You are on page 1of 28

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 92, NO.

Dll, PAGES 13,345-13,372, NOVEMBER 20, 1987

Global Trendsof Measured SurfaceAir Temperature


JAMES HANSEN

NASA GoddardSpaceFlightCenter,blstitutefor SpaceStudies,New York

SERGEJ LEBEDEFF

SigmaData Services
Corporation,New York

We analyzesurfaceair temperaturedata from availablemeteorological stationswith principalfocuson


the period 1880-1985.The temperaturechangesat mid- and high latitude stationsseparatedby lessthan
1000 km are shownto be highly correlated;at low latitudesthe correlationfalls off more rapidly with
distance for nearby stations.We combine the station data in a way which is designedto provide
accuratelong-termvariations.Error estimatesare basedin part on studiesof how accuratelythe actual
station distributionsare able to reproducetemperaturechangein a global data set producedby a three-
dimensional general circulation model with realistic variability. We find that meaningful global
temperature change can be obtained for the past century, despite the fact that the meteorological
stationsare confinedmainly to continentaland island locations.The resultsindicate a global warming
of about 0.5ø-0.7øCin the past century,with warming of similar magnitudein both hemispheres;the
northern hemisphereresult is similar to that found by severalother investigators.A strong warming
trend between 1965 and 1980 raised the global mean temperaturein 1980 and 1981 to the highestlevel
in the period of instrumentalrecords.The warm period in recent years differs qualitativelyfrom the
earlier warm period centeredabout 1940; the earlier warmingwas focusedat high northern latitudes,
while the recent warnting is more global. We present selectedgraphsand maps of the temperature
change in each of the eight latitude zones. A computer tape of the derived regional and global
temperaturechangesis availablefrom the authors.

1. INTRODUCTION
globalclimateforcings,
suchas increasing
atmospheric
CO2.
Surface air temperature has been measured at a large In this paper we use the temperaturerecordsof meteorolog-
number of meteorological stations for the past century, ical stations to obtain an estimate of global surface air
mainly at northern hemisphereland locations.These station temperature change, and we estimate the errors due to
data have been used by a number of investigators[e.g., incompletespatialcoverage.
14qllett,1950; Mitchell, 1961; Budyko, 1969; Vinnikov et al., Jones et al. [1986c] recently published an estimate of
1980; Yamamoto and Hoshiai, 1980;Joneset al., 1982, 1986a; global near-surfacetemperaturechangeobtainedby combin-
Jones and Kelly, 1983; Bradley et al., 1985] to estimate ing the surfaceair temperaturemeasurementsof meteorolog-
ical stations with marine surface air and surface water
temperature change, with appropriate caveats concerning
restrictionsof spatial coverage(cf. review by Wigleyet al. temperaturemeasurements.We comparetheir resultswith
[1986]). Analysis of ocean surfacetemperaturechangehas ours at the end of this paper; our global mean and hemi-
alsobeen made [Palttidgeand Woodruff,1981;Barnett,1984; sphericmean results are generallyin good agreementwith
theirs.
Folland et al., 1984] on the basisof ship data. Becausethe
land and ocean data sets each have their own problems In section 2 we define the surface air temperature data
concerningdata quality and uniformity over long periods set we employ, including illustration of the global distribu-
(see previouslycited referencesabove, especiallyBarnett tion of stations, and we estimate the area over which the
[1984] and Joneset al. [1986a]), it seemsbetter to analyze temperature change obtained from a given station is
the two data sets separately, rather than lumping them meaningful.In .section3 we describethe method we use to
togetherprior to analysis.Another valuable sourceof global combine the records of different stations,which is designed
temperature data is provided by the radiosondestations to retain temperature change information wi•il... minimizing
[Angell and Korshover, 1983]. This source includes data effects of incomplete spatial and tempera! coverage. In
through the troposphere and lower stratosphere but is section 4 we present detailed graphs of our results for
restrictedto the period from 1958to the present. global, hemispheric,zonal and regional temperaturechange.
Although it is safer to restrict temperature analysesto In section5 we make several checksof the significanceof
regions with dense station coverage, there is a great the inferred trends, for example, by using an artificial
incentive for trying to obtain estimatesof long term global globaltemperaturehistorygeneratedby a three-dimensional
temperature change. Such global data would provide the generalcirculationmodel to obtain a measureof the error
most appropriatecomparisonsfor global climate modelsand due to incomplete spatial coverage, by reanalyzing the
would enhance our ability to detect possible effects of northern hemisphere temperature trend using a station
distribution comparable to that available in the southern
hemisphere, and by omitting urban stations to test for
possible anthropogenicheat island effects. In section 6 we
Copyright1987by the American GeophysicalUnion.
compare the derived hemisphericand global temperature
Paper number 7D0578. changewith the recent results of Jones et al. [1986c] and
0148-0227/87/007D-0578
$05.00 AngellandKorshover[1983;privatecommunication, 1987].

