Predictable signals associated with teleconnections

from tropical phenomena

Franco Molteni,

Stephanie Johnson, Chris Roberts, Retish Senan,

Tim Stockdale, Frederic Vitart (ECMWF, Reading, U.K.)

with contributions from the PRIMAVERA project partners (www.primavera-h2020.eu)

October 29, 2014

Teleconnections from DJF tropical rainfall in Pacific and Indian Oceans

2

WCIO

40E-90E

10S-10N

Nino4w

160E-150W

10S-10N

ERA-Int.

Local correlation between rainfall and SST

(Molteni et al., Climate Dyn 2015)

October 29, 2014

Seasonal forecasts and multi-decadal historical simulations at ECMWF

• a) Seasonal fc. System 5 (Seas5)

– IFS cy43r1 Tco 319 (~32 km grid) L91 + NEMO v3.4 ORCA 0.25 deg. Z75+ LIM2 sea-ice

– 7-month forecasts, 13-month fc. From Feb/May/Aug/Nov

– Ensemble size: operational fc.: 51 members, re-forecasts: 25 members

– Re-forecast period: Jan 1981 – Dec 2016 (36 years), IC from ERA-interim + ORA-S5

– Verification datasets: ERA-Interim, GPCP v2.3 precipitation

• b) Multi-decadal historical simulations for the EU PRIMAVERA project, following HighResMIP (ECM-hist)

– High res.: IFS cy43r1 Tco 399 (~25 km grid) L91 + NEMO v3.4 ORCA 0.25 deg. Z75 + LIM2 sea-ice

– Low res.: IFS cy43r1 Tco 199 (~50 km grid) L91 + NEMO v3.4 ORCA 1.0 deg. Z75 + LIM2 sea-ice

– CMIP6 forcing fields (GHG, aerosol, ozone, …)

– 1950-2014 started from 50-yr spin-up (1950 forcings)

– Additional runs: AMIP integrations using HadISST2 data (SST + sea-ice conc.), 1950-forcing control

– CERA20C (1950-2010), precipitation data calibrated w.r.t. GPCP v2.3 in overlapping period

3

October 29, 2014

Seas5: teleconnections from DJF tropical rainfall in Pacific and Indian Oceans

4

ERA-Int Seas5 IFS as Seas5, obs. SST

WCIO

40E-90E

10S-10N

Nino4w

160E-150W

10S-10N

October 29, 2014

Questions:

• Are the teleconnections robust in the observational record?

• Do we understand the dynamical mechanisms that connect heating anomalies in the Indian and Pacific Ocean to North Atlantic variability?

• Are the deficiencies shown in System 5 reproduced in multi-decadal runs of the ECMWF model, when the model has reached its own attractor?

• Is the difficulty in reproducing the Indian Ocean teleconnection common to other GCMs?

• How do teleconnections (and errors in modelling them) relate between the seasonal and sub-seasonal scale?

5

October 29, 2014

Covariance with NAO index based on 500 hPa height in Nov-Dec and Jan-Feb

6

October 29, 2014

Tropical rainfall covariances based on 2-month means

7

WCIO DJ

Nino4w ND

Nino4w JF

Z 500

WCIO DJ

Nino4w ND

Nino4w JF

T 850

October 29, 2014

CERA20C 1951-2010 - CERA20C 1981-2010 - ERA-In./GPCP 1981-2010

8

WCIO

DJ

Nino4w

ND

Nino4w

JF

October 29, 2014

Sampling uncertainty in teleconnection: prob. density function of NAO index

9

Results from

bootstrap estimate:

200 60-yr samples

with years randomly

selected from

CERA20C data

October 29, 2014

Heat transport by planetary waves: Fletcher & Kushner 2011, 2013

10See also: Garfinkel et al. 2014

3-variable NAO model: dynamics and physical feedbacks(Molteni and Kucharski, Clim. Dyn. 2019)

• Vorticity advection

• Divergence of meridional heat

transport

• Thermal dissipation due to surface

heat fluxes

• Relaxation towards forced

equilibrium state driven by long-

wave radiative damping

From NAO diagnostics:

γ ≈ σ ≈ 2, κ ≈ 0.5, B* ≈ 12

U* = 0

October 29, 2014

CERA20C ECMWF historical ensembles (1950-2010, 6 members)

AMIP low-res. AMIP high-res. Coupled high-res.

