Isaac Held, Beijing, 2011 Thank you for your invitation and kind hospitality !

Seminars:

Monday: Time Scales of Global Warming

Tuesday: Simulating the climatology, interannual variability, and trends of tropical cyclone genesis

Wednesday: The hydrological cycle and global warming

Thursday: Shifting latitude of surface westerlies – a case study in utilizing a hierarchy of climate models (understanding climate by starting with comprehensive models and gradually removing layers of complexity)

Friday: Problems in quasi-geostrophic dynamics (understanding climate by starting with very idealized models and gradually adding layers of complexity)

Probing the fast and slow components of global warming by returning abruptly to pre-industrial forcing

Held, Winton, Takahashi, Delworth, Zeng, Vallis, J. Clim 2010

Importance of Ocean Heat Uptake Efficacy to Transient Climate Change

Winton, Takahashi, Held, J. Clim, 2010

Time scales of climate responses, climate sensitivity, and the recalcitrant component of global warming

Isaac HeldBeijing, 2011

Uncertainty in climate sensitivity has not been reduced appreciably in past 30 years

2 well-known assessments reach similar conclusions :“Charney report” (1979) IPCC/AR4 (2006)

Equilibrium global mean surface temperature warming due to doubling of CO2

is most probably in the range 1.5-4.5 K

Knutti+Hegerl, 2008

Assorted estimates of equilibrium sensitivity

23

Ultra-fast SlowFast Ultra-slow

Months(Atmosphere)

a few years(mixed layer)

Multiple centuries(deep ocean)

Time scales of climate response

Equilibrium climate sensitivity:Double the CO2 and wait for the system to equilibrate

Transient climate response:Increase CO2 1%/yr and examine climate at the time of doubling

t

CO2 forcing

Heat uptake by deep ocean

W/m2

Typical setup – increase till doubling – then hold constant

After CO2 stabilized, warming of near surface can be thought of as due to reduction in heat uptake

T response

11

2 43 51

2

1.5

2.5

Equilibrium sensitivity

CMIP3/AR4 models

Transientresponse

Not well correlated across models – equilibrium response brings into play feedbacks/dynamics (especially in subpolar oceans)

that are suppressed in transient response19

Increase CO2 by 1%/yr ; global mean warming at the time of doubling = Transient Climate Response (TCR)

Histogram of TCR/TEQ

for AR4 models

Response of global mean temperature in GFDL’s CM2.1 to instantaneous doubling of CO2

Equilibrium sensitivity 3.4KTransient response 1.5K

€

T = (1.5K)e−t /(4 yrs)

Fast response

Slow responseevident onlyafter 80 yrs

20

€

cF

dT

dt= −βT − γ(T − TD ) + F

cD

dTD

dt= γ(T − TD ) forcing

Heat exchangebetween mixed layerand deep ocean

Mixed layerHeat capacity

Deep ocean heat capacity

€

T = TD =F

βin equilibrium

€

cF

dT

dt= −βT − γ(T − TD ) + F

cD

dTD

dt= γ(T − TD )

€

τF =cF

β + γForcing varies on time scales longer than

€

⇒ T ≈F

β + γ+

γTD

β + γ

€

⇒ cD

dTD

dt= −

βγ

β + γTD +

γ

β + γF

€

⇒ TD ≈ 0 T ≈F

β + γ

TCR /TEQ ≈β

β + γ

€

τF =cF

β + γForcing varies on time scales longer than

€

τD =cD

β

β + γ

γand time scales shorter than

€

“Intermediate regime”

OLR

SW down

SW up

total

Forcing computed from differencing TOA fluxes in two runs of a model (B-A)B = fixed SSTs with varying forcing agents; A fixed SSTs and fixed forcing agents

51

Temperature change averaged over 5 realizations of coupled model

52

€

CdT

dt= F −αT; α =1.6 Wm−2 /K;

C

α= 4years

Fit with

53

Forcing with no damping

Forcing (with no damping) fits the trend well, if you use transient climate sensitivity, which takes into account magnitude/efficacy of heat uptake

54

Observations (GISS)

GFDL’s CM2.1 with well-mixed greenhouse gases only

Global mean temperature change

46

“It is likely that increases in greenhouse gas concentrations alone would have caused more warming than observed because volcanic and anthropogenic aerosols have offset some warming that would otherwise have taken place.” (AR4 WG1 SPM).

Observations (GISS)

GFDL’s CM2.1 with well-mixed greenhouse gases only

Global mean temperature change

46

A1B-CM2.1

€

T ≈F

β + γ+

γTD

β + γ

€

⇒ T ≈γTD

β + γ

Return instantaneously to pre-industrial forcing ( F = 0)

the “Recalcitrant” warming

Relaxation to recalcitrant warming

5 years

3 years

Normalized to unity over the globe

Normalized to unity over the globe

Fast

Slow“Recalcitrant”

Control drift

Sea level response due to thermal expansion

Sea level response mostly recalcitrant

€

CdT

dt= F − βT ≡ N

TEQ = F β

N /F =1− T /TEQ

N/F

T/TEQ

The simplest linear model

If correct, evolution should be along the diagonal

15

Suppose you have two forcing agents C02 and B (something else)

leading to radiative forcing FC02 and FB

.

