Skamarock

Information about Skamarock

Published on September 25, 2007

Author: Nickel

Source: authorstream.com

Content

Resolving Clouds in Atmospheric Models:  Resolving Clouds in Atmospheric Models Bill Skamarock NCAR/MMM Slide2:  Clouds in the Atmosphere:  Clouds in the Atmosphere Weather: Precipitation – rain, snow, hail Wind, radiation, visibility Climate Moisture redistribution and precipitation – hydrological cycle Radiation Chemistry/Air-Quality: Chemical processing (acid rain) Ozone chemistry Transport of pollutants Wet deposition Representation of Clouds in Atmospheric Models:  Representation of Clouds in Atmospheric Models Large-scale models: h andgt; 30 km The effects of the clouds are diagnosed (parameterized) from the predicted water vapor field precipitation vertical transport and redistribution of moisture and heat radiative effects turbulence Representation of Clouds in Atmospheric Models:  Representation of Clouds in Atmospheric Models Meso-scale models: 8 km andlt; x andlt; 30 km The effects of the clouds are partially prognosed from predicted fields: water vapor, cloud water and ice, and frozen and liquid precipitation. Some portions of the cloud effects are still diagnosed (parameterized). some precipitation some vertical transport and redistribution of moisture and heat turbulence Representation of Clouds in Atmospheric Models:  Representation of Clouds in Atmospheric Models Cloud-scale models: 100 m andlt; x andlt; 8 km The effects of the clouds are entirely prognosed from predicted fields: water vapor, cloud water and ice, and frozen and liquid precipitation. Problems with Modeled Clouds:  Problems with Modeled Clouds Large-scale models (clouds completely diagnosed): Poor diagnosis of cloud type, composition, and precipitation. Clouds and cloud-systems do not know about vertical wind shear. Implications: (1) Large uncertainty in climate-model predictions (2) A key limiting factor for weather-forecast accuracy Slide8:  Mobile AL Radar Meso-/Cloud-Scale Model (WRF) Hurricane Katrina Reflectivity at Landfall 29 Aug 2005 14 Z 4 km WRF, 62 h forecast Slide9:  Realtime WRF 4 km BAMEX Forecast Composite NEXRAD Radar Reflectivity Forecast 12 h forecast Initialized 5/24/03 00Z Slide10:  Realtime WRF 4 km BAMEX Forecast Composite NEXRAD Radar Reflectivity Forecast 12 h forecast Initialized 5/24/03 00Z Vertical Velocity at z = 5 km, t = 5 h:  Vertical Velocity at z = 5 km, t = 5 h Along-line cell spacing ~ 6 to 8 Dh until Dh andlt; 500 m (cell diameter is 3 to 4 km in converged solutions) (Courtesy of G. Bryan, NCAR/MMM) Simulations using x = 4 km to x = 250 m:  Simulations using x = 4 km to x = 250 m Weak-shear case: Vertical cross-section of tracer concentration at 6 h (not a line-average). x = 4000 m x = 1000 m x = 250 m (Courtesy of G. Bryan, NCAR/MMM) Surface rain rate, weak shear:  Surface rain rate, weak shear 250 m solution close to convergence 1, 2, 4 km solutions over-predict precipitation. (Courtesy of G. Bryan, NCAR/MMM) Problems with Cloud Models:  Problems with Cloud Models Solutions do not statistically converge until h andlt; O(100 m) - turbulence problem When will our applications get there? (assume comp. speed doubles every 18 months) Climate - not in my lifetime Weather - global (state-of-the-art h ~ 25 km) 36 years (maybe in my lifetime) Weather - regional (state-of-the-art h ~ 7 km) 19 years (hopefully in my lifetime – but will I be retired?) Cloud Models:  Cloud Models Cloud models solve the 3D Euler equations and transport equations for water vapor and liquid/solid water species with subgrid models for turbulence and other models (parameterizations) for everything else (moisture phase