1Hadley Centre for Climate Prediction and Research

Bias Adjustments to Historical SSTChris Folland, Hadley Centre, Met Office

• Characteristics of uncorrected SST data

• Collection of sea water

• Models of heat losses from sea temperature buckets

• Calculated bias corrections

• Tests of bias corrections

• Conclusions

2Hadley Centre for Climate Prediction and Research

Annual anomalies from 1951-80 average of uncorrected SST(solid) & corrected NMAT (dashed) for (a) Northern (b) Southern

Hemisphere, 1856-1992.

Collocated 5o x 5o SST & NMAT values were used.

3Hadley Centre for Climate Prediction and Research

Monthly anomalies of uncorrected SST (solid)and corrected NMAT (dashed) during 1939-48

for three regions covering the globe

4Hadley Centre for Climate Prediction and Research

Power spectrum of uncorrected SST anomalies from 1951-80, forNorthern hemisphere north of 20oN in 1901-40. Insert shows band-

pass filtered annual cycles of SST anomalies for 1922-61.

5Hadley Centre for Climate Prediction and Research

Running 15-year average ratio of annual to totalvariance, F, for uncorrected SST in zones 50-60oN,

40-50oN, 30-40oN & 20-30oS.

6Hadley Centre for Climate Prediction and Research

(a) German metal & leatherbucket, UK Met Office, blackinsulated bucket & UK Met

Office Mk II canvas seatemperature bucket

(b) ship’s wooden bucket,1891 (courtesy of Scottish

Maritime Museum).

(a)

(b)

7Hadley Centre for Climate Prediction and Research

Conceptual model of heat transfers affectinguninsulated sea temperature buckets.

8Hadley Centre for Climate Prediction and Research

Conceptual model of a wooden bucket.

9Hadley Centre for Climate Prediction and Research

Basic heat loss equation for Canvas Bucket

dQ/dt = hr A(Tb-Ta) + hsens A (Tb-Ta + 1.7(eb - ea)) - f(Qs) A

dQ/dt = rate of loss of heat

hr = radiative heat transfer coefficient

A = bucket surface area

Tb = bucket surface temperature

Ta = ambient air temperature

hsens = sensible heat transfer coefficient

eb = saturated vapour pressure at bucket surface temperature

ea = vapour pressure of ambient air

Qs = short wave radiative flux

10Hadley Centre for Climate Prediction and Research

Calculated cooling of Mk II canvas bucket & as observed byAshford (1948) in a wind-tunnel after 1 minute.

f = 0, no evaporation from base, f = 1, free evaporation frombase; f = 2 free evaporation from base & interior water surface.

11Hadley Centre for Climate Prediction and Research

Observed & modelled cooling of Mk II canvas bucket after10 minutes exposure on a ship in tropical North Atlantic.

Dotted lines are 95% confidence limits

12Hadley Centre for Climate Prediction and Research

Basic heat loss equation for insulated bucket

BUCKET WALLS & BASE

* Finite thermal conductivity of bucket walls & base.* Wooden buckets, and possibly, leather-covered metal buckets.

Unsteady heat conduction for one dimensional heat transfer in thex direction in a wall with no internal heat source described by:

∗t/∗τ =(k/∆c) (∗2t/∗x2)

t=temperature, k=thermal conductivity, τ =time, ∆=density &c=specific heat. An exterior film of water 0.1mm thick is included.

TOP OPEN WATER SURFACE

* Same model as used for canvas bucket

13Hadley Centre for Climate Prediction and Research

(a) Correctionscalculated usingcanvas bucketmodels C1-C4(dashed) and C6-C8(light solid) so as todiffer with minimumvariance fromcorrections calculatedusing model C5(heavy solid

(b) using woodenbucket model W3(dashed).

o CCorrection

14Hadley Centre for Climate Prediction and Research

(a) F ratio curve forlatitude band 30-40oN. 1911-40 usingmodel C5.

(b) running 15-yearaverage F ratios for30-40oN with modelC5 corrections forvarious integrationtimes.

15Hadley Centre for Climate Prediction and Research

Final bias corrections o(C), June (left) & December (right)1860, 1900 and 1940.

16Hadley Centre for Climate Prediction and Research

Annual Northern & Southern Hemisphere averagecorrections compared with a posteriori corrections of Jones

et al. (1991) based on land surface air temperatures

17Hadley Centre for Climate Prediction and Research

Annual anomalies from 1951-80 average of corrected SST(solid), uncorrected SST (dotted) & corrected NMAT (dashed).

Smoothed with 21-term binomial filter.

Northern HemisphereNorthern

Hemisphere

SouthernHemisphere

Globe

18Hadley Centre for Climate Prediction and Research

GLOBE 1872-1998

Diffe

ren

ce fro

m 1

946-6

5 a

vera

ge

o C

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1870 1890 1910 1930 1950 1970 1990

OBSERVED (JONES ET AL, 2001)

SIMULATED, CORRECTED SST SIMULATED, UNCORRECTED SST

a

r(1872-1941)=0.58*

NORTHERN HEMISPHERE, 1872-1998

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1870 1890 1910 1930 1950 1970 1990

b

r(1872-1941)=0.56*

SOUTHERN HEMISPHERE, 1872-1998

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1870 1890 1910 1930 1950 1970 1990

c

r(1872-1941)=0.49*

TROPICS, 1872-1998

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1870 1890 1910 1930 1950 1970 1990

d

r(1872-1941)=0.75*

EUROPE, 1872-1998

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1870 1890 1910 1930 1950 1970 1990

e r(1872-1941)=0.56*

AUSTRALIA, 1872-1998

-1.6

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1870 1890 1910 1930 1950 1970 1990

f

r(1872-1941)=0.32

Simulated annual landsurface air

temperatureanomalies using

HadAM2b atmosphericmodel.

