Using Cyberinfrastructure to Study the Earth

1
Using Cyberinfrastructure to Study the Earths
Climate and Air Quality
Don Wuebbles Department of Atmospheric
Sciences University of Illinois,
Urbana-Champaign February 2005
2
Special Acknowledgements

• Donna Cox
• John Martirano
• Lorne Leonard
• Stuart Levy
• Katharine Hayhoe
• And the students that contributed to the
  visualization project

3
Performance Data for MOZART-3
  MPI     OpenMP      wall-clock per

4
WACCM-3 with chemistry

• Global climate model from ground to 140 km
• Integrates a stiff system of coupled nonlinear
  differential equations over time
• Domain 63 chemical species (or more), 66
  vertical layers, 30 minute time step
• 96 IBM power4 cpus full simulation year takes
  about 6.6 wall clock hours (72 longit. x 46
  latitud.)
• 96 IBM power4 cpus full simulation year takes
  about 27.7 wall clock hours (144 longit. x 96
  latitud.)

5
Geosciences CI Challenges

• Enormously complex human-natural system
• Vast temporal (sec to B yrs) and spatial (microns
  to 1000s of km) scales
• Highly nonlinear behavior
• Massive data sets
• physical and digital
• static/legacy and dynamic/streaming
• geospatially referenced
• multidisciplinary and heterogeneous
• open access

6
Geosciences CI Challenges

• Massive computation
• weather, space weather, climate, hydrologic
  modeling
• seismic inversion
• coupled physical system models
• Inherently field-based, visual disciplines with
  the need to manage information for long periods
  of time
• Bringing advanced CI capabilities to education at
  all levels
• Connecting the last mile to operational
  practitioners

7
Atmospheric Science Research

• Huge growth in observational and model data
  volumes and data streams
• Huge growth in model complexity, resolution and
  sophistication requiring better initialization,
  analysis and visualization techniques
• Growing application of data assimilation as a
  tool for providing homogeneous data sets and
  extracting maximum information content from
  observations
• Distributed, intelligent observing systems

8
Atmospheric Science Education

• Integration of collaboration tools and
  distributed data capabilities into the
  atmospheric sciences curriculum
• Real-time weather data and archives of climate
  data now commonly available in the classroom
• Digital libraries assemblies of high-quality
  teaching and learning resources with tools to
  enable exploration of large atmospheric science
  data sets

9
Cyberinfrastructure and Science Analysis
Computing / Grid Services
Archive
Access/ Delivery
Information Technology Services
10
Climate Data Visualization Development

• 2 stages
• Parallel Data Processing
• Prepare visualization data
• E.g., absolute temperature change over period
• Relative changes (departures / anomalies)
• Surface and altitude data
• Various projections (e.g., globe versus map)

11
Climate Data Visualization Development

• Parallel Visualization
• Partiview on tiled display wall
• Multigigabytes visualizaions with separate views
• 40 node display wall at NCSA
• 15 node display wall at Atmospheric Sciences
• Arrange variables spatially for comparison and
  analysis
• 3-D and temporal interactivity
• Can be embedded in Virtual Director
• Allows for collaborators to see results in real
  time
• Recording and playback of spline-based navigation

12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
2000
17
2100
18
A Few Findings from our project

• Need to learn each others language
• Just what is a rendering anyway?
• Undergraduate student help not reliable for
  project completion
• Good workers, but did not seem concerned with
  whether they completed their tasks
• The NCSA folks were great!
• Very helpful and always concerned about how they
  could help me