12

October 29, 2014

Taylor diagrams for PRIMAVERA ensembles (1950-2010)AMIP low-res. AMIP high-res Coupled high-res.

13

N. Atlantic/

Europe

80W-40E

30N-85N

N. Pacific/

N. America

160E-80W

30N-85N

October 29, 2014

Zonal-mean and wave component of the Indian Ocean teleconnection

14

October 29, 2014

Norm. error of teleconnections in AMIP/coupled PRIMAVERA runs 1950-2010

15

Models:

1. CMCC-CM2

2. CNRM-CM6

3. EC-Earth 3

4. HadGEM3 GC3.1

5. MPI-ESM 1-2

6. ECMWF-IFS

7. ECMWF ens.6 members

AMIP Low resol. ≈ 100 km

AMIP High resol: 20 - 40km

Coupled High-res:

atmos: 20-40km,

ocean: ~1/4 degree

WCIO DJ

Nino4w ND

Nino4w JF

Z 500

N. Atlantic/Europe N. Pacific/N. America

October 29, 2014

Telecon. with NINO4w rainfall: CERA20C vs. PRIMAVERA coupled h.r. runs

16

October 29, 2014

Telecon. with WCIO rainfall: CERA20C vs. PRIMAVERA coupled h.r. runs

17

October 29, 2014

Telecon. with WCIO rainfall: CERA20C vs. single ECMWF ens. members

18

October 29, 2014

Errors in MJO phase frequency and Z500 teleconnections in SEAS5

From F. Vitart (SAC 2018

Special Topic paper on SEAS5)

MJO phases as in

Wheeler & Hendon 2004

19

October 29, 2014

Diagnostics of MJO propagation and teleconnections

20

∂u200/∂λ

- ∂u850/∂λ

[10S-10N]

u500 [40N-65N]

NAO

October 29, 2014

MJO propagation and teleconnections from 75 E

21

October 29, 2014

MJO - u500 / u200 / u100 teleconnections from 90 E

22

NAE sector:

135W - 45E0.5 m/s0.7 m/s0.6 m/s

October 29, 2014

Statistics from 300-winter runs of the 3-variable NAO model (Molteni & Kucharski 2019)

U* = 0 U* = 0.75

mean U’ 0.02 1.03

mean A 0.03 1.10

mean B 3.54 3.29

st.dev. A 2.57 2.25

23

Probability density

function for

monthly-means of

the NAO-like wave

component A in

runs with U* = 0

and U* = 0.75 m/s

1.10 / 2.57 = 0.43

October 29, 2014

Inter-decal variability in HadISST and CERA20C: 1981/2010 – 1951/1980 (1951/1980)

24

October 29, 2014

Summary

• Most of the errors in the simulation of interannual & intraseasonal teleconnections found in the

ECMWF seasonal forecast SEAS5 are also present in multi-decadal historical runs of the

ECMWF coupled model. On both scales, simulations with prescribed, observed SST produce

Indian Ocean teleconnections in much better agreement with observations.

• Teleconnections associated with late-winter rainfall anomalies in the Nino4 region are well

simulated in the historical runs of the PRIMAVERA models, while those originated from the

Indian Ocean are poorly represented despite their robustness in the observational record.

• On the sub-seasonal scale, deficiencies in the ECMWF model teleconnections are related to

problems in representing a coherent MJO propagation across the tropical Indian and West

Pacific oceans in long integrations of the coupled model.

• 30-yr inter-decadal differences in tropical rainfall and NH geop. height in CERA20C appear to

be consistent with seasonal-scale teleconnections during the satellite era.

• Despite progress in our theoretical understanding of the link between Indian Ocean heating

anomalies and North Atlantic variability, the simulation of teleconnections and inter-decadal

variability originated from the Indian Ocean shows significant deficiencies in many state-of-the-

art climate models. 25