Teaching Computational Thinking:
Examples from Weather and Climate
Modeling
“Essentially, all models are wrong,
but some models are useful.”
- George E. P. Box (1951)
Teresa Eastburn & Randy Russell
National Center for Atmospheric Research
University Corporation for Atmospheric Research
NSTA Denver, December 12, 2013
Computational Thinking
Solving problems, designing systems, and
understanding human behavior by drawing on the
concepts fundamental to computer science.
~ Jeannette Wing, Carnegie Mellon
Integrating the power of human
thinking with the capabilities of
computers.
~CSTA
Steven Gilbert
NSTA Press
Here’s What We’ll Be Covering
1. What is a climate model, why are supercomputers needed,
and what do they do and not do?
2. The Systems Game – Why systems thinking matters
3. What’s the difference between a weather model
vs a climate model (initial value problem vs.
a boundary value problem)?
4. Chaos Theory
5. Climate simulations for your
you and your students to
explore
Spark – science education at NCAR
National Center for Atmospheric Research
in Boulder
NCAR Mesa Lab in Boulder
Public and School Group Visits
spark.ucar.edu/visit
spark.ucar.edu/workshops
spark.ucar.edu/events/workshopcomputational-thinking-nsta-regional-2013
Evolution of Climate Models
Credit: Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4):
Working Group 1: Chapter 1, page 99, Fig. 1.2
Climate Model Components
Credit: UCAR (Paul Grabhorn)
Climate Model Components
Credit: UCAR
Progress in climate models
occurs as a result of:
Like a sturdy
3-legged
stool
• Observations
• Theory
• Numerical Modeling
THEORY
“Science presumes that
things and events
in the Universe occur in
consistent patterns
that are comprehensible
through careful,
systematic study.”
~ AAAS
Models are today’s
tech test tube for the Earth system.
Image source
adaption:
NOAA
Images adapted from K. Dickson, NOAA
Climate Models = Virtual Earth
• Now we can model various components
(parts or subsystems) in the Earth system
(atmosphere, ocean, sea ice, land physics…)
and how they will interact and respond over
time to a natural or human-made forcing
agent.
Atmosphere Circulation & Radiation
Sea Ice
Ocean Circulation
Land Physics
Resolution: What Does It Mean?
Improving Resolution of Climate Models
Grid Cell Sizes
• 1990s (T42)
• 200 x 300 km
• 120 x 180 miles
• 2000s (T85)
• 100 x 150 km
• 60 x 90 miles
Credit: Warren Washington, NCAR
Improving Resolution of Climate Models
Credit: Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4):
Working Group 1: Chapter 1, page 113, Fig. 1.4
Vertical Resolution of Climate Models
Vertical Layers
• 1990s
• 10 layer atmosphere
• 1 layer “slab” ocean
• 2000s
• 30 layer atmosphere
• 30 layer ocean
Credit: UCAR
Horizontal and Vertical Grid
Horizontal and Vertical Grid
Hexagonal Grid and Sub-grids
Credit: UCAR (Lisa Gardiner)
spark.ucar.edu/sites/default/files/SystemInMotionMaster.pdf
Using Models in Education
“Essentially, all models are wrong,
but some models are useful.”
- George E. P. Box (1951)
Weather vs Climate Projections
Physics is Physics, Right?
Why do we think we can make meaningful
100 year climate projections when we can’t
forecast the day-to-day weather a month from
now? Initial Value Problem vs Boundary Value Problem
Weather Model vs Climate Model
Compare and Contrast
Differences (and similarities) between
Weather vs. Climate Models
• Area Covered (scale)
• Resolution – distance (spatial) and time (temporal)
• Timespan covered by model runs
• Impacts on computing resources needed,
time required to run models
Weather Model vs Climate Model
Area Covered
Weather Model – up to about
continental size scale
Climate Model – global size scale
Larger area requires either more computing
power/time or lower resolution (spatial and/or
temporal)
Weather Model vs Climate Model
Resolution and Precision
Weather Model
• resolution typically about 3-10 km
• timesteps of hourly to 6 hours, forecast for next 3-4 days
Climate Models
• resolutions from about 25-30 km up to 100 (or a couple hundred) km
• running computer models can take days or weeks, which would be
impractical for weather models
Precision – why Wx forecast for Christmas is suspect, but temperature
next July is reliable (relationship to chaos)
Weather Model vs Climate Model
Timeframe
Climate Projection – decades to centuries
or longer
(climate is usually defined as at least 30
years of observations)
Weather Forecast – hours to days
(up to about 10 days)
Resolution: Spatial & Temporal (Time)
• Timesteps can be a few minutes to 12 hours or more
• Durations can be hours to centuries
Resolution and Computing Power
Double resolution – increase number of nodes – more calculations!
One Dimension
2 times as many nodes
Two Dimensions
4 times as many nodes
Resolution and Computing Power
What if we increase model to three dimensions (space) plus time?
Resolution and Computing Power
What if we increase model to three dimensions (space) plus time?
16 times as many nodes – 16x computing power required!
This is why we need supercomputers!
Chaos
• Chaos – 10-day forecast reliability limit
• Ensemble runs of models – tipping points – arctic
ice melt – sea ice and open water albedo images
• Why Wx forecast for Xmas is suspect, but
temperature next July is reliable (relationship to
chaos)
Climate Forcings
Source: Meehl et al
NCAR
Which of the following cannot be
addressed by a physical climate model?
1. How would Earth’s average surface temperature be expected
to change if carbon dioxide doubled?
