ArticlePDF Available

Interdecadal changes of surface temperature since the late nineteenth century

Authors:

Abstract

We present global fields of decadal annual surface temperature anomalies, referred to the period 1951-1980, for each decade from 1881-1890 to 1981-1990 and for 1984-1993. In addition, we show decadal calendar-seasonal anomaly fields for the warm decades 1936-1945 and 1981-1990. The fields are based on sea surface temperature (SST) and land surface air temperature data. The SSTs are corrected for the pre-World War II use of uninsulated sea temperature buckets and incorporate adjusted satellite-based SSTs from 1982 onward. The generally cold end of the nineteenth century and start to the twentieth century are confirmed, toegether with the substantial warming between about 1920 and 1940. Slight cooling of the northern hemisphere took place between the 1950s and the mid-1970s, although slight warming continued south of the equator. Recent warmth has been most marked over the northern continents in winter and spring, but the 1980s were warm almost everywhere. -from Authors
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 99, NO. D7, PAGES 14,373-14,399, JULY 20, 1994
Interdecadal changes of surface temperature since the late
nineteenth century
D. E. Parker, • P. D. Jones,: C. K. Foiland, and A. Bevan a
Abstract. We present global fields of decadal annual surface temperature anomalies,
referred to the period 1951-1980, for each decade from 1881-1890 to 1981-1990 and
for 1984-1993. In addition, we show decadal calendar-seasonal anomaly fields for the
warm decades 1936-1945 and 1981-1990. The fields are based on sea surface
temperature (SST) and land surface air temperature data. The SSTs are corrected for
the pre-World War II use of uninsulated sea temperature buckets and incorporate
adjusted satellite-based SSTs from 1982 onward. Our results extend those published in
the 1990 Intergovernmental Panel on Climate Change Scientific Assessment and its
1992 supplement. We assess the impact of various sources of error in the fields.
De, spite poor data coverage initially and around the two World Wars the generally cold
end of the nineteenth century and start to the twentieth century are confirmed, together
with the substantial warming between about 1920 and 1940. Slight cooling of the
northern hemisphere took place between the 1950s and the mid-1970s, although slight
warming continued south of the equator. Recent warmth has been most marked over
the northern continents in winter and spring, but the 1980s were warm almost
everywhere apart from Greenland, the northwestern Atlantic and the midlatitude North
Pacific. Parts of the middle- to high-latitude southern ocean may also have been cool in
the 1980s, but in this area the 1951-1980 climatology is unreliable. The impact of the
satellite data is reduced because the record of blended satellite and in situ SST is still
too short to yield a climatology from which to calculate representative anomalies
reflecting climatic change in the southern ocean. However, we propose a method of
using existing satellite data in a step toward this target. The maps are condensed into
global and hemispheric decadal surface temperature anomalies. We show the
sensitivity of these estimated anomalies to alternative methods of compositing the
spatially incomplete fields. Running decadal zonal means and annual global and
hemispheric time series are also shown. Finally, we discuss some salient features in
terms of observed atmospheric circulation changes and of the results of climate model
integrations with increasing atmospheric greenhouse gases.
1. Introduction
In the First Scientific Assessment of the
Intergovernmental Panel on Climate Change (IPCC),
Folland et al. [1990] concluded that despite great
limitations in the quantity and quality of the available
historical temperature data, the evidence pointed to a real
but irregular warming on a hemispheric and global scale
over the last 130 years. Included were world maps in
color showing combined land and oceanic surface
temperature anomalies for the decades 1950-1959,
1Hadley Centre for Climate Prediction and Research,
Meteorological Office, Braeknell, England.
2Climatic Research Unit, University of East Anglia,
Norwich, England.
Published in 1994 by the American Geophysical Union.
Paper number 94JD00548.
0148-0227/94/94JD-00548505.00
1967-1976, and 1980-1989. The land air temperatures
were based on an update of Jones et al. [1986a, hi, and
the sea surface temperatures (SSTs) were derived from the
United Kingdom Meteorological Offiee's data bank
[Bottomley et al., 1990].
Foiland et al. [1992] extended these results in the
IPCC First Science Report Supplement, showing decadal
annual and calendar-seasonal fields for 1981-1990, and
providing improved and updated annual hemispheric and
global surface temperature anomaly series for 1861 to
1991. The geographical coverage of SSTs in the earlier
years, particularly between the 1870s and the First World
War, had been substantially improved by supplementing
the Bottomley et al. [1990] SSTs with those from the
Comprehensive Ocean-Atmosphere Data Set (COPsDS)
[Woodruff et al., 1987]. The systematic instrumental
corrections to SST had also been developed further by the
use of improved models of the heat exchanges between
seawater buckets and their environment and by a
reassessment of the history of the use of canvas and
14,373
14,374 PARKER ET AL.: INTERDECADAL CHANGES OF SURFACE TEMPERATURE
wooden buckets. Full details will be published elsewhere
[FoEand and Parker, 1994].
Here, we provide more details of the techniques used
to blend and to quality control the SST and land air
temperature data bases (section 2) and extend the results
of Folland et al. [1990, 1992] by presenting decadal
surface temperature anomaly fields and zonal averages
from 1881 to the present (section 3) and by incorporating
satellite-based SSTs from 1982 onward.
Our fields are based on substantially more information
than those shown by Hansen and Lebedeff [1987, 1988],
who used only land station data and a few fixed-position
ocean weather ships. Moreover, Hansen and Lebedeff
[1987, 1988] did not correct their data for biases,
whereas the land data we use [Jones et al., 1986a, b;
Jones and Briffa, 1992] have been extensively examined
and, where necessary, corrected for heterogeneities which
may have resulted from, for example, station moves or
urban warming [e.g. Jones et al., 1989, 1990]. We also
provide initial estimates of random errors, remaining
instrumental biases, and sampling errors in the fields over
both land and ocean. We also assess the systematic
effects of deficient coverage by using several methods to
estimate global and hemispheric average temperature
anomalies (section 4). Finally, we discuss some major
features in the fields in terms of observed changes in the
atmospheric circulation of the extratropical northern
hemisphere (section 5). However, owing to the need for
substantial further development of the mean sea level
pressure and marine surface wind database, full
SUPPLEMENT MOHSST4 [
WITH COADS SST J
I
i REMOVE CLIMATOLOGICALLY I
I UNLIKELY SST DATA
I I
APPLY INSTRUMENTAL
CORRECTIONS TO SST
SPATIAL
QUALITY-CONTROL
OF SST
TEMPORAL
QUALITY-CONTROL
OF SST
UP TO
1981
1982 - 1993
I
USE GISST AFTER REMOVING
UNRELIABLE AREAS
BLEND SST WITH LAND
SURFACE AIR TEMPERATUREE
CALCULATE DECADAL
FIELDS
Figure 1. Data processing and analysis.
consideration of the influence of atmospheric circulation
on temperature variations is left to a later date.
2. Data
Our main source of in situ SST data was a further
updated version of the quality controlled Meteorological
Office historical sea surface temperature data set
(MOHSST4, Bottomley et al. [1990]). Where these data
were deficient, we used in situ SST data based on the
Comprehensive Ocean-Atmosphere Data Set (COADS)
2 ø x 2 ø area analyses [Woodruff et al., 1987]. Additional
quality controls were applied while blending the data.
The augmented in situ SST database (MOHSSTS) was
then corrected for instrumental biases and further quality
controlled. MOHSST5 up to 1981 and the Meteorological
Office global sea ice and sea surface temperature (GISST)
data set [Parker et al., 1994] thereafter were then
combined with the land surface air temperature data of the
Climatic Research Unit (CRU), University of East Anglia
[Jones et al., 1986a, b; Jones, 1988]. Figure 1 shows the
basic stages in the data processing and analysis.
2.1. Creation of Meteorological Office Historical
Sea Surface Temperature (MOHSST5) Data Set
Monthly 5 ø latitude x 5 ø longitude area SSTs were
taken from MOHSST4 if available; otherwise, gaps were
filled using 5 o x 5 o area data calculated from the COADS
2 ø x 2 ø area analyses. MOHSST4 already incorporated
most of the original ships' observations used to create the
Massachusetts Institute of Technology (MIT) database
used only in preanalyzed form in the maps in the work of
Bottomley et al. [1990], so the MIT database was not
used explicitly. Any SSTs <-1.8øC in the combined
MOHSST4-COADS were set to -1.8øC. Values for the
Caspian Sea were omitted because they appear to be
unreliable.
After application of instrumental corrections (section
2.3) the monthly fields were subjected to spatial and
temporal quality control as follows:
1. All 5 ø x 5 ø area values were expressed as
anomalies from the GISST 5 ø latitude x 5 ø longitude
monthly climatology for 1951-1980 [Parker et al., 1994].
We regard the GISST climatology as somewhat more
reliable than the MOHSST4 climatology [Bottomley et
al., 1990]. The extra data input to GISST will have
reduced, though certainly not eliminated, the influence of
the highly interpolated and older Alexander and Mobley
[1976] climatology that was used in MOHSST4 in parts
of the southern ocean and some other data-sparse areas.
2. Monthly 5 ø x 5 ø area anomalies were set to
missing if they exceeded 8øC in magnitude within the
area 10øN-15øS, 70ø-170øW, and 6øC elsewhere. The
extended limit was chosen so as to accept the largest
anomalies observed during E1 Nifio events in the eastern
tropical Pacific. The general limit was only intended as a
check on gross errors: many smaller errors will have been
removed by the quality controls described below,
PARKER ET AL.: INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,375
especially in data-rich areas and in regions of small
natural variability such as the tropical West Pacific.
3. For each 5 ø x 5 ø area the average anomaly for the
surrounding eight 5 ø x 5 ø areas was calculated provided
at least four of these contained data. This surrounding
average was then substituted in the target 5 ø x 5 ø area if
the existing anomaly was missing or differed from it by
more than 2.25øC. This limit was found, after extensive
empirical tests on data throughout the past century, to
result in rejection of most of the unlikely SSTs but
retention of the more feasible extreme values. This
assessment was based mainly on the spatial coherence of
adjacent 5 ø x 5 o area monthly anomalies: the process was
intended to remove, as far as possible, unrepresentative
or erroneous values often derived from an individual
ship. More stringent limits over-smoothed strong
anomalies especially in the eastern tropical Pacific during
E1 Nifio events. SSTs were retained when there were too
few neighboring data for this test to be carried out. Thus
the quality-controlled MOHSST5 may be less reliable in
data-sparse areas, but we retained the data so that decadal
values could still be estimated on the basis that the
remaining relatively large effectively random monthly
errors were likely to cancel. See, however, process 7
below; also see section 4.
4. For each 5 ø x 5 o area the average anomaly for the
previous and subsequent months was calculated, provided
both these values were available. This average was then
substituted in the 5 o x 5 o area if the existing anomaly was
missing or differed from it by more than 2.25øC. Again,
SSTs were retained when there were too few data to allow
this test to be carried out.
5. Processes 3 and 4 were repeated twice, i.e., carried
out 3 times altogether.
6. The 5 ø areas which initially had missing values
were reset to missing.
7. Any remaining anomalies differing by more than
2.25øC from the average of neighboring values were set
to missing even when there was only a single neighbor
available.
8. Despite the improved climatology used in process 1
above, we deleted the following areas in the Pacific sector
of the southern ocean where the anomalies appeared to
have remained unreliable: 165øE to 160øW south of
60øS; 160øW to 120øW south of 50øS; and 120øW to
90øW south of 55øS. The fields shown by Folland et al.
[1990, 1992] exhibit some unlikely anomalies in the
southern ocean, almost certainly because there were too
few data in 1951-1980 to form a reliable climatology, and
therefore too much reliance had been placed on the
Alexander and Mobley estimates. However, our analysis
appears to have improved in the Atlantic and Indian
sectors of the southern ocean.
2.2. Incorporation of Global Sea Ice
and Sea Surface Temperature Data Set (GISST)
From 1982, blended satellite and in situ SSTs from
the globally complete GISST data set [Parker et al.,
1994] were used in the decadal fields, zonal means, and
time series instead of MOHSST5. However, owing to
uncertainties in the 1951-1980 climatology in the
southern ocean and near ice edges, monthly 5 o x 5 o area
values north of 60øN and south of 50øS were deleted
whenever the corresponding MOHSST5 values were
missing.
2.3. Instrumental Corrections
to the Sea Surface Temperature (SST)
Geographically, historically, and seasonally varying
instrumental corrections as derived by Folland and
Parker [1994] were applied to all the $ST data obtained
for the period up to December 1941, to compensate for
the use of uninsulated (canvas or metal) or partially
insulated (open-top wooden) buckets. Because of heat loss
the canvas buckets in particular gave lower SST values,
especially in winter and in parts of the tropics, than the
more recently used engine intake thermometers, hull
sensors, and insulated buckets. The corrections
compensate for sensible, latent, and radiative heat
transfers: strictly speaking, they are corrections relative
to the average mix of observations made in 1951-1980.