changes, radiation, land-surface, ocean-surface, etc.) Generally speaking, there are 2 flavors: (1) Semi-Implicit (implicit treatment of acoustic and gravity waves) usually found in global models on lat-long grids – pole problem. (2) Explicit (explicit treatment of acoustic and gravity waves) some form of splitting is usually used to advance acoustic and gravity waves with a shorter timestep. Slide16:  Terrain-following hydrostatic pressure vertical coordinate Arakawa C-grid 3rd order Runge-Kutta split-explicit time integration Conserves mass, momentum, entropy, and scalars using flux form prognostic equations 5th order upwind or 6th order centered differencing for advection Limited area (not global) WRF-ARW (more info - http://www.mmm.ucar.edu/wrf/users/) Why Explicit :  Why Explicit Explicit time integration with splitting is more efficient than implicit solvers (operations for a given level of accuracy). Solver needs little tuning for application at different grid resolutions and problem sizes. Easily parallelized for SM, DM and SM/DM architectures. Slide18:  Time Integration in ARW 3rd Order Runge-Kutta time integration advance Amplification factor Slide19:  Time-Split Runge-Kutta Integration Scheme dt is the RK3 timestep acoustic timestep (in this case dt/4) Slide20:  Time-Split Runge-Kutta Integration Scheme In DM applications: A small amount of data is communicated within each acoustic step. Slide21:  Time-Split Runge-Kutta Integration Scheme In DM applications: A small amount of data is communicated within each acoustic step. A larger amount is data is communicated after each RK substep. Parallelism in WRF: Multi-level Decomposition:  Model domains are decomposed for parallelism on two-levels Patch: section of model domain allocated to a distributed memory node Tile: section of a patch allocated to a shared-memory processor within a node; this is also the scope of a model layer subroutine. Distributed memory parallelism is over patches; shared memory parallelism is over tiles within patches Single version of code for efficient execution on: Distributed-memory Shared-memory Clusters of SMPs Vector and microprocessors Parallelism in WRF: Multi-level Decomposition Logical domain 1 Patch, divided into multiple tiles Inter-processor communication WRF Software Framework Overview:  WRF Software Framework Overview Implementation of WRF Architecture Hierarchical organization Multiple dynamical cores Plug compatible physics Abstract interfaces (APIs) to external packages Performance-portable Slide24:  Courtesy of J. Michalakes; see http://box.mmm.ucar.edu/wrf/WG2/bench/ for more info Petascale Computing and Clouds:  Petascale Computing and Clouds Many effects of clouds on climate and weather are largely unknown/uncertain (observations lacking, models at coarse resolution have poor representation of clouds). Most important problem confronting dynamicists and modelers today. Cloud-resolving (Dh ~ O(100 m)) simulations of cloud systems are needed to understand cloud dynamics and to improve parameterizations - a petascale computing challenge. cloud- mixing eddies clouds cloud systems planetary waves synoptic systems meters to 100’s meters 102 - 104 meters 105 - 106 meters andgt;106 meters Petascale Computing and Clouds:  Petascale Computing and Clouds Split-explicit cloud models are easiest to scale to peta-computing - no global data exchange or implicit solver needed, numerics are not scale dependent. We can scale our problems to bigger machines. Questions: What will new machine architectures look like? Will we maintain efficiency with scaling and changes in machine architecture? What code architecture changes will be needed? Other problems: load balancing, analysis, I/O.

