Tropical variability in ACCESSCM 2 simulations Harun Rashid

Tropical variability in ACCESS-CM 2 simulations Harun Rashid, Climate Science Centre, CSIRO Oceans and Atmosphere

Model experiments and observations • Preindustrial control simulations: • CM 2 -pi. Cont (1850 -2185) • GC 3. 1 -LL (1850 -2349) • Historical simulation: • CM 2 -Hist (1850 -2014) • Variables: • Sea surface temperatures (SSTs) • Zonal wind stress (Taux) • Thermocline depth (Z 20) • Observations: • Had. SST, ERA-40/ERA-Int and PEODAS (1960 -2017) • Stats: Mean, variance, EOFs and spectral analysis

Time mean SST, Taux and Z 20. SST: Cold tongue bias in the western to central Pacific in both models Taux: Westerly bias in the central to eastern Pacific Z 20: Shallower CM 2 thermocline, with a shallow patch to the south

Temporal standard deviations SST: Max SST variance is confined to the Nino 3 region (east Pacific) for CM 2 Taux: Smaller Taux variance in the tropics than in obs Z 20: Also smaller tropical Z 20 variance

Zonal profiles of time means and variability Similar cold SST bias in the western and centraleastern Pacific Easterly (westerly) bias in the (western) eastern Pacific The simulated SST variance is too strong in the eastern Pacific Too little taux variability in the Niño-4 region Too shallow thermocline depth in CM 2 Niño-3 Z 20 variability is weaker than obs

EOFs of Pacific SST anomalies The 1 st EOF is meridionally too confined in both CM 2 models The 2 nd and 3 rd EOFs are not properly simulated in CM 2

EOFs of zonal wind stress anomalies Too little equatorial Taux variability The leading EOFs are dominated by subtropical variances, unlike in obs

EOFs of thermocline depth anomalies The leading eastwest mode is not pronounced in CM 2 models (appears as the 2 nd mode? ) The 1 st EOF in CM 2 looks more like the recharge-discharge mode (2 nd mode in obs)

Seasonality in the Nino-3 region

Power spectra

Seasonal phase locking Correct seasonality, except for the Nino 3 index

Asymmetries in ENSO and IOD (using PDFs of Nino 3. 4 and IOD indices) OBS: El Ninos are stronger than La Ninas (+ve skewness) Models: La Ninas are stronger than El Ninos (-ve skewness) Positive IODs are stronger than the negative IODs (+ve skewness)

Lag-regression SST (shades) and D 20 (contours) anomalies regressed onto the Nino-3 index D 20 anomalies (Kelvin waves) propagate too fast in ACCESS-CM 2

Summary • The simulated ENSO events are too strong and biennial in CM 2 -pi. Cont, but slightly better in the historical run. • The mean thermocline depth is shallower than observed in the equatorial Pacific • ENSO and IOD seasonality is correctly simulated • The leading Taux EOF is not correctly simulated • The leading Z 20 EOF is the recharge-discharge mode • Negative ENSO asymmetry in the models • Too fast Kelvin wave propagation may be responsible for the shorter ENSO periods

Cross-correlations and coherence The taux forcing of SST is realistic, but the SST feedback is weak in CM 2 SST, Taux and Z 20 in CM 2 are strongly coupled at the biennial frequency The simulated taux 4 -D 20 lag correlations are not realistic