13,345
13,346 HANSEN
ANDLEBEDEFF:
GLOBAL
TEMPERATURE
TRENDS

1870 1900
9O , ,

60

50

-50

- 6O

: ,

-9O

1950 1960
9O

6O

3O

-3O

-6O

-90
-180 -120 -60 0 60 120 180 -180 -120 -60 0 60 12:0 180

Fig. 1. Global distributionof meteorologicalstationswith surfaceair temperaturerecordsfor the four indicateddates.


A circle of 1200-kin radius is drawn around each station.

The method of temperature analysisdescribedhere was continuation,Monthly ClimaticData of the World (MCDW),
previously used by Hansen et al. [1981], who presented published by The National Oceanic and Atmospheric
global, northern latitude, southernlatitude, and low latitude Administration (NOAA). The full data set is continuously
temperature change for the period 1880-1978. The present updated and is available in digital form from the National
paper represents a more complete documentationof the Center for AtmosphericResearch(NCAR), as describedby
method of analysis,includesadditional stationsin the basic Jenne[1975]andSpangler andJenne[1980].
data set, and includesa more comprehensive presentationof The data set available from NCAR changeswith time as
results.The temperature changesobtained in the two studies late station reports are obtained, additional stations are
are in closeagreement. added, and errors in existingstation records are found and
There are several features of our surfaceair temperature corrected.Our presentanalysisis basedon the NCAR data
study which we believe justify its publication,despite the set as of June 1984 to which we added data for 1984 and
existenceof the other studies mentioned previously.Our 1985obtainedfrom MCDW whichis preparedby the National
method of analysisis designedto utilize fully information Climatic Data Center of NOAA in cooperationwith the
from stationswith incompletespatialand temporalcoverage. World MeteorologicalOrganization.However,we emphasize
We obtain a quantitative estimate of errors due to in- that we have worked with several versions of the NCAR
completestation coverage.We show that meaningfulglobal data set in the past 7 years; none of the general con-
temperaturechangecan be obtained from only the meteor- clusionswhich we draw in this paper have varied as station
ological station data; thus we avoid the ambiguityinherent data were added or corrected.
in combining sea surface temperatureswith surface air The principal limitation of this data set for global or
stationdata as well as the difficultiesencounteredwith any
henrisphericanalysis is the incomplete spatial coverage,
marine (surface air or sea surface) temperaturesdue to illustratedin Figure 1 for four dates.Although the number
temporal changesin the nature of ships.Our presentation and geographicalextent of recordingstationson land areas
also includes some novel results, for example,long-term increasedstronglybetween 1870 and 1900, there were still
changesin the seasonalcycle. large areas in Africa and South America, and all of
Antarctica, without coverage in 1900. Substantial station
2. STATION TEMPERATURE RECORDS
data for Antarctica begins in the 1950s.Large ocean areas
The principal data sourcesfor surfaceair temperatureat remainwithout fixedmeteorologicalstationsat all times.
meteorologicalstations are the World Weather Records Another indication of station coverage is provided by
(WWR), publishedby the SmithsonianInstitutionand their Figure 2, which dividesthe global surfaceinto 80 equal area
HANSEN
ANDLEBEDEFF:
GLOBAL
TEMPERATURE
TRENDS 13,347

9O

__ 1900 ,• .,=,•..•_ 8 1888

6O
•• 43• 1752
1860
1881 1881 //-"•-1881
/

--1921 18 1836 1•81 1949 1854•1057•186• • 1875 1870•187• 1937


3O

18831906 1878'11871'1863
1921188818921901kJ882
1869186618801921lg41
--7 7 •B--•',42'•54
3 '73 7 • '4• 69' 30 7 7
1

