Transcript Open_Climate_SC_2010_20101104x

“Community Climate Models:
Is a New Paradigm of Model
Development Possible?”
Richard B. Rood
Cell: 301-526-8572
2525 Space Research Building (North Campus)
[email protected]
http://aoss.engin.umich.edu/people/rbrood
November 17, 2010
Supercomputing 2010, New Orleans, LA
Opportunity
• Climate projections offer us historical
opportunity. We have, at least for some
part of our future, a credible vision of what
the future will be like. It is up to us
whether or not we take advantage of this
opportunity.
An essay, perhaps.
An Interesting Problem (motivator 1)
• A couple of years ago one of my students
brought me the following problem
– Client wants to build a set of desalination
plants in western Africa.
• Purpose: Irrigation of Sahara
• He has read that hurricanes are somehow affected by the Sahara.
• Can he get some value for this?
• Does he incur liability?
Outline
• Motivating Examples (2)
• How do we address problems of climate variability and
change?
• Comparing weather and climate
• Reframing, perhaps
• What stands in our way
• A final remark
• “I'm up on the tight wire, one side's ice and
one is fire …” Leon Russell
Tight wire 1
• I will be talking about “open.”
– Climate is a political issue, and there is a lot
of talk about the “openness” of the climate
community. There is, for example, discussion
about how to make IPCC assessments more
open.
– I am not, here, talking about the political
issues and a political flavor of inclusivity and
transparency.
Another interesting problem (motivator 2)
• I want to protect New Orleans from a
Category 5 hurricane, that occurs during a
100 year flood, in 2075.
– Some key things I need to know
•
•
•
•
•
What is sea level rise?
What is land subsidence?
How has the coast been engineered?
What do I need to protect?
How much money do I have?
Think about these motivating problems
• They make sense.
• For less than $200.00 an hour someone will answer them for you.
– What knowledge base will they use?
– What knowledge base “should” they use?
• There is stunning complexity
–
–
–
–
–
–
Built environment
Land-use policy
Environmental engineering projects
Wealth
Local environment
Climate change
• There exists a substantial knowledge base across many disciplines
Two paths to an answer
Path 1
• I call a national lab and say I have an interesting
problem. I develop a relationship with a scientist who
agrees it is interesting. The scientist takes the idea
forward and tries to convince colleagues that it is a
scientifically interesting problem. Management
reluctantly agrees that scientist can pursue the problem
as a side interest, if scientist can find some funding. But
no “special” experiments will be run.
– Sponsor has funding: How does lab accept funding from Dubai
Industries?
– Sponsor does not have funding: Write proposals
Two paths to an answer
Path 2
• Find a smart consultant who has a copy of
– Ecological Climatology, Gordon Bonan
• Use MAGICC/SCENGEN: It has been widely used and
will have a branded level of credibility.
– (http://www.cgd.ucar.edu/cas/wigley/magicc/ )
• Augment MAGICC/SCENGEN analysis with localized
information from Climate Wizard
– http://www.climatewizard.org/
Resolution
• Which solution path is more likely?
Consider the West African Desalinization Motivator
• The question makes sense?
– Regional scale alteration of the land-surface to address a
fundamental issue of a consuming population.
– Surely will have climate consequences outside of the immediate
region
• Global?
– Many similar scale energy projects and water engineering
projects are being discussed
– What is the climate impact?
• Positive and negative
• Risk and liability
Doesn’t this make sense?
• Access to a validated climate model with a
configurable land-surface to perform
climate impact assessments.
– Offers a path to direct evaluation of the
questions being posed.
• Access to expert interpretation and
translation.
Let’s think about weather for some moments
How often do we ask, I want to
change the surface of the Earth
or the composition of the
atmosphere to see how it would
change the weather forecast?
Access here mostly scientists and
software engineers.
What goes in here is
mostly settled.
From a prediction point of view, here is the product.
Same approach for climate prediction and
projection? Maybe …
Surface will change appreciably
over course of projection.
Composition will change.
We might decide to change the
surface or composition to impact
the climate?
Or need to know the climate
impact of major changes?
Access here is needed to answer
these questions. There may not
be a scientist gatekeeper.
What goes in here is not
settled.
From a prediction point of view, here is the product.
Uncertainty: Weather and Climate
• Weather
– Ability to quantify
uncertainty based on
many forecasts and
many instantiations of
observed state
– Uncertainties related
to known model –
observation
shortcomings
• Climate
– Limited ability for
quantitative analysis of
forecasts
– Modes of variability not all
known or modeled (atmocn-lnd-ice-life)
– Forcing is changing
– Dependent on human
activity and policy
– Processes not included
What is the meaning of a climate projection?
Scientists perception of useful.
Users perception of usability.
Users of Climate Projections
Model Projections
Scientific
Investigation
Applications
Some would like to think of a nice linear path like this.
Perhaps would like to think scientific investigation is more
or less done, and that we can hand this off to application’s
experts. Perhaps create a separation between research
and operations. But we have made this mistake, a
dichotomy between research and operations, many times
before, and we are not required to do it again.
Perhaps a more realistic flow
NM
nM, nS, nDC
“product”
Model Projections
“uncertainty”
Scientific
Investigation
NS
Tailoring 1
(e.g. downscaling)
NT1
“product”
“uncertainty”
Applications
NA
“product”
Tailoring 2  n
(e.g. downscaling)
NTn
“uncertainty”
The Point: Enormous
complexity. NA >> NM, and
NM is unmanageable.
Fundamental Flaws in this Analysis
• There is implied, here, a push of information from climate
models to applications.
– But this is naïve,
– And the numbers don’t work.
• Too large, Too slow
• There is, still, implied a barrier between research and
applications, with research limited to, perhaps, “climate
science.”
