Contribution to a new NSF Science
and Technology Center (top)
New Science & Technology Center Proposal
David Randall (Colorado State University) is the Principal
Investigator on a proposal to NSF for a new Science & Technology
Center (STC). The preliminary proposal survived the
first round of the selection process and a full proposal
was invited. A key part of the proposal is the super-parameterization
approach, which was initially developed in MMM. Wojciech
Grabowski, Andrew Heymsfield, Chin-Hoh Moeng, and Mitchell
Moncrieff contributed
to
the science
plan of the proposed STC on Multi-Scale Modeling
of Atmospheric Processes. Specifically, they were
invited
to community-wide workshops in Fort Collins (October
2002) and Kauai (May 2003) where presentations and
subsequent discussion set the scene for the scientific
content of the proposal. Another workshop is planned
for late 2003 in Fort Collins to work on detailed
content of the full proposal due February 2004.
Cloud systems
on short (diurnal) time scales (top)
Dynamical model of traveling warm-season organized
precipitation over the U.S. continent
Mitchell Moncrieff used
one of his nonlinear dynamical models of precipitating
deep convection in shear to
explain, in the simplest possible way, the travel of
sequences of convection observed by Richard
Carbone and
collaborators and simulated in two spatial dimensions
by Changhai Liu and Moncrieff (ASR
2002). Such simplification is essential if organized
convection is ultimately
to be parameterized in large-scale models. The model
of the steering level regime of convective organization
captures the observed 14 m/s travel speed of the precipitation
sequences (i.e., 7 km or 300 mb-steering level). This
suggests the sequences have a simple dynamical basis,
which is a fundamental requirement if they are ultimately
to
be parameterized. Hypotheses are now being formulated
for evaluation against using the three-dimensional
simulations reported below.
Cloud-resolving simulations of warm season precipitation
sequences
Liu and Moncrieff extended their cloud-resolving simulations
of warm season convection to three spatial dimensions.
The model used longitudinally uniform orography and
land-use conditions averaged from 34N to 40N latitude.
It was forced with the diurnally-varying boundary conditions
and large-scale forcing derived from an MM5 ensemble
prediction. The radar-derived statistics by Carbone and
collaborators were realized approximately. Typically,
isolated convection that developed over the Rockies
in late morning subsequently aggregated into several
mesoscale cloud clusters that traveled onto the Plains
during the evening. The meridionally
averaged time-space diagram of the distribution of
precipitation was broadly comparable to its two-dimensional
counterpart.
Daytime convective development over land
The diurnal cycle of precipitating convection is typically
poorly represented in weather prediction and climate
models that rely on convective and boundary-layer parameterizations.
In such models, convection often transitions too quickly
from shallow to deep, and surface rainfall peaks a
few hours too early. Grabowski performed
idealized numerical simulations based on observations
of convective
development over the TRMM/LBA (Tropical Rainfall Measuring
Mission/Large-scale Biosphere-Atmosphere Experiment)
field project in Rondonia, Brazil in February 1999.
Field observations and an ensemble of benchmark high-resolution
simulations are being compared with cloud-system-resolving
models applied in two- and three-dimensional configurations
and with simple models applying convective and boundary-layer
parameterizations (single-column models). This project
is a contribution to the GEWEX (Global Energy and Water
Cycle Experiment) Cloud System Study (GCSS) Working
Group 4 (Precipitating Convective Cloud Systems, chaired
by Wojciech Grabowski),
to improve the representation of deep precipitating
convection
in weather prediction and
climate models.
The set of numerical simulatuions representing benchmark
for model intercomparison is described at http://box.mmm.ucar.edu/gcss-wg4/model/intercomp.html
Explicit simulations of warm season precipitation
sequences and comparisons
with radar observations
Liu, Moncrieff,
John Tuttle, and Carbone compared
the explicit ten-day (20-30 July 1998) simulation
of US warm season convection
performed using
a triply-nested version of MM5 to the radar observations
and the nine-km grid-resolution simulations applying
cumulus parameterizations. The inner domain covers
an area of 2000 km x 1200 km (approximately from 110W
to 88W and from 34N to 45N latitude) with a three-km
grid increment. The Kain-Fritsch convective parameterization
was applied in the outer domain, but convection was
explicit in the middle and inner domains. The observed
convective generation over the Rockies and subsequent
eastward propagation were qualitatively reproduced
during most of the ten-day period. On average, the
propagation speed, zonal span and duration of rainfall
streaks are slightly under-predicted. Another deficiency
is that the simulated rainfall intensity is too weak
in comparison with the WSR-88D estimate. In spite of
these discrepancies, the explicit simulation is far
more realistic than the coarse-grid simulations that
apply convective parameterizations.
|
|
| Figure
22: The
diurnal rain-rate Hovmöller diagrams averaged
over the
20-29 July 1998 period. The upper and lower panel
correspond to the 3-km-resolution explicit simulation
using MM5 and the radar observations, respectively.
