Earth system model limitations

1
Earth system model limitations
Graeme L Stephens Dept Atmospheric Sciences
Colorado State University Ft Collins, CO USA
2

• Climate system models seek to represent a range
  of complex processes.
• Models are an important tool in developing our
  understanding of the Earth climate system.
• Observations are essential for testing and
  building this understanding
• What has emerged are four levels of model
  evaluation
• comparison to observed climate change (global
  and regional)
• comparison to observed modes of variability on
  many scales
• evaluation of critical climate processes
• evaluation of abrupt system changes (such as
  shifts in weather regimes,)

Examples from IPCC FAR
3
Atmosphere-ocean general circulation climate
models (AOGCMs)
Resolved scale O(100-200km)
Intrinsic scale - Few kms ? mircoscale
Unresolved, parameterized processes largely
determine model sensitivities through feedbacks
and even determine the amount, distribution and
change in essential climate parameters - notably
precipitation.
4
Simulating observed global climate change
On the global scale agreement is impressive but
suspicions about the degree of tuning of
feedbacks linger
5
Water vapor feedback (process)
Two (related) modes to this feedback
6
Observed Elements Water vapor feedback
Clear-sky ocean obs
SST
W
Clear-sky OLR
A measure of feedback 0.0027/0.0012
2.3 ? 1.0
i.e. we have observed a strengthening of the
clear-sky Greenhouse effect - 1/3 of it is
due to CO2 increases, and 2/3 is due to water
vapor increase (a positive feedback)
7
Water vapor feedback in the FAR models

• From observation, f2.3?1.0 - model feedback
  strength is similar so it appears on the global
  scale the feedback in models is credible

AR4-1pctto2x

• Stronger feedback exists in models with amplified
  UT moistening (UT warming)

f

• The warmest models are those that exhibit greater
  UT moistening

8
The good
On the global-mean scale - model projected
changes of selected climate parameters
(temperature, water vapor) and feedbacks between
them appear to be consistent with observed
changes over the past 20-100 yrs
Global warming
sign
9
Regional patterns of climate
FAR
Agreement on this scale is less convincing -
credible for some parameters (like surface air
temperature) but not for others (like
precipitation)
10
Projected precipitation changes by FAR Models

• Increasing levels of greenhouse gaseswarm the
  climate and lead to increasesin very heavy
  precipitation events in a global mean sense

Why Global increases to precipitation are
determined by changes to the energetics of the
atmosphere (and less so by increased water vapor
in the air). Why this change occurs as an
increase in the heavier events is not understood.
11
Observed changes

• From 1908-2002
• Total annual precipitation across the contiguous
  U.S. increased 7
• Precipitation falling in very heavy daily events
  increased by 20
• Warmer climates get more rainfall in extreme
  events compared to colder climates - this is
  consistent with model projections

10
Confidence Index
0
Courtesy T Karl
12
Not so good FAR Projected regional changes
Difference maps - years 60-70 minus years 1-10
These projections suggest that the haves get more
and the have-nots get less - these patterns are
largely determined by the patterns of changing
atmospheric circulation
Change in precip (mm/day)
13
The bad FAR Projected regional changes
Agreement on the regional scale is approching a
credible level for some parameters (like surface
air temperature). For others (like precipitation)
there is still much to be understood (why more
intense, how much does large-scale circulation
control regional precip changes, etc) before
confidence can be assigned
Global Precip increases
Global intensity
14
How well is the link between dynamics and
precipitation understood modeled?
The MJO is a mode of variability on the timescale
of 30-60 days and dramatically effects global
weather The MJO encapsulates many of the
couplings between the physics (convection,
radiation, etc) and large-scale dynamics thought
important to key climate feedbacks.
15
(No Transcript)
16
850 hPa zonal wind anomaly derived from weather
analysis
17
After March 1st the model is run in free mode -
ie this is the model predicted state the MJO
disappears
18
The nature of tropical convection in AOGCMs
AOGCMs indiscriminately produce deep convection
everywhere all the time - the precipitation of
models is also heavily biased to deep convection
whereas the new observations tell us
otherwise As a consequence convection cannot
become organized by the large-scale flows of the
MJO, does not add heat to the atmosphere in the
correct way, and cannot feedback on the MJO to
sustain it.
19
The ugly
New observations are beginning to reveal the
unrealistic nature of convection in models and
hint at why fundamental modes of variability are
not predicted. The representation of the
hydrological cycle, and precipitation (and cloud)
physics specifically, is disturbingly simple
being defined by empirical, conceptual
parameterizations that do not resolve the basic
physical processes of importance.
20
Model evolution in the coming decade
In an attempt to put the important
hydrological-related processes on firmer ground,
the evolution of atmospheric models is moving
from coarse-scale (O(200km)) to global cloud
resolving models of O(few kms).
NICAM example 3.5 km on the Earth Simulator
21
Backups
22
Dynamics and the bulk transport of heat poleward
23
For the most part, AOGCMs well represent these
transports there is a significant improvement
between TAR FAR
24
Soden and Fu, 1995
UTH correlates with frequency of deep convection
(tropics) Therefore the feedback amplifier
presumably relates to convection in some way
25
(No Transcript)
26
(No Transcript)