Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models

F. Lott

^{1}

,

S. Denvil^{2}, N. Butchart^{3}, C. Cagnazzo^{4}, M. A. Giorgetta^{5},

S. C. Hardiman^{3}, E. Manzini^{5}, T. Krismer^{5}, J-P Duvel^{1}, P. Maury^{1}, J. F. Scinocca^{6}, S.

Watanabe^{7}, and S. Yukimoto^{9}

{1} Laboratoire de Météorologie Dynamique, Paris, France

{2} IPSL, Paris France, {3} Met Office, Exeter, United Kingdom {4} ISAC-CNR, Roma, Italy {5} MPI Meteorologie, Hamburg, Germany {6} CCCMA , University of Victoria, Canada {7} JAMSTEC, Yokohama, Japan

{8} CRD-Met. Res. Institute, Tsukuba, Japan.

List of models (and data) used: CMIP5 with stratosphere, the 4 that has a QBO and 4+1 without to compare with.

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

●

Motivation:

–

An important step in CMIP5 is that many models represent the stratosphere : We need to diagnose how well they do, and in the

stratosphere the day to day variability is dominated by Kelvin and Rossby- gravity wave packets

● How well are they represented ?

● What make the differences between models ?

–

CMIP3 models present a large spread in their representation of convection variability (ENSO, MJO, Convectively Coupled Waves (Straub Haertel and Kiladis 2010). If the same is true for CMIP5:

● Does it impact the largest scales stratospheric waves or does the dynamical filtering ?

● Does other sources of waves balance tropospheric defects in large scale organization of equatorial precipitations ?

The method only used the fact that the gravest equatorial waves are characterized by one dynamical field being of uniform sign over the Equatorial Band and at a

given longitude:

u, T, and Z for the Kelvin waves (n=-1) and v for the Rossby Gravity waves (n=0).

10 °N -1 0 °S

10 °N -1 0 °S

In ERA40 and NCEP see the climatology in

Lott et al.~(2009)

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Rossby-Gravity Kelvin

Meridional wind at Eq. In September 1995 (ERA40, CI=2m/s))

From Lott et al. (JAS 2009)

In the low equatorial stratosphere, the day to day variability is dominated by Kelvin and Rossby Gravity wave packets : We need to have them correct in the physical space (not only in the

Spectral space), since they impact advection, de-hydratation, i.e. not only the QBO.

Zonal wind at Eq. In February in Feb 1999 (ERA40, CI=2m/s))

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Spectral analysis of 50hPa fields averaged over the Eq. Band and filter used

The same filter is applied to all models

½ power point of F ½ power point of F

Right: zonal wind spectra in the eastward propagating direction

and

Kelvin wave band-pass filter (in spectral space)

u ̃ (λ , ϕ , z , d )= ∑

s=1 s=nlon/2

∑

n=1 n=nday/2

F ( s , ω

_n

) ̂ u ( s , ϕ , z , ω

_n

) e

^{i(sλ−ωnd)}

Frequency (ω, Cy/day)

Filtered zonal wind

(u)

build from the zonal wind double Fourier transform

(û

^):

Left : meridional wind spectra in the westward propagating direction

and

Rossby-Gravity wave band-pass filter (in spectral space)

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Kelvin wave index:

K = Max ⏟

0< λ<360

[ ^{20 1 ∫}

⁻¹⁰¹⁰

^{u d ̃ φ} ]

KW= 1

20∫₋₁₀^{10 u d̃ ϕ} Rossby gravity wave index:

RG = Max ⏟

0<λ <360

[ ^{20 1 ∫}

⁻¹⁰¹⁰

^{v d ̃ φ} ]

Zonal mean zonal wind

u

Larger when u<0 : westward QBO

Larger when u>0 : eastward QBO

An exampl of dynamical filtering:

waves propagate

better when

u

and the wave phase speed have opposite signs

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Symmetric precipitations : spectra (lines)

And

coherency with U850 (color) Increasing precipitation variability

Convectively coupled waves

A large variability in precipitations

spectra

With QBO Without QBO

Frequency (Cy/day)

0.1

0.01

Frequency (Cy/day)Frequency (Cy/day)Frequency (Cy/day)Frequency (Cy/day)

0.1

0.01

0.1

0.01

0.1

0.01

0.1

0.01

ERAI and GPCP MRI

MPI-MR MPI-LR

MPI-MR

CMCC IPSL5A

IPSL5B HadGEM2-MR

MIROC Can-ESM2

-10 -8 -6 -4 -2 2 4 6 8 10 -10 -8 -6 -4 -2 2 4 6 8 10

Some models have : CCEWs strong enough to affect the precipitation

Variablity

CCEWs but not strong enough to impact the PREC

spectra (only visible in the coherency) No CCEWs at all The PREC variability varies a lot from one model

to the other (as in CMIP3, see Straub Hertel Kiladis 2013)

Wavenumber Wavenumber

CMCC

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Kelvin waves

All models have Kelvin waves (see eastward panel)

More substantial signals are for the models with QBO, which is confirmed when we compare the 2 MPI runs (Take care of the CIx4 factor in

The models with QBO only!!!).