-3O --1876 1941 1942 1660 1856 1943 18,57 1885 1951 19171 1841 1859

-- IBB8 1904 1949 1949 1864


0 0 x3• 3 I 3 0 6
-6O • 69 70 71 72 73 74 75 76

-- 45
• • 2 1• lo
-90 I I I T7 I I I I TM I I I Y I I I I aø
- 180 - 120 -60 0 60 120 180
Fig.2. Totalnumberof stations
in eachof 80 regions
of equalarea,andthedatewhencontinuous
coverage
beganfor
eachregion.A boxidentification
numberis givenin the lowerright-handcornerof eachbox.

"boxes,"the full dimensionof a box side being about 2500 reflectsthe dominanceof mixingby large-scaleeddies.At
km. The total number of stations within each box and the low latitudesthe most activeatmospheric
dynamicalscales
date at which continuousstation coveragebegan for that are smaller, but apparently there are also substantial
box is given in Figure 2. Most oceanboxescontain at least coherenttemperaturevariationson very large scales(for
a few island stations, but five of the 80 boxes have no example, due to the quasi-biennialoscillation, Southern
stations.Note that the fraction of the hemisphericarea Oscillation,and E1 Nifio phenomena),
whichabcountfor the
within each of the four latitudezones,startingat the pole, slighttendencytowardpositivecorrelations
at largestation
is 10, 20, 30 and 40%, respectively.Thus the Antarctic separations.
latitude band, which had practicallyno temperaturerecords We examinedthe dependenceof the correlationson the
until the secondhalf of the twentieth century,represents directionof the line connectingthe two stations.For the
5% percent of the global area. The neighboringlatitude regionsfor which this checkwas performed,the United
band, which also has poor station coverage,representsan Statesand Europe, no substantialdependenceon direction
additional10% of the globalarea. was found. For example,in these regionsthe average
Before defining a procedure for extractinglarge-area correlationcoefficientfor 1000-kmseparationwas found to
temperaturechangefrom measurements, it is important to be within the range 0.5-0.6 for each of the directions
have a quantitativemeasureof the size of the surrounding definedby 45ø intervals.We did not investigate
whetherthe
area for which a givenstation'sdata may providesignificant correlationsare more dependent on direction at low
informationon temperaturechange.For this purposewe latitudes.
computed the correlation coefficient between the annual The number of stations as a function of time is shown in
mean temperaturevariationsfor pairs of stationsselectedat Figure 4, as is the percent of the global surfacethat has a
random from among the station pairs with at least 50 station located within 1200 kin. The 1200-km limit is the
common years in their records.The distribution of correla- distance at which the average correlatioq coefficient of
tion coefficientsas a function of station separationis temperature variations falls to 0.5 at middle and high
shownin Figure3 for the samelatitudezonesas in Figure latitudes and 0.33 at low latitudes.Note that the global
2. At middle and high latitudesthe correlations approach coveragedefinedin this way doesnot reach 50% until about
unity as the station separationbecomessmall;the correla- 1900;the northernhemisphere obtains50% coveragein about
tions fall below 0.5 at a stationseparationof about 1200 1880 and the southernhemispherein about 1940.Although
km, on the average.At low latitudesthe mean correlationis the number of stationsdoubled in about 1950, this increased
only0.5 at smallstationseparation.The distanceoverwhich the area coverageby only about 10%, becausethe principal
strongcorrelations are maintainedat high latitudesprobably deficiencyis ocean areaswhich remain uncoveredeven with
13,348 HANSEN
ANDLEBEDEFF:
GLOBAL
TEMPERATURE
TRENDS

