Reconstructing El Ni

1
Reconstructing El NiñoSouthern Oscillation
(ENSO) using high-resolution palaoarchives

• Joëlle Gergis
• PhD graduate, School of Biological Earth
  Environmental Sciences, UNSW, Sydney Australia
• In collaboration with
• Anthony Fowler, Tree-Ring Laboratory, University
  of Auckland, New Zealand
• Karl Braganza, National Climate Centre, Bureau of
  Meteorology, Melbourne Australia
• Scott Mooney, Biological Earth Environmental
  Sciences, UNSW, Sydney Australia
• James Risbey, Marine Research, CSIRO, Hobart
  Australia
• Scott Power, Bureau of Meteorology Research
  Centre, Melbourne Australia

2
Why Study El Niño-Southern Oscillation (ENSO) ?

• Largest source of inter-annual climate
  variability affects 60 of the planet
• Modulates high-impact climate extremes
  droughts, floods, cyclones, bushfires

3
What was ENSO like in the past ?

• Relatively limited understanding of
  pre-observational ENSO
• How have the frequency and magnitude of ENSO
  event cycles changed?
• What were the societal impacts of past ENSO
  events?

• Use past ENSO behaviour to help refine
  predictive capability of GCMs
• Current IPCC models do not simulate ENSO very
  well
• It is still unclear how ENSO will respond to an
  increasingly warm world

4
Research Objectives
Aim to assess how unusual recent ENSO
variability has been in the context of the past
five centuries

1. Review the use of observational ENSO indices used
   for proxy calibration/reconstruction

2. Capturing regional ENSO signals from proxy
records West Pacific example NZ Kauri
tree-ring record

• ENSO reconstructions results (A.D. 1525-2002)
  using tree-ring, coral, ice documentary records
• Discrete event analysis (Quinn, Ortlieb)
• EOF analysis (Mann, Stahle, DArrigo)

5
Defining instrumental ENSO for palaeoclimatic
studies
Event DEFINITION critical for proxy calibration/
transfer function development

• TWO components of the ENSO system
• Atmosphere pressure (Southern Oscillation)
• Ocean temperature (Niño Region SSTs)

• Differences in ENSO event signatures and El
  Niño/ La Niña phase transitions
• - Nov 2007 east Pacific SST cooling, relatively
  minor western Pacific signal, SOI only weakly
  positive

6
Instrumental Records of ENSO SSTs SOI
Event capture comparison Hanley et al. (2003),
Journal of Climate 161249-1258

• SOI has lower La Niña event capture than most
  SSTs
• Classic South American El Niño region 12
  notably poor (lowest) event capture
• Niño 3 good El Niño sensitivity, but La Niña
  sensitivity lower than other SST indices
• West Pacific SSTs (Niño 3.4 and Niño 4) most El
  Niño sensitive
• Niño 3.4 best event capture of BOTH phases (Niño
  4 less La Niña sensitivity)

7
Instrumental ENSO Classification Issues

• - Little widespread consensus on the definition
  of ENSO- issue is non-trivial
• - Atmospheric oceanic components can be out of
  phase (decoupled) magnitude of event
  anomalies differ
• - Limited attempts to combine oceanic
  atmospheric indices for palaeoclimate applications

8
Coupled ENSO Index (CEI)

• - Simultaneously indicates ocean (Niño 3.4 region
  SSTs) atmospheric (SOI) anomalies
• - SOI SST combined so anomalies indicated by
  each component index are preserved
• - Coupled (synchronous) events indicated by
  amplification of magnitude
• - New composite index BOTH ENSO components
  maintained gt amplitude preserving calibration
• - Common basis for comparison of proxy records
• (Gergis and Fowler (2005), International
  Journal of Climatology, 2515411565)

9
High-resolution ENSO proxy records

• Instrumental records too SHORT to resolve
  decadal-centennial ENSO variability
• High resolution proxies record seasonal and
  annual climate variability
• Complementary aspect of ENSO e.g marine vs.
  terrestrial, tropics vs. mid-latitudes
• Relatively limited Southern Hemisphere and
  Western Pacific coverage

