The cloud resolving storm simulator
Download
1 / 70

The Cloud Resolving Storm Simulator: Large-scale Parallel Computations - PowerPoint PPT Presentation


  • 323 Views
  • Uploaded on

Asia-Pacific Advanced Network. 27 August 2003. The Cloud Resolving Storm Simulator: Large-scale Parallel Computations. HPC application to Meteorology in Japan. TSUBOKI Kazuhisa SAKAKIBARA Atsushi. Hydrospheric Atmospheric Research Center, Nagoya University, Japan. OUTLINE.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about 'The Cloud Resolving Storm Simulator: Large-scale Parallel Computations' - Angelica


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
The cloud resolving storm simulator large scale parallel computations l.jpg

Asia-Pacific Advanced Network

27 August 2003

The Cloud Resolving Storm Simulator:Large-scale Parallel Computations

HPC application to Meteorology in Japan

TSUBOKI Kazuhisa

SAKAKIBARA Atsushi

HydrosphericAtmospheric Research Center, Nagoya University, Japan


Slide3 l.jpg

OUTLINE

  • Description of the numerical model: the Cloud Resolving Storm Simulator(CReSS)

  • Numrical Experiment of tornado within supercell

  • Numerical Experiment of typhoon

  • Numerical experiment of snow cloud over the sea

  • Summary


Introduction l.jpg
Introduction

  • Convective clouds and storms are highly complicated

    systems of flows and hydro-meteors. Their structure

    and dynamics are determined by a nonlinear interaction

    between the fluid dynamics and the cloud microphysics.

  • In order to simulate an evolution of a convective cloud

    storm, it is essential to formulate cloud physical

    processes as well as the fluid dynamic and thermo-

    dynamic processes.


Slide5 l.jpg


Purposes l.jpg
Purposes

  • A thunderstorm produces many types of severe weather: heavy rain, hail storm, downburst, tornado and so on.

  • The purposes of this study are to develop a cloud

    resolving model, the Cloud Resolving Storm Simulator

    “CReSS” and its parallel computing to simulate cloud-

    scale to storm-scale phenomena.

  • The simulation of the thunderstorm will clarify the

    characteristics of dynamics and evolution and

    contribute to the storm-scale prediction.


Targets of cress are l.jpg
Targets of CReSS are ...

to perform large-scale numerical experiments of mesoscale storms and prediction experiments of real severe weather systems

  • within a very large domain, ~500 km × 500 km

  • with very high resolution, 1 km ~10 m in horizontal

  • with detailed cloud microphysics and

  • with optimization for high performance parallel

    computers (including the Earth Simulator)


Characteristics of cress l.jpg
Characteristics of CReSS

  • CReSS is formulated in the non-hydrostatic and

    compressible equation system.

  • Coordinate system is a terrain-following in atwo or

    three dimensional geometry.

  • Prognostic variables are:

    • three-dimensional velocity components

    • perturbation of pressure

    • perturbation of potential temperature

    • subgrid-scale turbulent kinetic energy, TKE

    • mixing ratios for water vapor and several types of

      hydrometeors


Slide9 l.jpg

  • Finite difference method for the spatial discretization.

  • Time integration is mode splitting method.

    • Large time step: the leap-frog with the Asselin time filter.

    • Small time step : explicit both in horizontal and vertical or explicit in horizontal and implicit in vertical.

  • Advection is a fourth order scheme.

  • Turbulence is the first order closure and 1.5 order

    closure with TKE.

  • Numerical smoothing is the second or forth order

    computational mixing.


Slide10 l.jpg


Slide12 l.jpg

  • Initial conditions are:

    • horizontally uniform field from upper air sounding or

      theoretical function

    • three-dimensionally inhomogeneous data

  • Lateral boundary conditions are:

    • rigid wall, periodic, zero normal gradient

    • wave-radiating

    • externally forced time-dependent condition

  • Ground model and surface processes have been

    implemented.

  • Parallel processing is performed by the Message Passing

    Interface, MPI.


Slide13 l.jpg

Decomposition and communication


Overview of the tornado tatsumaki l.jpg
Overview of the tornado (TATSUMAKI)

The tornado occurred on 24

September 1999 was one of the most

Intense tornadoes in Japan. It

occurred during day time. They

took many pictures and videos.