But suppose the global mean temperature responses TC02 and TB

are not proportional to the the radiative forcing

Following Hansen, define efficacy B (using CO2 as a standard)

€

B =TB /FB

TC 02 /FC 02

Efficacy can orten be understood in terms of the spatial structure of the response, Coupling of surface with troposphere is weaker in high latitudes => harder to radiate away a perturbation

=> Radiative restoring strength is weaker for responses thatare larger in higher latitudes

=> Forcings with stronger high latitude responses have larger efficacy

Forcings with stronger high latitude responses have larger efficacy

Think of heat uptake as a forcing – ie

replace F = T + H or T = F + H

with T = F + H H with H > 1

Equivalently,

T = TF + TH = F/ - H/H

With H = /H

€

cF

dT

dt≈ 0 ≈ −βT − γ(T − TD ) + F = −βT − H + F

CM 2.0

CM 2.1

Efficacy

Eff

icie

ncy

Heat uptake = T ; = efficiency of heat uptake

Cooling due to heat uptake = T ; = efficacy of heat uptake

Knutti+Hegerl, 2008

Assorted estimates of equilibrium sensitivity

23

(GFDL CM2.1 -- Includes estimates of volcanic and anthropogenic aerosols, as well as estimates of variations in solar irradiance)

Models can produce very good fits by including aerosol effects, but models with

stronger aerosol forcing and higher climate sensitivity are also viable (and vice-versa) 45

Observational constraints

•20th century warming•1000yr record •Ice ages – LGM•Deep time

•Volcanoes•Solar cycle•Internal Fluctuations

•Seasonal cycle etc36

Observed total solar irradiance variations in 11yr solar cycle (~ 0.2% peak-to-peak)

42

Camp and Tung, 2007 => 0.2K peak to peak(other studies yield ~0.1K)

Seems to imply largesensitivity

4 yr damping time

1.8K (transient) sensitivity

Only gives 0.05 peak to peak

43

Global mean cooling due to Pinatubo volcanic eruption

Range of ~10 ModelSimulationsGFDL CM2.1

Courtesy of G Stenchikov

Observationswith El Ninoremoved

Relaxation time after abrupt cooling contains information on climate sensitivity

40

Yokohata, et al, 2005

Low sensitivity model

High sensitivity model

Pinatubo simulation

41

Response to pulse of forcing (volcano), F(t):

2-box model:

€

Tdt =Fdt∫

β + γ0

fast

∫ Tdt =Fdt∫β0

∞

∫

Stenchikov, et al 2009

€

T0(1− e−t /τ )

T0 = 2.5K • yrs

τ = 4.2yrs

Near surface air temperature response (20 member ensemble)Courtesy of Stenchikov, et al

Forcing

TOA flux

Responsewith exponential fit

Wm-2yr

Integrated forcing and response

CM2.1 Pinatubo summary-- fast response --

€

Tdt =Fdt∫

β + γ0

fast

∫ €

Tdt = 2.8 Kyr0

fast

∫

5.02.8

2.2

CM2.1 Pinatubo summary-- fast response --

Forcing (W/m2)yr

€

Tdt =Fdt∫

β + γ0

fast

∫

€

Fdt∫

€

Tdt0

fast

∫

€

Tdt0

fast

∫

Heat uptake (W/m2)yr

Radiative restoring (W/m2)yr

€

Tdt = 2.8 Kyr0

fast

∫

Pinatubo =>

~ 1.0 (W/m2)/K

~ 0.8 (W/m2)/K

1%/yr CO2 increase =>

~ 1.7 (W/m2)/K

~ 0.7 (W/m2)/K

Can we use interannual variability to determine the strength of the radiative restoring?

Model results (CM2.1) raise some roadblocks

Longwave regression across ensemble (collaboration with K. Swanson)

LW

Wm-2K-1

All-forcing20th century

61

year

Following an idea of K. Swanson, take a set of realizations of the 20th century from one model, and correlate global mean TOA with surface temperature across the ensemble

Longwave regression across ensemble, collaboration with K. Swanson

All-forcing20th century

A1B scenario

62

LW

Wm-2K-1

Longwave regression across ensemble, collaboration with K. Swanson

63

LW

Wm-2K-1

Estimate of noise in this statistic from 2000yr control run

Longwave regression across ensemble, collaboration with K. Swanson

Well-mixedgreenhouse gases only

64

LW

Wm-2K-1

Longwave regression across ensemble, collaboration with K. Swanson

Independent set of 10 A1B runs

65

LW

Wm-2K-1

Longwave regression across ensemble, collaboration with K. Swanson

Independent set of 10 A1B runs

But we can fit the models 20th century simulations without time-dependence in OLR-temperature relationship!

May be telling us that ENSO is changing, but with no obvious connection to global sensitivity

65

LW

Wm-2K-1

I suspect that:

Transient climate sensitivity can be constrained more tightly that it currently is, despite the uncertainty in

aerosol forcing

Volcanic responses may play a central role in tightening this constraint, along with the observed

warming trend

Less hopeful about use of interannual variability

Solar cycle response has some mysteries

Thank you for listening

21st century emissions commitment

AR4/IPCC

Shortwave regression across ensemble, following K. Swanson 2008

Wm-2K-1

All-forcing20th century

Following an idea of K. Swanson, take a set of realizations of the 20th century from one model, and correlate global mean TOA with surface temperature across the ensemble

56

Shortwave regression across ensemble, following K. Swanson 2008

All-forcing20th century

A1B scenario

Wm-2K-1

Is this a sign of non-linearity? What is this?

57

Shortwave regression across ensemble, following K. Swanson 2008

All-forcing20th centuryWm-2K-1

A1B scenario

90%

Estimate of noise in this statistic from 2000yr control run58

Shortwave regression across ensemble, following K. Swanson 2008

Wm-2K-1

Well-mixedgreenhouse gases only

59

Shortwave regression across ensemble, following K. Swanson 2008

Wm-2K-1

Independent set of 10 A1B runs

60