development and evaluation of a high resolution oceanic general circulation model (ogcm)

Development and Evaluation of a High Resolution Oceanic General

Circulation Model (OGCM)

Yongqiang YU, Hailong Liu, Pengfei Lin,

Xuehong Zhang, Huier Mo

LASG/IAP/CAS

Guilin, China, October 28th-31st , 2013

Outline

1. Introduction

2. Model Description

3. Numerical and Parallel Computing

Schemes

4. Large-scale Circulation

5. Marginal Sea Circulation near China

6. Summary

10.6% /10 years

Global Mean Surface Air Temperature Change (IPCC, 2013)

How to reduce the Model Uncertainties

Increase spatial resolution of climate

system models

Improve Numerical Methods

Improve the subgrid physical

parameterization schemes

Introduce new component models

Spectrum of Kinetic Energy in the ocean

Outline

1. Introduction

2. Model Description

3. Numerical and Parallel Computing

Schemes

4. Large-scale Circulation

5. Marginal Sea Circulation near China

6. Summary

Ocean Model Development Historyat LASG/IAP

4-Level (4X5) OGCM by Zhang et al. (1989) Convex-C120

20-Level (4X5) OGCM by Chen et al. (1997) Convex-C3210

30-Level (2X2) GCM by Jin et al. (1999) SGI workstation

LICOM1.0 (0.5X0.5) by Liu et al. (2003) SGI and IBM

LICOM2.0 (0.1X0.1) by Liu et al. (2012) and Yu et al. (2012)

(TH-1A)

Approximations and Equations

z

I

c

1F

td

Td

p0T

u0

Fx

p1vf

td

ud

pST ,,

0z

w

v

gz

p

SFtd

Sd

v0

Fy

p1uf

td

vd

Barotropical Equations

y

V

x

U

t

z0

Xdzx

p11Vf

x

zzHg

t

U 0z

H

C

0x,bx

0

00

Ydzy

p11Uf

y

zzHg

t

V 0z

H

C

0y,by

0

00

Sub-Grid Physical Parameterizations

Turbulent Mixing schemes (Canuto,

2003,2007)

Solar radiation penetration scheme

Bottom drag scheme

Bi-Harmonic diffusion/viscosity

Outline

1. Introduction

2. Model Description

3. Numerical and Parallel Computing

Schemes

4. Large-scale Circulation

5. Marginal Sea Circulation near China

6. Summary

Z-Coordinate and Latitude-Longitude

grid

Leap-frog time integration

Barotropic and Baroclinic Split

Technology

2-order central difference for

momentum advection

Two-step shape-preserving scheme

for temperature and salinity

advection

Implicit scheme for vertical

diffusion/viscosity

2-D MPI Parallel Computing

14

Speed-up in TianHe-1A

Zonal filter in high latitudes lead to imbalance

Serial IO becomes key bottleneck

Experiment Design of LICOM2.0

12-year spin-up forced NCEP climatological flux

2000-2007 simulation by daily forcing form

NCEP Heat Flux and QuickSCAT wind stress

A 60-year integration with NCEP reanalysis.

1 model year/ 2 days

Outline

1. Introduction

2. Model Description

3. Numerical and Parallel Computing

Schemes

4. Large-scale Circulation

5. Marginal Sea Circulation near China

6. Summary

Standard Deviation of Sea Level Height

SST Snapshot near Western Boundary

Observed and Simulated zonal cuurent in the tropic Pacific

Kuroshio Path from 2000 to 2007

MOC FC Ek MO

mean std mean std mean std mean Std

RAPID/MOCHA 18.7 4.7 31.7 2.8 3.5 3.2 -16.5 3.15

LICOM 16.6 6.3 26.3 3.5 3.6 3.5 -13.2 5.7

5-daily mean integrated transports

MOC=FC+EK+MO

MOC ， r=0.62

FC ： Florida Current transport ， r=0.43

EK ： Ekman transport r=0.96

MO ： upper mid-ocean transport ， r=0.3

RAPID (dash line) model (solid line)

CMIP3 models (from Sarojini et al.

2011) the black line is OBS

Power Spectrum of 5-daily mean transports

LICOM

hardly reproduce the 30 to 40 days period

LICOM successfully represent this feature

Velocity profile of ITF (m/s) .

