Transcript Climate Change And The Pacific Northwest

The Unknown
And The Unexpected:
Climate Change And
The Pacific Northwest
Mark R. Abbott
College of Oceanic and
Atmospheric Sciences
Oregon State University
Overview
Global scale climate
Modeling and climate
Climate and the Northwest
New ways to approach the unknowable
Historical Records of CO2
CO2 And Temperature
Past And Future Rise In Sea Level
20,000 years ago
2200? ( + 5 meters)
Timeline Of Climate
Model Development
Figure courtesy W. Washington, NCAR
Impacts Of Spatial Resolution On The Types
Of Terrain That Are Included In Models
Figure courtesy S. Hostetler, USGS/OSU
What Was Predicted For Last Winter?
What Actually Happened?
An Impact Of Climate Change?
Climate Change
Impacts Depend
On Variability
Variability may
have more
human impact
than mean state
Examples
European
heat wave
Niger drought
Figure courtesy A. Mix, OSU
The Five “Currencies”
Of The Earth System
Energy
Water
Carbon
Money
Information
Next-generation Earth system models
must account for all of these currencies
Innovation In Earth System
Models The “Old” Way
Increased temporal and spatial resolution
Data assimilation and near real-time
observatories and sensor networks
Increasing coupling of components
and increasing richness of models
Impact on HPC requirements has been a
focus on increased capacities for computation
as well as on post-production analyses
Can the community obtain the HPC resources
it needs in a rapidly-changing market?
New Demands On Climate Models
Not just increased resolution and more physics
Risk assessments and scenarios
Impacts of climate change on
human migration patterns
Rare but high impact scenarios
Feedbacks between climate
and socioeconomic processes
Past as a poor predictor of the future
Need to create diverse and resilient knowledge
networks to develop and test scenarios
Adaptive management
rather than rigid protocols
System Architectures
The Intel and AMD Approach
General purpose systems for
a wide array of applications
CPU
CPU
MPI
6 month technology cycle
Science community started rolling
their own clusters because of lower costs
Increasingly Complex Architectures
Intel and AMD stalled at 3 GHz because of
packaging issues (i.e., heat dissipation)
Now going multi-core at lower frequency
OpenMP
Core 1
Core 2
Core 3
Core 4
Core 1
Core 2
Core 3
Core 4
OpenMP
MPI
Increasingly complex programming model
Low efficiencies because fundamental models require
lots of shared memory (and hence message passing)
Focus on communications infrastructure
Hard to take generalized commodity
systems and support innovation
New architectures (CPU/GPU) now entering the fray
Is This Innovation?
Community models
Good for production runs
and post-production analyses
Good for education
Not so good for adding new components
and new physics and biogeochemistry
Supercomputer centers
Good for production runs of very large models
Not especially efficient for exploration
of new models and new approaches
New architectures make it increasingly difficult to code
We need more than a big data
center with VT 100 terminals
Shifting The Context For IT
Transforming our workflows
Real-time, continuous, adaptive
Shifting the balance of data and knowledge
Collaborative, networked, interdisciplinary
Not just bigger and faster
A collection of adaptable, dynamic services
Increased customization in a commodity world
HPC-enabled components live
everywhere, not just data centers
End-to-end, user experience is critical
Not just in the office or lab
A New Approach?
Overcoming language and cultural
barriers to enable collaboration
Merging data streams and sensors
Developing and testing models
Seeking out underlying and emergent
rules to make projections and predictions
This looks more like a
network gaming metaphor
New capabilities include
Preservation, provenance, collaboration,
accountability, and reputation
Shifting The Community
From data delivery to knowledge services
The Wikipedia model for smart mobs
Networks of potential resources
Innovation will come from the fringes
More “democracy” in the system
What used to be expensive and
complex is now available to everyone
Need for standard, open frameworks
Simple and extensible
Not driven solely by science requirements
But still require enormous amount of work
Convergences
New IT capabilities will bring
more integration among diverse
communities, data sets, and models
New demands for climate models
and climate services will require
these capabilities
Need a resilient approach to the
complexities and uncertainties of
the future
© 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other
countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to
changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of
this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.
Microsoft Research
Faculty Summit 2007