The shading scale is in the unit of mm h-1. |
Convective cloud systems in the Indian monsoon
Someshwar Das (visitor from the National Center for
Medium Range Weather Forecasting, India), Moncrieff and Liu began modeling and comparing convective cloud
systems in the Indian and North American monsoon season.
The research objective is to examine commonalities
and differences in the large-scale environment, organization,
structure, and evolution behavior of mesoscale convective
systems in the two geographic locations. A ten-day
ten-km-resolution simulation of an Indian monsoon depression
and associated convection during 18-28 June 2002 was
conducted using a triply-nested MM5 configuration.
Diagnosis of these results and a 2-km cloud-resolving
simulation are ongoing.
|
|
| Figure
23: Right, an infra-red image of a major multiscale
cloud
cluster
during
the onset of the 1992
summer monsoon in India. Left, a numerical simulation
using MM5 with parameterized
convection at 30-km grid-resolution. |
Effects of tropospheric moisture and stratification
on the mode of cumulus convection
Tetsuya Takemi (Osaka University, Japan) and Liu investigated
the effects of temperature and moisture distributions
on the trimodal characteristics of tropical convection
(i.e., shallow cumulus, cumulonimbus, and cumulus congestus)
by analysis of observational data over the tropical
western Pacific during 1999-2001 and idealized cloud-resolving
numerical experiments. The observational analysis revealed
that shallow cumulus was prevalent
in the extreme dry episodes and a pronounced peak for
cumulonimbus clouds was identified in the rainy periods
while congestus clouds with tops between four and eight
km were abundant in the moderate dry (little rain)
and rainy episodes. They found that the development
of cumulus cloud types was more relevant to tropospheric
moisture profiles than temperature profiles. Cloud-resolving
simulations suggested that dry and stable layers in
the middle troposphere inhibited cloud growth through
dry-air entrainment and cloud detrainment enhancement
respectively and were responsible for the observed
abundant populations of cumulus congestus.
Small-scale turbulent mixing in convective clouds
In collaboration with Miroslaw Andrejczuk and Szymon
Malinowski (both from Warsaw University, Poland), Grabowski and Piotr
Smolarkiewicz extended a modeling study of
decaying moist turbulence reported last year. This
problem is important for radiative transfer through
clouds, initiation of precipitation in warm (i.e.,
ice-free) clouds, and parameterization of small-scale
and microscale processes in larger-scale models. In
the moist case, kinetic energy of small-scale motion
originates not only from the classical downscale energy
cascade, but is also generated/enhanced. The new
set of simulations showed that previously reported
results obtained in preliminary low-spatial resolution
simulations are confirmed by high-resolution (direct
numerical simulation type) simulations with improved
representation of cloud microphysics. In addition,
a range of initial conditions further extended the
previous findings. This work validated the conclusions
of the preliminary study and improved the accuracy
of the predictions.
|
|
| Figure
24: Results from numerical simulations of cloud/clear
air interfacial
mixing
in a decaying moist turbulence setup. The volume
considered is about
0.3 cubic meter and half of the volume is initially
filled by cloudy filaments.
Detailed representation of cloud microphysics with
16 classes of cloud droplets is applied. Panels
show the evolutions
of the dissipation rate of the turbulent kinetic
energy
(TKE) derived
either from the evolution of the volume-integrated
TKE
(red curves marked
dTKE) or from the theoretical prediction based
on the evolution of
the volume-integrated enstrophy (blue curves).
The upper, middle,
and lower panels represent results from simulations
using 643 grid points. The three simulations
illustrate transition from LES-type
simulation in the 643 case (where TKE
dissipation occurs to a large
degree through the model numerics) toward DNS-type
simulation in the 643 case,
where the TKE dissipation is resolved. |
Effects of turbulence on the collision rate of cloud
droplets
Lian-Ping Wang (University of Delaware) and Grabowski (who
is an Adjunct Professor at Delaware), continued their
investigation of the effects of turbulence on
the collision of cloud droplets when droplet inertia,
gravity, and turbulence microstructure are all considered.
