visualization of a new generation of climate simulations with avizo green

Visualization of a new generation of climate

simulations with Avizo Green

Michael BöttingerDeutsches KlimarechenzentrumGerman Climate Computing Center

1

Michael Böttinger, DKRZ

Michael Böttinger, DKRZ

2

DKRZ – who we are High Performance Computing Center National Service Facility for Climate- and Earth System

Research Limited (non for profit) Company (Shareholder: MPG: 53%,

University HH: 27%, AWI: 10%, HZG: 10%)

Michael Böttinger, DKRZ

• IBM "Power6“: 264 Nodes • 8448 Cores• 158 TeraFlops (sustained)• System memory 20 TByte

• Infiniband Interconnect (4xDDR)

• 6 PByte GPFS file system

• Optimized for Climate Modeling

• >100 PByte Archive Capacity

Top500 (06/2009): Rank 27

High Performance Computing

3

Top500 (09/2011): Rank 98

Michael Böttinger, DKRZ

4

The Climate System

Michael Böttinger, DKRZ

Conservation Laws (Momentum, Energy, Mass) State Equations Radiation Boundary Conditions: i.e. Earth Geometry,

Astronomical Parameters, Distribution of Land and Sea

Differential Equations

5

Atmosphere Model

Michael Böttinger, DKRZ

Atmosphere Models

6

... approximate solution: finite differences method

using a 3D-computational grid

Climate Models – Data Structures◦ Georefererenced data -> Mapping◦ Relatively coarse 3D grids, some quantities only 2D◦ Grids: rectilinear, curvilinear, rotated, irregular …◦ Multivariate / scalar and vector fields◦ Special values◦ Time axis: very long time series◦ Coupled models: distinct grids◦ Multi run experiments (ensembles)

Michael Böttinger, DKRZ

Michael Böttinger, DKRZ

8

MomentumEnergyWater

Land Surface

Coupler

Sun Atmosphere

Ocean

concentrations(GHG, SO4)

Source: MPI-M

A coupled climate model

Michael Böttinger, DKRZ                                          

Scenarios forenergy production,

land use, etc.Emission scenarios

for CO2

and othergases/aeros.

Carbon Cycle ModelChange in the

concentration ofCO2 and other

gases/aeros.

Climate ModelClimate change(Temperature,

Precipitation,…)

IPCC AR4 / CMIP3

9

?

Earth System Model

Michael Böttinger, DKRZ

10

Atmosphere

Vegetation / Land

Ocean: Physics Biogeochem.

Coupler

Simulations performed by DKRZ 2010/2011

Earth System Model MPI-ESM: land biosphere, ocean-

biogeochemistry (developed by MPI-M)

Computational demand: factor 60 compared to IPCC AR4

13 000 simulated model years

Results: 650 TB raw data, 60 TB in ESG

Simulations for CMIP5 / IPCC AR5

Michael Böttinger, DKRZ

11

Base experiments

◦ e.g. the recent history historical 1850 to 2005.

Projections (Scenarios)

◦ 2005 to 2100, and 2005 to 2300, 3 different scenarios

Decadal climate forecasts

Role of the carbon cycle with respect to climate change

Long history

◦ e.g. the time 850 to 1850 or the last glacial maximum

CMIP5 Experiments with MPI-ESM

Michael Böttinger, DKRZ

12

Michael Böttinger, DKRZ

IPCC AR5 SimulationsIntroduction: RCP Scenarios

RCP = Representative Concentration Pathways

13

Michael Böttinger, DKRZ

IPCC AR5 SimulationsIntroduction: RCP Scenarios

14

Michael Böttinger, DKRZ

IPCC AR5 SimulationsProjected Temperature Changes

15

Michael Böttinger, DKRZ

IPCC AR5 SimulationsProjected Temperature Changes

16

Michael Böttinger, DKRZ

17

Ensemble Mean Temperature Change relative to 1986-2005

Michael Böttinger, DKRZ

18

Michael Böttinger, DKRZ

19

Michael Böttinger, DKRZ

20

Simulated Sea Ice Concentration for September (blue) and March (white)

Michael Böttinger, DKRZ

21

Michael Böttinger, DKRZ

22

Simulated Sea Level Change (without possible contributions of land ice)

Michael Böttinger, DKRZ

23

Michael Böttinger, DKRZ

24

Michael Böttinger, DKRZ

25

Michael Böttinger, DKRZ

26

Information and Data Access

Michael Böttinger, DKRZ

27

Earth System Grid

IS-ENES Data Access

http://www.dkrz.de/Klimaforschung/konsortial/ipp-ar5/

STORM is a consortium effort aiming at multi-century climate

change integrations with a high resolution climate model

Beyond IPCC AR5 simulations

Michael Böttinger, DKRZ

28

Michael Böttinger, DKRZ

29

1/10 Deg MPI-OM

Michael Böttinger, DKRZ

30

Avizo with domain specific extensions, i.e.

NetCDF CF-1.0 Reader◦ supports regular, rectilinear, rotated, curvilinear and irregular grids◦ Long time series: streaming for time dependent data◦ Caching in memory

Map Projections◦ Cylindric Equidistant, Spherical, Mollweide etc.

„Earth“ Module◦ Textures, LOD techniques◦ Topography / Bathymetry◦ Continental and political Outlines

HW-based Bump Shading for 2D Slices Vector Vis Methods Volume Rendering for rectilinear grids

Michael Böttinger, DKRZ

Avizo Green

High visual quality of the resulting imagery

(relatively) easy to use.

DKRZ contributions to Avizo Green:

◦ Avizo Green Tutorial (see http://www.dkrz.de/Nutzerportal-

en/doku/halo/sw/avizo/avizo)

◦ Enhanced modules, i.e. DKRZ-Earth, BumpSliceDOI (under development)

Next generation climate model: unstructured grid -> developed for

DKRZ, now included in Avizo7 (funded by DKRZ and MPI-M)

More stability needed

Scalability: Performance problems with large grids

Status and Challenges

Michael Böttinger, DKRZ

32

Merci!

boettinger@dkrz.de

http://www.dkrz.de

33

Michael Böttinger, DKRZ

MPI-ESM LR: T63L47/GR15L40

◦ Atmosphere: horizontally ~200 km, 47 height levels (up to 80 km

height)

◦ Ocean: variable resolution between 12 and 150 km, 40 depth levels

MPI-ESM MR: T63L95/TP04L40

◦ Atmosphere: horizontally ~200 km, 95 height levels (up to 80 km

height)

◦ Ocean: variable resolution at 0.4° (~40 km), 40 depth levels

Model Configurations

Michael Böttinger, DKRZ

34