup (complete)
• Dataset downloads (partially complete)
• netCDF/GRIB interface:
– sample AIRS level III data file ingested into database
• Python processing scripts:
–
–
–
–
Scripts to read model data (complete)
Statistical metric functions (partially complete)
Re-gridding code (in progress)
Display code (in progress)
Jet Propulsion Laboratory
California Institute of Technology