This is an important problem because the impact of
cloud turbulence on microphysical processes (warm rain
initiation in particular) remains ambiguous. Previous
results suggested that interaction between cloud droplets
and turbulent flow could significantly enhance geometric
collision kernel. Work focuses on the development of
computational techniques to address the impact of small-scale
turbulence on collision efficiency between colliding
droplets and on the development of accurate techniques
to solve the stochastic collection equation (SCE).
These two projects involve graduate students Orlando
Ayala and Yan Xue (University of Delaware)
Convectively
generated tropical ice clouds (top)
Interactions among cloud microphysics, surface processes,
and radiative
transfer in subtropical shallow convection
Using a cloud-resolving model, Wojciech
Grabowski and
Gregory McFarquhar (University of Illinois, Urbana-Champaign)
examined indirect and semi-direct effects of aerosols
on the local water and energy budget by modeling how
aerosols affect the macrophysical (e.g., cloud cover)
and microphysical properties (e.g., liquid water paths)
of trade wind cumuli, diurnal cycle, and cloud radiative
forcing. Simulations are initialized using vertical
profiles of temperature, moisture, and velocities measured
during the Indian Ocean Experiment (INDOEX) and surface
fluxes estimated using sea surface temperatures estimated
from the microwave imager (TMI) of the Tropical Rainfall
Measuring Mission (TRMM) satellite. Aerosol properties
are determined using estimates of their microphysical
and optical properties obtained during INDOEX.
The
dependence of the modeled properties on surface
fluxes of heat and moisture, on aerosol absorptive
properties,
on cloud condensation nuclei concentrations, and
on the vertical distribution of aerosols is being
investigated and evaluated by examining vertical profiles
of radiative
heating induced by aerosols.
Hurricane ice microphysics
Andrew Heymsfield and Aaron
Bansemer, along with S.L.
Durden
and T. Paul Bui (both of NASA/Ames Research Center),
used measurements from the NASA DC-8 collected in
Hurricane
Humberto
(during the 2001 Convection And Moisture EXperiment
(CAMEX-4) to study microphysical properties. The
measurements assessed ice growth processes
between the 0 and –50C levels. They combined
the microphysical data with coincident, nadir-viewing,
multi-wavelength Doppler radar measurements. They developed
a conceptual model of Humberto ice microphysics.
|
|
| Figure
25: Conceptual model of Humberto ice microphysics shows
high concentrations of small ice particles within
and around the updrafts at temperatures near -40oC
and below. |
High concentrations
of small ice particles were observed within and around
the updrafts at temperatures near -40oC and below.
The origin of these particles is homogeneous ice nucleation.
Aggregates (some larger than seven mm) dominated the
larger sizes and were attributed to the growth of particles
heterogeneously nucleated in the updrafts. Aggregation,
which began in the hurricane eye-wall, continued from
upper levels to the melting layer. Rain in the lower
regions extended up to the six-seven km levels, with
graupel above. The authors fitted gamma-type
size distributions to the particle size distribution
measurements. The slope of the fitted exponential size
distribution was distinctly different close to the
eye compared to outside that region.
Cloud systems on
long time scales (top)
Impact of free-tropospheric moisture on the large-scale
organization of tropical convection
Wojciech Grabowski and Mitchell
Moncrieff investigated
interactions between equatorially-trapped disturbances
and tropical convection using a global model that
applies either super-parameterization or the Emanuel
convection parameterization. The modeling setup was
a constant-SST aquaplanet maintained in radiative-convective
quasi-equilibrium. With the super-parameterization,
robust coherent structures occurred with deep convection
at the leading edge and strong surface westerly winds
to the west (westerly wind bursts), resembling the
Madden-Julian Oscillation (MJO). The coupling among
deep convection, free-tropospheric moisture, and
the large-scale flow (the moisture-convection feedback)
was found essential for MJO-like structures. When
large-scale fluctuations of convectively-generated
free-tropospheric moisture were removed on a time
scale of a few hours, the MJO-like systems did not
develop and, if already present, disintegrated rapidly.
Weak MJO-like structures developed in simulations
applying the standard Emanuel parameterization. The
MJO dramatically increased in strength when the amount
of precipitation in the parameterization falling
outside clouds was doubled. This study highlights
the key role of free-tropospheric moisture in large-scale
organization of tropical convection, which may explain
why large-scale models struggle with the MJO.
|
|
| Figure
26: Results from the constant-SST aquaplanet
simulation using Emanuel convective parameterization
in its standard (left panels) and modified (right
panels)
configurations. Upper panels show Hovmöller
diagrams of the surface precipitation at the equator.