• There is implied a one-way communication of
information, from climate-knowledge providers to
climate-knowledge customers.
• There is implied a hierarchical structure, with one action
leading to the next action down a definable decision tree.
Re-framing
• We have many of the ingredients that are
needed to provide both useful and usable
knowledge-based climate information.
– We have some successful experiences.
• To scale these successful experiences, to utilize
these ingredients more effectively, to build that
which is missing, we must re-frame how we view
the end-to-end system necessary for climatechange problem solving.
An attempt at re-framing (1)
• Push of climate information: There is great intellectual
capital in the applications community, and we need to
provide foundational climate knowledge in a way that this
community can co-evolve with pulls on the knowledge
base and incorporation of communities expertise and
knowledge back into the knowledge base. We need to
co-generate solutions.
• Complexity: In a traditional institutional manner, we
cannot manage, control or even provide the knowledge.
We need to strive for self-organizing, self-correcting
communities that form structures and advance from an
experiential base.
An attempt at re-framing (2)
• Implied one-way communication: There must be
equivalent communication from the application side back
to the climate-science side. Shared problem-solving
language needs to be developed.
• Hierarchical Structure: The relation between sciencebased knowledge and it’s use in applications is complex
- and it changes. It is strongly influenced by externalities
that render the role of climate information as
subordinate.
– Rather than thinking of hierarchical relations, perhaps, biological
relations are more useful, objects that interact in different ways
at different times for different purposes (this challenges the org
chart).
• Attention to interactions and interfaces
Open Climate
• The importance of utilizing the information
provided by climate projections in concert with
the complexity, including the breadth, depth, and
distribution of intellectual resources, frames the
need for open community approaches that allow
the emergence of solution paths and
development of tools to accelerate and improve
our ability to use climate projections to plan for
climate change.
Open Communities, Open Source
• An attempt to define a little
– Open source means different things to
different people.
• To some it means that code is available
• To some it means a certain type of license
• In this case it represents a culture
Open Communities, Open Source, Open Innovation
• “The open source model includes the concept of
concurrent yet different agendas and differing
approaches in production, in contrast with more
centralized models of development such as those
typically used in commercial software companies. A main
principle and practice of open source software
development is peer production by bartering and
collaboration, with the end-product, source-material,
"blueprints" and documentation available at no cost to
the public” (Wikipedia, 20101117)
The Cathedral and the Bazaar
Focus on climate models
The problems we face require tailoring here.
NM
nM, nS, nDC
“product”
Model Projections
“uncertainty”
Scientific
Investigation
NS
Tailoring 1
(e.g. downscaling)
NT1
“product”
“uncertainty”
Applications
NA
“product”
Tailoring 2  n
(e.g. downscaling)
NTn
“uncertainty”
The Point: Enormous
complexity. NA >> NM, and
NM is unmanageable.
Tight wire 2
• Isn’t the climate community the archetype
for community software? Haven’t we done
that problem?
– i.e. Community Earth Science Model
Paul Edwards: A Vast Machine
An excellent history of weather and climate
as prototype community.
• Yes, but ….
Tight wire 2
• Yes, but ….
– This form of community evolved when the
problem was smaller and resources more
centralized.
•
•
•
•
Models for the community
Models with community participation
Stunningly successful
Still, fundamentally, institutional
– Institutional and cultural inertia
Paul Edwards: A Vast Machine
An excellent history of weather and climate
as prototype community.
There is a need
• We have made the case that the surface
of the planet will warm, ice will melt, sea
level will rise, and the weather will change.
• We cannot, then say, wait until the models
are “good enough” by our unknown
standards, then we will provide numbers
for your use.
• The role of institutions and gate keeping
must change.
What stands in the way?
• From Free and Open Source Software for
Geospatial (FOSS4G) Conference:
– “Professor (Andy) Pitman made an impassioned plea
for more involvement by open source developers in
climate models …” (link)  open source is not part of
the climate modeling culture.
• Something special about climate model
software?
– Easterbrook on Climate Model Software
What does stand in the way
• Culture and History
• Belief that climate software is special?
• Misunderstanding of how “science” fits into
software development and problem solving.
• Good software
• Transparent tuning and model validation
• Governance of communities
– How are decisions made?
• …
A moment with tuning and validation
• When talking with scientists, we always evolve
or devolve to a conversation of model validation
and tuning. That this is impossible without
experts and institutional control. This comes off
as a, de facto necessary, arcane process that
cannot be explained, much less codified.
• Is this true?
– If yes … then think of the implications of a “yes”
answer.
Addressing Validation and Uncertainty Description
• Deconstruct the problem of validation
– Monitoring, Quality Assessment, Forecasting, Systems
Validation, Scientific Validation
• Write a validation plan, including how to determine when
a model is validated
• Separate development and validation
– Make it scientifically robust
• Deconstruct uncertainty
– Describe uncertainty
– Relate uncertainty to validation
• Relate uncertainty and validation to next development
cycle.
In the end
• Development of open innovation, open source, open …
is critical to accelerate, to improve, and to develop the
use of state-of-the-art climate information.
– If this is not done, then the scientific investigation will be marginalized.
• There are many successes and assets, some of which
have been moving towards open communities. Often,
still, anchored in the community of scientists with the
idea of providing “science.”
– This is too restrictive.
– Can models be built by communities rather than for the community?
• Cultural transformation
• Have I been pushed of the tight wire?
Some Basic References
• Intergovernmental Panel on Climate Change
– IPCC (2007) Working Group 1: Summary for Policy
Makers
• Spencer Weart: The Discovery of Global
Warming
• Paul Edwards: A Vast Machine
• Rood
– Rood Climate Change Class
• Reference list from course
– Rood Blog Data Base