Interdecadal Changes of 30-Year SST

1
Interdecadal Changes of 30-Year SST Normals
during 1871-2000 Yan Xue Climate Prediction
Center, NCEP/NWS Thomas M. Smith and Richard W.
Reynolds National Climatic Data Center,
NESDIS for the 27th Annual Climate Diagnostics
and Prediction Workshop held in Fairfax,
Virginia on October 21-25, 2002
2
Motivation

• A 30-year SST normal includes climatological mean
  and standard deviation, and is defined in a
  30-year base period.
• 30-year SST normals used at CPC
• 1951-1979 normal
• 1961-1990 normal
• 1971-2000 normal
• 30-year normals recommended by WMO
• 1951-1980, 1961-1990 and 1971-2000,
• Interdecadal changes of the 30-year normals in
  the past 130 years (1871-2000).

3
Review

• Interannual variability associated with ENSO is
  high during 1880-1920 and 1960-1990 and low
  during 1920-1960 (Kestin et al. 1998
  Mestat-Nunez and Enfield 2001).
• Pacific Decadal Oscillation (PDO) in the North
  Pacific (Mantua et al. 1997).
• Decadal and interdecadal variability related to
  the North Atlantic Oscillation in the North
  Atlantic (Kushnir 1994).
• Decadal and interdecadal variability in the
  tropical Atlantic.
• Decadal and interdecadal variability in the
  Indian Ocean.

4
Approach

• Calculate climatological mean (CM) and standard
  deviation (SD) in sliding 30-year windows in the
  past 130 years.
• (1871-1900, 1881-1910, 1891-1920, )
• Use two independent SST analyses.
• Make conclusions based on the consistent results
  from the two SST analyses.

5
Goals

• Primary goals
• Estimate interdecadal variability in the global
  ocean.
• Estimate interdecadal changes of interannual
  variability in the global ocean.
• Estimate interdecadal changes of seasonal
  persistence.
• Secondary goals
• Provide justification on why the 30-year SST
  normal needs to be updated every 10 years.
• Estimate the uncertainties in the two SST
  analyses.

6
Data

• Extended Reconstructed Sea Surface Temperature
  (ERSST) by Smith and Reynolds (2002).
• Period 1854-2000 Resolution 2O x 2O
• Fitting COADS in situ observations to a set of
  EOFs
• Analyzing low- and high-frequency variations
  separately.
• Hadley Centres Sea Ice and Sea Surface
  Temperature (HadISST) by Rayner et al. (2002).
• Period 1871-1999 Resolution 1O x 1O
• Combination of global SST and sea ice
  concentration.
• Using reduced-space optimal interpolation.

7
Methodology

• Seasonal Climatological Mean (CM)
• Calculating monthly CM by averaging monthly SST
  over appropriate 30-year base periods.
• Averaging monthly CM over appropriate seasons.
• Seasonal Standard Deviation (SD)
• Deriving monthly anomalies by subtracting monthly
  CM from monthly data.
• Deriving seasonal anomalies by averaging monthly
  anomalies over appropriate seasons.
• Deriving seasonal SD from seasonal anomalies.
• Seasonal Persistence (P)
• Auto-correlation of seasonal anomalies at a
  two-season lag over appropriate 30-year base
  periods.

8
Standard Deviation of Seasonal CM
Standard Deviation of the eleven 30-year seasonal
CMs in 1871-2000.
9
Standard Deviation of CM in SON
PDO
NAO
INDIAN
SATL
NINO3
10
INDIAN Index (45O-100OE, 10OS-25ON)
Departures from the seasonal CM averaged over all
the 30-year base periods during 1871-2000.
Years
Differences between HadISST and ERSST
Years
Seasons
11
PDO Index (160OE-160OW, 35ON-45ON)
Departures from the seasonal CM averaged over all
the 30-year base periods during 1871-2000.
Years
Differences between HadISST and ERSST
Years
Seasons
12
NAO Index (60OW-20OW, 40ON-60ON)
Departures from the seasonal CM averaged over all
the 30-year base periods during 1871-2000.
Years
Differences between HadISST and ERSST
Years
Seasons
13
SATL Index (35OW-10OE, 22OS-2ON)
Departures from the seasonal CM averaged over all
the 30-year base periods during 1871-2000.
Years
Differences between HadISST and ERSST
Years
Seasons
14
NINO3 Index (150OW-90OW, 5OS-5ON)
Departures from the seasonal CM averaged over all
the 30-year base periods during 1871-2000.
Years
Differences between HadISST and ERSST
Years
Seasons
15
Average season SD
HadISST-ERSST
16
Seasonal SD of NINO3

• Variance is largest in NDJ and smallest in MAM.
• Variance is intermediate in 1871-1920, small in
  1930-1950 and large in 1960-2000.
• Seasonality of variance is moderate in
  1871-1920, weak in 1930-1950 and strong in
  1960-2000.

1971-2000
1961-1990
1951-1980
Average in 1871-2000
Average in MAM
Years
Average in NDJ
Seasons
17
Seasonal SD of Darwin SLP
1971-2000
1961-1990
1951-1980
Average in 1871-2000
Average in MAM
Years
Average in NDJ
Seasons
18
Seasonal Persistence
19
Seasonal Persistence of NINO3

• Persistence is largest in JJA and smallest in
  DJF.
• Seasonality of persistence is weak in 1871-1910,
  moderate in 1920-1950, strong in 1955-1975, and
  moderate in 1975-2000.
• Predictability is high in 1871-1920, low in
  1920-1940, and high in 1950-2000.

1951-1980
P0.4
Average in DJF
Years
Average in JJA
Seasons
20
Summary

• The interdecadal changes of seasonal CM are
  prominent (0.3-0.6OC) in the tropical Indian
  Ocean, the middle latitude North Pacific and
  North Atlantic, most of the South Atlantic and
  the Sub-Antarctic Front.
• The tropical Indian and Atlantic show a warming
  trend, and the middle latitude North Pacific and
  North Atlantic show multidecadal oscillations.
• The seasonal SD is prominent (0.5-1.3OC) in the
  eastern equatorial Pacific, the middle latitude
  North Pacific and North Atlantic.
• The seasonal SD of NINO3 varies interdecadally,
  intermediate during 1885-1915, small during
  1910-1965 and large during 1965-2000.
• The seasonal P is largest in the eastern
  equatorial Pacific, the tropical Indian and
  Atlantic.
• The seasonal P of NINO3 varies interdecadally,
  high during 1885-1920, low during 1920-1940 and
  high during 1950-2000.

21
Summary

• ERSST appears resolves climate signals better and
  contains less noise than HadISST does.
• In the middle latitude North Pacific and North
  Atlantic, HadISST is biased toward warm in summer
  and cold in winter relative to ERSST. This
  difference appears too large to be accounted for
  by analysis errors and should be examined in the
  near future.