Differences from the corrections used by Bottomley et al.
[1990] include more rigorous mathematical formulation
and better historically based specification of the models of
wooden buckets, an exposure time of 4 min for these
buckets [Maury, 1858], and an assumption, based on
comparisons with independently corrected
contemporaneous tropical night marine air temperatures,
that 80% of buckets were wooden in 1856 [see also
Maury, 1858] with a linear change to all uninsulated by
1920. By contrast, Bottomley et al. [1990] assumed that
only 25% of buckets were wooden in 1856, changing
linearly to all uninsulated by 1905. The corrections also
differ from those used by Jones and Briffa [1992] who
assumed a linear change from all wooden to all
uninsulated buckets by 1905 and that the wooden buckets
were perfectly insulated, whereas our wooden bucket
models allow some heat transfer from the water surface,
and a small amount through the walls using the nonlinear
heat conduction equation. We have not applied
instrumental corrections to SST data from January 1942
onward, though we recognize that future research may
specify the need for this [Folland et al., 1993; Kent et
al., 1993]. Any future corrections to recent data could
affect the earlier corrections, because the latter are
calculated relative to the average characteristics of
uncorrected data for 1951-1980. See also section 4.2 for
a brief assessment of remaining instrumental biases.
Examples of the corrections, nominally those
applicable to June and December 1940, are illustrated in
Plate 1. These corrections are generally the largest in the
historical record for their calendar month. The largest,
positive corrections are required in early winter over the
Gulf Stream and Kuroshio where warm water, cold dry
air, and strong winds cause rapid evaporative heat loss
from the buckets. Corrections are large in the tropics
14,376 PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE
90N
60N
50N
30S
60S
90S
o Convos bucket corrections (øC) for 7ms-' ship, June 1940
I I I I I !
135E 180 135W 90W 45W 0 45E 90E
b. Convos bucket corrections (øC) for 7ms-' ship, December 1940
-1.2 -0.9 -0.6 -0.5 0 0.5 0.6 0.9
Hate 1. Canvas-bucket corrections for 7 ms -1 ship: (a) June 1940; (b) December 1940.
because of the high rates of evaporation from buckets
when $ST is high. Canvas buckets require larger
corrections than open-top wooden buckets, generally by a
factor of about 4 to 5 [Folland and Parker, 1994] because
of the partial insulation and generally greater size of the
latter. Later data also need larger corrections than earlier
data for a given type of bucket because of the greater
ventilation on board the faster-moving ships. There are
some small negative corrections where warming of the
bucket is calculated to occur, this being mainly where the
mean air temperature around the bucket exceeds the mean
SST and evaporation is small, eg near Newfoundland in
The corrections to SST are well supported by
independent but colocated night marine air temperature
data, as discussed by Bottomley et al. [1990] and Folland
and Parker [1994].
2.4. Blending of SST With Land Air Temperatures
The SST data had been created for areas bounded by
multiples of 5 ø of latitude and 5 ø of longitude (e.g.,
50øN-55øN, 5øW-10øW), but the land air temperature
data had been created for points on a 5 ø latitude x 10 ø
longitude grid (e.g., 55øN, 10øW; 55øN, 20øW, etc.)
(Figure 2). So in our blended 5 ø x 5 ø resolution data set,
which used the SST grid, there were a maximum of two
land values associated with a given 5 ø x 5 ø area, e.g.,
land values at b and e in Figure 2 could contribute to area
B during blending. We chose not to weight colocated
gridded anomalies of SST and land air temperature
exactly by the relative proportions of sea and land in a
grid area, because this would have reduced the influence
of valuable and reliable land stations on small, remote
islands, so we gave more weight to island data using the
expression
PARKER ET AL.: INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,377
b r-
A
B
G
D
E
F
G
e f
H
g h
J
Figure 2. Combination of land and marine data grids: a,b, ..., i are the positions on a 5 ø latitude x
10 ø longitude grid, where a land value may be available. A,B, ..., J are 5 ø x 5 ø areas for which a sea
surface temperature (SST) value may be available.
0=1,2)
t b = (1)
wd• , + n/•
where
t b = blended temperature anomaly in a 5 ø x 5 ø area;
t, monthly SST anomaly with respect to 1951-1980
for the same area;
monthly gridded land air temperature anomaly with
respect to 1951-1980 at ith comer of the 5 ø x 5 ø
area on a 10 ø meridian (e.g., at b or e for area B in
Figure 2);
4 if there is an SST value, otherwise w, is zero;
proportion of sea in the area;
lifps < land0ifp,= 1;
number of comers with land data. For complete
data, n will be 2; because of gaps it can be 1.
tli
W 8
The p, prevents overweighting of SSTs near coastlines
and the /• prevents the influence of land data from
extending into purely oceanic 5 ø x 5 ø areas. Land data
from a single very small remote island will have a
weighting of 0.2 in a 5 ø x 5 ø area when combined with
SST data, as p, is almost 1.
During the blending process we rejected, subjectively,
unlikely land air temperature anomalies at 10øN, 90øE up
to 1900 (derived from a station on the Andaman Islands)
and at 20øS, 50øE for 1911 to 1930 (from a station on
Madagascar).
2.5. Computation of Decadal Fields
The monthly combined SST-land air temperature data,
which were all expressed as 5 ø x 5 ø area anomalies
relative to 1951-1980 climatology, were averaged into
3-month seasons as defined below. A seasonal anomaly
for a box was, however, accepted if it was based on as
little as 1 month's data. Over land, this situation was
uncommon because stations generally operate on a regular
basis, and even if it occurred, the monthly value will
have been based on essentially complete daily data. Over
the ocean the criterion is justified by the known
substantial persistence of SST anomalies [Bottomley et
al., 1990]. However, here the monthly data could, in
practice, be based on a single observation, even though in
many such cases the quality controls described above are
likely to have smoothed or removed unrepresentative
values. So, to confirm our approach, we calculate in
section 4.1 statistically based estimates of the random
errors of the seasonal 5 ø x 5 ø area anomalies calculated in
this way.
To create decadal annual anomaly fields (Plates 2 to
13), we used seasonal anomaly fields for January to
March, April to June, etc., of the relevant years. For
each 5 ø x 5 ø area at least 6 out of a possible 20 seasons'
data, drawn from at least three separate years, had to be
available in each half of the decade, otherwise the area
was assigned as having missing data for that decade. For
5 ø x 5 ø areas with sufficient data, all the available
seasonal anomalies were averaged within each
half-decade, then the two half-decadal anomalies were
averaged to yield the decadal anomaly. For decadal
calendar-seasonal fields (Plates 14 and 15), we used
December to February for boreal winter (or austral
summer), etc., with a minimum of three seasons' data out
of a possible five in each half of the decade and the same
averaging procedure. So in Plate 14 we used December
1935 to February 1936 for the first boreal winter of
1936-1945. In section 4.3 we discuss the likely sampling
errors in the decadal fields resulting from missing
seasonal data. We also discuss there the effects of using
stricter criteria for data availability on coverage and on
estimated decadal hemispheric and global mean surface
temperature anomalies.
To show the beneficial effects of the COADS data,
Figure 3 illustrates the coverage which would have been
obtained in a corresponding annual decadal analysis for
1901-1910 without these data. This coverage should be
compared with that indicated in Plate 4 below. The
14,378 PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE
o
o
w
z z z o
o o o o o o•
o
o--
:p .
,
o
i
i,i
o
z z z o l.n •/• o• ø•
o o o o o o'-
o
Z Z Z 0 U") (/) U")
0 0 0 0 0 0'-
PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,379
o
w
w
z z z 0 •Y) •.0 •0•
0 0 0 0 0 0 •-
o
w
Z Z Z 0 03 (13 t.,J}
0 0 0 0 0 0"-
O3 •0 I• I• •0 O3
I
ß
,
!
Z Z Z 0 03 03 03 ('"1
0 0 0 0 0 0 •'-
0') q:) I'0 i,c) cO 03
0
o
14,380 PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE
I
o
z) .
o
0'-
m,m
0
0'-
0
m,m
0
0'-
0
m,m
PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,381
Z Z Z 0 03 I./3
0 0 0 0 0
L•J
•/• •/• c• o•
o o o'-
L•
0
. .7
Z Z Z 0 03 I,/3 03 0'1
0 0 0 0 0 0'--
14,382 PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE
z z z
o o o
ß
La.I
•/1 to t/• ø•
o o o'-'
o•
z
o
L•
PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,383
90N
60N
30N
30S
60S
i I i i i i i i i i l
..... •n.--
90S
120E 150E 180 150W 120W 90W 60W 30W 0 30E 60E 90E 120E
Figure 3. The coverage which would have been obtained in the analysis in Plate 4 (1901-1910),
without the Comprehensive Ocean-Atmosphere Data Set (COADS) sea surface temperatures.
increase in oceanic coverage of seasonal data for this
decade was over 20% of the area of the world's ocean
[Folland et al., 1992]. The extra data, which were
mainly in the eastern Pacific, raised estimates of seasonal
global average surface temperature anomalies for this
decade by typically 0.05øC, because the eastern Pacific
was less anomalously cold then elsewhere (Plate 4). There
were similar improvements in coverage in 1881-1900 and
smaller improvements in all other decades [Folland et al.,
1992].
3. Results
3.1. Surface Temperature Anomaly Fields
Plates 2 to 13 show the decadal annual surface
temperature anomaly fields for 1881-1890, 1891-1900,
..., 1981-1990, and finally for 1984-1993. We also show
decadal calendar-seasonal fields for 1936-1945, near the
peak of the mid-twentieth century warmth, and for the
very warm recent decade 1981-1990 (Plates 14 and 15).
Table 1 lists the hemispheric and global anomalies
computed by cosine (latitude)-weighting of the 5 ø x 5 ø
area decadal anomalies constituting each plate.
We regard these maps as only a preliminary attempt to
estimate decadal patterns of temperature change, so only a
brief description is given. Much more attention is given
to an error analysis (section 4), to determine the
reliability of our estimates.
Before 1900 (Plates 2 and 3), many continental areas at
middle and high latitudes of the northern hemisphere had
decadal annually averaged anomalies colder than -0.5øC.
Most oceanic regions had, typically, anomalies of
-0.25øC, but parts of the western North Atlantic and the
eastern Pacific were about 0.25øC warmer than their
1951-1980 climatology, and anomalies in the southern
tropical Indian Ocean were also slightly positive. Between
1901 and 1920 (Plates 4 and 5) there was almost uniform
coldness on an annual average, typically -0.25øC to
-0.5øC relative to 1951-1980, except in the eastern
tropical Pacific and the southern tropical Indian Ocean
where positive anomalies approached 0.25øC. Oceanic
coverage fell temporarily during the First World War
(Plate 5).
Plates 6 and 7 document the progress of the early to
mid-twentieth century warming. This was strongest over
Alaska, Greenland, the North Atlantic, Scandinavia, and
parts of Siberia, and anomalies approached + 1 øC from
Greenland to northwestern Siberia in 1931-1940. A
similar pattern was shown by Jones and Kelly [1983] and
extended by Jones and Briffa [1992] to include most of
the midlatitude and tropical oceans. The marine data used
here confirm the results of Jones and Briffa [1992] and
show in particular that the midlatitude North Pacific
remained cold, with anomalies as low as -0.5øC.
Coverage over parts of the oceans was reduced again
during the Second World War (Plate 8).
By 1951-1960 (Plate 9) a slight fall in global average
temperatures had commenced (Table 1): the northern
hemisphere oceans remained relatively warmer than those
of the southern hemisphere. By 1971-1980 (Plate 11),
however, this interhemispheric contrast had reversed. The
most recent warming (Table 1) has been strongest over
the northern hemisphere continents at middle and high
latitudes (Plates 12 and 13), whereas much of the
central and western North Atlantic north of 50øN and
much of the central midlatitude North Pacific have
cooled. Note that the use of GISST rather than
MOHSST5 has very slightly reduced the estimated overall
warmth since 1982 (Table 1). The cold early twentieth
century, the warmth of the mid-twentieth century, and the
recent overall warming are discussed further, in terms of
northern hemisphere atmospheric circulation anomalies,
in section 5.