Some sensitivity to the convection, confirmed when we compare

the two IPSL runs.

Between models, the effect of the convection is not as pronounced

as in Horinouchi et al.~2003.

Zonal wind at the equator spectra at 50hPa

ERAI MRI

MPI-MR MPI-LR

CMCC IPSL5A

HadGEM2-MR IPSL5B

MIROC Can-ESM2

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Kelvin waves

Hovmoller of u at the Equator

Lag (days)Lag (days)Lag (days)Lag (days)Lag (days)

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

-180 -90 0 90 180

Longitude (0 is arbitrary) MIROC Can-ESM2 HadGEM2-MR HadGEM2-MR IPL5B (Low precip variability) CMCC IPL5A (High precip variability) MPI-MR MPI-LR ERAI MRI -180 -90 0 90 180

-180 -90 0 90 180

Longitude (0 is arbitrary) -180 -90 0 90 180

-180 -90 0 90 180

-180 -90 0 90 180 -180 -90 0 90 180

-180 -90 0 90 180

-180 -90 0 90 180

-180 -90 0 90 180

Kelvin wave composites at 50hPa

amplitude are the The strongest for the models

with a QBO

and largest precipitation variability.

Without QBO With QBO

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Kelvin waves

Hovmoller of u at the Equator

Lag (days)Lag (days)Lag (days)Lag (days)Lag (days)

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

-180 -90 0 90 180

Longitude (0 is arbitrary) MIROC Can-ESM2 HadGEM2-MR HadGEM2-MR IPL5B (Low precip variability) CMCC IPL5A (High precip variability) MPI-MR MPI-LR ERAI MRI -180 -90 0 90 180

-180 -90 0 90 180

Longitude (0 is arbitrary) -180 -90 0 90 180

-180 -90 0 90 180

-180 -90 0 90 180 -180 -90 0 90 180

-180 -90 0 90 180

-180 -90 0 90 180

-180 -90 0 90 180

Kelvin wave composites at 50hPa

amplitude are the The strongest for the models

with a QBO

and largest precipitation variability.

The eastward phase speeds are quite correct

Without QBO With QBO

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Kelvin waves

Latitude 15N 0 15S

15N 0 15S

15N 0 15S

15N 0 15S

15N 0

15S-90 -45 0 45 90

Longitude (0 is arbitrary) -90 -45 0 45 90

Longitude (0 is arbitrary)

15N 0 15S

LatitudeLatitudeLatitudeLatitude

Can-ESM2 MIROC

HadGEM2-MR IPL5B (Low precip variability)

IPL5A (High precip variability)

CMCC

MPI-MR MPI-LR

ERAI MRI

-90 -45 0 45 90 -90 -45 0 45 90 -90 -45 0 45 90 -90 -45 0 45 90 -90 -45 0 45 90 -90 -45 0 45 90

-90 -45 0 45 90

All models have realistic Kelvin waves, when compared to ERAI

(which probably underestimates them, Ern et al. 2008 and after)

in the models without QBO, the zonal mean Zonal winds are westward all the time (U<0 not shown), which is the

favourable Situation for Kelvin waves vertical propagation. Hence it is not

surprise that all the models simulate well the Kelvin waves.

But models with QBO have stronger Kelvin waves, clearly due to their increased vertical resolution since in

them the Kelvin waves are filtered out half of the time

Kelvin wave composites at 50hPa u, v, and T at 0-day lag

Without QBO With QBO

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Kelvin waves

f

ERAI MRI

MPI-MR MPI-LR

CMCC IPSL5A

HadGEM2-MR IPSL5B

MIROC Can-ESM2

Meridional wind spectra at the equator and at 50hPa Models with a QBO have a much

Larger Rossby-GW signature (remember that

model without QBO have westward Winds, a situation not favorable

for waves with westward phase Speeds)

(Take care of the CIx4 factor in The models with QBO except

HadGEM2!!!).

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Rossby Gravity waves

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5

Lag (days)Lag (days)Lag (days)Lag (days)Lag (days) -10v 10

5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -120 -60 0 60 120 -10

Longitude (0 is arbitrary)

-120 -60 0 60 120 -120 -60 0 60 120

-120 -60 0 60 120 -120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120 Longitude (0 is arbitrary)

Rossby Gravity wave composites at 50hPa

Hovmoller of V at equator

MRI ERAI

MPI-MR MPI-LR

CMCC IPL5A (High precip variability)

HadGEM2-MR

HadGEM2-MR IPL5B (Low precip variability)

MIROC Can-ESM2

With QBO Without QBO

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Rossby Gravity waves

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5

Lag (days)Lag (days)Lag (days)Lag (days)Lag (days) -10v 10

5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -120 -60 0 60 120 -10

Longitude (0 is arbitrary)

-120 -60 0 60 120 -120 -60 0 60 120

-120 -60 0 60 120 -120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120 Longitude (0 is arbitrary)

Rossby Gravity wave composites at 50hPa

Hovmoller of V at equator

MRI ERAI

MPI-MR MPI-LR

CMCC IPL5A (High precip variability)

HadGEM2-MR

HadGEM2-MR IPL5B (Low precip variability)

MIROC Can-ESM2

With QBO Without QBO

Eastward phase speed but

Westward group velocity as expected for RGWs.