Makassar

Lombok

Ombai

TimorLifatomola

Strait or passage

LICOM2.0 (Sv)

Mean

INSTANT (Sv)

Mean

Makassar Strait

-10.8 (77%) -11.6 (82%)

Lifamatola Passage

-3.2 (23%) -2.5 (18%)

Lombok Strait -2.7 (18%) -2.6 (17%)

Ombai

Strait

-3.1 (21%) -4.9(33%)

Timor Passage -9.2 (61%) -7.5 (50%)

In/out -14/-15 (-1) -14.1/-15 (-0.9)

LICOM1.0 (Sv)

Mean

HYCOM (Sv)

Mean

-8.0 -7.3 (78%)

- -2.0 (22%)

-5.6 (44%) -2.4 (18%)

-6.5 (51%) -5.1 (38%)

-0.7 (5%) -5.9 (44%)

-/12.8 -9.3/-13.4 (-4.1)

Annual mean volume transport

Time series of Volume transport

Outline

1. Introduction

2. Model Description

3. Numerical and Parallel Computing

Schemes

4. Large-scale Circulation

5. Marginal Sea Circulation near China

6. Summary

Depth-Longitude section of observed and simulated temperature along 35N

Jan Apr

Jul Oct

Luzon transport in upper(0-500), middle (500-1500) and deep layers (<1500).

Summary

Compared with coarse resolution of the ocean model LICOM, the eddy-resolving version well reproduces the meso-scale eddies as observed.

The model doesn’t only reproduce equatorial current system such as EUC, ETC etc, but also oceanic circulation in the North Pacific.

The model can simulate the large meander of Kuroshio, as well as its interannaul variability.

The marginal sea circulation near China are also reproduced by the eddy-resolving model.

The mass transport along 26.5N and Indonesian straits are well reproduced.

Ongoing Work

Sub-grid Parameterizations: Tidal mixing,

wave mixing, internal wave mixing etc.

Dynamical Core: Time integration scheme

Tripole Coordinate: reduce load imbalance

Parallel Optimization : MPI + Open MP

High resolution coupled Ocean-Atmosphere

Model

Next Step

在 LICOM 模式中引进三极坐标 改进湍流混合方案 引进潮汐混合、波浪混合、底边界层混合方

案等 进行高分辨率耦合试验 .

Time mean temperature transects at 26.5ºN

LICOM can’t resolve the Bahamas at the western boundary, and the depth of Florida Straits is too shallow compared to the real topography.

WOA09

LICOM

LICOM-WOA09

LICOM can reproduce the observational temperature structure

Schematic of MOC monitoring array at 26ºN

Observational Strategy Modeled Strategy

Meridional overturning circulation (MOC)

MOC=FC+EK+MO zonally and vertically integrated meridional volume transport above 1000m

Florida Current transport (FC)

Cables, velocity profile hori ： west boundary to 78.8W

vert ： surface to the depth of model topography

Ekman transport (EK)

Mid-ocean transport (MO) Density difference between the profiles at the boundaries

MO=MOC-EK-FC

from Rayner et al. ， 2011

QuickScat

meridional heat transport

pQ c v dxdz

observation model

mean 1.34PW 1.14PW

std 0.39PW 0.44PW

R=0.68

Underestimated by 0.2PW

Purpose: evaluate the ability of the eddy-resolving ocean circulation model LICOM to reproduce the variabilities of AMOC on the time mean scale and the time scales less than one year against the high frequency observational data (RAPID data) along 26.5 ºN in Atlantic.

Mass and Heat Transport along 26.5N in North Atlantic

Qu(2000)

Fang(2003)

Lan(2004)

Cai(2005)

Zhao(2007)

Chen(2009)

Wang(2009)

High resolution LICOM

Jan. -6.1 -11.27 -5.53 ≈-5.6 -9.5 -21.03 ≈-6.8 -7.97

Apr. -4.11 -2.5 ≈-3.0 -3.1 -2.01 ≈-3.0 -0.18

Jul. -0.9 -2.4 -0.29 ≈-3.1 -1.6 -9.08 ≈-2.4 -0.47

Oct. -7.65 -2.24 ≈-5.2 -9.2 -17.96 ≈-5.2 -5.97

Ann. -2.4 -6.40 -2.69 -4.27 -5.7 -13.05 -4.5 -3.76

Volume Transport across Luzon Strait （ Unit ： Sv)

Observed and Simulated SSH Anomaly in 2005 and 2007

OBS (2005)OBS (2007)

Model (2005) Model (2007)

Zonal Current along 160E

Dec Mar

SepJun

Deep layer transport across Luzon strait

LICOM

SODA

Zonal Jets in the subtropics