3.2. Zonally and Decadally Averaged Anomalies
We summarize the long-term changes firsfly by
showing a latitude-time section of zonally and decadally
averaged surface temperature anomalies from 1881 to
1993 (Figure 4). The GlSST-blended satellite and in sire
14,384 PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE
Table 1. Hemispheric and Global Decadal Surface Temperature Anomalies (øC) (With Respect to
1951-1980) Computed by (1) Area-Weighted Averaging of 5 ø x 5 ø Area Decadal Anomalies Constituting
the Fields Shown in Plates 2 to 15 or (2) Averaging of Annual Anomalies From Figure 5
Northern Hemisphere Southern Hemisphere Globe
1 2 1 2 1 2
Annual
1881-1890 -0.23 -0.25 -0.15 -0.15 -0.19 -0.20
1891-1900 -0.23 -0.23 -0.13 -0.15 -0.18 -0.19
1901-1910 -0.30 -0.30 -0.29 -0.29 -0.29 -0.30
1911-1920 -0.31 -0.31 -0.18 -0.19 -0.26 -0.26
1921-1930 -0.10 -0.08 -0.20 -0.20 -0.15 -0.14
1931-1940 0.05 0.06 -0.05 -0.05 0.01 0.01
1941-1950 0.07 0.08 -0.02 -0.03 0.04 0.04
1951-1960 0.06 0.06 -0.04 -0.05 0.02 0.01
1961-1970 0.01 0.01 -0.04 -0.04 -0.01 -0.01
1971-1980 -0.05 -0.05 0.08 0.07 0.01 0.01
1981-1990 a 0.19 0.19 0.24 0.24 0.21 0.21
1981-1990 b 0.18 0.18 0.23 0.23 0.20 0.21
1984-1993a 0.20 0.20 0.25 0.26 0.23 0.23
1984-1993 b 0.19 0.19 0.25 0.25 0.22 0.22
Northern Hemisphere Southern Hemisphere Globe
1 Only 1 Only 1 Only
Seasonal
1936-1945
Dee.-Feb. 0.12 0.04 0.09
March-May 0.10 0.04 0.08
June-Aug. 0.14 0.06 0.11
Sept.-Nov. 0.21 0.00 0.15
a b a b a
1981-1990
Dec-.Feb. 0.26 0.25 0.27 0.26 0.26
March-May 0.24 0.22 0.24 0.24 0.24
June-Aug. 0.14 0.12 0.24 0.24 0.19
Sept.-Nov. 0.13 0.12 0.23 0.22 0.18
1984-1993
Dee.-Feb. 0.28 0.26 0.27 0.26 0.27
March-May 0.27 0.25 0.27 0.27 0.27
June-Aug. 0.16 0.14 0.26 0.26 0.20
Sept.-Nov. 0.11 0.10 0.23 0.23 0.17
0.25
0.23
0.18
0.17
0.26
0.26
0.19
0.16
a Using Meteorological Office Historical Sea Surface Temperature (MOHSST5) data set.
b Using Global sea Ice and Sea Surface Temperature (GISST) data set from 1982.
PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,385
90N
75N
60N
45N
30N
15N
155
3os I'-o.•
45S ' -
60S
755
90S
1890
i i i
1905 1920
1935
Decode Ending
1950 1965 1980
Figure 4. Decadal annual zonal mean surface temperature anomalies,
1951-1980.
1881-1993, relative to
data were used from 1982. The section was constructed
from the blended monthly land and marine data by first
calculating annual 5 ø x 5 ø area anomalies as a simple
average of the available monthly anomalies, then zonally
averaging, and finally taking running 10-year averages
for 1881-1890, 1882-1891, etc. All the major features
discussed in section 3.1 are evident, especially the general
coldness before 1920, the warmth north of 60øN in the
mid-twentieth century, the reversal o f the
interhemispheric contrast between 1951-1960 and
1971-1980, and the recent overall warming.
3.3. Hemispheric and Global Anomaly Time Series
We also show annual time series of anomalies, relative
to 1951-1980, for the hemispheres and globe since 1861
(Figure 5a, 5b, 5c). The solid smoothed curves were
created from the annual anomalies using a 21-point
binomial low-pass filter. These series differ somewhat
from those presented by Folland et al. (1992), which are
summarized by the dashed binomially smoothed curves:
the latter used an average of an earlier version of
MOHSST5 and the corrected SST data used by Jones and
Briffa [1992]. The latter gives cooler anomalies in the
nineteenth century, mainly because of slightly different
corrections (section 2.3). We have also used the blended
satellite and in situ GISST data from 1982, because we
consider that the extra coverage yields a more reliable
result even though we have continued to omit some
high-latitude oceanic areas, as described in section 2.2.
The binomially smoothed dotted curves, based on a blend
of corrected night marine air temperatures [Bottomley et
al., 1990] and land air temperatures, largely confirm the
longer-term changes shown by the blended SST and land
air temperatures. The night marine air temperature data
were corrected independently of the SST data from the
mid-1890s onward [Bottornley et al., 1990; Folland and
Parker, 1994].
Our method of computation is the same as in the work
of Folland et al. [1992]: area-weighted averaging of
seasonal 5 ø x 5 ø area anomalies, followed by averaging
of the four seasons for the hemisphere or globe. The
global annual series shown in Figure 5 is thus the average
of all available data and not the average of the two
hemispheric series, as was computed by Jones and Briffa
[1992]. The decadal averages of the annual values in
Figure 5 are included in Table 1 and differ slightly from
the decadal averages in Table 1 calculated by areal
averaging of the decadal annual 5 ø x 5 ø area anomalies.
Different methods of processing the data implicitly treat
missing areas differently and therefore can give small
differences sufficient to affect noticeably the "ranking" of
years, or even decades, according to their warmth.
Therefore the exact ranking of years according to their
warmth is not meaningful. See Section 4.5 for further
discussion regarding the impact of different analysis
methods on decadal hemispheric and global averages.
4. Error Analysis
4.1. Random Errors
We began by estimating the standard errors a, of the
individual seasonal 5 ø x 5 o area SST anomalies on which
the decadal anomalies are based. To do this, we first
calculated root-mean-square (RMS) differences between
seasonal 5 ø x 5 ø area MOHSST5 SST anomalies, created
as in section 2.5, in a given 5 ø latitude band and with
their centers separated by 5 ø, 10 ø, and 15 ø longitude.
The results were then extrapolated, by linear regression,
to yield estimates of the RMS for zero separation, s o.
There is no formal basis for a particular functional fit,
e.g., linear or exponential, but the RMS differences at
these separations were in general not far from a linear
function of distance. The RMS differences were
calculated for 5 ø x 5 ø areas in zonal bands because the
14,386 PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE
<•
o
z
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
1860
a) Northern hemisphere averages 1861-1995
1880 1 go0 1920 1940 1960 1980 2000
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
1860
b) Southern hemisphere averages 1861 - 1995
1880 1900 1920 1940 1960 1980 2000
c) Global averages 1861-1995
0.6
o
z
-0.4•
-0.6
-o.8i
1860
1880 1900 1920 1940 1960 1980 2000
Figure 5. Combined annual land air and sea surface temperature anomalies relative to 1951-1980
(bars and solid smoothed curves). The dashed smoothed curves are corresponding results from
Folland et al. [1992]. The dotted smoothed curves are based on corrected night marine air
temperatures [Bottomley et al., 1990] and land air temperatures. The smoothed curves were created
from the annual anomalies using a 21-point binomial low-pass filter. (a) Northern hemisphere; (b)
southern hemisphere; (c) globe.
zonal coherence of SST anomalies generally exceeds the
meridional coherence [Bottomley et al., 1990; Folland et
al., 1991; Clancy et al., 1992; see also Briffa and Jones,
1993], yielding a greater range over which RMS
differences increase with distance and thereby making the
estimates of s o more reliable. We next assumed that
errors in each 5øx 5 ø area are independent, so that s o
represents the RMS difference between nominally
colocated independent seasonal 5 ø x 5 ø area samples.
Thus So/xf2 is an estimate of the standard error a, of the
seasonal 5 ø x 5 ø area SSTs. The procedure is similar to
that used when analyzing correlated point rainfall data to
find the uncertainty in point values owing to gauge errors
and local effects [O'Connell et al., 1977]. The resulting
random errors a a in the decadal annual fields were then
taken to be a. lx/'n where n is the number of constituent
seasons.
Estimates of a, over land were made by the same
method using the CRU data but using 10 ø longitude
increments because this was the resolution of the original
gridded data (section 2.4); aa was then calculated from a•
in the same way.
Although we have not shown fields of a•, we note that
our results for the oceans for 1981-1990, without satellite
data, were broadly consistent with the results of Folland
et al. [1993], who compared different centers' operational
SST analyses for 1982-1984 only, and also with those of
Trenberth et al. [1992J, who considered the random
PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,387
90N
a b
605
90S
1201[ 1501[ 180 150W 120W 90W 60W 50W
c
90N
60N
0 50œ 60œ 90E 120E 120E 150E 180 150W 120W 90W 60W ,.lOW 0 50E 60E 90E 120œ
3OS
60S
90S
120E 150E 180 150W 120W 90W 60W 50W
0 50E 60E 90E 120E 120E 150E 180 150W 120W 90W 60W 50W 0 50E 60E 90E 120E
e f
60N
!
30S
60S
90S
120E 150E 180 150W 120W 90W 60W 30W 0 50E 60E 90E 120E
120E 150E 180 150W 120W 90W 60W SOW 0 $OE 60E 90E 120E
Figure 6. (a) Estimated random errors in decadal annual average surface temperature anomalies for
1881-1890; (b) as Figure 6a but for 1971-1980; (c) estimated sampling errors in decadal annual
average SST anomalies for 1881-1890; (d) as Figure 6c but for 1971-1980; (e) estimated combined
random and sampling errors in decadal annual average surface temperature anomalies for 1881-1890;
(f) as Figure 6e but for 1971-1980. All contours are at 0.05øC intervals.
errors and unrepresentativeness of individual SST
observations along with the typical numbers of data
available. For 1981-1990, without satellite data, we
estimated that seasonal a s < 0.2øC over much of the
North Atlantic and eastern North Pacific but that a s was
of the order of 0.5øC over the poorly sampled southern
ocean, underlining the need for complementary satellite
data, homogenized to be consistent with the available in
situ SSTs. For 1881-1890, only a few areas, mainly in
the Atlantic, had a s < 0.2øC; much of the far northern
and western North Atlantic, the extratropical western
Pacific, and the far southern Atlantic and Indian Oceans
had a s around 0.5 øC.
Combined decadal fields of oceanic and land a d for
1881-1890 and for 1971-1980 are shown in Figures 6a
and 6b. Where both oceanic and land-based estimates of
a,• were available, the square root of the average of the
squares of the two estimates was taken. For 1881-1890,
a d < 0.05 o C over the main shipping lanes, but 0.05 o C
< a d < 0.1 øC over most other oceanic regions and the
better sampled land areas such as Europe and eastern
North America. Values in excess of 0.2øC over some
land areas may indicate unreliable, sparse, or
unrepresentative station data (early instrumental
exposures were more varied [Chenoweth, 1993; Parker,
1994]) but may also have resulted from real regional scale
topographical effects. This is clearer in the field of a a for
1971-1980 (Figure 6b) which has maxima in the
Greenland region. For this later period, a a < 0.05øC
over substantial portions of the oceans and a a < 0.1 øC
over most of the globe.
For decadal anomaly fields confined to particular
calendar seasons, the random errors will be
approximately 2a a because the sample size is about one
quarter of that for decadal annual anomalies. The oceanic
errors may be greater than 2a a in winter before 1942,
14,388 PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE
however, because of enhanced meteorologically induced
day-to-day and interannual variations in the cooling rate
of uninsulated buckets. Such variations are not taken
account of in the corrections in Plate 1, which were
calculated assuming a climatologically averaged
environment for the buckets for a given calendar month
and 5 ø x 5 ø area [Folland and Parker, 1994] because data
were inadequate for more detailed calculations. Separate
winter half-year estimates of a s over land (not shown)
also suggest greater errors in some regions than in
summer.
4.2. Instrumental Biases
Instrumental biases in SST are likely to relate mainly to
shortcomings in the bucket corrections applied to the
pre-1942 data. On a global average, the uncertainty
arising from bucket corrections is likely to exceed
_+0.1 øC in the nineteenth century [Folland et al., 1992]
but is probably less than this later; but in regions where
the corrections themselves are large, e.g., the Gulf
Stream and Kuroshio in winter, the uncertainty will be
considerably more. These uncertainties are discussed by
Folland and Parker [1994]. Some uncertainties regarding
more recent measurement techniques are discussed by
Folland et al. [1993] and by Kent et al. [1993].
For the land air temperatures, stations showing
systematic urban warming have been eliminated from the
analysis as far as possible [Jones et al., 1986a, b].
However, changes of thermometer exposure are likely to
have resulted in regional and local biases. Relative to
modem data, archived temperatures for some stations in
the tropics, particularly in India, may be biased by up to
0.4øC too warm until the 1930s [Parker and Folland,
1991; Parker, 1994]; and there may also be national
biases of _+0.2øC in late nineteenth century data for
middle and high latitudes [Parker, 1994]. At individual
stations, biases may have reached _+0.5øC [Chenoweth,
1992]. Globally, however, biases are likely to have
remained within _+0.1 øC on a decadal timescale in view
of the overall agreement with marine data (Figure 7.9 of
Folland et al. [1990]).