Related presentations


Other presentations created by Nickel

Final Capital Budgeting
18. 06. 2007
0 views

Final Capital Budgeting

culturaldiversity
15. 06. 2007
0 views

culturaldiversity

copyright
19. 09. 2007
0 views

copyright

Heynderickx sampex
25. 09. 2007
0 views

Heynderickx sampex

regression
25. 09. 2007
0 views

regression

toke wci fragmodel 3
25. 09. 2007
0 views

toke wci fragmodel 3

PACIS05
25. 09. 2007
0 views

PACIS05

WSC06
25. 09. 2007
0 views

WSC06

bh us 02 veeneman wireless
25. 09. 2007
0 views

bh us 02 veeneman wireless

NOAAg
25. 09. 2007
0 views

NOAAg

phrase EM NAACL presentation
25. 09. 2007
0 views

phrase EM NAACL presentation

LCmodels
25. 09. 2007
0 views

LCmodels

models in geography
25. 09. 2007
0 views

models in geography

hawaii 0401
23. 08. 2007
0 views

hawaii 0401

Albrecht Durer
23. 08. 2007
0 views

Albrecht Durer

Brian McClaren Seminar
17. 08. 2007
0 views

Brian McClaren Seminar

Parables of Jesus
17. 08. 2007
0 views

Parables of Jesus

Girish Sant
17. 08. 2007
0 views

Girish Sant

Safer PI ISIS May 00 Seattle
25. 09. 2007
0 views

Safer PI ISIS May 00 Seattle

DOOR South Asia
07. 08. 2007
0 views

DOOR South Asia

8 2 Participants Maldives
07. 08. 2007
0 views

8 2 Participants Maldives

Maldives Country Presentation
07. 08. 2007
0 views

Maldives Country Presentation

Tsunami Relief Efforts
07. 08. 2007
0 views

Tsunami Relief Efforts

safta2
07. 08. 2007
0 views

safta2

maldives khaleel3
07. 08. 2007
0 views

maldives khaleel3

193
07. 08. 2007
0 views

193

Lay sermon Jesus is the only way
17. 08. 2007
0 views

Lay sermon Jesus is the only way

Eschatology 2
17. 08. 2007
0 views

Eschatology 2

AFD 060714 020
23. 08. 2007
0 views

AFD 060714 020

Finder Workshop
18. 06. 2007
0 views

Finder Workshop

acctpresentation
23. 08. 2007
0 views

acctpresentation

Mobile Fun for a New Generation
18. 06. 2007
0 views

Mobile Fun for a New Generation

Medwoche
18. 06. 2007
0 views

Medwoche

MAGIC
18. 06. 2007
0 views

MAGIC

Lezione 2
18. 06. 2007
0 views

Lezione 2

leonardo
18. 06. 2007
0 views

leonardo

Kostenbetrachtung
18. 06. 2007
0 views

Kostenbetrachtung

kore einfuehrung
18. 06. 2007
0 views

kore einfuehrung

KBSBCH d Dez2003
18. 06. 2007
0 views

KBSBCH d Dez2003

13 triossi
18. 06. 2007
0 views

13 triossi

Foot Ankle Injuries
18. 06. 2007
0 views

Foot Ankle Injuries

Jesus as Messiah
17. 08. 2007
0 views

Jesus as Messiah

New Educ Prog Trng Vis
18. 06. 2007
0 views

New Educ Prog Trng Vis

kailas karthikeylan
07. 08. 2007
0 views

kailas karthikeylan

Montes qmapp
18. 06. 2007
0 views

Montes qmapp

mind akad 2004
18. 06. 2007
0 views

mind akad 2004

METRO PARMA
18. 06. 2007
0 views

METRO PARMA

Lezione 10
18. 06. 2007
0 views

Lezione 10

Sprites Presentation
19. 09. 2007
0 views

Sprites Presentation

dy scheler
15. 06. 2007
0 views

dy scheler

DenmarkPresentation
15. 06. 2007
0 views

DenmarkPresentation

DeckertFINALPM
15. 06. 2007
0 views

DeckertFINALPM

Curtiss Murphy
15. 06. 2007
0 views

Curtiss Murphy

moggi telefono
18. 06. 2007
0 views

moggi telefono

emotional branding2007
15. 06. 2007
0 views

emotional branding2007

Ein Gen Enzym
15. 06. 2007
0 views

Ein Gen Enzym

E Economists in the Cartoons
15. 06. 2007
0 views

E Economists in the Cartoons

lorenz
18. 06. 2007
0 views

lorenz

cmodel intro
25. 09. 2007
0 views

cmodel intro

Laboratoriodicittadi nanza
18. 06. 2007
0 views

Laboratoriodicittadi nanza

DibuzSupremeCourt
15. 06. 2007
0 views

DibuzSupremeCourt

LMclerran fest
17. 08. 2007
0 views

LMclerran fest

prm tutorial 2003 02 20
25. 09. 2007
0 views

prm tutorial 2003 02 20

fonasaygestion
18. 06. 2007
0 views

fonasaygestion

ncd survll dr shah
07. 08. 2007
0 views

ncd survll dr shah

Lezione 4
18. 06. 2007
0 views

Lezione 4

polymeric composite
25. 09. 2007
0 views

polymeric composite

CS662b
15. 06. 2007
0 views

CS662b

patterson barcodes zoocodes
19. 09. 2007
0 views

patterson barcodes zoocodes

Presentations sde small islands
07. 08. 2007
0 views

Presentations sde small islands

20070605 Bulte
07. 08. 2007
0 views

20070605 Bulte

Manual TARRAGON
18. 06. 2007
0 views

Manual TARRAGON

Day1TUMGMSeafax
07. 08. 2007
0 views

Day1TUMGMSeafax

kodak memory studio
07. 08. 2007
0 views

kodak memory studio

Jack Carlsen
07. 08. 2007
0 views

Jack Carlsen