Whither ENSO?Assessing El Niño/Southern Oscillation risks for the coming decades

Andrew WittenbergNOAA/GFDL

Earth's dominant year-to-year climate fluctuation:

NOAA/CPC

El Niño

Normal

How will ENSO behave in the coming decades?

CMIP5 & AR5 – new focus on near-term (30yr) projections.

ENSO drives/confounds global SI-to-decadal variability.

If we knew the next decade would have: - mega-ENSO → insure, invest, prepare, monitor & model - no ENSO → different investments, e.g. rebuild habitats

Both historical & paleo recordssuggest past modulation of ENSO

Vecchi & Wittenberg (WIREsCC 2010)

Historical SSTA (ERSST.v3)

Palmyra corals(Cobb et al.,Nature 2003)

Multiproxy reconstructions:e.g. Emile-Geay et al.

(2011abc, subm.)

Projected ENSO changes (CMIP3/AR4)

correl(SST trend of 1%/yr, SST.PC1 of PICTRL)10S-10N, 120E-80W

Yamaguchi & Noda (JMSJ 2006)

std(

SLP.

PC1

of S

RES.

A2 (2

051-

2100

))/ s

td(S

LP.P

C1 o

f 20C

3M)

30S-

30N,

30E

-60W

van

Olde

nbor

gh e

t al.

(OS

2005

)

CM2.1

Weak/ambiguousnear-term

anthropogenicimpacts on ENSO

Intrinsicmodulation

Reviews:Meehl et al.

(IPCC-AR4 2007)

Guilyardi et al.(BAMS 2009)

Vecchi & Wittenberg (WIREs CC 2010)

Collins et al.(Nature Geosci. 2010)

4000-year pre-industrial control run 1860 atmospheric composition, insolation, land cover 220yr spinup from 20th-century initial conditions big investment: 2 years on 60 processors

Delworth et al., Wittenberg et al., Merryfield et al., Joseph & Nigam (JC 2006), Wittenberg (GRL 2009)Zhang et al. (MWR 2007); van Oldenborgh et al. (OS 2005); Guilyardi (CD 2006); Reichler & Kim (BAMS 2008)

Kug et al. (JC 2010), Vecchi & Wittenberg (WIREsCC 2010), Collins et al. (Nature Geosci. 2010)

1990 control (300yr), 2xCO2 (600yr), 4xCO2 (600yr)

GFDL CM2.1 global coupled GCM atmos: 2°x2.5°xL24 finite volume ocean: 1°x1°xL50 MOM4 (1/3° near equator) 2hr coupling; ocean color; no flux adjustments ENSO & tropics rank among top AR4-class models SI forecasts; parent of GFDL AR5 models (ESM2M, ESM2G, CM3, CM2.5)

What sort of ENSO do we simulate?

These are from asingle run with

unchanging forcings.

strong, skewed,long period,

eastward propagating(1980s & late 1990s)

weak, biennial, “Modoki”(early 1990s & 2000s)

regular &westward propagating

(1960s & 70s)

20 centuries of simulated NINO3 SSTsannual means & 20yr low-pass

Wittenberg (GRL 2009)

Modulation of NINO3 SST power spectrum

2000yr mean

(e.g. satellites, TAO) (e.g. reconst SST)

Wittenberg (GRL 2009)

Centuries of weak or strong ENSOs

CM2.1 mean state hardly differs between active/inactive ENSO centuries

Will forecasts capture anything at all?

How predictable are decades of extreme ENSO?

Will forecast ensembles differ between epochs?

°C°C

Will forecasts capture the intensity of the epoch?Will forecasts capture events, if not their timing?Will forecasts track the control?Tiny perturbation:+0.0001C at one gridcell (equator, 180W, top 10m)NINO3 SSTA

“Perfect” ensemble reforecasts

Some members resemble the control.(forecasts with minimum NINO3 SST RMS error over each decade)

°C°C

“Perfect” ensemble reforecasts

Other members look nothing like the control.(forecasts with maximum NINO3 SST RMS error over each decade)

°C°C

“Perfect” ensemble reforecasts

These are what perfect forecasts look like!(perfect model, near-perfect initial conditions, 40 members)

°C°C

“Perfect” ensemble reforecasts

Summarize the ensemble PDF with quartiles.25th and 75th percentiles of NINO3 SSTA, from 40 members

°C°C

“Perfect” ensemble reforecasts

Quartiles “forget” initialization after a few years.gray: 95%-bands for control quartiles, from 5000 resampled 40-ensembles