Lower
panels show spatial distributions of surface precipitation
and surface zonal winds at day 60 (left) and day
50 (right). Precipitation rate larger than 1.5
and 15 mm day-1 is shown using light
and dark shading, respectively. Zonal winds are
shown using solid and dashed contours for positive
and negative values, respectively, with contour
interval of 10 ms-1 starting from 5
ms-1. |
Analytic representation of the multiscale organization
of tropical convection
Moncrieff quantified
the pivotal role of precipitating convection organized
on mesoscales in tropical intraseasonal
dynamics by a nonlinear theoretical-dynamical model.
Two interlocked circulations are the building blocks;
one represents organized convection in the vertical
plane and the other a two-layer large-scale open gyre.
The gyre in the lower layer is Rossby-like, whereas
in the upper layer it is driven by outflow from organized
convection. This property distinguishes MJO-like systems
from convectively-coupled systems. Despite its extreme
simplification, the archetypal approximation of the
general formulation represents morphology, momentum
transport, and equatorial super-rotation associated
with tropical cloud systems realized by Grabowski's (2001) super-parameterization. Transport of zonal momentum
in the vertical and meridional directions
is shown to be a key process. Implications for the
parameterization of organized convection are discussed.
|
|
| Figure
27: Nonlinear theoretical-dynamical model of mature
MJO-like
systems based on
coupling an analytic mesoscale parameterization
of organized convection
(Moncrieff 1992) to a two-layer large-scale equatorial
beta-plane model. Red trajectories
represent the mesoscale parameterization in the
vertical
plane at the Equator.
Yellow trajectories represent a Rossby-gyre-like
gyre
(rotationally-balanced
dynamics) in the lower layer. Blue trajectories
are driven by organized outflow from
the mesoscale parameterization (divergent flow
balance) in the upper layer.
The theory verified against Grabowski's (2001)
super-parameterization in
regard to the meridional and vertical transports
of zonal momentum as well
as equatorial super-rotation. |
Diagnostic analysis of multiscale processes of U.S.
warm season precipitation.
The temporal variability of the warm-season precipitation
over North America was investigated by Hsiao-ming Hsu
(RAP), Moncrieff,
Wen-Wen Tung (ASP), and Changhai
Liu by
using the eight-year (1996 to 2003) composite rainfall
data
derived from WSR-88D. Richard
Carbone et
al. (2002) identified the diurnal pattern of rain sequences
based on this
dataset. Applying discrete Fourier transform and continuous
wavelet transform to the rain-rate time series, early
results indicate that a -1.3 scaling law appears over
the frequencies higher than the diurnal frequency,
while over lower frequencies the variance varies strongly
from year to year. General structure-function analysis
likewise suggests such a scaling law.
|
|
| Figure
28: Wavelet-based
statistical analysis of sequences of precipitation
over the
US continent during the warm season for an 8-year
period. High-frequency
behavior (< 1-day) is punctuated
by a remarkably consistent-4/3 slope. Low-frequency
behavior shows significant interannual variability.
The diurnal cycle of precipitation shows up as
a pronounced
peak. |
Simulation of the Intertropical Convergence Zone:
effects of sea surface
temperature gradients
Liu and Moncrieff extended
their work on explicit simulations of the intertropical
convergence zone (ITCZ)
by adding sea surface temperature (SST) gradients on
an aqua-planet. A series of 100-day two-dimensional
simulations on an equatorial beta plane were performed,
corresponding to various latitudes of the maximum SST.
The meridional convective distribution
and the ITCZ structure are strongly modulated by the
prescribed SST perturbations. First, the convective
activity peaks near the maximum SST in the quasi-equilibrium
state. Second, a single ITCZ occurs when the maximum
SST is near the equator. In contrast, a double ITCZ
occurs when the maximum SST is displaced more than
ten degrees away from the equator. Third, the strongest
ITCZ occurs when the SST maximizes about 1,500 km from
the equator or at the equator. The dependence of the
ITCZ intensity on the maximum SST position has important
implications on the observed discontinuous latitudinal
migration of the ITCZ (i.e., monsoon onset and retreat).
Tropical convection, sea surface temperature, and
cloud-interactive radiation
Sea surface temperature (SST) forcing has a strong
influence on observed tropical convection, but there
is considerable variation in the relationship of convection
to the underlying SST distribution due to complex atmospheric
interactions. In order to unambiguously quantify the
correlation between the spatial structure of tropical
convection and the spatial distribution of SST, cloud-resolving
two-dimensional simulation of convection on an f-plane
was conducted by Liu and Moncrieff. Early results reveal
that the strongest convective activity often occurs
at a few hundred kilometers (typically, 200-400 km)
from the center of warm pools. This is an alternative
interpretation for the observations that peak convection
is commonly located several degrees of latitude toward
the equator of the maximum SST in some tropical regions.