In models without QBO, the RGWs packets stay at the same place, the eastward intrinsic group

speed balancing the westward advection.

This is not the case in Models with a QBO, and

where the RG waves packets travel over very large distances.

This long distance travel is particularly pronounced

in the UKMO model!

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Rossby Gravity waves

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5

Lag (days)Lag (days)Lag (days)Lag (days)Lag (days) -10v 10

5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -120 -60 0 60 120 -10

Longitude (0 is arbitrary)

-120 -60 0 60 120 -120 -60 0 60 120

-120 -60 0 60 120 -120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120 Longitude (0 is arbitrary)

Rossby Gravity wave composites at 50hPa

Hovmoller of V at equator

MRI ERAI

MPI-MR MPI-LR

CMCC IPL5A (High precip variability)

HadGEM2-MR

HadGEM2-MR IPL5B (Low precip variability)

MIROC Can-ESM2

With QBO Without QBO

Eastward phase speed but

Westward group velocity as expected for RGWs.

In models without QBO, the RGWs packets stay at the same place, the eastward intrinsic group

speed balancing the westward advection.

This is not the case in Models with a QBO, and

where the RG waves packets travel over very large distances.

This long distance travel is particularly pronounced

in the UKMO model!

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Rossby Gravity waves

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5

Lag (days)Lag (days)Lag (days)Lag (days)Lag (days) -10v 10

5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -10

10 5 0 -5 -120 -60 0 60 120 -10

Longitude (0 is arbitrary)

-120 -60 0 60 120 -120 -60 0 60 120

-120 -60 0 60 120 -120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120

-120 -60 0 60 120 Longitude (0 is arbitrary)

Rossby Gravity wave composites at 50hPa

Hovmoller of V at equator

MRI ERAI

MPI-MR MPI-LR

CMCC IPL5A (High precip variability)

HadGEM2-MR

HadGEM2-MR IPL5B (Low precip variability)

MIROC Can-ESM2

With QBO Without QBO

Eastward phase speed but

Westward group velocity as expected for RGWs.

In models without QBO, the RGWs packets stay at the same place, the eastward intrinsic group

speed balancing the westward advection.

This is not the case in Models with a QBO, and

where the RG waves packets travel over very large distances.

This long distance travel is particularly pronounced

in the UKMO model!

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Rossby Gravity waves

Rossby-gravity waves composites at 50hPa:

All models have Rossby gravity waves, but the more realistic and

large signals are in models with QBO.

In them, the composite technique tends to pick dates when the QBO is eastward (U>0) which

is a favourable situation for RGWs vertical propagation.

As situation with U>0 never occurs in models without a QBO,

it is natural that these models do not reproduce

the RGWs as well as the models with QBO do.

Latitude 15N

0 15S

Latitude 15N

0 15S

LatitudeLatitudeLatitude 15N

0 15S

Latitude

15N 0 15S

15N 0 15S

15N 0 15S 15N 0 15S 15N 0 15S

-90 -45 0 45 90

Longitude (0 is arbitrary)c

-90 -45 0 45 90

Longitude (0 is arbitrary)c

Can-ESM2 MIROC

-90 -45 0 45 90 -90 -45 0 45 90

HadGEM2-MR IPL5B (Low precip variability) -90 -45 0 45 90

15N 0 15S 15N 0 15S

IPL5A (High precip variability)

CMCC

-90 -45 0 45 90

MPI-MR MPI-LR

-90 -45 0 45 90

ERAI MRI

-90 -45 0 45 90 -90 -45 0 45 90 -90 -45 0 45 90

Without QBO With QBO

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Rossby Gravity waves

The High top CMIP5 models simulate realistic aspects of the stratospheric large- scale Kelvin and RG waves.

They represent them better than the tropospheric convectively coupled waves.

There is nevertheless a large spread among the models, and those with a Quasi- Biennial Oscillation (QBO) produce larger amplitude waves than the models without For the RG waves this follows that models without a QBO never have

u>0

^{, a}

situation that is favorable to the propagation of westward propagating waves..

For the Kelvin waves, larger amplitudes in the presence of a QBO is counter intuitive because Kelvin waves are expected to be larger when

u<0

and this is always satisfied in models without QBO. We attribute the larger amplitude to the fact that models tuned to have a QBO require finer vertical resolution in the stratosphere.

Models with large precipitation variability tend to produce larger amplitude waves.

The effect is not as pronounced as found in previous studies.

In fact, even models with weak precipitation variability still have quite realistic stratospheric waves : (i) other sources can be significant or (ii) the dynamical filtering mitigates the differences in the sources between models.

Lott et al.~(JGR 2014)

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Other sources in models : Exemple of the Composite of EP fluxes between a model and in re-analysis (Maury and Lott ACP 2014)

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.

Other sources in reality : Exemple of the Composite of RGWs in re-analysis, when QBO winds underneath are negative

(Stratospheric reloading in Maury and Lott ACP 2014)

Kelvin and Rossby gravity wave packets in the lower stratosphere of CMIP5 models.