The overall global contribution of ocean and land
instrumental biases to uncertainties in decadal averages is
expected to have remained within _+0.15øC since the
1880s. However, further quantification is desirable. Note
that the statement by Folland et al. [1992] that global
warming has been between 0.3øC and 0.6øC since the
nineteenth century reflected a subjective assessment of the
combined effects of instrumental biases and sampling
errors. Our results are consistent with that estimate.
4.3. Local Sampling Errors
Marked local sampling errors may particularly affect
the oceanic portions of the decadal annual fields because
up to 70% of the constituent seasonal SST anomalies
could be missing in a given 5 ø x 5 ø area (section 2.5).
The sampling error-variance a2• associated with omission
of seasonal data from the decadal annual averages will be
[Parker, 1984]
a2• = a2(1 +2 [(n-1)(r•a•(1-p)+r•p)
np n
+ (n-2)(r2ot2(1-p) + r2P ) +'" ])
-a2(1 + 2[(n-1)r 1 + (n-2)r 2 + '" 1) (2)
n n
where
a= standard deviation of the individual seasonal
anomalies within the decade (e.g., January
to March 1981, April to June 1981, ...,
October to December 1990 );
n maximum number of seasons, i.e., 40;
p probability that a season has data: here
0.3 _<p _< 1;
r•, r2, . .., lag 1, lag 2, ..., autocorrelation coefficients
of the individual seasonal anomalies within
the decade;
oq, ot 2, ..., lag 1, lag 2, ..., autocorrelation coefficients
of the time series of individual seasons' data
availability/•,/•2, '",
/• = 1 when data exist;
/• = 0 when data are missing.
Equation (2) expresses the effect of the reduction in the
sampled number of degrees of freedom resulting from the
omission of data from an autocorrelated time series.
In applying equation (2) we terminated the summations
at r 4 because the autocorrelations generally became small
as the lag was increased to four seasons. Examples of the
application of equation (2) are as follows:
1. If data are available in each season, a• -- 0.
2. If data are available in the minimum permissible
number of individual seasons (i.e., np = 12) and r•, r 2,
..., are all very small, then the number of degrees of
freedom falls from 40 to 12; a2• is then a2/12 - a2/40, so
that if a = 0.5øC, a• = 0.12øC.
3. Large a• could occur in the eastern tropical Pacific
with, typically, a = 1 øC, r• = 0.6, r 2 = 0.4, r 3 = 0.2,
r 4 = 0. In this case, if np = 12 and c• = 0.8, c• 2 = 0.6
and % = 0.4 (the condition that data be spread through
at least 3 years in each half-decade limits or), then a• =
0.26 øC.
Fields of a• for decadal annual averages of SST for
1881-1890 and 1971-1980 are shown in Figures 6c and
6d. Even in the earlier decade, few areas fulfilling the
conditions for having a decadal annual mean anomaly
have a sampling error, as defined above, exceed•g
0.1 øC. Our estimates of a• may, however, themselves be
subject to uncertainty where sample sizes are small, e.g.,
in the southern ocean, or where rather more persistence
than allowed for here further reduces the number of
PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,389
degrees of freedom, e.g., in the eastern tropical Pacific
[Zwiers and yon Storch, 1993].
Corresponding fields of a• for decadal calendar-
seasonal averages of SST (not shown) have similar
patterns. For example, values of a• for December to
February 1971-1980 are very similar to Figure 6d,
though enhanced to typically 0.1øC in the southern
ocean. Values of a• for December to February
1881-1890 generally exceed those in Figure 6c, especially
over the Pacific where a e is typically 0.1 to 0.15øC, but
they remain generally less than 0.05øC over the Atlantic
and Indian Oceans. We ignored sampling errors over land
because stations generally operate on a regular basis
(section 2.5).
4.4. Combined Random and Sampling Errors
The decadal annual random errors a a (Figures 6a and
6b) dominated the decadal annual sampling errors a•
(Figures 6c and 6d) so that the combined errors,
x/'(aa2+0'e 2) (Figures 6e and 6f), were similar to the
random errors. In view of the above discussions, the
same will have been true of decadal calendar-season
combined errors (not shown) which will have
approximated to their random errors, i.e., to about twice
the values shown in Figures 6a and 6b (section 4.1). The
dominance of a a over a e is also implicitly confirmed by
the sampling studies in section 4.5 below.
Smoothing the 5 ø x 5 ø area anomalies into 10 ø x 10 ø
area anomalies would, if the 5 o x 5 ø area anomalies were
complete, yield local values with halved random errors.
However, because the sampling errors are spatially
correlated, the reduction in combined errors would be by
less than the factor of 2 otherwise expected.
4.5. Systematic Effects of Deficient Coverage of
Data on Hemispheric and Global Mean Surface
Temperature Anomalies
Frozen-grid experiments [Jones et al., 1986a, b;
Bottomley et al., 1990; Folland et al., 1990], in which
data coverage is restricted to that available in some past
period, suggest that varying data coverage has produced
biases of the order of _+0.05øC on decadal timescales in
global and hemispheric surface temperature series, such
as those we show in Figure 5. In individual years, early
in the record, biases of _+0.2øC can be attributed to these
coverage changes. Frozen-grid experiments, however,
take no account of regions which have never had data.
Madden et al. [1993] avoided this difficulty by sampling
globally complete model-generated fields for January at
the locations of present and past data. Although February
has the maximum variability of northern hemisphere land
surface air temperature anomalies [Jones, 1994] and,
consequently, of monthly global surface temperature
anomalies, the variability in January is almost as great.
Madden et al.'s results indicated typical sampling errors
of 0.2øC to 0.3øC for individual January global mean
surface temperature anomalies in the late nineteenth
century, decreasing to less than 0.1 øC by the
mid-twentieth century, and thus support the frozen-grid
experiments.
Here, we carried out a different sensitivity experiment.
We recalculated the decadal annual fields, firstly with a
minimum of 2 months' data required to constitute a
season but with the same criteria for numbers of seasons
as described in section 2.5; secondly, still with a
minimum of 2 months' data constituting a season and also
with a minimum of 10 seasons' data, drawn from at least
four separate years, in each half of the decade; and
thirdly, requiring all 3 months' data to constitute a season
and a minimum of 15 seasons' data, drawn from all five
years, in each half of the decade. The resulting coverages
for 1881-1890 and 1971-1980 are shown in Figure 7. As
anticipated (section 2.5), the coverage over land declined
little, except over Africa and Amazonia in 1971-1980, as
the criteria were tightened; but the oceanic coverage
became progressively sparser especially in the Pacific in
1881-1890 (Plate 2 and Figures 7a, 7c, 7e) and in the
southern ocean in 1971-1980 (Plate 11 and Figures 7b,
7d, 7f). We then used these more limited fields to
recalculate the decadal annual hemispheric and global
mean surface temperature anomalies, which are shown in
Table 2 alongside the original values from Table 1. The
differences are mainly 0.02øC or less after 1910, but for
earlier decades they sometimes exceed 0.05øC for the
slacker criteria and 0.1 øC for the tightest criterion. The
loss of coverage makes these decades appear to become
colder, as expected in view of the warming impact of the
addition of COADS data (section 2.5), since the areas
which benefited from the COADS data, e.g., the eastern
Pacific, now lose coverage (Plate 2 and Figures 7a, 7c,
7e). In view of the generally small local sampling errors
permitted by the coverage criteria used in Plates 2 to 13
(section 4.3), we consider that these criteria allow both
improved coverage and more representative hemispheric
and global average anomalies.
Karl et al. [1994] suggest that area-weighted averaging
of anomalies for small grid areas can sometimes yield
hemispheric and global average anomalies which are
inferior to those derived by averaging larger areas, as was
done, for example, by Hansen and Lebedeff [1987,
1988]. Using larger areas effectively gives more weight
to isolated data. To test the influence of the spatial
averaging technique on the hemispheric and global
decadal annual averages listed in Table 1, we recomputed
these averages by first calculating area-weighted averages
of all available 5øx 5 ø area decadal annual anomalies
within Hansen and Lebedeff's boxes and then averaging
the resulting subregional decadal annual anomalies with
weighting proportional to the total area of each
subregion. Table 3 shows no differences exceeding
0.02øC after 1950 but some hemispheric differences of up
to 0.05øC for earlier decades. Global differences never
exceed 0.03 øC. We repeated the computations again,
using 30 ø latitude x 60 ø longitude regions (60ø-90øN,
14,390 PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE
90N
a b
60N
3ON
3O$
6O$
90S
120E 150E 180 150W 120W 90W 60W 50W 0 50E 60œ 90E 120E
120E 150E 180 150W 120W 90W 60W 50W 0 50E 60œ
90N
c d
6ON
3ON
30S
6O$
905
120E 150E 180 150W 120W 90W 60W 50W 0 50E 60œ 90E 120E
120E 150œ 180 150W 120W 90W 60W 30W 0 30E 60E
90N
6ON
30N
3OS
60S
e f
90S
120E 150E 180 150W 120W 90W 60W SOW 0 30E 60E 90E 120E 30W 0 30E 60E
9ON
60N
SON
0
50S
60S
90S
120E 150œ 180 150W 120W 90W 60W
Figure 7. The decadal annual coverage which would have been obtained (a) for 1881-1890 with
numbers of seasons as in section 2.5 and at least 2 months' data to constitute a season; (b) as Figure
7a but for 1971-1980; (c)for 1881-1890 with at least 10 seasons' data, drawn from at least four
separate years, in each half of the decade, and at least 2 months' data to constitute a season; (d) as
Figure 7c but for 1971-1980; (e) for 1881-1890 with at least 15 seasons' data, drawn from all five
years, in each half of the decade, and 3 months' data to constitute a season; (f) as Figure 7e but for
1971-1980.
gOE 120œ
90E 120œ
90E 120E
180ø-120øW, etc) at the intermediate stage, with very
similar results (Table 3). Note, however, that the region
south of 60øS has no 5 ø x 5 ø area decadal annual
anomalies before 1950 (Plates 2 to 8), so that all three
spatial averaging techniques omit this region.
Although in most decades the area of missing data is
greater in the southern hemisphere than in the northern
hemisphere, this imbalance is smaller than suggested by
Plates 2 to 13 because the Mercator projection
overemphasizes the areas of missing data in high southern
latitudes. Thus the global decadal annual average
anomalies in Table 1 are always within 0.02øC of the
simple average of the two individual hemispheric values.
5. Influence of Atmospheric Circulation
A full discussion of the influence of atmospheric
circulation on the decadal temperature anomaly fields
requires both a reliable global mean sea level (msl)
pressure database and a consideration of seasonal
temperature anomaly fields. The former has yet to be
developed in a reliable form for any extensive historical
period; therefore we restrict ourselves to events in the
northern hemisphere north of 20øN, especially the North
America-North Atlantic-Europe sector for which existing
data are the most reliable. Also, we consider three
particular epochs: the early twentieth century when the
PARKER ET AL.: INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,391
14,392 PARKER ET AL.: INTERDECADAL CHANGES OF SURFACE TEMPERATURE
Table 3. Hemispheric and Global Decadal-Annual Surface Temperature Anomalies (øC) (1951-1980) Computed
From Decadal Fields Using Different Areal Averaging Techniques
Northern Hemisphere Southern Hemisphere Globe
1 2 3 1 2 3 1 2 3
1881-1890 -0.23 -0.28 -0.27 -0.15 -0.15 -0.15 -0.19 -0.22 -0.22
1891-1900 -0.23 -0.25 -0.28 -0.13 -0.15 -0.12 -0.18 -0.20 -0.20
1901-1910 -0.30 -0.32 -0.30 -0.29 -0.29 -0.29 -0.29 -0.30 -0.29
1911-1920 -0.31 -0.34 -0.34 -0.18 -0.21 -0.20 -0.26 -0.28 -0.27
1921-1930 -0.10 -0.05 -0.08 -0.20 -0.22 -0.23 -0.15 -0.13 -0.15
1931-1940 0.05 0.10 0.08 -0.05 -0.09 -0.10 0.01 0.02 0.00
1941-1950 0.07 0.11 0.10 -0.02 -0.03 -0.02 0.04 0.05 0.05
1951-1960 0.06 0.07 0.06 -0.04 -0.03 -0.04 0.02 0.02 0.01
1961-1970 0.01 0.01 0.01 -0.04 -0.04 -0.05 -0.01 -0.02 -0.02
1971-1980 -0.05 -0.05 -0.05 0.08 0.08 0.08 0.01 0.02 0.02
1981-1990 a 0.19 0.20 0.20 0.24 0.23 0.24 0.21 0.21 0.22
1981-1990 b 0.18 0.18 0.18 0.23 0.22 0.23 0.20 0.20 0.21
1984-1993 a 0.20 0.21 0.21 0.25 0.24 0.25 0.23 0.23 0.23
1984-1993 b 0.19 0.19 0.19 0.25 0.23 0.24 0.22 0.21 0.22
1, Values taken from Table 1. 2, Values estimated from Plates 2 to 12 by calculating area-weighted averages of 5 ø x 5 ø area
anomalies within the boxes used by Hansen and Lebedeff[1987, 1988] and then averaging the resulting regional anomalies with
weighting proportional to the total area of each region. 3, As 2 but using 30 ø latitude x 60 ø longitude regions (60ø-90øN,
180ø-120øW, etc.) at the intermediate stage.
a Using MOHSSTS.
b Using GISST from 1982.