The maximum convection is further displaced by several
hundred more kilometers from the warm pool if the radiative
heating gradients between cloudy and clear regions
are excluded.
Parameterization of deep convection (top)
Heavy precipitation events in India
Someshwar Das (visitor
from NCMKWF, India), Changhai
Liu, Mitchell Moncrieff, and Jimy
Dudhia collaborated
with L. Prabhawati and K. Sowjanya (both Andhra University,
India) in an investigation of heavy precipitation
episodes over
the west coast
of India using MM5, run
on real-time operational basis over India. The initial
and lateral boundary conditions were specified from
the operational global T80 model of the National Center
for Medium Range Weather Forecasting (NCMRWF). Experiments
were carried out to quantify the sensitivity of heavy
rainfall simulations to different convection and cloud
microphysics schemes at 30- and 10-km resolutions.
Results indicated that at these resolutions, the rainfall
forecasts are less sensitive to the varying convection
and microphysics parameterizations, but it is very
important to have correct initial conditions and environment,
emphasizing the importance of continuous FDDA and/or
3DVAR. Early results were presented at the IUGG conference
held at Sapporo, Japan during 1-7 July 2003.
Improved parameterizations of microphysics and the
planetary boundary layer
Dudhia worked with
visitors Song-You Hong and Jeong-Ock Lim (both of Yonsei
University, South Korea) on a microphysics
parameterization for WRF that has a new way to represent
ice crystal concentrations in a single-moment scheme.
The scheme development is ongoing, using results from
field programs provided by Andrew
Heymsfield with
regard to parameterizing ice and snow properties in
a bulk
sense as a function of temperature (also noted in
WRF Model Physics). In addition, Song-You
Hong continued development of
a
new planetary boundary layer scheme for WRF that explicitly
represents PBL top entrainment processes. Preliminary
implementations in WRF for real-time forecast efforts
shows some improvement compared to the popular medium
range forecast (MRF) scheme in MM5 and WRF. In future
the new scheme will likely replace the MRF scheme in
WRF.
Implementation of super-parameterization in the Community
Climate System Model (CCSM)
Michal Ziemianski (postdoctoral
fellow visiting from the Institute of Meteorology and
Water Management,
Poland), in collaboration with Wojciech
Grabowski, Moncrieff, and William
Collins (CGD), continued his investigation
of cloud-radiative interactions over the tropical western
Pacific warm pool using super-parameterization and
the Community Atmospheric Model (CAM) of the Community
Climate System Model (CCSM). Uncoupled simulations,
where the super-parameterization was driven off-line
by CAM tendencies but no feedback was included in CAM,
illustrated uncertainties with CAM's standard cloud
and convection parameterizations. These aspects were
further highlighted in coupled simulations, where CAM's
deep convection parameterization was replaced by super-parameterization.
This resulted in a significant improvement in the temporal
and spatial variability of warm pool convection as
well as the tropopause height. In particular, the diurnal
cycle of convective precipitation had realistic phase
and amplitude when compared to observations. Additionally,
the coupled experiment realized more realistic structures
of the ITCZ and the large-scale organization of convection
in MJO-like systems
|
|
| Figure
29: Investigation
of cloud-radiative interactions over the tropical western
Pacific warm pool using super-parameterization and
the Community Atmospheric Model (CAM) of the Community
Climate System Model (CCSM). Uncoupled simulations,
where the super-parameterization was driven off-line
by CAM tendencies but no feedback was included in CAM,
illustrated uncertainties with CAM's standard cloud
and convection parameterizations. These aspects were
further highlighted in coupled simulations, where CAM's
deep convection parameterization was replaced by super-parameterization.
This resulted in a significant improvement in the temporal
and spatial variability of warm pool convection as
well as the tropopause height. |
An improved framework for super-parameterization
Grabowski improved his super-parameterization approach.
A physically-based coupling of horizontal momentum
between the large-scale and cloud-scale models replaced
the relaxation method applied previously. A new time-stepping
algorithm for the coupled system was developed in which
large-scale and cloud-scale model calculations were
separated, improving the computational efficiency.
The small-scale model can be aligned in any direction,
the orientation can be different in different columns,
and it can change during a simulation. The improved
approach was applied to the large-scale organization
of moist convection on a constant-SST aqua-planet studied
previously using the original super-parameterization.
The results show a strong MJO-like system, in agreement
with the previous formulation. Inclusion of surface
friction, not considered in the previous formulation,
reduces the strength of westerly wind burst west of
the maximum surface precipitation to the level comparable
to real-world MJOs.
|
