The Computational Future for Climate Change Research

1
The Computational Future for Climate Change
Research

• Warren M. Washington
• National Center for Atmospheric Research
• Boulder, Colorado

June 2005
2
Climate Change Computational Future

• Components Atmosphere, Land-Vegetation, Ocean,
  Sea Ice, Surface Hydrology, Biogeochemical cycles
  (e.g. carbon cycle) including aerosols, and ice
  sheet model
• Simulations of recent past and future climate
  including the Inter-governmental Panel on Climate
  Change (IPCC) simulations
• Higher resolutions
• Computational Efficiency
• Various Computational Paradigms

3
DOE and NSF support has lead to major
improvements in Community Climate System Model
version 3 (CCSM3.0)

• Complete system
• New portability for vector and Linux
  supercomputers systems
• Improved physical processes
• Improved methods to run IPCC climate-change
  experiments on variety of computers and
• Flexibility to simulate climate over a wide range
  of spatial resolutions with greater fidelity.

4
Timeline of Climate Model Development
Ice Sheet
5
Model Resolutions
R15
T42
500 km
300 km
75 km
T170
T85
150 km
6
(300 km)
(150 km)
(75) km
(37 km)
7
(No Transcript)
8
View of the Parallel Ocean Program (POP) Model
Horizontal Grid at 2/3o Resolution
Note high resolution in North Atlantic and near
equator
9
(No Transcript)
10
(No Transcript)
11
Sea-ice Concentration Climatology (1979-1999)
12
Global and Regional Climate Aspects Using a
Climate Model

• El Niño/La Niña
• Monsoons
• North Atlantic Oscillation
• Arctic Oscillation

13
Sample Trajectory Figures from the 0.1o POP model
showing complex eddies structures

• From
• Julie McClean, Mathew Maltrud, Frank Bryan and
  Deterlina Ivanova
• Naval Post Graduate School
• Los Alamos National
  Laboratory
• National Center For
  Atmospheric Research

14
Numerical Trajectories Salinity particles
released in the GIN Sea
15
Numerical Trajectories Salinity particles in
the North Atlantic
16
Numerical Trajectories Salinity particles near
Indonesian Seas
17
Flow through Indonesian Seas
18
River Transport
Coarse Resolution
19
Title slide
Mt Pinatubo eruption in the Philippines, June 15,
1991. Gases and solids injected 20 km into the
stratosphere.
From Church, White, Arblaster
20
(No Transcript)
21
(No Transcript)
22
Barnett et al. SCIENCE, July 8th, 2005
23
Climate Change Scenarios At any point in time,
we are committed to additional warming and sea
level rise from the radiative forcing already in
the system. Warming stabilizes after several
decades, but sea level from thermal expansion
continues to rise for centuries. Each emission
scenario has a warming impact. (Meehl et al.,
2005 How much more warming and sea level rise?
Science, 307, 1769-1772)
24
(Meehl et al., 2005, GRL, submitted)
25
(No Transcript)
26
Climate Change Animation
The following animation depicts warming of the
Earth's surface as simulated by the CCSM. It
shows both regional changes and an average of the
entire globe. The data represents the
temperature difference from the 1870-1900 period,
and is an average of five separate CCSM
experiments, from 1870 to 2100. The future period
is derived from the IPCC A1B scenario. The gray
shading around the global average shows the
spread of the ensemble members. Created by Gary
Strand of NCAR using the Ferret software package.
27
By the end of the 21st century, greater warming
occurs at high northern latitudes and over the
continents in the scenario simulations. We are
already committed to about another half a degree
of warming over North America by the year 2100.
28
(No Transcript)
29
Changes in N. Atlantic Currents
30
Change of Extremes

• Heat waves, cold snaps
• Floods, droughts
• First freeze dates, hard freeze frequency
• Precipitation intensity
• Diurnal temperature

31
Pioneering Computation by Ming and Droegemeier of
Un. Of Oklahoma. showing what happens with good
cloud physics and a high enough resolution

• Domain 50 km by 16km
• ?x, ?y25 m, ?z20 m
• 1/3 billion grid points

32
This is what you get with a 25 m grid and the
right physics!
33
Earth System Model Interactions
Climate and Sea Level
Atmospheric Composition
Atmospheric Chemistry
Climate
Ocean Carbon Cycle
Ocean - temperature - sea level
Ecosystems
Human Activities
Other Human Systems
Terrestrial Carbon Cycle
Energy System
Unmanaged Ecosystem
Agriculture, Livestock and Forestry
Coastal System
Crops and Forestry
Hydrology
34
Ongoing and Future Developments