North Atlantic westerlies were strong [Lamb, 1972; Lamb
et al., 1973; Parker and Folland, 1988; Hense et al.
1990], the period 1931-1950 encompassing the peak of
the midcentury warmth, and the warm decade 1981-1990.
5.1. 1901-1930 (Plates 4, 5, and 6)
There was increased westerliness and cyclonicity over
the midlatitude North Atlantic, especially in winter and
spring (Figure 8; see also Figure 4 of Parker and
Folland, [1988]). In Figure 8 all areas north of 70øN,
along with 60ø-70øN, 90øW westward to 120øE, have
been omitted because the analysis is suspected to be
biased there [see RodewaM, 1950; Trenberth and
Paolino, 1980]. In 1901-1910 the strengthened winter
westerlies penetrated into Eurasia, yielding positive
temperature anomalies (seasonal map not shown). This
strengthened atmospheric circulation might also have
been expected to produce cooling of the ocean surface in
the midlatitude North Atlantic, via enhanced vertical
mixing, southeastward Ekman drift, and latent and
evaporative heat transfers into the atmosphere [Bjerknes,
1964]. However Plates 4 to 6 do not suggest that surface
temperature anomalies in this area of the globe were more
negative than elsewhere: possibly the Gulf Stream had
been strengthened by the Ekman drift, or more heat had
been advected by the atmosphere (see also section 5.2). It
is also possible that the strengthened north-south surface
temperature gradients in the northern North Atlantic
(Plates 4 and 5) had resulted in a stronger atmospheric
circulation (see also Deser and Blackmon [ 1993]).
5.2. 1931-1950 (Plates 7 and 8)
The marked anomalous warmth over high northern
latitudes contrasted with persistent negative anomalies
over the midlatitude North Pacific in 1931-1940 (Plate 7).
The latter feature was also present in the warm decade
1981-1990 (Plate 12) and in 1984-1993 (Plate 13) but the
North Atlantic anomalies differed greatly between
1931-1940 and 1981-1990 or 1984-1993. Mean sea level
pressure anomaly fields relative to 1951-1980 (similar to
those for 1936-1945 (Figure 9) discussed in section 5.4
below) indicate weakly enhanced westerlies and/or
PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,393
DJF
1901-19!0 MAM 1901-1910
'..
,
JJA 1901-1910
SON 1901-1910
•,• . o o ..... , ...... ... --
Figure 8. Mean sea level pressure anomalies for the extratropical northern hemisphere, 1901-1910,
relative to 1951-1980 climatology. Contours every 1 hPa. Data source: Meteorological Office
interpolation of United States Weather Bureau analyses.
cyclonicity at most times of the year in 1931-1940 over
the midlatitude North Pacific (see also Van Loon and
Williams [1976] and Trenberth [1990]), suggesting that
atmospheric circulation anomalies may have, in part,
forced the SST anomaly field shown in Plate 7 by
increased southeastward Ekman drift, increased mixing in
the ocean, increased heat transfers into the atmosphere,
and reduced solar radiation. Enhanced cyclonicity
(relative to 1951-1980) is also indicated over the Davis
Strait (west of Greenland) in all seasons in 1931-1940 and
1936-1945. The resulting anomalous Ekman drift may
have contributed to a strengthening of the Gulf Stream
and thereby to the anomalous warmth in the central North
Atlantic: in addition, variations in advection of heat to
and by the atmosphere may have contributed to these
longer-term changes [Bjerknes, 1964].
5.3. 1981-1990 (Plates 12 and 15) and 1984-1993
(Plate 13)
Many of the oceanic anomaly patterns of the previous
decade (Plate 11) became more accentuated in 1981-1990,
14,394 PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE
DJF 1936- 1 94-5
./
0
MAM 1936-194-.5
i j
0
JJA 1 9,56- 1 94-5
SON
,,
.,
ß
ß
ß
1936-194-5
Figure 9. As Figure 8 but for 1936-1945. Data source: as Figure 8 until June 1939, then
Meteorological Office analyses.
with an additional warm area in the eastern Pacific where
there were major El Nifio-Southem Oscillation events in
1982-1983 and 1986-1987. Pronounced annual mean
warmth over northern Eurasia and northwestern North
America in 1981-1990, exceeding +0.75øC in many
areas, resulted from positive anomalies exceeding + 2 øC
in winter and exceeding + IøC in spring (Plate 15,
Foiland et al. [1992]), whereas anomalies in summer and
autumn were more mixed. The cold central North Pacific
and warm northwe, stem North America in winter are
consistent with the deepened Aleutian low (Figure 10 and
Trenberth [1990]); and the winter warmth over northern
Eurasia is consistent with strengthened westerlies
penetrating from the North Ariantic [U. S. Department of
Commerce, 1991]. Marked warmth over the Antarctic
Peninsula is in accord with Jones [1990]. See also
section 5.4 below.
The patterns of the decadal annual anomaly field for
1984-1993 (Plate 13) are very similar to those for
1981-1990, despite the cooling influence of the 1991
PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,395
DJF
/
/
/
-.
1981 1990 1981-1990
,
/
/
/
MAM
/
/
/
JJA
, ,,
.,
,
19al 1990
SON
1981 1990
Figure 10. As Figure 8 but for 1981-1990. Data source: blended analyses of the Meteorological
Office, National Center for Atmospheric Research and Scripps Institute of Oceanography.
eruption of Mt. Pinatubo [Robock and Mao, 1992].
Annual mean warmth now exceeded 1 øC over parts of
northwestern North America as well as eastern Siberia,
but the far northeast of Canada was a little colder.
5.4. Comparison
Decade 1936-1945
of 1981-1990 With the Warm
The peak of the mid-twentieth century global warmth
occurred around 1936-1945, so Plate 14 shows the
seasonal anomaly features for 1936-1945, for comparison
with those for 1981-1990 in Plate 15. Northern Eurasian
anomalies differed substantially between the two periods;
in particular, winter and spring were generally cold in
1936-1945 (below -0.5øC in some areas) but were warm
in 1981-1990, while eastern Europe was warm in summer
(up to + IøC) in 1936-1945 but cold (down to -0.25øC)
in 1981-1990. Northwestern North America was
anomalously warm in winter and spring in 1936-1945,
exceedMg +IøC in some areas, as in 1981-1990; but it
14,396 PARKER ET AL.' INTERDECADAL CHANGES OF SURFACE TEMPERATURE
90N
75N
60N
45N
30N
15N
0
15S
30S
45s
60S
75S
90S
120E
150E 180 150W 120W 90W 60W 30W 0 30E 60E 90E 120œ
0 0.5 1 1.5 2 10
Figure 11. Variance ratio of seasonal surface temperature anomalies, March to May 1974-1993,
relative to 1954-1973. At least 60% of seasons had data.
was also warm in autumn in 1936-1945 but cold (below
-IøC) in 1981-1990. All these similarities and
differences are consistent with anomalous advection by
the atmospheric circulation (Figures 9 and 10). Jones
and Briffa [1992] obtained similar results when
comparing 1981-1990 with 1931-1940. For the areas of
overlapping data the correlations between the seasonal
fields in Plates 14 and 15 were 0.05 (December to
February), 0.27 (March to May), 0.21 (June to August)
and 0.12 (September to November), reflecting the major
differences between the patterns along with some regional
similarities, e.g., in northwestern North America in
spring.
The recent anomalous coldness in the North Atlantic
near Greenland and in parts of the southern ocean
resembles results obtained by some coupled
atmosphere-ocean models forced with steadily increasing
atmospheric greenhouse gases [Manabe et al., 1991,
1992; Meehl et al., 1993]. Furthermore, the North
Atlantic coldness was not a feature of the warmth of the
1930s and 1940s (Plates 7, 8, and 14) when the
anthropogenic greenhouse gas burden was relatively
small. However, the data in the southern ocean are not
yet adequate for assessing past changes in that area
(section 2.1). Also, especially in view of palaeoclimatic
evidence for marked natural variations in the North
Atlantic [e.g., Lehman and Keigwin, 1992] the causes of
the recent pattern there, which has lasted about 25 years,
remain unproven. Thus the recent North Atlantic cold
anomaly, with the accompanying warmth in the South
Atlantic, may indicate a natural multidecadal fluctuation
rather than a long-term trend.
5.5. Recent Interannual Variability
It is often speculated that climate variability may have
changed. Any changes in the variability of the
atmospheric circulation are likely to have affected the
variability of surface temperature. We have therefore
calculated the variances of seasonal surface temperature
anomalies for the 20-year period 1974-1993 for each
calendar season separately and compared them with the
variances for the previous 20-year period 1954-1973. The
variance of the anomalies for each season and period was
calculated about the mean for that season and period. We
used MOHSST5 throughout because the introduction of
the smoother GISST [Parker et al., 1994] in 1982-1993
would have yielded an artificial lowering of variance in
the later period. Figure 11 illustrates the variance ratios
for March to May. Areas of 5 ø latitude x longitude are
blank unless at least 60% of seasons, i.e., 12 in each
period, had data. Values of 2.20 and 0.45 are,
approximately, the thresholds for locally statistically
significant increases and decreases at the 95 % confidence
level where data are complete, assuming 19 ø of freedom
for each period. In Figure 11, 7 % and 5 % of the data
area lie outside those thresholds, but these percentages
may have been inflated by the incomplete data record in
some regions (e.g., the southern ocean) and may not
represent real changes of variance. The average ratio in
Figure 11, calculated using a z transformation, is 1.08.
Averages for the other seasons ranged between 1.04 and
1.11, and the fields (not shown) were similar in general
character. We do not consider that these results imply
any systematic increase of variability between the earlier
and the later period.
6. Conclusions
Our results essentially confirm the earlier works of
Jones [1988], Jones and Kelly [1983], Jones et al.
[ 1986a, b], Jones and Briffa [1992], Hansen and Lebedeff
[1987, 1988], Vinnikov et al. [1987, 1990], and
BottomIcy et al. [1990] as updated and combined in the
PARKER ET AL.: INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,397
IPCC reports [Folland et al., 1990, 1992]. The fact of
global warming in the past century is beyond dispute even
though the precise amount is certainly not. On theoretical
grounds a likely contributory cause of the warming is the
rise in greenhouse-gas concentrations, but despite some
similarities between the most recent oceanic surface
temperature anomalies and those modeled by Manabe et
al. [1991, 1992] and Meehl et a/.[1993], it is definitely
premature to ascribe all or most of the warming to this
particular cause [Houghton et al., 1992].
Importantly, our results relating to large-scale trends
are relatively insensitive to reasonable coverage or
sampling criteria and to variations in the techniques used
to composite the data, e.g., averaging 5 ø x 5 ø area values
into subregions and then averaging these to yield
hemispheric and global values. Thus we do not confirm
the sensitivities suggested by Karl et al. [1994].
We have combined, in a preliminary manner, the $STs
from the Meteorological Office and COADS "in situ"
databases and thereby gained substantial geographical
coverage in the late nineteenth and early twentieth
century, with lesser but still useful gains in recent
decades. The benefits we have gained from blending the
conventional SST data stress the great value that could
accrue from blending the entire databases (i.e., including
air temperature, pressure, wind, etc.), observation by
observation, eliminating duplicates and applying
appropriate quality controls [Parker, 1992]. In addition,
many marine data have not yet been digitized especially
for the years of and around the two World Wars [Worm
Meteorological Organization (WMO), 1990], and there
are plans to add these to the blended compilation [Elms,
1992; Komura and Uwai, 1992; Parker, 1992].
Satellite SST data from 1982 have somewhat
consolidated the coverage, but the full benefits of these
data have not been realized in our analyses because we
have not been able to construct a sound climatology for
1951-1980 in parts of the southern ocean and in a few
areas near the Arctic ice limit. This problem can possibly
be circumvented by creating a more recent climatology
incorporating satellite SSTs. This climatology would need
to be at least 15 years long, to minimize adequately the
aliasing effects of short-term climatic variability.
