CLOUD in CLIMATE MODELS

1
CLOUD in CLIMATE MODELS

• Chin-Hoh Moeng
• MMM/NCAR/UCAR
• CMMAP

2
Outline

• My journey into the atmospheric sciences
• My research interests/responsibilities
• Cloud in climate models at CMMAP
• Center for Multi-scale Modeling of
  Atmospheric Processes

3
Journey

• BS degree, Taiwan 1972
• MS, South Dakoda School of MT 1974
• Ph.D, UCLA 1979
• Postdoc at NASA Goddard 1980-1982
• NCAR scientist since 1982
• Deputy Director of CMMAP since 2006

4
Research interests

• Turbulence in the
• Planetary Boundary Layer
• Clouds and their effect in climate

5
Major responsibilities

• Perform good research
• Publish good papers
• build up national
• international reputation

6
Clouds Are Central to the Earth Sciences

• Climate change
• Weather
• The water cycle
• Global chemical cycles
• The biosphere

Diff. cloud types (deep cloud, shallow cumulus,
low-level stratus, cirrus) play diff. roles !!
Slide from Randall
7
For example marine (low-level) stratocumulus
off California coast
This cloud reflects a lot of solar energy back
to space!
8
Deep cloud

• vertical transport
• latent heating
• precipitation

9
...The modeling of clouds is one of the weakest
links in the general circulation modeling
efforts. --Charney et al., National Academy
Report, 1979
...Cloud processes are most important for
determining ... temperature. Considerable
uncertainty remains this represents a
significant source of potential error in
climate simulations. --
IPCC Third Assessment (2001)
IPCC Intergovernmental Panel on Climate Change
Slide from Randall
10
Todays IPCC Models (Fourth Assessment 2007)
Zonally annually averaged reflected solar
radiation

light grey lines from models
obs.
Slide from Bjorn Stevens
11
The problem is multiple scales.
Cloud Particles
GCM grid 200 km
All clouds and turbulence are unresolved and
are parameterized.
12
All unresolved processes are represented by
single column models (or called
parameterizations) at every grid column.
200 km
200 km

A discretized globe resolved unresolved
processes

13
Clouds are complicated.
Radiation
Deep convection
Microphysics
Shallow conv.
Turbulence
Surface processes
They are crudely represented in GCMs. Progress of
climate prediction has been stalled.
14
Center for Multi-Scale Modeling of Atmospheric
Processes
led by Dave Randall


Take advantage of tera-computing to better
represent clouds in GCMs.
15
Create Multi-scale Modeling Framework (MMF)



MMF to better represent deep cloud system in GCMs
Note small cloud/turb. remain unresolved.
16
The most straightforward way to represent clouds
is to explicitly resolve them,using a global
cloud-resolving model (GCRM).

• But

17
Why not using a global cloud-resolving model
(GCRM)?

• Current climate-simulation models have
• grid columns with 200 km

• A global model with 2 km will have
  grid
• columns and about times shorter

• CPU of a GCRM is thus times larger

We are almost there! Frontier Research Center in
Japan is performing the 1st GCRM.
18
A bridge to GCRM climate simulation
GCRM climate
Super-Parameterization
Current climate models
Slide from Dave Randall
19
Super-Parameterization (Prototype MMF)
This idea was proposed and first tested by
Wojciech Grabowski.
Slide from D Randall
20
Super-Parameterization
Using super-paramerization in GCMs has shown
better results by CMMAP scientists. Searching for
improvements. Also
21
Improve the unresolved stuff

• Small clouds
• Turbulence
• Cloud microphysics
• Radiation

22
Unresolved processes in MMF
Radiation
Microphysics
Shallow conv.
Turbulence
How to include them in MMF or CRM?
23
To study relation between deep small cloudswe
performed a benchmark simulation (a virtual
cloud systm)
large/deep clouds
small clouds turb.
global scales
Super-LES resolves deep, small clouds and
turbulence!
24
? 2048 x 2048 x 256 grid points over
a 205 km x 205 km x 27 km domain. ? CMMAP
just finished the very first super-LES
ever performed. ? It used 400,000 CPU hours
on 2048 processors. ? It generated 5
tera-bytes dataset!
Super-LES of a tropical deep cloud system
A snap shot of the computer generated cloud-top
height ?
25
deep clouds
Computer-generated cloud field
N ?
shallow clouds
E?
205 km ( a GCM grid cell)
from Marat Khairoutdinov
from Marat
26
Animation Vertically integrated liquid water
amount
Total model domain a GCM grid mesh. Total
simulation time a day.
27
(No Transcript)
28
Why atmospheric sciences?

• Exciting science and not too difficult.
• Fundamental, yet relevant to society.
• Wide scales of motions to work on.
• Many disciplines to choose from.
• Modeling, observation, theory.
• Job market is never hot, but steady.