• Higher resolution, especially important near
  mountains, river flow, and coast lines
• Full hydrological coupling including ice sheets
• Better vegetation and land surface treatments
  with ecological interactions
• Carbon and other biogeochemical cycles

35
New York Times 8 June 2004
36
Melting Greenland Glacier This lubricates bottom
of ice. Photo by Roger Braithwaite and Jay
Zwally, NASA
37
Lightning natural source and combustion
anthropogenic of nitrate
denitrification
microbes
Coupled Carbon-Nitrogen dynamics

• Strong feedback between decomposition and plant
  growth soil mineral nitrogen is the primary
  source of nitrogen for plant growth. Nitrogen
  fixing bacteria/algae very important, however,
  limited field and laboratory data on their role.
• Can result in a shift from carbon source to
  carbon sink under warming scenario.

P.E. Thornton, NCAR
38
Climate Model Problems with Supercomputer Systems

• Computers not balanced between processor speed,
  memory bandwidth and bandwidth between processors
  including global.
• More difficult to program and optimize
• Hard to get I/O out of computers efficiently and
  computer facilities need to have expanded
  archival capability in the petabyte range.
• Little relationship between theoretical
  performance and performance on actual working
  climate model programs.

39
What SciDAC has done for the Climate Modeling
Science

• It has brought new ideas and computational
  techniques to the climate modeling community.
  (The days of a small group of people putting
  together a state-of-art climate model are
  limited.)
• It has helped with BER to bring together the
  expertise of DOE laboratories, universities, and
  centers into a coordinated effort. Innovation
  still alive and needs to be protected!
• It has provided enhanced computing capability to
  the nations climate modeling community but
  further increased are needed.
• Together we are providing scientifically credible
  information to the national and international
  policymakers.

40
News Event
Joint Science Academies statement on 7 June
2005 there is now strong evidence that global
warming is occurring and the scientific
understanding of climate change is now
sufficiently clear to justify nations taking
prompt action
Climate models have helped determine future
environmental and energy strategies by
performing many what if simulations. Bottom
lineSciDAC has improved climate models and
their projections.
41
The End

• This project involves the efforts of over 300
  scientists and computational experts at DOE
  Laboratories, NCAR, and the Universities.

42
Next IPCC Schedule2007-9

• Need to get ready
• Petabyte data sets

?
43
How do we make progress?

• Finer resolution
• Improved physical processes
• High performance supercomputers

44
Analysis of Millennium Temperatures
45
Climate models can be used to provide information
on changes in extreme events such as heat
waves Heat wave severity defined as the mean
annual 3-day warmest nighttime minima event
Model compares favorably with present-day heat
wave severity In a future warmer climate, heat
waves become more severe in southern and western
North America, and in the western European and
Mediterranean region Meehl, G.A., and C. Tebaldi,
2004 Science, 305, 994--997.
Observed
Model
Future
46
(No Transcript)
47
(No Transcript)
48
CCSM3 (Meehl et al., 2005, J. Climate
special issue paper)
49
(No Transcript)
50
David Pierce Scripps Institution of Oceanography
51

52
(No Transcript)
53
Change of Extremes

• Heat waves, cold snaps
• Floods, droughts
• First freeze dates, hard freeze frequency
• Precipitation intensity
• Diurnal temperature

54
CCSM3 (Meehl et al., 2005, J. Climate
special issue paper)
55
Climate Change for 2100
56
Science Problems (From G. Assmar, NASA)

• Climates of past geologic periods 102-106
  years).
• Causes of ice-age climates.
• Effects of alternative parameterizations of
  various physical processes in the atmosphere,
  ocean, land surface, sea ice, and snow.
• Interactions of atmospheric circulations with
  photochemistry, especially the carbon cycle.
• Interactions of the atmosphere and oceans with
  sea ice and continental ice sheets.
• El Ni\no, the Southern Oscillation, the Walker
  Circulation, North Atlantic and Arctic
  Oscillation.
• Interannual and intra-annual variability.
• Vegetation changes (for example, those produced
  by deforestation).
• Causes of changes in monsoon circulations.
• Causes of droughts and the formation of deserts.
• Effects of volcanic eruptions.
• Effects of \cotwo and other trace gases.
• Climate predictions on monthly, seasonal, and
  longer time scales.
• Effects of natural and anthropogenic aeroso