Another technique that could improve marine
climatologies for a recent 30-year period would be to
blend in situ data for such a period with a background
field based on GISST for the period when satellite data
are used in GISST. This should remove most of the
erroneous gradients of SST in the southern ocean evident
in the Alexander and Mobley [ 1976] climatology and may
provide a reasonable, provisional, truly global 30-year
marine climatology.
Satellite data can also provide reliable sea ice limits
back to 1978 and possibly to 1973 [Gloersen and
Campbell, 1991]. In due course these newer forms of
data, when blended with in situ data, will start to enhance
greatly our ability to monitor and assess decadal and
longer-term climatic changes.
Acknowledgments. T. Basnett, A. Colman,
R. Hackett and M. Jackson carried out substantial
computing and graphics work for this paper. The
development of the land air temperature database was
supported by the U.S. Department of Energy,
Atmosphere and Climate Division, under grant
DE-FG02-86ER60397. The development of the GISST
and mean sea level pressure databases were supported by
the UK Department of the Environment under contract
PECD/7/12/37. The development of GISST was also
supported by the Commission of the European
Communities under Contract EPOC-0003-C(MB). We
are grateful to the referees, particularly T.Karl, for
valuable comments.
References
Alexander, R.C., and R.L. Mobley, Monthly average
sea-surface temperatures and ice-pack limits on a 1 ø global
grid, Mon. Weather Rev., 104, 143-148, 1976.
Bjerknes, J., Atlantic air-sea interaction, in Advances in
Geophysics, vol. 10, edited by H. E. Landsberg and J. van
Mieghem, 82 pp., Academic, San Diego, 1964.
Bottomley, M., C. K. Folland, J. Hsiung, R. E. Newell, and
D. E. Parker, Global Ocean Surface Temperature Atlas
(GOSTA), 20+iv pp. and 313 plates, Joint Meteorol.
Off./Mass. Inst. of Technol. Proj., supported by U.S. Dep.
of Energy, U.S. Nat. Sci. Found., and U.S. Off. of Nav.
Res., funded by UK Depts. of Energy and Environment,
Her Majesty's Stationery Office, London, 1990.
Briffa, K. R., and P. D. Jones, Global surface air temperature
variations during the twentieth century, 2, Implications for
large-scale high-frequency palaeoelimatie studies, Holocene,
3, 77-88, 1993.
Chenoweth, M., A possible discontinuity in the U.S. historical
temperature record, J. Clim., 5, 1172-1179, 1992.
Chenoweth, M., Nonstandard thermometer exposures at U.S.
co-operative weather stations during the late nineteenth
century, J. Clim., 6, 1787-1797, 1993.
Clancy, R.M., J.M. Harding, K.D. Pollak, and P. May,
Quantification of improvements in an operational
global-scale ocean thermal analysis system, J. Atmos.
Ocean. Technol., 9, 55-66, 1992.
Deser, C., and M. L. Blackmon, Surface climate variations
over the North Atlantic Ocean during winter: 1900-1989,
J. Clim., 6, 1743-1753, 1993.
Elms, J. D., Status of NCDC keying of historical marine data,
in Proceedings International COADS Workshop, Boulder,
Colo., 13-15 Jan. 1992, edited by H. F. Diaz, K. Wolter,
and S. D. Woodruff, pp.37-45, NOAA/ERL Spec. Publ.,
Environ. Res. Lab., 1992.
Folland, C. K., and D. E. Parker, Correction of instrumental
biases in historical sea surface temperature data,
Q. J. R. Meteorol. Soc., in press, 1994.
Folland, C.K., T.R. Karl and K. Ya. Vinnikov, Observed
climate variations and change, section 7, in Climate Change,
the IPCC Scientific Assessment, edited by J. T. Houghton,
14,398 PARKER ET AL.: INTERDECADAL CHANGES OF SURFACE TEMPERATURE
G. J. Jenkins and J. J. Ephraums, pp. 195-238, Cambridge
University Press, New York, 1990.
Foiland, C.K., J. Owen, M.N. Ward, and A.W. Colman,
Prediction of seasonal rainfall in the Sahel region using
empirical and dynamical methods, J. Forecasting, 10,
21-56, 1991.
Folland, C.K., T.R. Karl, N. Nicholls, B.S. Nyenzi,
D. E. Parker, and K. Ya. Vinnikov, Observed climate
variability and change, section C, in Climate Change 1992-
The Supplementary Report to the IPCC Scientific
Assessment, edited by J. T. Houghton, B. A. Callander and
S. K. Varney, pp.135-170, Cambridge University Press,
New York, 1992.
Foiland, C.K., R.W. Reynolds, M. Gordon, and
D. E. Parker, A study of six operational sea surface
temperature analyses, J. Clim., 6, 96-113, 1993.
Gloersen, P., and W. J. Campbell, Recent variations in Arctic
and Antarctic sea-ice covers, Nature, 352, 33-36, 1991.
Hansen, J., and S. Lebedeff, Global trends of measured surface
air temperature, J. Geophys. Res., 92, 13,345-13,372,
1987.
Hansen, J., and S. Lebedeff, Global surface temperatures:
Update through 1987, Geophys. Res. Lett., 15, 323-326,
1988.
Hense, A., R. Glowienka-Hense, H. von Storch, and
U. St•hler, Northern Hemisphere atmospheric response to
changes of Atlantic Ocean SST on decadal time scales:
A GCM experiment, Clim. Dyn., 4, 157-174, 1990.
Houghton, J. T., B. A. Callander, and S. K. Varney (Eds.),
Climate Change 1992- The Supplementary Report to the
IPCC Scientific Assessment, xii + 200 pp., Cambridge
University Press, New York, 1992.
Jones, P. D., Hemispheric surface air temperature variations:
Recent trends and an update to 1987, J. Clim., 1,654-660,
1988.
Jones, P. D., Antarctic temperatures over the present century -
A study of the early expedition record, J. Clim., 3,
1193-1203, 1990.
Jones, P. D., Hemispheric surface air temperature variations: A
reanalysis and an update to 1993, J. Clim., in press, 1994.
Jones, P. D., and K. R. Briffa, Global surface air temperature
variations during the twentieth century, I, Spatial, temporal
and seasonal details, Holocene, 2, 165-179, 1992.
Jones, P.D., and P.M. Kelly, The spatial and temporal
characteristics of Northern Hemisphere surface air
temperature variations, J. Climatol., 3, 243-252, 1983.
Jones, P.D., S.C.B. Raper, R.S. Bradley, H.F. Diaz,
P.M. Kelly, and T. M. L. Wigley, Northern Hemisphere
surface air temperature variations, 1851-1984, J. Clim.
Appl. Meteorol., 25, 161-179, 1986a.
Jones, P. D., S.C. B. Raper, and T. M. L. Wigley, Southern
Hemisphere surface air temperature variations, 1851-1984,
J. Clim. Appl. Meteorol., 25, 1213-1230, 1986b.
Jones, P. D., P.M. Kelly, C. M. Goodess, and T. Karl, The
effect of urban warming on the Northern Hemisphere
temperature average, J. Clim., 2, 285-290, 1989.
Jones, P.D., P. Ya. Groisman, M. Coughlan, N. Plummer,
W-C. Wang, and T. R. Karl, Assessment of urbanization
effects in time series of surface air temperature over land,
Nature, 347, 169-172, 1990.
Karl, T.R., R.W. Knight, and J.R. Christy, Global and
hemispheric temperature trends: Uncertainties related to
inadequate spatial sampling, J. Clim., 7, in press, 1994.
Kent, E. C., P. K. Taylor, B. S. Truscott, and J. S. Hopkins,
The accuracy of Voluntary Observing Ships' meteorological
observations - Results of the VSOP-NA, J. Atmos. Ocean.
Technol., IO, 591-608, 1993.
Komura, K., and T. Uwai, The collection of historical ships'
data in Kobe Marine Observatory, Bull. Kobe Mar. Obs.,
211, 19-29, 1992.
Lamb, H. H., British Isles weather types and a register of the
daily sequence of circulation patterns 1861-1971, Meteorol.
0•. Geophys. Mere., 116, 85 pp, 1972.
Lamb, H. H., P. Collison, and R. A. S. Ratcliffe, Northern
Hemisphere monthly and annual mean-sea-level pressure
distribution for 1951-66, and changes of pressure and
temperature compared with those of 1900-39, Meteorol. 0•.
Geophys. Mere., 118, 36 pp, 1973.
Lehman, S. J., and L. D. Keigwin, Sudden changes in North
Atlantic circulation during the last deglaciation, Nature, 356,
757-762, 1992.
Madden, R. A., D.J. Shea, G.W. Branstator, J.J. Tribbia,
and R. O. Weber, The effects of imperfect spatial and
temporal sampling on estimates of the global mean
temperature: Experiments with model data, J. Clim., 6,
1057-1066, 1993.
Manabe, S., R.J. Stouffer, M.J. Spelman, and K. Bryan,
Transient responses of a coupled ocean-atmosphere model to
gradual changes of atmospheric CO2, I, Annual mean
response, J. Clim., 4, 785-818, i991.
Manabe, S., M.J. Spelman, and R.J. Stouffer, Transient
responses of a coupled ocean-atmopshere model to gradual
changes of atmospheric CO 2, II, Seasonal response, J.
Clim., 5, 105-126, 1992.
Maury, M.F., Explanations and Sailing Directions to
Accompany the Wind and Current Charts, vol. 1, pp.
383+51 plates, edited by W.A.Harris, Washington, D.C.,
1858.
Meehl, G.A., W.M. Washington, and T.R. Karl,
Low-frequency variability and CO 2 transient climate
change, Clim. Dyn., 8, 117-133, 1993.
O'Conneil, P.E., M.A. Beran, R.J. Gurney, D.A. Jones,
and R. J. Moore, Methods for evaluating the UK raingauge
network, Inst. Hydrol. Rep. 40, Wallingford, England,
1977.
Parker, D. E., The statistical effects of incomplete sampling of
coherent data series, J. Climatol., 4, 445-449, 1984.
Parker, D. E., Blending of COADS and UK Meteorological
Office marine data sets, in Proceedings International
COADS Workshop, Boulder, Colo., 13-15 Jan 1992, edited
by H. F. Diaz, K. Wolter and S. D. Woodruff, pp 61-72,
NOAA/ERL Spec. Publ., Environ. Res. Lab., 1992.
Parker, D. E., Effects of changing exposure of thermometers at
land stations, lnt. J. Climatol., 14, 1-31, 1994.
Parker, D.E., and C.K. Folland, The nature of climatic
variability, Meteorol. Mag., 117, 201-210, 1988.
PARKER ET AL.: INTERDECADAL CHANGES OF SURFACE TEMPERATURE 14,399
Parker, D.E., and C.K. Foiland, Worldwide surface
temperature trends since the mid-19th century, in
Greenhouse-Gas-Induced Climatic Change: ,4 Critical
Appraisal of Simulations and Observations, edited by
M. E. Schlesinger, pp. 173-193, Elsevier, 1991.
Parker, D.E., C.K. Foiland, A. Bcvan, M.N. Ward,
M. Jackson, and K. Maskell, Marine surface data for
analysis of climatic fluctuations on interannual to century
timescales, in Natural Climate Variability on Decade to
Century Timescales, edited by D. G. Martinson, K. Bryan,
M. Ghil, M.M. Hail, T.R. Karl, E.S. Sarachik,
S. Sorooshian, and L. D. Talley, National Academy Press,
Washington, D.C., in press, 1994.
Robock, A., and J. Mao, Winter warming from large volcanic
eruptions, C7eoph. Res. Lett., 12, 2405-2408, 1992.
Rodewald, M., On the question of atmospheric pressure
conditions in the Arctic (in German), Hamb. Meteorol. Amt
Nordwestdtschl. Ann. Meteorol., $ (9/10), 284-290, 1950.
(English translation available at the National Meteorological
Library, Brackncil, England.)
Trenberth, K.E., Recent observed interdecadal climate
changes in the Northern Hemisphere, Bull. Am. Meteorol.
$oc., 71,988-993, !990.
Trenberth, K.R., and D.A. Paulino, Jr., The Northern
Hemisphere sea-level pressure data set: Trends, errors and
discontinuities, Mon. Weather Rev., 108, 855-872, 1980.
Trenberth, K. R., J. R. Christy, and J. W. Hurreil, Monitoring
global monthly mean surface temperatures, J. Clim., 5,
1405-1423, 1992.
U.S. Department of Commerce, Climate Assessment- •1
decadal review 1981-1990, 109 pp., 1991.
van Loon, H., and J. Williams, The connection between trends
of mean temperature and circulation at the surface, I,
Winter, Mon. Weather Rev., 104, 365-380, 1976.