These are samples of the pictures

of the tornado. A dark funnel

reached to the ground from the

cloud. The horizontal diameter of

the tornado was about 500 m.


Doppler radar observation of the tornado producing super cell l.jpg
Doppler radar observation of the tornado-producing super-cell

  • Horizontal displays of radar echo of the

    super-cell with a trace of the tornado

    (tatsumaki) observed on 24 September 1999

    in the Tokai District, Japan. The tornado

    occurred at the central part of the a hook-

    shaped echo.

  • Vertical cross-section of radar echo of the

    super-cell. A vault structure is significant.

  • Horizontal display of Doppler velocity. A

    signature of mesoscale cyclone was observed.


Experimental design of simulation supercell and tornado l.jpg
Experimental design of simulation- Supercell and Tornado -

  • domain 48km × 48km × 12km

  • horizontal grid size 75m

  • vertical grid size 25 ~200m

  • grid numbers 603 × 603 × 63

  • integration time 4 hours

  • time increment large: 0.5s, small: 0.1s

  • microphysics the bulk cold rain type

  • initial condition Shionomisaki soundingdata at 09JST, 24 Sept.

  • initial disturbance warm bubble

  • boundary condition the wave-radiating type

  • platform SR8000, 8nodes


The 2 dimensional horizontal display of the simulated super cell l.jpg
The 2-dimensional horizontal display ofthe simulated super-cell

CReSS successfully simulated the

super-cell observed in the Tokai

District. The simulated super-cell

was in quasi-steady state around

1.5 hours after the initial time.

The 2-dimensional horizontal display will show its evolution and

movement. The color level is the mixing ratio of precipitation, the

thick red lines are intense vorticity and white lines are vertical

velocity at a height of 1000 m. The thick blue lines indicate the

surface gust front. Arrows are the surface horizontal velocity.


Close view of the simulated tornado in horizontal l.jpg
Close view of the simulated tornado in horizontal

  • Vertical vorticity at a height of

    100m (a) when the tornado was

    generated and (b) in the mature

    stage.

  • Pressure perturbation at a

    height of 100m (a) when the

    tornado was generated and (b)

    in the mature stage.

The velocity field is in the cyclostrophic balance with

the pressure field.


Close view of the simulated tornado in vertical l.jpg
Close view of the simulated tornado in vertical

  • Vertical vorticity and pressure

    perturbation.

  • Vertical vorticity and vertical

    velocity.


The 3 dimensional movies of the tornado l.jpg
The 3-dimensional movies of the tornado

Since the tornado within the super-

cell has a highly 3-dimensional

structure, we made the 3-dimensional

movies of the tornado obtained from

the simulation experiment.

The movies shows vertical vorticity

larger than 0.15 /s and the surface

temperature perturbation within the

box of 13 km in horizontal and

12 km in vertical. The period of

display is about 2 hours.


The 2 dimensional movie of vertical vorticity l.jpg
The 2-dimensional movie of vertical vorticity

In order to examine a detailed process of the development of the

tornado, we will show a 2-dim. movies of vertical vorticity which

focus on the stage of the tornado genesis. The display is a

projection of the tornado on the west-east vertical plane.

The movies shows that the

tornado developed from the lower

level to the upper level.



Experimental design of typhoon 0111 l.jpg
Experimental Design of - Typhoon 0111 -

  • domain 400km × 400km × 14km

  • horizontal grid size 4000m

  • vertical grid size 400m

  • grid numbers 103 × 103 × 35

  • integraion time 12 hours

  • time increment large: 3s, small: 1.5s

  • topographyincluded

  • surface processincluded

  • sea surface temp.contant at 300.16 K

  • micropysics the bulk cold rain type

  • initial condition JMA RSM analysis at 03Z Aug.05 2001

  • boundary condition JMA RSM analysis

  • platform VPP5000, 1node


The 3d animation of typhoon 0111 l.jpg
The 3D animation of Typhoon 0111

The simulation of Typhoon 0111 is

visualized in 3-dims. All calculation

domain is displayed. The color levels

show surface pressure perturbation.

The brightly white parts indicate

clouds and gray parts are rain.

The animation shows the typhoon which was located to the

west of the domain at the initial time moved into the domain

smoothly through the west boundary.