57
Pacific Decadal Oscillation (PDO)
Warm phase
Cold phase
From University of Washington
58
Ensemble Simulations
59
PCM Observed ENSO pattern
60
(No Transcript)
61
(No Transcript)
62
(No Transcript)
63
(No Transcript)
64
From M. Prather University of California at Irvine
65
(No Transcript)
66
We are already committed to about as much warming
as we experienced in the 20th century by the year
2100, but over three times as much sea level rise
67
(No Transcript)
68
(No Transcript)
69
(No Transcript)
70
(No Transcript)
71
Status of CCSM3

• Bill Collins
• National Center for Atmospheric Research
• Boulder, Colorado

72
Topics

• Release of CCSM3
• Simulations for IPCC and paleoclimates
• Improvements in the climate simulation
• Systematic challenges in the simulations
• Extensions to coupled chemistry/climate
• Extensions to higher spatial resolution and
  process fidelity

73
(No Transcript)
74
Release of CCSM3

• Release Date June 23, 2004
• Contents
• Code
• Input data sets
• Scripts for compilation and execution
• Documentation
• URL http//www.ccsm.ucar.edu/models/ccsm3.0/
• Number of downloads to date Counter

75
Supported Resolutions and Platforms

• Scalar Platforms
• IBM SP
• SGI Origin
• Linux (PGI)
• Compaq

• Vector Platforms
• NEC (Earth Simulator)
• Cray X1

T85x1
T42x1 FV 2?2.5
T31x3 .
IPCC
Resolution
AMWG, CVWG, Paleo,
BGCWG, Paleo
Dynamics
76
CCSM3 Control Experiments
1990 1780 1870 1CO2 1CO2 2x 1CO2 4x 20th C.
T31x3 748 406 TBD 171 104 107 TBD
T42x1 869 408 110 168/ 214 152 151 TBD
T85x1 611 280 114/ 160 151 44 2000 (5x)
Yellow NCAR Systems Red Earth Simulator
TBD To Be Done
?
77
Distribution of CCSM3 Control Runs

• Central site Earth System Grid
• DOE project to integrate major centersfor
  supercomputing and analysis
• CCSM3 output available
• Current contents Perpetual 1990 runs
• Averaged data sets Open access
• Original history files SCD account holders
• Access point https//www.earthsystemgrid.org/

78
CCSM3 Documentation

• New software manuals
• Users guides
• Code references
• New NCAR technical notes
• CAM
• CSIM
• CLM
• DVGM (Dynamic Global Vegetation Model)
• New technical report for POP

79
Special Issue of Journal of Climate

• Objectives
• Describe CCSM3 to the climate community
• Document CCSM3 for the IPCC 4th Assessment Report
• Topics
• Overview of CCSM
• Description of features climate state for each
  component
• Climate sensitivity
• Response of CCSM to paleo, pre-industrial
  conditions
• Major modes of coupled variability
• Editor Dave Randall
• Deadline November 2004
• Contents 29 papers proposed by seven working
  groups

80
Special Issue of IJHPCA(International Journal of
High Performance Computing Applications)

• Guest editors
• John Drake (ORNL)
• Phil Jones (LANL)
• Tom Henderson (NCAR)
• Schedule
• May 2004 Call for papers
• Oct. 2004 Deadline for papers
• Fall 2005 Publication of special issue (V.
  18,3)
• Major topics
• Software engineering for climate models
• Performance and portability of climate model codes

81
The Experimental Configuration

• Three phases
• Pre-industrial (1870)
• 20th Century (1870-2000)
• Emissions Scenarios
• SRES Scenarios
• Commitment (20th C. CO2)
• 2000-2100 (5 runs)
• 2100-2200 (1 run)
• A1B and B1 Scenarios
• 2000-2100 (5 runs)
• 2100-2200 Const (5 runs)
• 2200-2300 Const (1 run)
• A2 Scenario
• 2000-2100 (5 runs)

