Tropical Oceanic Influences on Global Climate

Prashant. D. Sardeshmukh @ noaa.gov

Climate Diagnostics Center, CIRES, University of Colorado

and Physical Sciences Division/ESRL/NOAA

Thanks to: Joe Barsugli, Gil Compo, and Sang-Ik Shin

GEOSS meeting Quebec 15 September 2008

Some Basic Questions

1. What parts of the climate system need to be observed and modeled most accuratelyto improve regional climate predictions ? The tropical oceans ?

2. Is there more sensitivity to SST changes in some tropical regions than others ?

3. What is the best way to estimate such sensitivities ? One way is to perturb an atmospheric GCM with the same prescribed localized SST anomaly at different locations and assess its impact.

4. Are the responses to SST linear enough to justify such an analysis ? The notion of “sensitivity” becomes murky if the response to SST fields is not the sum of the responses to the SSTs at individual locations, because we cannot then speak unambiguously of the effect of a local SST change in isolation.

5. Finally, since everything in the climate system depends on everything else to some degree, can the sensitivities to SST changes be summarized in a meaningful way ?

IPCC

Summary for Policy

Makers

2007

To whatdegreehas the oceanic warminginfluencedthe continentalwarming ?

Figure SPM.4

Global and Continental Temperature Change

Most Recent Surface Temperature Change (1991-2006 minus 1961-1990)

Observed

Simulated by ECHAM AGCMwith prescribedobserved SST changes (but no GHG changes)

As above, butsimulated byNASA AGCM

From Compo and Sardeshmukh Climate Dynamics 2008

The IPCC models have not captured recent 50-yr regional climate trends

Observed Trends

IPCC model simulated trends (Coupled models)

(From Shin and Sardeshmukh 2008)

Figure currently unavailable

The IPCC models have not captured recent 50-yr regional climate trends

On the other hand, atmospheric GCMs with prescribed observed SSTs (either globally or just in the tropics) have done much better in this regard.

Observed Trends

IPCC model simulated trends (Coupled models)

Atmospheric Model simulated trends (Uncoupled models with prescribed observed SSTs)

(From Shin and Sardeshmukh 2008)

Figure currently unavailable

The IPCC models have also not captured recent 50-yr Tropical SST trends

(From Shin and Sardeshmukh 2008)

Figure currently unavailable

Evidence for the Dominance, Linearity, and Low-Dimensionality of Tropical SST influences on regional climates around the

globe

Dominance : Tropical SSTs account for most of the global SST-forced responses

Linearity : The response to tropical SSTs is approximately linear; y = Gx

Low-Dimensionality: G is a “low-dimensional” linear operator

G = U S VT = s1 u1 v1T + s2 u2 v2

T + . .

In other words, tropical forcing/global response relationships can be characterized by just a few forcing-response singular vector pairs

SBS-1

To demonstrate the important role of the tropical oceans, we have used the NCAR/CCM3 atmospheric GCM, which realistically captures the observed Land-

Average 850 mb Temp. changes with prescribed observed global SST changes (and without prescribed GHG changes)

Blue CurveObserved 850 mb land-averagetemperature(NCEP/NCARreanalysis)

Red CurveEnsemble-mean response to global SSTs with + 1 sigma of ensemblespread

An Array of localized SST patches

A comprehensive analysis of atmospheric sensitivity to Tropical SST

“Fuzzy Green’s Functions” :

Global CCM3 responses to an array of localized tropical SST anomaly patches

(Barsugli, Shin, and Sardeshmukh, Climate Dynamics, 2006)

BSS

CCM3’s GLOBAL MEAN temperature and precipitation responses to observed global SST changes over the last 50 years (GOGA runs) are well captured by linearly combining the responses to our tropical SST patches

This is an important result. It demonstrates both the dominance and linearity

of Tropical SST influences on the global mean climate.

Winter (DJF) Summer (JJA)

850 mbTemperature

Precipitation

Red CurvesResponse to global SSTs

Black CurvesLinear responseTo tropical SSTs

BSS

Local correlations of our “linear reconstructions” with the CCM3’s annual-mean global SST-forced responses over the last 50 years

These high correlations demonstrate the dominance and linearity of Tropical SST influences also on REGIONAL climate changes

BSS

Sensitivity of GLOBAL MEAN Temperature and Precipitationto SST increases at different tropical locations

(Barsugli, Shin, Sardeshmukh Clim. Dyn. 2006)

This implies that Global Warming will be very sensitive to the precise pattern of tropical ocean warming

Temperature Precipitation

“Low-Dimensionality” of DJF teleconnections

Top and Middle Panels

The “response” part of the SV pairs: 500 mb z, land surface temperature, and land precipitation

These are “optimal” response patterns

Bottom Panels:

The “forcing” part of the SV pairs:Tropical SST

These are “optimal” SST forcing patterns for generating responses with the largest amplitude

The first two forcing-response Singular-Vector (SV) pairs of G account for 72% of the structure of G

Dominant SST sensitivity patterns are VERY DIFFERENT from

the dominant pattern of observed interannual SST variability DJF 57%

MAM 30%

JJA 40%

SON 41%

EOF 1 of Observed SST 32%

There is much greater sensitivity to SST changes in the Central Pacificand the Warm Pool.

The sensitivity is generally opposite to SST changes in the Indian and Western Pacific portions of the Warm Pool .

Summary

1. Tropical SSTs account for most of the global responses to global SST variations.

2. The response to tropical SSTs is approximately linear; y = Gx

G contains all the sensitivity information.

3. G is a “low-dimensional” operator : That is, tropical forcing/global response relationships can be characterized by just a few “optimal” forcing-response Singular Vector pairs

Our analysis reveals the largest sensitivity - often with opposite signs - to SST changes in the Extended Tropical Warm Pool area of the largest recent and projected SST trends.

Climate models need to simulate and predict SST changes accurately in this area, and also to correctly represent the global atmosphere’s sensitivity to them.

Gij =∂yi / ∂x j