The ocean component of climate models: what resolution and parameterisations are needed?

1

The ocean component of climate models: what resolution and

parameterisations are needed?

Richard Wood

Hadley Centre for Climate Prediction and Research, Met Office, Bracknell, UK

(Thanks to Malcolm Roberts)

2

Atlantic Ocean Heat Transports

HadCEM

HadCM3

3

Do atmospheric processes set ENSO period? (1)

(Guilyardi et al. 2002)

4

Do atmospheric processes set ENSO period? (2)

(Guilyardi et al. 2002)

5

Do ocean processes set ENSO amplitude? (1)

(Meehl et al. 2001)

6

Do ocean processes set ENSO amplitude? (2)

(Meehl et al. 2001)

7

Temperature drifts - ocean only model

0m 147m

289m 520m

A: PCM B: add KPP mixing C: add GM D: add Visbeck et al.(Gent et al . 2002)

8

Temperature drifts - coupled model

0m 147m147m

289m

520m

AC: PCM AD: add KPP mixing + GM + Visbeck et al.(Gent et al . 2002)

9

Do we need eddy resolving climate models?

Dynamical length-scale - baroclinic Rossby radius - 20-50 km in ocean at mid-latitudes, of order 1000 km in atmosphere

Most climate models have ocean resolution more coarse than 1°

Oceanic processes possibly important for climate:

– boundary currents and ocean eddies - 30-50 km - heat and freshwater transport

– flow through topographic channels (20-50 km) and impact on thermohaline circulation

– tropical dynamics and representation of equatorial waves

– convective regions often pre-conditioned by eddies

– water mass ventilation may be resolution dependent

– coastal polynyas - 10’s km - regions of sea-ice formation

10

Thermocline ventilation in two ocean models

Tracer

PV

1° 1/3°

(Cox 1985)

11

A systematic study of the effects of ocean resolution on climate simulations

Results from 3 coupled models:

HadCM3 Ocean resolution 1.25°x1.25°xL20

HadCEML Ocean resolution 1.25°x1.25°xL40

HadCEM Ocean resolution 0.33°x0.33°xL40

All ocean models are coupled to the identical atmosphere, resolution 2.5°x3.75°xL19.

No flux adjustments used

12

Dependence of global heat balance on ocean advection scheme and resolution

13

HadCEM: model drifts

14

SST errors

HadCEM

HadCM3

15

North Atlantic salinity drifts in 3 coupled models

HadCEM

HadCEML

HadCM3

0m

2000m

4000m

0m

2000m

4000m

0m

2000m

4000m

0 yrs 150 yrs

-0.45 psu +0.45 psu

16

HadCEM: Atlantic Overturning

17

HadCEM: Ocean Heat Transports

18

Atlantic Ocean Heat Transports

HadCEM

HadCM3

19

Pacific Ocean Heat Transports

20

Instability Waves in Tropical Pacific (HadCEM)

(Maybe these can be parameterised?)

Does the atmosphere model see them?

21

Salinity and velocity at 260m in HadCEM

(Parameterise these!)

22

Global mean SST response to increasing CO2 in 3 coupled models

23

The future…?

24

Impact of inhomogeneous diapycnal mixing

Homogeneous

Inhomogeneous

Hor. Average (INHM)

(Hasumi & Suginohara 1999)

25

Mesh adaptivity: flow over seamount

(Courtesy Matthew Piggott, Imperial College)

26

Supercomputer performance

(Source: Earth Simulator Centre)

27

Current technological issues

Typical atmosphere GCM has about 6x the computational cost of typical ocean GCM per gridpoint

A multicentury ocean integration at 0.1° resolution is a significant cost, even on a 40 Tflop computer

Inclusion of biogeochemical processes can increase the cost of atmosphere and ocean dramatically

Mass data storage and methods of data sharing may become the limiting technologies

28

Summary

Ocean-only modelling studies give valuable insights into model sensitivities - but coupling brings surprises!

Sensitivity to some parameterisations/model formulations can be resolution-dependent

Increasing resolution doesn’t automatically give a better climate simulation

Large scale climate response fairly insensitive to ocean resolution. Need compatible atmospheric and ocean resolutions?

Future: higher resolution (->parameterisations?), diapycnal mixing, biogeochemistry, more flexible numerical methods, (technological limiters) …

… and more integration with observations! Balance between model diversity and spreading too thinly

29