CLIVAR-WGOMD 8 th panel meeting UK Met Office Hadley Centre, Exeter, UK 30April – 01 May 2009

CLIVAR-WGOMD 8th panel meeting

UK Met Office Hadley Centre, Exeter, UK

30April – 01 May 2009

Current status of the Coordinate Ocean-ice Reference Experiments

(COREs) in Japan

Hiroyuki Tsujino

Meteorological Research Institute / Japan Meteorological Agency

collaboration with

M. Hirabara, H. Nakano, G. Yamanaka, T. Motoi, H.Ishizaki (MRI/JMA)

contributions from

T. Suzuki, Y. Komuro, Y. Sasaki (JAMSTEC)Outline:

1. Some preliminary results from CORE-II related activities in MRI.

2. Solution / idea to deal with specific questions asked.

Requested issue 1:

Summarize use of interannually forced simulations in Japan, and detail experimental design.

Institutional efforts in Japan

• Meteorological Research Institute

- CORE-II v2 for 1º x 1/2º global model (used in CMIP5)

- status: on-going

• CCSR Univ. of Tokyo / JAMSTEC

- CORE-II v1 for 1/4º x 1/6º global model (used in CMIP5)

- status: done, CORE-II v2 planned

• JAMSTEC Earth Simulator Center

- CORE-I for 1/30º x 1/30º North Pacific model, considering CORE-II

- status: highly-likely, CORE-II v2 planned

Meteorological Research Institute CORE-II v2 for 1 x 0.5 CMIP-type global model

Model settings

• global (tri-pole) 1º x 0.5º x (L50+BBL)

• oceanic component of MRI-CGCM used for CMIP5

Initialization and integration period (on-going)

• initial state: spun-up (accelerated 5000 years and synchronized 500 years)

• period: two cycles of de-trended CORE-II ver. 2 and one cycle of CORE-II ver.2

Objectives

• baseline hindcast simulation of the latter half of 20C to facilitate CMIP5 expt.

• to understand interannual variability and trend in the ocean

under “true” atmospheric forcing

a. CORE-I-500yr

initial: PHC3.0 (no ice)forcing: cnyf-v1period: 500yrs

purpose:• check with Griffies et al. (2009)• parameter sensitivity tests

b. CORE-II-moncl-millennia

initial: “a” or PHC3.0 (no ice)forcing : monthly climatology of ciaf-v2period : 50 yrs (synchronous) ~ 5000 yrs (accelerated) ~ 500 yrs (synchronous) terminated when drift becomes smallpurpose:• obtain balanced state to initialize “c”

c. CORE-II-20C

initial ：” b”forcing ： 1860-1957 de-trended ciaf-v2 x 2 1958-2006 ciaf-v2period ： 1860-2006 (-2008)

purpose ：• reproduce oceanic state of the latter half of 20C

Our experimental design and recommendation for CORE-II expt.

Notes:- One of main targets should be long-term trends of the simulated field.- Off-line bio-geochemical model will be run using the flow fields.- There remain concerns regarding discrepancy between re-constructed (monthly climatology) and original forcing. Do they really yield similar mean simulated states?- Fresh water flux and salinity restoring adjustment would be needed during the long-term integration.- How should fresh water and salinity restoring adjustments be treated in the interannual run? (Fresh water adj. fixed during integration; salinity adj.) - This recommendation assumes use of eddy-less models.

Requested issue 2: Provide recommendations for CORE-II experimental design.

A long-term spin-up is needed because

- “drift” should desirably be removed from simulated fields to detect “trends”

- deep circulation of the Pacific Ocean, which is of particular interest to

Japanese modelers, needs more than a few thousand years to establishWhy is monthly climatology used to force the model?

- to take a long model time step (e.g., > 6hr; original forcing time interval)

- short-time variability in forcing should be absent in a distorted physicsHow to make CORE-II-monthly climatology and how to calculate fluxes

- monthy climatology for all items & scalar wind speed

- we presently calculate fluxes in the same way as the 6-hourly data, except

- scalar wind speed is based on absolute wind speed

- wind stress is calculated using scalar wind speed and relative wind vectorHigh resolution (eddy-permitting/resolving) models might be initialized by a spun-up state of a lower-resolution model and integrated using monthly climatology for 20 yrs, and then switched to CORE-II interannual expt.

Model Settings- Meteorological Research Institute Community Ocean Model

(MRI.COM, Ishikawa et al. 2005)

English document is almost completed and will be released by the end of 2009

- tracer advection scheme ：

Second Order Moment (Prather, 1986) for non-accelerated run

UTOPIA and QUICKEST (Leonard 1979, 1993) for accelerated run

- sea ice part: 5-categories, 1-layer ice, EVP dynamics, MPDATA advection

- neutral physics:

GM （ 250 m2s-1 / 100 km x gridsize ） and isopycnal diffusion (1000 m2s-

1)

- viscosity:

Smagorinsky viscosity with scaling factor 4.0,

anisotropic tensor (direction of the flow:1 / normal to the flow 0.2)

- SSS restore ： 50m / 365day except for coastal gridpoints with sea ice

- fresh water flux adjustment: global correction factor for P+R, adjusted annually

- tidal mixing parameterization: Kuril Islands (St. Laurent and Simmons 2006)

for the sake of North Pacific intermediate salinity minimum

Test 1.CORE-II-ciaf and CORE-II-moncl as a continuation of CORE-I (500yr)(Note: the model version is changed from the 500th year)

Atlantic MOC at 45N Southern Ocean MOC

Pacific Deep MOCDrake Passage transports

CORE-II-moncl: monthly climatology for all items & scalar wind speed

CORE-II-ciaf and CORE-II-moncl might be expected to yield similar mean fields

Black:

CORE-I

Red:

CORE-II-moncl

(non-accel.)