10
Western Pacific ENSO proxy New Zealand Kauri
tree-ring record

• 17 modern sites (living trees)
• 11 sub-fossil sites (swamp material)

11
New Zealand Kauri tree-ring record

• NZ Kauri record now longest tree-ring record in
  the Western Pacific 3,726 yrs continuous
• Calendar dated 1724 B.C. - A.D.2002
• Long-lived species rare only handful of ENSO
  sensitive tree-ring records available globally

• Provides long-term history of ENSO from Western
  Pacific
• Important counterpart to East Pacific tree-ring
  records from SW USA Mexico e.g Stahle et al.
  (1998), DArrigo et al. (2005)

12
Kauri response to ENSO indices

• Kauri shows good statistical relationships with
  ENSO indices correlations 0.6 over the
  1950-2002 period
• (Fowler et al. (2007) International Journal of
  Climatology, in press)

• - Peak sensitivity of SON and DJF (peak maturity
  of ENSO)
• Stronger relationship with (western pole) SOI
• Correlations with CEI stronger than any of the
  component indices

13
Reconstructing the CEI using Kauri tree-rings
Spring (SON) CEI regressed using kauri
tree-rings (r 0.45, 1871-2002)

• - Every proxy IMPERFECT lt50 instrumental
  variance common
• - Regional signals only a PART of the large-scale
  ENSO signal
• Complicated by which ENSO index used for
  calibration (SOI, SSTs, CEI ??)
• Differences in phase sensitivity e.g. Kauri has
  a stronger El Niño signal

14
Kauri master chronology (A.D. 15502002)

• - Distinct 5070 year cycle evident in 31-yr
  moving variance
• Baseline shift in variance around 1870 (end of
  Little Ice Age) the start of large-scale
  (global) industrialisation
• Late C20th appears unusual (5 widest rings occur
  post 1982)

15
Multiproxy approaches to ENSO reconstruction
- Single proxy analysis limited each record has
unique REGIONAL event signature - MULTIPLE
records allows us to look at GLOBAL patterns of
ENSO variability - Large-scale spatial patterns
event magnitudes (Gergis et al., 2006)
Dec 1997-98 SST anomalies NCEP/Reynolds SST
analysis
N.B selection based on published ENSO sensitivity
continuous record length
16
Reconstructing Discrete ENSO events

• Records assessed by hits and misses against
  instrumental record (CEI, each season) gt proxy
  skill
• Percentile analysis no loss of variance,
  outliers usually excluded from regression
  maintained
• Calibration shows differences in
  phase-sensitivity 2 separate reconstructions
  based on proxy subsets
• Tree-ring records best overall performers
  (replication gt exact dating), uncertainties in
  coral dating??

(Gergis Fowler, 2006 Gergis Fowler, Climatic
Change in review)
17
ENSO Event lists
Notable phase sensitivities at single and
multiple proxy level.
92 El Niño 82 La Niña events reconstructed
since A.D. 1525

1. This study is the 1st to use extensive East West
   Pacific ENSO proxies
2. Range of (quantified) event magnitudes
3. Reconstruction quality assessed
4. Most comprehensive La Niña event list compiled to
   date
5. Considerably expands existing historical El Niño
   chronologies (South American Quinn records)

Gergis Fowler, 2006 Gergis Fowler, Climatic
Change in review
18
Decadal trends in ENSO event magnitude

• Quality-adjusted Magnitude (MQ) time-series
  takes replication and proxy skill into account
• - Five percentile based magnitude classes
  calculated for each phase (outliers maintained
  rather than truncated using regression)

• C16th -17th overall reduction in ENSO activity
  during the Little Ice Age
• C18th -19th 65 of events classed as
  weak-moderate
• C20th 43 of all extreme events, 30 of total
  extreme event years are post-1940

19
Reconstructing ENSO indices using EOF analysis

• Empirical Orthogonal Function (EOF) analysis
  used to decompose proxy data into leading modes
  of co-variability based on current (t), lead
  (t-1) and lag (t1) relationships
• Best results with CEI DJF (t0) r -0.72 and
  SON (t1) r -0.67
• 53 of C20th DJF CEI variance 48 of DJF SOI
  and 49 MAM Nino 3.4 SSTs
• Peak maturity of ENSO conditions best captured
  by coupled anomalies retained in CEI
• Monte Carlo synthetic data testing used for
  verification