82
Experimental Design
Forcing 20th Century 21st-23rd Century
Greenhouse Gases Observed SRES
Ozone Trop MOZART Strat Solomon Trop MOZART scaled by O3 TAR forcing Strat Solomon
Sulfate Aerosols SO2 Smith/Wigley SO2 SRES
Carbon Aerosols Population Scaling SO2 Scaling
Sea-salt Dust Year 2000 values Year 2000 values
Volcanic Aerosols Ammann (2003) Year 2000 values
Solar Variation Lean (1995) Year 2000 values
Indirect Effects None None
83
Aerosol Optical Depths
Sulfates
Black Organic Carbon
Strat. Volcanics
84
Comparison of CCSM3 and Global Surface
Temperatures
CCWG/Arblaster
85
Component Models

• Atmosphere
• Land-Vegetation
• Ocean
• Sea Ice
• Biogeochemical Cycles (e.g. Carbon cycle)
• Ice Sheet
• Regional Model Coupling

86
Projections for the 21st Century
A1B Scenario
B1 Scenario
Commitment
?
Kyosei Consortium
87
Transient Climate Sensitivity from CCSM3
88
SST Biases in W. Coastal Regions
?
89
Equilibrium Sensitivity from CAM3 SOM
Kiehl and Shields
90
Equilibrium Climate Sensitivity
Kiehl and Shields
91
Cloud Radiative Response to 2xCO2
Kiehl and Shields
92
CCSM3 Surface TemperatureGlobal/Annual Mean
Coupled
Slab ocean
- 2.8C
- 5.7C
- 5.9C
Otto-Bliesner
93
Land Use as a Climate Forcing
Much of the present-day natural vegetation has
been cleared for agricultural land. By 2100,
there is likely to be further expansion of
agricultural land in North America, South
America, Africa, and Southeast Asia What is the
land use forcing relative to other natural and
anthropogenic forcings?
Undisturbed
1970
2100
Participants LMWG, CCWG, and NCAR Biogeosciences
and Assessment initiatives
94
Simulations for Present-Day ConditionsCCSM3 vs.
CCSM2
95
Effects of the Ocean Diurnal Cycle
Observations
CCSM3 with Cycle
CCSM3 without Cycle
96
Effects of Chlorophyll Absorption
Heating By Chlorophyll
?
97
Effects of Resolution on Sea-Ice Thickness
T85?1
T42?1
T85?1 ? T42?1
98
Effects of New CAM Physics on Temperatures
CAM2 T42
NCEP
CAM3 T42
CAM2 T42
CAM2 - NCEP
CAM3 CAM2
99
Improved Cloud Response to ENSO
CAM3 T42 AMIP
CAM3 T42 AMIP
CAM2 T42 ERBE
ERBE
100
Improvements in Surface Radiation over Sea-Ice
Shortwave Flux
Observations
Longwave Flux
101
Semi-Annual SST Cycle
Anomaly Observations
Anomaly T85?1
102
Periodicity of ENSO
NCEP Monthly Nino 3.4 ?0.82 K
T85?1 Monthly Nino 3.4 ?0.77 K
?
103
Double ITCZ Issue
104
Excessive Ice in FV Coupled Runs
FV2?2.5?1
T85?1
FV2?2.5?1?T85 ?1
105
Continental Precipitation Biases
Large dry precipitation biases in southeast U.S.,
Amazonia, and SE Asia could adversely affect the
terrestrial carbon cycle
106
Continental Temperature Biases
Large (6-10ºC) winter warm temperature biases in
the Arctic could adversely affect the terrestrial
carbon cycle
107
Multi-Century Coupled Carbon/Climate Simulations
2.0
14.1
13.6
-2.0
Net CO2 Flux (Pg C/yr)
Surface Temp.

• Fully prognostic land/ocn BGC and
  carbon/radiation
• Atm-Land 70 PgC/yr ?? Atm-Ocean 90 PgC/yr ??
• Net Landocean 0?1 PgC/yr
• Stable carbon cycle and climate over 1000y
• Projection of climate change on natural modes
• Detection attribution
• Future climate projections/fossil fuel
  perturbations

Doney and Fung
108
Response of Terrestrial Carbon to Interactive
Nitrogen Deposition
A2 Scenario
2000
2100
Wet deposition
NEE response to 1 C step change
Dry
Sink
Coupled C-N model
Nitrogen deposition likely to increase in
future Including nitrogen in terrestrial carbon
model can change sign in carbon dioxide flux!
C-only model
Source
?
Mahowald and Thornton
109
Simulations with CLM-CN for 275 years
Leaf carbon pool (gC/m2) Captures the essential
global details of canopy distribution
Net Ecosystem productivity (gC/m2d) Boreal zone
still accumulating carbon.
CLM3-CN operating in C-only mode for rapid
spinup (Peter Thornton, NCAR)
110
Offline Transport Modeling in CCSM3
Offline CAM3
MOZART