Vinnikov, K. Ya., P. Ya. Groisman, K.M. Lugina and
A. A. Golubev, Variations in Northern Hemisphere mean
surface air temperature over 1881-1985, (in Russian).
Meteorol. Hydrol., 1, 45-53, 1987.
Vinnikov, K. Ya., P. Ya. Groisman, and K.M. Lugina,
Empirical data on contemporary global climate changes
(temperature and precipitation), J. elire., 3, 662-677, 1990.
Woodruff, S. D., R. J. Slutz, R. J. Jenne, and P.M. Steurer, A
comprehensive ocean-atmosphere data set, Bull. Am.
Meteorol. $oc., 68, 1239-1250, 1987.
World Meteorological Organization (WMO), Report of meeting
of experts on Climate Change Detection Project,
Niagara-on-the-Lake, 26-30 November 1990, WCDP-13,
WMO/TD/418, 112 pp., 1990.
Zwiers, F. W., and H. yon Storch, Taking serial correlation
into account in tests of the mean, Max Planck Inst. j•r
Meteorol. Rep. 120, 23 pp., 1993.
A. Bcvan, C. K. Foiland, and D. E. Parker, Hadley Centre
for Climate Prediction and Research, Meteorological Office,
London Road, Brackncil, Berkshire RG12 2SY, England.
P. D. Jones, Climatic Research Unit, University of East
Anglia, Norwich, England.
(Received August 4, 1993; revised February 19, 1994;
accepted February 22, 1994.)
... Tout d'abord, les niveaux de CO2 observés dans l'atmosphère ont presque doublé depuis la révolution industrielle (IPCC, 2014). L'augmentation de la teneur en CO2 atmosphérique conduit au renforcement de l'effet de serre (Parker et al., 1994;Jones et al., 1999) et, par conséquent, au réchauffement climatique qui accompagne le changement climatique (CC) (Luterbacher et al., 2004;Zeng et al., 2008;Barriopedro et al., 2011;Lewis et al., 2011). Le CC impacte aussi le cycle de l'eau à tous les niveaux. ...
... Le CO2 est catalogué comme le principal élément qui contribue au renforcement de l'effet de serre (Parker et al., 1994;Jones et al., 1999). Les émissions anthropiques de CO2 ont des effets clairs et bien documentés depuis de nombreuses années, comme la hausse de la quantité de carbone présente dans l'atmosphère et contenue dans les océans (Keeling and Peng, 1995). ...
Thesis
Full-text available
Karst hydrosystems are very sensitive environments to anthropogenic activities (AA) and hydroclimatic variations, such as those caused by Climate Change (CC). Within the Critical Zone (CZ), these environments are the most reactive and the most vulnerable to these forces, which can alter the rapid internal transfer of dissolved elements, including contaminants. These karst systems are therefore excellent candidates to study the impact of CC and AA on water resources and on biogeochemical cycles of elements, especially C and S, strongly disturbed by land use changes, as well as by atmospheric deposition linked to AA (acid deposition). The objective of this thesis is therefore to understand the hydro-biogeochemical functioning of karst systems in temperate environments and to identify the impact of CC and AA. These questions were addressed at different spatial and temporal scales in the Baget karst watershed (BC, Ariège Pyrenees, forest basin, multilithological and quasi-pristine) and in about twenty other karst watersheds at the regional scale (Pyrenees and Massif Central). These basins have a robust database for studying their hydrogeochemical response to these impacts in the long or medium term. In BC, since the 1970s, hydroclimatic (precipitation, temperature, discharge), hydrochemical (major elements) and isotopic (13C, 34S) monitoring has allowed: (i) to identify and quantify the contribution of atmospheric, biological, anthropic and lithological sources to the fluxes of dissolved elements exported by the river; (ii) to determine the main environmental factors controlling the concentration and fluxes of dissolved elements exported by the river, such as lithology, drainage, epikarst, temperature and land use; (iii) establish the dissolved and particulate matter balance, the intensity of mechanical and chemical erosion, and the consumption of CO2 by chemical weathering of rocks and; (iv) highlight in-stream mechanisms such as calcite precipitation and CO2 outgassing, through the innovative development of 13CDIC-based mixing diagrams. In addition, the high frequency analysis allowed: (i) to identify the importance of the different types of flows (quick-response, subsurface and baseflows) on water quality; (ii) to establish a typology of floods according to their nature, intensity and element behavior; (iii) to quantify the importance of major hydrological events on the total annual fluxes of suspended solids (90%) and elements in solution (>50%) for less than one third of the time of the year. The evolution of major element concentrations in the BC is globally comparable to the trends observed in the Pyrenees (5 basins studied) and the Massif Central (15 basins). The increase in Ca+Mg (+5 µeq.L-1.yr-1) and alkalinity (+9 µeq.L-1.yr-1) in the BC drainage water over the last 40 years is statistically linked to the increase in temperature (+0.03 °C.yr-1), the decrease in flow (-4 L.s-1.yr-1) and the closure of the environment (recovery of the forest by +0.05 Km2.yr-1). This quantification was made possible by determining an original indicator of landscape evolution by image analysis since the 1940’s. Moreover, the decrease in sulphates (-2.2 µeq.L-1.yr-1) in the rivers reflects the decrease in acidic atmospheric deposition observed in French rural areas. This continuous decrease suggests that the destocking of sulphur from the soils, originating from acid atmospheric inputs, is still occurring today.
... Climate and Society at Columbia University (Parker et al., 1994;Reynolds and Smith, 1994;Kaplan et al., 1998). Atmospheric pressure, humidity, and horizontal and vertical wind components fields from 1949 to 2021 were obtained from the National Centers for Environmental Prediction (NCEP)-National Centers for Atmospheric Research (NCAR) reanalysis dataset (Kalnay et al., 1996;Kistler et al., 2001) with 5°spatial resolution. ...
Thesis
Full-text available
Riverine floods are one of the most common and devastating natural hazards worldwide. Between 1998 and 2017, floods were the most frequent type of disaster and the one that affected the largest number of people worldwide. Besides, positive trends of the frequency, magnitude, and financial impacts related to them from 1970 to 2020 were reported. Further, due to anthropogenic climate change, the frequency and magnitude of flooding are expected to increase. This motivates the urgent need for skillful flood projection and early warning systems to reduce or at least mitigate flood-related impacts in areas with high populations. To enable this, two components are essential - a robust modeling framework that can provide skillful forecasts of river flows with their attendant uncertainties at the short-lead time (i.e., up to 10 days lead) and seasonal forecasts before the start of the flood season (mid-lead time). The first component helps with the evacuation of the population before the occurrence of floods, while the second can help with planning and mitigation strategies before the wet season. Traditional models – physical and statistical - for streamflow projections fall short, especially in providing uncertainties to enable risk-based decision-making. Motivated by these gaps and needs, this dissertation developed a novel Bayesian Hierarchical Network Modeling framework to project daily streamflows and extremes on a river network. The modeling framework provides ensembles of flows and robust estimates of uncertainties for short and mid-term streamflow and extremes forecasting. The rainfed Narmada River Basin in Central India is selected as a testbed for this framework, where floods occur during the peak monsoon season of July-August. This translates into four main contributions: (1) Development of a Bayesian hierarchical network model (BHNM) for daily streamflow modeling and forecasting. The daily flows at each gauge are modeled as a Gamma distribution with parameters varying in space and time. The model uses the spatial dependence induced by the river network topology and hydrometeorological variables from the upstream contributing area between the covariates’ gauges. (2) Adaptation of the BHNM framework for simultaneous post-processing of daily streamflow obtained from a physically based model at multiple sites. (3) Implementation of the Bayesian hierarchical framework developed in (1) by incorporating information from multiple sources to provide real-time daily ensemble streamflow forecasts for the peak monsoon season at different lead times (1- to 10-day lead time). (4) Modification of the Bayesian hierarchical model (BHM) to model and forecast the monsoon seasonal frequency and magnitude of streamflow extremes for several lead times. In this implementation, the marginal distributions of seasonal extremes at each location are modeled as a Generalized Extreme Value (GEV) distribution with nonstationary parameters along with a Gaussian elliptical copula to capture the dependence structure. The applications show very good skills in modeling and forecasting modes. The BHNM framework and its various adaptations and applications make for novel methodological and application contributions to literature. The framework is general in that it can be adapted to any river basin worldwide. The developed Bayesian framework offers a robust approach for flood projection at multiple time scales and the associated risks, which will be of critical use in enabling operational, planning, and mitigation strategies for flood hazards.
... The monthly anomaly data were combined: in data set for 1856-1981, they were calculated based on vessel observations from the British Meteorological Bureau (https://www.metoffice.gov.uk; Parker et al. 1994;Kaplan et al. 1998); data after 1981 were taken from the US National Center for Environmental Prediction (https://www.ncep.noaa.gov), with the use of the optimum interpolation algorithm combining shipborne observations and remote sensing of the ocean surface on the grid 1x1° (Reynolds and Smith 1994). ...
... The monthly anomaly data were combined: in data set for 1856-1981, they were calculated based on vessel observations from the British Meteorological Bureau (https://www.metoffice.gov.uk; Parker et al. 1994;Kaplan et al. 1998); data after 1981 were taken from the US National Center for Environmental Prediction (https://www.ncep.noaa.gov), with the use of the optimum interpolation algorithm combining shipborne observations and remote sensing of the ocean surface on the grid 1x1° (Reynolds and Smith 1994). ...
... GST consists of global land surface air temperature (LSAT), which is the 2 m air temperature observed by land weather stations, and sea surface temperature (SST) observed by ships, buoys, and Argos. However, there are large uncertainties in the temperature data observed by weather stations, ships, buoys, and Argos in long-term observations, including uncertainties due to uneven spatial and temporal distribution of sampling (Jones et al., 1997;Brohan et al., 2006) and uncertainties due to stations, environment, and instrumentation changes (Parker et al., 1994;Parker, 2006;Trewin, 2012;Kent et al., 2017;Menne et al., 2018;. Nevertheless, several countries and research teams have applied different homogenization methods to generate a series of representative homogenized global land-sea surface temperature gridded datasets, including the Met Office Hadley Centre/Climatic Research Unit global surface temperature dataset (HadCRUT) (Morice et al., 2012), Goddard Institute for Space Studies Surface Temperature (GISTEMP) (Hansen et al., 2010;Lenssen et al., 2019), the NOAA's NOAA Global Temperature (NOAAGlobalTemp) (Vose et al., 2012;Zhang et al., 2019;Huang et al., 2020), and Berkeley Earth (BE) (Rohde et al., 2013a;Rohde and Hausfather, 2020), which serve as benchmark data for monitoring and detecting GST changes and related studies. ...
Article
Full-text available
Global surface temperature observational datasets are the basis of global warming studies. In the context of increasing global warming and frequent extreme events, it is essential to improve the coverage and reduce the uncertainty in global surface temperature datasets. The China global Merged Surface Temperature Interim version (CMST-Interim) is updated to CMST 2.0 in this study. The previous CMST datasets were created by merging the China global Land Surface Air Temperature (C-LSAT) with sea surface temperature (SST) data from the Extended Reconstructed Sea Surface Temperature version 5 (ERSSTv5). The CMST 2.0 contains three variants: CMST 2.0 − Nrec (without reconstruction), CMST 2.0 − Imax, and CMST 2.0 − Imin (according to their reconstruction area of the air temperature over the sea ice surface in the Arctic region). The reconstructed datasets significantly improve data coverage, whereas CMST 2.0 − Imax and CMST 2.0 − Imin have improved coverage in the Northern Hemisphere, up to more than 95 %, and thus increased the long-term trends at global, hemispheric, and regional scales from 1850 to 2020. Compared to CMST-Interim, CMST 2.0 − Imax and CMST 2.0 − Imin show a high spatial coverage extended to the high latitudes and are more consistent with a reference of multi-dataset averages in the polar regions. The CMST 2.0 datasets presented here are publicly available at the website of figshare, https://doi.org/10.6084/m9.figshare.16929427.v4 (Sun and Li, 2021a), and the CLSAT2.0 datasets can be downloaded at https://doi.org/10.6084/m9.figshare.16968334.v4 (Sun and Li, 2021b). Both are also available at http://www.gwpu.net (last access: January 2022).
... Many SST products from a variety of institutions have been released (Parker et al., 1994;Kaplan et al., 1998;Reynolds et al., 2002;Ray et al., 2012), but in this study, we opted for ERSST.v4, which is the current, updated version of ERSST. ...
