The link between tropical convection and the Northern Hemisphere Arctic surface air temperature change on intraseaonal

and interdecadal time scales

Steven Feldstein

The Pennsylvania State University

September 11, 2012

Collaborators: Sukyoung Lee, Tingting Gong, Nat Johnson,, Changhyun Yoo, David Pollard, and Tim White

NOAA Global Historical Climatology Network Gridded surface air temperature

Source: JISAO, University of Washington

Warming trend is greater toward higher latitudes(1) Present-day observations

Source: Hoffert & Covey (1992, Nature)

Barron & Washington reconstruction data (1985, AGU)

Warmer climate hasa smaller equator-to-pole temperature gradient

(2) Past climates (from reconstruction)

Present day

Multi-model mean changes in surface air temperature (°C, left) and precipitation (mm/day) A1B Scenario, 2080-2090 relative to 1980-1999

IPCC (2007)

(3) Future projections (from models)

“Advances in climate modeling and reconstruction of paleoclimates from proxy data have resolved some controversies but have underscored areas

where understanding of greenhouse climates remains imperfect. Latitudinal temperature gradients are particularly problematic.”

from Warm Climates in Earth History (Huber et al. 2002)

surplus solarterrestrial

deficit

deficit Poleward heat flux

Poleward heat flux

Baroclinic eddy heat flux = -Diffusivity x mean gradient

What other mechanisms?

• CO2 warming causes tropical convective heating to be more localized • Convective heating excites Rossby waves• Rossby waves propagate poleward, transporting westerly momentum

equatorward: eastward acceleration in tropics• Atmosphere adjusts toward a balanced state• Poleward heat transport by the wavesEastward acceleration

Proposed mechanism for polar amplification:Tropical Rossby wave source, poleward wave propagation, &

high-latitude adiabatic warming

pole

heig

ht

equator

X

westward acceleration

Adiabatic warming

Poleward Rossby propagation

Inspiration of the hypothesis: Rossby waves excited by a tropical heat source cause equatorial superrotation [Saravanan 1993]

Upper level zonal wind of normal state

Upper-level zonal wind of a superrotating state

longitude

latit

ude

Wave source

Inspiration of the hypothesis: Rossby waves excited by a tropical heat source cause equatorial superrotation [Saravanan 1993]

Eddy heat flux of a superrotating state

Eddy heat flux of normal state

Upper level zonal wind of normal state

Upper-level zonal wind of a superrotating state

In a superrotating state, there must be a process other than baroclinic eddy heat flux that maintains the weak temperature gradient

Test of the hypothesis for the Cretaceous climates

• Coupled atmosphere-mixed layer ocean GCM (GENESIS)• 50-m oceanic mixed layer with diffusive heat flux• Atmospheric GCM: T31 (3.75o), 18 vertical levels• CO2: 4 X pre-industrial value 1120 ppmv• CTL run: no prescribed tropical heating• EXP run: local tropical heating – compensating cooling• Heating/cooling profile: maximum in the mid-troposhere

longitude

equator

Warm pool heating (150 W/m2)

Compensating cooling (-50 W/m2)

tropical perturbation run - control run

Surface temperature 250-hPa height and winds

To test the hypothesis, consider thermodynamic energy equation

heat fluxconvergence

adiabatic warming

radiative cooling

diabatic heating

heat flux convergence stationary eddies

temperature change

heat flux convergence transient eddies

adiabatic warming

Change in stationary eddy heat flux

Change in stationary eddy heat flux convergence

0.3 K/day;

for g = 20 day, dT ~ 6 K

Change in transient eddy heat flux

Change in transient eddy heat flux convergence

0.3 K/day;

for g = 20 day, dT ~ 6 K

change in vertical velocity (dP/dt) change in adiabatic warming

0.4 K/day;

for g = 20 day, dT ~ 8 K

change in vertical velocity (dP/dt) change in stationary eddy momentum flux convergence

EE W W

what drives high-latitude adiabatic warming?

January zonal mean surface air temperature

Tropical heating contrast (W/m2)

Main findings of the Cretaceous modeling study

1. Localized tropical heating (net zero heating) can warm the Arctic.

2. Winter Arctic warming is driven by dynamic warming (adiabatic warming driven by stationary eddy momentum flux + stationary eddy heat flux), and possibly by downward IR (cloud cover increases).

3. Within the range of 0-150 W/m2, zonal mean Arctic temperature increases 0.8oC per 10 W/m2 for CO2 level of 4 X PAL (Cretaceous);

increases 0.3oC per 10 W/m2 for CO2 level of 1 X PAL (present)

4. Above 150 W/m2, the Arctic warming saturates

5. Tropical upper tropospheric zonal wind becomes more strongly super-rotating.

Does localized tropical convective heating increase as the climate warms?

mm/day/century

Climatology

Linear trend (1979-2002)

(1982-2001) minus (1958-1977)

Testing the hypothesis with ECMWF reanalysis

Surface temperature 250-hPa streamfunction

250-hPa eddy streamfunction Convective precipitation

To test the hypothesis, consider the thermodynamic energy equation

heat fluxconvergence

adiabatic warming

radiative cooling

diabatic heating

heat flux convergence

temperature change

adiabatic warming

diabatic heating

Surface heat flux

(1982-2001) minus (1958-1977)

Testing the hypothesis with ECMWF reanalysis

Surface temperature Downward IR flux

Horizontal temperature advection + Adiabatic warmingDirect consequence of Rossby wave dynamics

The findings so far:

Dynamical processes & downward IR radiative flux warm the Arctic

Questions:

1. Are these two processes linked?

2. How does the inter-decadal time-scale Arctic warming occur through Rossby wave dynamics which takes place on intraseasonal time scales?

Daily evolution associated with the 250-hPa streamfunction trend pattern

250-hPa streamfunction Downward IR flux Surface temperature

Spatial correlations between trend patterns & daily patterns

(day)

Tropical precip => atm circulation (Rossby waves) => IR warming, surface warming

Surface Air Temperature Response to Madden-Julian Oscillation

MJO Phase 1 MJO Phase 5

Interdecadal changes in MJO phase frequency

Interdecadal changes in surface air temperature due to MJO

Main findings: from (1958-1979) to (1980-2001)

1. Tropical convective precipitation (thus heating) became more confined into the Indo-Pacific warm pool region.

2. Winter Arctic warming is driven by dynamic warming and downward IR.

3. This decadal time-scale Arctic warming occurs through changes in the frequency of occurrence of a small number of intraseasonal-scale teleconnection patterns.

4. The warm pool convection leads the teleconnection patterns by 3-4 days; the teleconnection patterns lead the IR flux by 2-3 days;

5. Interdecadal changes in the Madden-Julian Oscillation contribute toward polar amplification.

MJO phase 1 MJO phase 5

Passive tracer evolution