Green:

CORE-II-ciaf

Test 2. Long-term (2000yr) accelerated spin-up using CORE-II-moncl forcing

Why monthly? … to take a long time step (8hr).

Short-time variability in forcing should be absent in a distorted physics

Atlantic MOC at 45N Southern Ocean MOC

Pacific Deep MOCDrake Passage transports

Grey:

CORE-I

Orange:

CORE-II-moncl

(non-accel.)

Brown:

CORE-II-moncl

(accelerated)

MOC after 2000 yrs of CORE-II-moncl

OBS(COBE-SST) MODEL

Tropical Indian Ocean SST Anomalies (Yamanaka 2008; GRL)

Tropical Indo-Pacific Ocean SST

ERA-40

JRA-25

CORE

OBS

CORE

8July2008 version

CORE

5March2009 version

Results from CORE-II-ciaf (1): Tropical SST

CORE-II is superior to reanalyses

annual mean sea ice area 1979-2007(solid: CORE-II-ciaf simulation, dashed: HadISST)

(blue: sea ice extent, red: actual area)

correlation ：extent: 0.81, actual area: 0.90

Results from CORE-II-ciaf (2): Arctic sea ice

Sea ice seems to have adjusted by 1979.

Shouldn’t CORE-II be extended to the year 2007?

upper: HadISST,

lower: CORE-II-ciaf simulation extended to 2007 using Japanese re-analysis

JAMSTEC / CCSR Univ. of Tokyo - CORE-II version 1 for 1/4 x 1/6 global model

Model setting

• global (rotated pole) 1/4 x 1/6 x L50+BBL

• oceanic component for the decadal prediction expt. in the context of CMIP5 Initialization and integration period

• initial state: climatological temperature and salinity (WOA01)

• period : CORE-II version 1 (1958-2004) 1-cycle

Objective

• check reproducibility of the latter half of 20C

• provide physical fields to marine bio-geochemical models

Known problems possibly related to forcing:• low SST along the Equator (too weak radiation?)• thin ice around North Pole (too strong short wave radiation?)• divergent wind-driven ice drift around the Antarctica (too strong wind?)

1. What are the key scientific uses for interannually forced global ocean-ice simulations?

- to understand ocean’s response (adjustment) to “true” atmospheric forcing

- attribution of recent ocean-climate events to trends or natural variability

2. How does CORE-II fit into the spectrum of coupled climate modelling (e.g., IPCC “historical experiments”) and reanalysis projects such as those addressed in GSOP?

- CORE-II could serve as a baseline simulation for 20C expt. of CMIP5

- MRI assimilation group plans a reanalysis expt. using CORE-II in the near future

3. Can we identify interesting scientific questions that will make a CORE-II project of scientific interest to the panel?

- direct comparison with WOCE re-visit observations

- detect of trends

5. What further metrics and diagnostics are of scientific interest for CORE-II simulations?

- zonally averaged linear trends of temerature and salinity

- some EOFs for SST, SSH, upper layer heat content

6. What observational datasets and reanalysis products should be used to evaluate the CORE-II simulations?

- a carefully QC’ed upper layer temperature data

by Ishii et al.(2006) would be available

4. What are the baseline metrics required to assess the simulation integrity?

- to compare climatological state with WOCE and re-visited WOCE sections

- SST, SSH, upper layer heat content

8. Would it be possible to construct key atmospheric forcing fields for the first half of the 20th century and merge this (constructed) forcing with CORE-II to generate a continuous, albeit not fully consistent, data set from year 1900 (or 1880)?

- Can objective-analysis SST data by JMA (COBE SST, 1x1, 1890-present) be utilized?

7. Should Large and Yeager (2008), using NCAR bulk formula (as in CORE-I), form the basis of the CORE-II benchmark simulation? What are the problems with this dataset that make certain groups use reanalysis products, or alternative approaches?

- “Yes” to the first question- radiation data before 1984 are replaced by satellite climatology- Arctic ice retreat event in 2007 could be a good test for the ocean-ice model

11. Are there any un-spoken issues that may handicap the goal of producing a common benchmark simulation (besides the magnitude of the effort)?

-

10. How should the ocean and sea ice models be initialized? What about spin-up time prior to the focused analysis period? What portion of the simulation should be analyzed, and what should be ignored (due to initialization shock)?

- simulations are desirably started from a quasi-steady state- if some repeated cycles are applied prior to the last cycle, the forcing for the cycles except for the last should be de-trended- about first 10 years of the last cycle would not be reliable- sea ice may be feasible after a short period even without careful initialization

9. What about the salinity boundary conditions? Should this remain a choice for each group, much as in CORE-I? Can we instead provide more specific guidance, and perhaps make another effort to unify the approaches?

- This should remain a choice of each group to guarantee simulation integrity of each group’s model, i.e., to be usable for scientific analyses.

12. How can we make the CORE-II simulation output more accessible to the broader community?

- making use of repository, like PCMDI, REOS would be useful

- publish an inter-comparison paper to notify this project to broader community

13. Is it feasible, and of interest, to aim towards a comparison paper to be written during 2010?

- Yes, our CORE-II expt. would be completed by early 2010.

- A new set of CORE-II expt. with some specific protocols would be possible.

14. How can we coordinate simulations leading up to a CORE-II paper (e.g., password protected Wiki page monitored by Anna)?

- Wiki would be useful. More “on-the-spot” discussion would be needed.