2. How much carbon dioxide and methane will humans add to
the atmosphere during each of the next five decades?
3. Can cosmic rays from the sun affect clouds and hence play an
important role in climate variability and change?
4. Is it possible to learn about past climate variations by
gathering data from holes drilled deep into the Earth’s crust?
5. All above can be addressed by physical climate science.
How will GHG vary?
F=Pxgxexfxd
•
•
•
•
F = total GHG emission rate
P = population size (global and/or national)
g = per capita gross world/domestic capital
e = energy use per $ of gross world/national
product
• f = GHG emissions per unit energy use
• d = deforestation effects
Ensemble Projections of Global Temperature
for Various Emission Scenarios
Future
Projections
Verses
Forecasts
Source: UCAR/NCAR
Climate Models help with…
DETECTION - Is the planet’s climate changing
significantly?
ATTRIBUTION – If so, what is causing the change?
Nature? Human Actions? Both?
PROJECTION – What does the future hold for
Earth’s climate?
Models in the Standards
Next Generation Science Standards
Greenhouse Effect Review
 CO2 absorbs heat in
the atmosphere
 When heat
accumulates in the
Earth system, the
average global
temperature rises
Increased CO2 & the Greenhouse Effect
 When the amount of carbon dioxide in the atmosphere increases,
average global temperature rises.
 Longwave radiation emitted by CO2 is absorbed by the surface,
so average global temperature rises
Emissions ->
More CO2 in Air ->
Higher Temperature
18°
15°
Climate Sensitivity - definition
Whenever the amount of carbon dioxide in the atmosphere
doubles, average global temperature rises by 3 degrees
Celsius.
18°
18°
15°
15°
Learning from the Past (ice cores)
Ice age
Ice age
Ice age
Ice age
CO2 Emissions – Where are we now?
In 2013, CO2
emissions are
around 10
gigatons (GtC)
per year (10,000
million tons in
units used on
this graph)
CO2 in Atmosphere – Where are we now?
 For hundreds of thousands of years,
400
CO2 varied between 180 and 280 parts
per million, beating in time with ice
ages
350
396 ppm in 2013
300
 Since the Industrial Revolution, CO2
has risen very rapidly to about 400
ppm today
250
ice
age
ice
age
ice
age
ice
age
200
150
-400000
-300000
-200000
year
-100000
0
Math of Climate Sensitivity
When the CO2 concentration in the atmosphere doubles,
temperature rises by 3°Celsius (about 5.4°F)
Examples:
 If CO2 rises from 200 ppmv to 400 ppmv,
temperature rises 3°C
 If CO2 rises from 400 ppmv to 800 ppmv,
temperature rises 3°C
 Note: as CO2 rises from 200 to 800 ppmv
(800 = 4 x 200),
temperature rises 6°C
( = 2 x 3 degrees, not 4 x 3 degrees)
Climate Sensitivity Calculator demo
spark.ucar.edu/climate-sensitivity-calculator
Climate Sensitivity Calculator Activity
Use the calculator (previous slide) to determine the expected temperature for the
various CO2 concentrations listed in column 1 of the table above (students fill in
column 2); then have them graph.
Advanced Climate Sensitivity Math
T = T0 + S log2 (C / C0)
T : new/current temperature
T0 : reference temperature (e.g. 13.7 degrees C in 1820)
S : climate Sensitivity (3 degrees C)
C : new/current atmospheric CO2 concentration
C0 : reference atmospheric CO2 concentration (e.g. 280 ppmv in 1820)
Example:
What is new temperature if CO2 rises to 400 ppmv (from 280 ppmv)?
T = T0 + S log2 (C / C0) = 13.7 + 3 log2 (400/280) = 13.7 + 3 log2 1.43
= 13.7 + 1.54
= 15.2 degrees C
Math of CO2 Emissions and
Atmospheric Concentration
Dry air mass of atmosphere = 5.135 x 1018 kg = 5,135,000 Gigatons
CO2 currently about 599 ppm by mass (395 ppmv) = 0.0599%
CO2 current mass = 0.0599% x 5,135,000 Gt = 3,076 Gt
CO2 current emissions = 9.5 GtC/year
(16 + 12 + 16) / 12
Atmospheric fraction = 45%
= 44/12 = 3.67
M = M0 + [0.45 x (3.67 x m)]
= 3,076 GtCO2 + [0.45 x (3.67 x 9.5 GtC/yr)]
= 3,076 + 15.7 GtCO2 = 3,092 GtCO2
GtC vs GtCO2
CO2 concentration = 3,092/5,135,000 = 602 ppm by mass
CO2 concentration = (602/599) x 395 ppmv = 397 ppmv
Poll: Rising Emissions
?
A
?
B
?
C
Poll: Rising Emissions
?
A
?
B
?
C
Poll: Emissions rise then
steady
?
A
?
B
?
C
Poll: Emissions rise then
fall
?
A
?
B
?
C
Very Simple Climate Model demo
spark.ucar.edu/simple-climate-model
Why does temperature continue to rise
as emission rate declines?
Atmosphere
CO2
Emissions
CO2 in Atmosphere
CO2 Removal by
Oceans & Plants
spark.ucar.edu/climate-bathtub-model-animations-flow-rate-rises-falls
spark.ucar.edu/imagecontent/carbon-cycle-diagram-doe
Contact Us
Teri Eastburn
eastburn@ucar.edu
303.497.1000
Randy Russell
rrussell@ucar.edu
303.497.1000