20
Spectral correspondence over the instrumental
period
Braganza et al., Climate Dynamics in review

• Power spectrum for instrumental DJF CEI
  reconstructions for the period 1871-1982.
• Reconstructions reproducing frequencies in
  classical 2-7 year ENSO band
• Significance at the 90 and 95 (dotted lines)
  level is indicated relative to estimated
  background AR1 noise, effective bandwidth after
  smoothing 2.66/N cycles/year.

21
Past ENSO amplitude and frequency modulation
Hovmoller of the 50 year moving (window) power
spectrum for R5 for 1525-1982. Effective
bandwidth after smoothing 2.66/N cycles/year.
Evidence of increased high frequency
variability over the past 2 centuries
22
Low frequency Inter-decadal Pacific Oscillation
(IPO) The IPO is the Southern Hemisphere
equivalent of the PDO
Proxy-ENSO indices R5 and R8 are shown with the
low frequency component of the SOI (Braganza et
al., Climate Dynamics, in review).
23
Complementary applications of ENSO reconstructions
- Event chronologies used for pre-instrumental
verification/cross-validation - Stahle, Rodbell,
Mann etc use Quinn chronologies from South
America to define ENSO event
reference years
NB Quinn record is a documentary record of ONE
(East Pacific) regions impact of El Niño
eventsWest Pacific La Niña NOT considered
Source Mann et al. (2000)
24
Summary
1) Multi-proxy reconstructions of ENSO still in
its infancyencouraging results but there are
many opportunities for resolving more information
on past ENSO behaviour
2) Further research needed to examine regional
synoptic features, propagation characteristics,
regional stationarities, event probabilities
etccomplementary techniques needed to target
specific issues and overcome biases in each method
3) Proxy reconstruction potential is maximised
by using a calibration index that accounts for
both ocean and atmosphere components of ENSO
4) Greatest skill in reproducing observed
climate variability was found using the Coupled
ENSO Index (CEI) reconstructions particularly for
the December-May period up to 53 variability
reproduced higher than skill published by Mann,
Stahle, DArrigo
5) Considerable improvements in ENSO
reconstruction are achieved from expanding the
representation of records from the western Pacific
6) Late C20th century high-frequency ENSO
appears anomalous in the context of the past 5
centuries it is likely that ENSO operates
differently under pre and post-industrial
background states modeling studies required to
assess ENSO forcings
7) ENSO is the dominant source of the Earths
inter-annual climate variability, improving our
understanding of past ENSO is critical for
determining future predictability needed to
mitigate the socio-economic impacts of future
ENSO events
25
Acknowledgements

• Many thanks to
• NOAA the International Pacific Research Center
  (IPRC) for financial assistance to participate in
  the workshop.
• Tree-ring Laboratory, University of Auckland,
  New Zealand Anthony Fowler, Gretel Boswijk, Drew
  Lorrey, Jonathan Palmer, Peter Crossley ad Jenny
  Lux.
• NZ Department of Conservation, Panguru
  Development Trust, Te Iwi o Te Rarawa and Te Iwi
  O Te Roroa communities (Northland NZ).
• Penny Whetton (CSIRO) and Ian Rutherfurd
  (University of Melbourne) for historical ENSO
  data Nile Flood Record, Berlage Teak Chronology,
  North China Rainfall data.
• Erica Hendy (Lamont Doherty Earth Observatory)
  for Great Barrier Reef coral luminescence master
  record.
• Pavla Fenwick (Lincoln University, New Zealand)
  for Pink Pine tree-ring data, South Island, NZ.
• Bruce Bauer (NOAAs World Data Center for
  Paleoclimatology/National Climatic Data Center.
  Paleoclimatology Branch) for data acquisition and
  assistance.
• Karl Braganza thanks Scott Power, Matthew
  Wheeler Harry Hendon (BMRC) for statistical
  advice.