Experiment of snow cloud bands over the japan sea the east sea l.jpg
Experiment ofSnow Cloud Bands over the Japan Sea(the East Sea)


Observation of snow clouds l.jpg
Observation of Snow Clouds

Visible satellite image (GMS) at

15 JST

JMA radar at a height of 2 km


Experimental design of snow cloud bands japan sea case on 5 january 2003 l.jpg
Experimental Design of Snow Cloud Bands- Japan Sea Case on 5 January 2003 -

  • domain 1080km × 720km × 10km

  • horizontal grid size 2000m

  • vertical grid size 300m

  • grid numbers 543 × 363 × 35

  • integraion time 18 hours

  • time increment large: 3s, small: 1s

  • topographyincluded

  • surface processincluded

  • sea surface temp.NCEP 1 week average

  • micropysics the bulk cold rain type

  • initial condition JMA RSM at 00Z Jan.05 2003

  • boundary condition JMA RSM

  • platform SR8000, 8node


Result of the simulation l.jpg
Result of the Simulation

Animation of horizontal display of

cloud mixing ratio (g/kg) and the

horizontal wind vectors at a height of

1350 m.

Animation of horizontal display of

precipitation mixing ratio (g/kg) and

horizontal wind vectors at a height of

450m.


Slide29 l.jpg

Animation of horizontal display of

vorticity (10e-3/s) and the horizontal

wind vectors at a height of 1350 m.

Animation of horizontal display of

vertical velocity (m/s) and the

horizontal wind vectors at a height of

1350 m.


Summary l.jpg
Summary

  • We are developing the Cloud Resolving Storm

    Simulator, CReSSfor simulations of cloud-scale to

    storm-scale phenomena.

  • Parallel computing is essentially important for this

    type of large simulations.

  • CReSS characteristic features and some result obtained

    in dry experiments were presented.

  • The result suggested that CReSS is capable to simulate a

    thunderstorm and a related phenomenon.


Slide31 l.jpg

  • The result of the simulation experiment of the supercell

    and the associated tornado was presented. CReSS

    simulated both disturbances with a uniformly fine grid

    spacing.

  • Prediction experiment of snow cloud over the sea showed that the

    detailed structure of roll convections and mesoscale disturbances of snow clouds.

  • The result suggested that CReSS is capable to simulate a

    thunderstorm and a related phenomenon.


Future plans of cress l.jpg
Future Plans of CReSS

  • Detailed cloud microphysical processes which resolve size distributions of hydrometeors

  • Parameterization of turbulence

  • Nesting in a coarse-grid model

  • Data assimilation of Doppler radar

  • Radiation of cloud



Summary35 l.jpg
Summary

  • We are developing CReSSfor simulations of clouds and storms with parallel computation.

  • CReSS simulated the tornado within the supercell with the uniformly fine grid spacing (75m) in the large domain.

  • The simulation shows that successive formation of tornado vortex within the maximum updraft of the supercell.

  • The tornado develops from the surface and extends to the upper level.

  • We infer that the tornado is caused by stretching of the lower-level vortex resulted from shear instability.


Experiment of snow cloud bands over the labrador sea l.jpg
Experiment ofSnow Cloud Bands over the Labrador Sea


Experimental design of snow cloud bands labrador sea case on 8 february 1997 l.jpg
Experimental Design of Snow Cloud Bands- Labrador Sea Case on 8 February 1997 -

  • domain 250km × 120km × 12km

  • horizontal grid size 500m

  • vertical grid size 200m

  • grid numbers 503 × 243 × 63

  • integraion time 6 hours

  • time increment large: 2s, small: 0.5s

  • topographyincluded

  • surface processincluded

  • sea surface temp.constatnt at 3 C

  • micropysics the bulk cold rain type

  • initial condition NCEP analysis

  • boundary condition NCEP analysis

  • platform SX-6, 4node


The result of the simulation l.jpg
The Result of the Simulation

Horizontal displays at height of 500m

  • vertical velocity

  • cloud mixing ratio (g/kg)

  • precipitation mixing ratio (g/kg)

    Vertical cross-sections in zonal direction

  • vertical velocity

  • cloud mixing ratio (g/kg)

  • precipitation mixing ratio (g/kg)

  • zonal velocity (m/s)

  • pressure (hPa)

  • mixing ratio of water vapor (g/kg)