Stratospheric climate and variability of the CMIP5 models

© University of Reading 2006www.met.reading.ac.uk/research/

stratclim/21 April 2023

Stratospheric climate and variability of the CMIP5 modelsAndrew Charlton-Perez, Mark Baldwin, Thomas Birner, Robert Black, Amy Butler, Natalia Calvo, Nicholas Davis, Edwin Gerber, Nathan Gillett, Steven Hardiman, Junsu Kim, Kirstin Krüger, Yun-Young Lee, Elisa Manzini, Brent McDaniel, Lorenzo Polvani, Thomas Reichler, Tiffany Shaw, Michael Sigmond, Seok-Woo Son, Matthew Toohey, Laura Wilcox, Shigeo Yoden, Bo Christiansen, François Lott, Drew Shindell, Seiji Yukimoto, Shingo Watanabe

Comparing high-top and low-top CMIP5 models

• Where are there broad differences between high-top and low-top models in CMIP5?

• Look at model performance vs. re-analysis for basic dynamical diagnostics

• Focus on historical runs of the models

• Validation against re-analysis (ERA-Interim and MERRA)

• Red = High-top; Blue = Low-top

Models usedHigh-Top Max

EMNo.

diagnostics

Low-top Max EM

No. of diagnostic

s

GFDL CM3 5 5 BCC-CSM1.1 3 3

GISS-E2-R 5 5 CCSM4 5 5

GISS-E2-H 15 3 CNRM-CM5 1 4

HadGEM2-CC

3 6 CSIRO-mk3.6.0

10 5

IPSL-CM5A-LR

4 4 GFDL-ESM2M 1 6

IPSL-CM5A-MR

1 4 GFDL-ESM2G 1 1

MIROC-ESM 3 5 HadCM3 10 4

MIROC-ESM-CHEM

1 6 HadGEM2-ES 4 6

MPI-ESM-LR 3 6 INMCM4 1 3

MRI-CGCM3 3 6 MIROC5 4 6

NorESM1-M 3 6

Metrics

Taylor plotThick – CMIP5 high-topThin – CMIP5 low-topThick dash – CCMVal-2Thin dash – CMIP3

90S-90N, 100-10hPa

Ovals – 2 s.d. sampling intervals for model ensemble

Temperature biases

Jet Biases

1975-2005 vs. ERA-Interim

Grey shading - 95% confidence intervalCircles – MERRA climatology

SSW Frequency/Variance

Variance of 60N and 50hPa DJF zonal mean zonal wind

Stratosphere-

Troposphere

coupling

Baldwin – ‘dripping paint’ NAM index plots for events when index drops below -3 standard deviations at 10hPa

Volcanic response

50hPa geopotential height anomaly for two winters following El Chichon and Pinatubo eruptions (only one OND for El Chichon)

Temperature Trends

Lower stratospheric temperature anomalies compared with RSS SSU datasetModels are processed to have the same weighting function as satellite channelShading shows approx. 5-95% sampling interval for ensemble

Volcano years excluded from bottom plot

Main conclusions

For models in the CMIP5 ensemble:

1. Mean climate biases in the lower stratosphere is similar for low-top and high-top ensemble,

2. Stratospheric variability is too weak in the low-top ensemble,

3. This has a potential impact on stratosphere-troposphere coupling which is also weaker in the low-top models,

4. The simulation of stratospheric temperature trends is similar in the low-top and high-top ensemble.

11

What next?

Some things to think about in the context of DynVar:

1. Can we make progress on persistent biases in both model ensembles (e.g. cold biases in lowermost stratosphere)?

2. Characterising differences in CMIP5 important, but how much does this reflect general low-top/high-top differences?

3. Link between variability and coupling worth exploring in more detail.

12

Implications for CMIP6/other experiments

• Likely that many/most models in CMIP6 will include a fully resolved stratosphere – low-top/high-top comparison outdated?

• How do we efficiently characterise stratospheric climate/variability in models.

• Which elements of the stratospheric climate is it necessary to simulate to capture stratosphere-troposphere coupling?

13

Extra Slides

14

Trends continued

1979-2011Top – obs.Middle – High-top (hatching where observed trends outside range of models)Bottom – Low-top (hatching where low-top and high-top different)

Tropopause

Seasonal cycle/Final Warming

Tropical Upwelling