Ensemble of 6 runs.

Red: forced withuncorrected SST

Blue: forced withcorrected SST

Black: observed

19Hadley Centre for Climate Prediction and Research

AnoC)

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

GLO

BE

N H

EM

ISP

HE

RE

S H

EM

ISP

HE

RE

TRO

PIC

S

EX

TRA

TRO

PIC

AL N

H

EX

TRA

TRO

PIC

AL S

H

EX

TRA

TRO

PIC

AL N

AM

ER

EU

RO

PE

EX

TRA

TRO

P. E

UR

AS

IA

EX

TRA

TRO

P. S

AM

ER

ICA

Differences between simulated land air temperature & observationsusing corrected (solid circles) and uncorrected SST (open circles)

1872-1941

20Hadley Centre for Climate Prediction and Research

Model test of SST bias corrections for extratropical NorthernHemisphere through annual cycle - ensemble means & uncertainties

Corrected SST to 1941

Uncorrected SST

Observed

Extratropical Northern Hemisphere January-February

Ensemble mean land surface air temperature anomalies

-1.6

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1872 1892 1912 1932 1952 1972 1992

Corrected SSTto 1941

Uncorrected SST

Observed

Extratropical Northern Hemisphere May-June

Ensemble mean land surface air temperature anomalies

-1.6

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1872 1892 1912 1932 1952 1972 1992

Corrected SSTto 1941

Uncorrected SST

Observed

Extratropical Northern Hemisphere September-October

Ensemble mean land surface air temperature anomalies

-1.6

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1872 1892 1912 1932 1952 1972 1992

Corrected SSTto 1941

Uncorrected SST

Observed

Extratropical Northern Hemisphere March-April

Ensemble mean land surface air temperature anomalies

-1.6

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1872 1892 1912 1932 1952 1972 1992

Corrected SSTto 1941

Uncorrected SST

Observed

Extratropical Northern Hemisphere July-August

Ensemble mean land surface air temperature anomalies

-1.6

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1872 1892 1912 1932 1952 1972 1992

Corrected SSTto 1941

Uncorrected SST

Observed

Extratropical Northern Hemisphere November-December

Ensemble mean land surface air temperature anomalies

-1.6

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

1872 1892 1912 1932 1952 1972 1992

21Hadley Centre for Climate Prediction and Research

Conclusions• Model of SST bias corrections integrates physical principles & datacharacteristics

• Reproduces wind tunnel measurements well & supported by tests on aship.

• Climate model forced with corrected & uncorrected SST supports theaccuracy of annual mean corrections

• The annual cycle of corrections was tested in extratropical NorthernHemisphere. Within limits of model & land surface air temperature data,observations well simulated after 1900, but somewhat less before then.

• Mix of uninsulated & wooden buckets needs more research before 1900,though full account needs to be taken of land data uncertainties.

• Improved SSTs and land temperatures should help refinements - majorchanges not expected.

22Hadley Centre for Climate Prediction and Research

Model test of SST bias corrections for Extratropical SouthernHemisphere through annual cycle - ensemble means & uncertainties

Corrected to 1941

Uncorrected

Observed

Extratropical Southern Hemisphere January-February

Ensemble mean land surface air temperature anomalies

-1.2

-0.8

-0.4

0.0

0.4

0.8

1872 1892 1912 1932 1952 1972 1992

Corrected to 1941

Uncorrected

Observed

Extratropical Southern Hemisphere March-April

Ensemble mean land surface air temperature anomalies

-1.2

-0.8

-0.4

0.0

0.4

0.8

1872 1892 1912 1932 1952 1972 1992

Correctedto 1941

Uncorrected

Observed

Extratropical Southern Hemisphere May-June

Ensemble mean land surface air temperature anomalies

-1.2

-0.8

-0.4

0.0

0.4

0.8

1872 1892 1912 1932 1952 1972 1992

Correctedto 1941

Uncorrected

Observed

Extratropical Southern Hemisphere July-August

Ensemble mean land surface air temperature anomalies

-1.2

-0.8

-0.4

0.0

0.4

0.8

1872 1892 1912 1932 1952 1972 1992

Correctedto 1941

Uncorrected

Observed

Extratropical Southern Hemisphere September-October

Ensemble mean land surface air temperature anomalies

-1.2

-0.8

-0.4

0.0

0.4

0.8

1872 1892 1912 1932 1952 1972 1992

Correctedto 1941

Uncorrected

Observed

Extratropical Southern Hemisphere November-December

Ensemble mean land surface air temperature anomalies

-1.2

-0.8

-0.4

0.0

0.4

0.8

1872 1892 1912 1932 1952 1972 1992