°C°C

Ensemble spread40-member interquartile range (IQR), with 95%-band from control

°C°C

“Perfect” ensemble reforecasts

Decadal statistics of ENSO

Can we predict the epoch-mean amplitude?Absolute value of NINO3 SSTA (degC)

°C°C

Decadal statistics of ENSO

Smoothed measure of ENSO activityNINO3 SSTA amplitude, smoothed with 4yr running mean

°C°C

Decadal statistics of ENSO

Ensemble forecasts of ENSO activitysmoothed NINO3 SSTA amplitude from 40 members

°C°C

Decadal statistics of ENSO

Activity distributions also “forget” the ICs{10,50,90}-percentiles of smoothed amplitude, with 95%-bands from control

°C°C

Long-term memory?

median6yr

10% >15yr

5-year waitmost common

recharge delay

consistentwith Poisson

But beyond 10 years?

Distribution of inter-event wait times suggests that NINO3

SSTA might have some memory beyond 5 years.

Even a purely memorylessENSO would give occasional

waits of 20 years or more,as seen in CM2.1.

Wittenberg (GRL 2009)

3822yr / 495 events= 7.7yr mean wait

ENSO events and their nearest neighbors

strong events more

isolated

weak EN, 3yrafter strong EN

strong EN, 4yrafter weak EN

Best hope for long-term ENSO predictability?NINO3 memory might last 5yr, following strong warm events.

Best hope for long-term ENSO predictability?NINO3 memory might last 5yr, following strong warm events.

Given enough years, we can say...

CM2.1 ENSO is too strong

CM2.1 ENSO is very sensitive to some parameters

Pre-industrial range of 100yr spectra

1990: ENSO strengthens, spectrum narrows

2xCO2: slightly shorter period than 1990

4xCO2: ENSO weaker than at 2xCO2

Can we extrapolate ENSO projections to reality?

weak ENSOs

?

Merryfield (JC 2006)

Vecchi & Wittenberg (WIREsCC

2010)

CM2.1

CM2.0

Futu

re E

NSO

am

plifi

catio

n

Width of wind stress response

The “most realistic”

pre-industrial ENSOs show amplification

at 2xCO2

Summary

1. CMIP3/AR4 projections were ambiguous for ENSO: a. Diverse responses to anthropogenic forcings b. ENSO modulation in models, historical/paleo records

2. 4000-year run of pre-industrial CM2.1: a. Strong intrinsic modulation of ENSO b. Extreme ENSO centuries: not due to climate shifts c. Extreme ENSO decades: - Multidecadal lulls consistent with memoryless ENSO - ENSO memory up to 5 years after strong warm events - After that, even perfect forecasts don't beat a memoryless PDF

3. With long enough ENSO records, we can still detect: a. Model biases & sensitivities to some parameters/forcings b. CO2 impacts (barely detectable with 100yr record) c. CO2 optimum for ENSO - A source of disparate model sensitivities? - How close is the optimum, and which side are we on?

Projecting ENSO risks for the coming decades

1. What is ENSO capable of on its own? - long runs, large ensembles - historical/paleo reanalyses, pseudoproxies - impacts of extreme events

3. Understand ENSO's sensitivities - primary controls, feedbacks & nonlinearities - diverse tests: forecasts, volcanoes, paleo, idealized - model diversity + physical understanding -> extrapolation to reality

4. Decadal forecasts - does precise initialization matter for ENSO? - intrinsic modulation may dominate ENSO behavior over our lifetimes

2. Improve models, understand & convey their uncertainties - metrics: robust, grounded in theory (ICMs), community-wide - AR5: new feedbacks, better resolution/physics -> different projections?

Reserve Slides

Last Glacial Maximum (20ka)tropical SST cools 3°C

TC deeper & more diffuse

Mid-Holocene (6ka)perihelion shifts from

Jan -> Oct;less SH seasonality

seasonal/ENSO confounding inpaleo proxies?

Spectrum of NINO3 SST

Wittenberg (GRL 2009)

less phase-locking of cold events & weak warm events

strong warm events peak in SON

warm events are stronger& rarer than cold events

CM2.1 ENSO peaksvs. calendar month

abs(NINO3) > 1 stddev

CM2.1 mean state hardly differs between active/inactive ENSO centuries

100yr-mean SST & trades are robust diagnostics for CM2.1

Inactive centuries have slightly warmer water in the west Pacific

Active/inactive centuries show no systematicdifference in the scaled anomaly patterns

Active/inactive centuries show no systematicdifference in the scaled anomaly patterns

Warm pool intraseasonals are slightlymore variable during active-ENSO centuries

Two extreme epochs

Activity spread40-member IQR of 4yr-smoothed amplitude, with 95%-band from control

°C°C