Thesis
Full-text available
Climate change is undeniable and constitutes one of the major threats of the 21st century. It impacts sectors of our society, usually negatively, and is likely to worsen towards the middle and end of the century. The agricultural sector is of particular concern, for it is the primary source of food and is strongly dependent on the weather. Considerable attention has been given to the impact of climate change on African agriculture because of the continent’s high vulnerability, which is mainly due to its low adaptation capac- ity. Several studies have been implemented to evaluate the impact of climate change on this continent. The results are sometimes controversial since the studies are based on different approaches, climate models and crop yield datasets. This study attempts to contribute substantially to this large topic by suggesting specific types of climate pre- dictors. The study focuses on tropical Africa and its maize yield. Maize is considered to be the most important crop in this region. To estimate the effect of climate change on maize yield, the study began by developing a robust cross-validated multiple linear regression model, which related climate predictors and maize yield. This statistical trans- fer function is reputed to be less prone to overfitting and multicollinearity problems. It is capable of selecting robust predictors, which have a physical meaning. Therefore, the study combined: large-scale predictors, which were derived from the principal component analysis of the monthly precipitation and temperature; traditional local-scale predictors, mainly, the mean precipitation, mean temperature, maximum temperature and minimum temperature; and the Water Requirement Satisfaction Index (WRSI), derived from the specific crop (maize) water balance model. The projected maize-yield change is forced by a regional climate model (RCM) REMO under two emission scenarios: high emission scenario (RCP8.5) and mid-range emission scenario (RCP4.5). The different effects of these groups of predictors in projecting the future maize-yield changes were also assessed. Furthermore, the study analysed the impact of climate change on the global WRSI. The results indicate that almost 27 % of the interannual variability of maize production of the entire region is explained by climate variables. The influence of climate predictors on maize-yield production is more pronounced in West Africa, reaching 55 % in some areas. The model projection indicates that the maize yield in the entire region is expected to decrease by the middle of the century under an RCP8.5 emission scenario, and from the middle of the century to the end of the century, the production will slightly recover but will remain negative (around -10 %). However, in some regions of East Africa, a slight increase in maize yield is expected. The maize-yield projection under RCP4.5 remains relatively unchanged compared to the baseline period (1982-2016). The results further indicate that large-scale predictors are the most critical drivers of the global year-to-year maize-yield variability, and ENSO – which is highly correlated with the most important predictor (PC2) – seems to be the physical process underlying this variability. The effects of local predictors are more pronounced in the eastern parts of the region. The impact of the future climate change on WRSI reveals that the availability of maize water is expected to decrease everywhere, except in some parts of eastern Africa.
... GST consists of global land surface air temperature (LSAT), which is the 2 m air temperature observed by land weather stations, and sea surface temperature (SST) observed by ships, buoys, and Argos. However, there are large uncertainties in the temperature data observed by weather stations, ships, buoys, and Argos in long-term observations, including uncertainties due to uneven spatial and temporal distribution of sampling (Jones et al., 1997;Brohan et al., 2006) and uncertainties due to stations, environment, and instrumentation changes (Parker et al., 1994;Parker, 2006;Trewin, 2012;Kent et al., 2017;Menne et al., 2018;. Nevertheless, several countries and research teams have applied different homogenization methods to generate a series of representative homogenized global land-sea surface temperature gridded datasets, including the Met Office Hadley Centre/Climatic Research Unit global surface temperature dataset (HadCRUT) (Morice et al., 2012), Goddard Institute for Space Studies Surface Temperature (GISTEMP) (Hansen et al., 2010;Lenssen et al., 2019), the NOAA's NOAA Global Temperature (NOAAGlobalTemp) (Vose et al., 2012;Zhang et al., 2019;Huang et al., 2020), and Berkeley Earth (BE) (Rohde et al., 2013a;Rohde and Hausfather, 2020), which serve as benchmark data for monitoring and detecting GST changes and related studies. ...
Preprint
Full-text available
Global surface temperature observational datasets are the basis of global warming studies. In the context of increasing global warming and frequent extreme events, it is essential to improve the coverage and reduce the uncertainty of global surface temperature datasets. The China global Merged Surface Temperature Interim version (CMST-Interim) is updated to CMST 2.0 in this study. The previous CMST datasets were created by merging the China global Land Surface Air Temperature (C-LSAT) with sea surface temperature (SST) data from the Extended Reconstructed Sea Surface Temperature version 5 (ERSSTv5). The CMST2.0 contains three variants: CMST2.0-Nrec (without reconstruction), CMST2.0-Imax, and CMST2.0-Imin (According to their reconstruction area of the air temperature over the sea ice surface in the Arctic region). The reconstructed datasets significantly improve data coverage, whereas CMST2.0-Imax and CMST2.0-Imin have improved coverage in the Northern Hemisphere, up to more than 95 %, and thus increased the long-term trends at global, hemispheric, and regional scales from 1850 to 2020. Compared to CMST-Interim, CMST2.0-Imax and CMST2.0-Imin show a high spatial coverage extended to the high latitudes and are more consistent with a reference of multi-dataset averages in the polar regions. The CMST2.0 datasets presented here are publicly available at the website of figshare, https://doi.org/10.6084/m9.figshare.16929427.v4 (Sun and Li, 2021a) and the CLSAT2.0 datasets can be downloaded at https://doi.org/10.6084/m9.figshare.16968334.v4 (Sun and Li, 2021b).
Article
The aim of this study is to examine the periodic structures of forest fire numbers in Türkiye and sunspot numbers. The main purpose of this study is to investigate the variability of forest fire numbers in Turkey and sunspot numbers using continuous wavelet transforms and global spectra method. The change of sunspots observed in the photosphere layer of the sun is one of its important activities affecting the world. The sunspot number is often used to show solar activity. Wavelet analysis reveals the frequency components of signals such as the Fourier transform yet wavelet transform is actually a advanced and well planned version of Fourier transform. Wavelet transform techniques approach has been applied to the periodic analysis of the number of forest fires and sunspot numbers in Turkey. In this context, in the period from 1937 to the end of 2020, the records of the current fire numbers of the General Directorate of Forestry (OMG) and the sunspots numbers for the same period have been evaluated. The periodic structure of the existing 84-year forest fire numbers data and the sunspot numbers data in the same period are analyzed using Continuous Wavelet transform (SDD) and Global Wavelet Spectrum (GDS) techniques. While the dominant periodic component for forest fire numbers was 6.60 years, 10.67 years periodic component is determined for sunspot numbers. Other long-year periodic components of both variables have also been determined. As a result of this study, it has been determined that the periodic structures of the number of forest fires and the number of sunspots in Türkiye are very different from each other.
Article
One way of deducing vertical shifts in the altitudinal distribution of Colombian high-altitude páramo environments is by inferring fluctuations in the height of the local freezing level. In our research, we are implementing two complementary approaches to reconstruct Late Pleistocene to Holocene changes in the freezing level height (FLH) in two of the most extensively glacier-covered areas of the northern Andes. We combined remote sensing and field-based geomorphological mapping with time-series reconstruction of changes in the altitude of the 0°C isotherm. Changes in the FLH were based on already-published ~30 kyr paleo-reconstructions of sea surface temperatures (SSTs) of the eastern tropical Pacific and the western tropical Atlantic, as well as on reconstructed long-term sea level changes and empirical orthogonal functions of present-day (historical) Indo-Pacific and tropical Atlantic SST anomalies. We also analyzed the probability distribution of air-sea temperature differences and the spatial distribution of grid points with SSTs above the minimum threshold necessary to initiate deep convection. We considered available historical near-surface and free air temperature data of ERA-Interim reanalysis products, General Circulation Model (GCM) simulations, weather stations, and (deployed by our group) digital sensors, to assess the normal Environmental Lapse Rates (ELRs) at the regional to local scale. The combined maps of glacial landforms and our reconstructed FLHs provided us with a well-founded inference of potential past glacier advances, narrowing down the coarse resolution of ice margins suggested by previous research efforts. The extent of the areas with temperatures below the freezing point suggested here for the summits of our main study site exceeds in magnitude the corresponding glacier icecaps and front advances proposed by previous studies. Conversely, our average lowest altitudes of the FLH for our comparative site are consistently above the main glacier-front advances previously suggested. Our results indicate that, compared to the maximum upward changes that likely took place over the past ca. 20,000 years in our two areas of interest, the observed (present-day) upward shifts of the FLH have occurred at a rate that significantly surpasses our inferred rates. Our study helps fill the gaps in understanding past climatic changes and present trends in the region of interest and provides some insights into analyzing the signals of natural and anthropogenic climate change.
Article
Full-text available
This study results from recommendations made by a 1984 WMO Expert Committee on Ocean-Atmosphere Interaction Relevant to Long-Range Forecasting. The committee suggested that comparisons be carried out between monthly sea surface temperature (SST) analyses routinely made in several different countries in near real time. Emphasis was placed on the improvement of such analyses for use in operational long-range forecasting, especially for initializing dynamical long-range forecasting models. Six different monthly averaged SST analyses have been compared. The extent to which the analyses agree on several space scales and for regions covering the global oceans is shown, together with estimates of the magnitude of various types of errors. Independent estimates of SST obtained from expendable bathythermographs indicate that the monthly mean Meteorological Office (UKMO), Climate Analysis Center (CAC) in situ, and CAC blended analyses showed small differences (biases) from the expendable bathythermograph data. The differences were near to or below the margins of statistical significance over the Northern Hemisphere and the Southern Hemisphere tropics. Apparent negative biases in the analyses were noted, however, in the extratropical Southern Hemisphere. The authors finish with a discussion of recent improvements to the accuracy and scope of SST analyses for both long-range forecasting and climate studies. These improvements include an integrated analysis of ice limit, in situ and satellite SST data, and the developing use of optimum interpolation as a method of SST analysis. 43 refs., 8 figs., 3 tabs.
Chapter
Sea surface temperatures (SSTs) for the period 1856 to the present have been corrected to compensate for the use of uninsulated buckets prior to the early 1940s. Trends in the corrected SST are consistent with trends in independently corrected nighttime marine air temperatures (NMAT). Global-scale patterns of variation of annual anomalies of SST and NMAT, as revealed by the first three covariance eigenvectors, are also in close agreement. The corrected SST anomalies are also compared with those of nearby coastal and island land air temperatures. Global-scale agreement is good except in the early 20th century when the land data were relatively warm by up to 0.2°C. Proposed causes are the siting of thermometers in open-sided thatched sheds in tropical regions at that time, along with a marked tendency to warm westerly atmospheric circulation over Europe in winter. Combined fields of SST and land air temperature are presented. The relative overall coldness of the late 19th century land air temperatures appears to have arisen from inner-continental and high-latitude regions, especially in winter. Combined fields do not yield full global coverage even in the 1980s, so satellite-based SST data need to be blended carefully with the ship-based observations if monitoring of global climate is to be complete.
Article
Global-scale analyses of ocean thermal structure produced operationally at the U.S. Navy's Fleet Numerical Oceanography Center are verified, along with an ocean thermal climatology, against unassimilated bathythermograph (bathy), satellite multichannel sea surface temperature (MCSST), and ship sea surface temperature (SST) data. Verification statistics are calculated from the three types of data for February–April of 1988 and February–April of 1990 in nine verification areas covering most of the open ocean in the Northern Hemisphere. The analyzed thermal fields were produced by version 1.0 of the Optimum Thermal Interpolation System (OTIS 1.0) in 1988, but by an upgraded version of this model, referred to as OTIS 1.1, in 1990. OTIS 1.1 employs exactly the same analysis methodology as OTIS 1.0. The principal difference is that OTIS 1.1 has twice the spatial resolution of OTIS 1.0 and consequently uses smaller spatial decorrelation scales and noise-to-signal ratios. As a result, OTIS 1.1 is able to...
Article
Climatological monthly ocean-surface temperatures obtained from the National Center for Atmospheric Research and from Fleet Numerical Weather Central are merged and interpolated onto a 1° global grid. Monthly distributions of the main ice packs of the Arctic and Antarctic are digitized from Fleet Weather Facility ice charts and Navy atlases, and are incorporated into the global arrays. Machine-analyzed maps of the resulting distributions for the months of January, March, May, July, September and November are presented to indicate the seasonal variations of temperature and ice extent.
Article
The hemispheric mean land-surface air temperature series developed for both hemispheres by Jones et al. are updated through 1987. The global temperature estimate for 1987 is the third warmest on record, 1858-1987. Recent warming over the last 20 years (1967-86) is shown to be centered on middle latitude belts in both hemispheres.
Article
Climatic variability results not only from the complex dynamics of the atmosphere but also from feedbacks involving the atmosphere, the oceans, the biosphere, and ice and snow. External influences such as solar changes, volcanic eruptions, and man-made pollution may also influence climate. Observed variations range in time-scale from daily weather fluctuations to pronounced interdecadal changes of temperature, rainfall and atmospheric circulation. The longer-term changes may involve changes in variability as well as in average conditions. Except for the very shortest time-scales, local and regional variations of weather and climate cannot be understood without considering the whole globe. -from Authors