• Comparison of passive tracer concentrations at
  500 mb
• Advection using NCEP fields for 30 days
• Tracer initialized to unity at 700 mb

Hess and Rasch
111
Prognostic Aerosols
Seasonal Cycle of Prognostic Dust Optical Depth

• Aerosol Species
• Sulfates
• Nitrates
• Dust
• Sea Salt
• BC and OC

Mahowald
112
Response of Ocean Biogeochemistry to Changes in
Iron Deposition
Changes in mineral aerosol deposition due to
climate change interacts with ocean uptake of
carbon through iron limitation and nitrogen
fixation ? changes ocean carbon uptake in future.
Mahowald
113
Differences in Ozone due to Dynamics
114
Reactive Tropospheric Ozone Chemistry
Lamarque
115
High Resolution Ocean Modeling
?
116
Entrain More Hydrologists In Model Development
Participants Dennis Lettenmaier (University of
Washington) Eric Wood (Princeton
University) LMWG and NCAR Water Cycle
initiative Gordon Bonan (NCAR) David Gochis
(NCAR) David Yates (NCAR) Keith Oleson (NCAR)
Goal The Variable Infiltration Capacity (VIC)
model is a leading macroscale hydrologic model.
We want to bring the VIC hydrology into the CLM
to improve the simulation of the hydrologic cycle
117
Improved SLP Teleconnections for ENSO
118
Conclusions

• CCSM3 has been released to the community.
• IPCC simulations are well underway.
• Analysis has begun for
• Characterization of the simulated climate
• Climate sensitivity
• Climate forcing for the 20th century
• Challenges ahead
• Process-oriented modeling of the climate
• Coupled chemistry/climate modeling

119
Sensitivity of Carbon to Dynamics
120
Global carbon budget, 275-year offline CLM spinup
GPP
NPP
AR
Interannual range in components
HR
NEE
Sym Description PgC/y
GPP Gross primary production 14.0
AR Autotrophic respiration 12.0
NPP Net primary production 6.5
HR Heterotrophic respiration 4.0
NEE Net ecosystem exchange 6.5
CLM3-CN operating in C-only mode for rapid spinup
(Peter Thornton, NCAR)
121
Meridional Overturning CirculationAtlantic Ocean
CCSM3
CSM1
LGM
Pre Industrial
Otto-Bliesner
122
Acceleration of CCSM3 CAM3
123
Vectorization of CLM
124
LGM (21K BP) Seasonal Cycles
125
Vectorization of CAM3
126
More Components

• Sea ice-viscous-plastic Hibler dynamics with
  relaxation, or elastic-viscous-plastic, 27 km
  resolution
• Parallel flux coupler ties components together

127
Relationship Between North Australian Rainfall
and the Decadal Pacific Oscillation
Observations correlation -0.8
PCM correlation -0.5
128
LGM (21K BP) Land-Ice Extent
129
(No Transcript)
130
Deep/Abyssal Northwest Atlantic
131
(No Transcript)
132
(No Transcript)
133
(No Transcript)
134
(No Transcript)
135
Major Improvements in CCSM3.0

• Atmosphere
• New treatments of the condensed water and
  geometry of clouds
• Improved treatments of cloud droplets and ice
  particles
• Improved representation of the interactions among
  water vapor, solar radiation, and terrestrial
  thermal radiation
• New treatment of the effects of aerosols on the
  reflection and absorption of solar radiation and
  
• New dynamical frameworks suitable for modeling
  atmospheric chemistry
• Coupler
• Improved performance and scalability on parallel
  supercomputers
• Faster multi-way communication among the
  component models and
• New communications infrastructure

136
Major Improvements in CCSM3.0

• Land
• New methods to enable simulation of the
  terrestrial carbon cycle
• New methods to enable simulation of dynamic
  vegetation and
• Improvements in land-surface physics to reduce
  temperature biases
• Ocean
• Improvements to the representation of the ocean
  mixed layer
• Inclusion of solar heating by chlorophyll and
• New infrastructure for studying vertical mixing
  in the ocean
• Sea-ice
• Improved schemes for horizontal advection of sea
  ice
• More realistic treatments of snow and ice
  reflectivity and
• Updated treatments for sea-ice thickness