Station Separation
(km) the greater number of stations.For the same reason, the
1.00 I000 2000 :5000 4000 5000 decreasein the number of stationsin the early 1960s,(due
!,.•.•.'. • latitudesl
64.:).-9(:•øN to the shift from Smithsonianto Weather Bureau records),
0.8? •
doesnot decreasethe area coveragevery much. If the 1200-
km limit described above, which is somewhatarbitrary, is
reducedto 800 km, the global area coverageby the stations
in recent decades is reduced from about 80% to about 65%.
In addition to the limitationsimposedby the incomplete
spatial coverage, there are temporal inconsistenciesin
1.0 * •* -
certain station records,which are caused,for example,by
changes in instrumentation, station location, observation
Q6 time, or environmental factors such as urban heat island
effects,as discussedin detail by Joneset al. [1986a]. We
screened the data only to eliminate gross errors; the
screening involved examination of the space and time
-0.4 * t' * '
.
' * : variability of the temperature deviations from their long-
term mean. Specifically,we (1) used the time history at
• .,,• _. latitudes23.6-44.•[•N
each station to identify instanceswhen the temperature
deviation was more than five standard deviations from the
long-termmean,and (2) examinedcolormapsof temperature
0.2 • •, •..• change, as illustrated later, to identify any station with a
trend greatly inconsistentwith its surroundings. These cases
were individuallyexaminedand usuallyled to the discovery
-0.4 -
-0.6 '
of a misplaced decimal or incorrect sign for the tem-
perature. In caseswhere obviouslybad data was discovered
1.0 - latHudes
0-2:5.6"N
this was reported to NCAR so that the original data set
could be corrected.Undoubtedly some bad data with small
OA • •)• errors escapedthis screening,but the very large number of
stations reduces the large-scale impact of such errors.
Later, we also test the importance of urban heat island
-0.2. ." .. -.... •.:,• • •,..,"Z•
,
;Z•:'•.;* effects by selectively eliminating city stations from the
-0'4 ., - •* -•.•.'. analysis,and we estimate the uncertainty in the global
-0.6 ' ' ' ' ,
- trends.Perhapsthe best indicationthat these problemsdo
,.o-.4•, ,•,,*'
0.8
" •
•.•. , ',.'
,
•'•
lati'iudes
0-2 .6"•
-
- not have a dominant effect on the results is provided by
exanfinationof the physicalnature of the geographicand
0.6 •' '•" • •'•;; temporalpatternsof the derivedtemperaturechange.

3. SPATIAL AVERAGING: BIAS METHOD

-o.2 ::•""'":"
-o.4-:•-"7; ....:•':½."
':t?, •.,.: :v'Z;'
. ' Our principal objective is to estimate the temperature
changeof large regions.We would like to incorporatethe
-0.6
information from all of the relevant available station
1.0 -
records. The essence of the method which we use is shown
0.8 latiludes
23.6-44.,
•'S_
O.6- • "• -- schematically
in Figure 5 for two nearby stations,for which
we want the best estimate of the temperature change in
o.4 •.•..%.,:½,
•.',) their mutual locale. We calculate the mean

for the period in common, and adjust the entire second


of both records

, * record(T2) by the difference(bias) ST. The meanof the


-0.4 >,:•, •: '.•': :" •* resultingtemperature records is the estimatedtemperature
change as a function of time. The zero point of the

,.o
0.8 • latitudes
•4.4-64.
temperaturescaleis arbitrary.
0.6 , -
I * A principal advantageof this method is that it usesthe
full period of common record in calculatingthe bias 8T
o.2 .. %: - between the two stations. Determination of 8T is the

-0.2
- OA
0
0[ I000
, '*,-'
*
2000

5000
-
4000
• -
-
5000
essence of the problem of estimating the area-average
temperature change from data at local stations.A second
advantageof this method is that it allows the information
StationSeparation
( km)
from all nearby stationsto be used, provided only that each
stationhave a period of record in commonwith another of
Fig. 3. Correlation coefficientsbetween annual mean temperature the stations. An alternative method commonly used to
changesfor pairs of randomlyselectedstationshavingat least 50
commonyears in their records.Each dot representsone station combine station records is to define 8T by specifyingthe
pair. The latitude zones are the same as those of the boxes in mean temperature of each station as zero for a specific
Figure 2. period which had a large number of stations,for example,
•SœN •U•D Lœ•œDœW:GLOBALTEMPERATURE
TRENDS 13,349

2000

1500

"-' I000 I00


o

80
E
z
60
500
40

o/Ioo -o

Northern
Hemisphere . • ,.
80 •

60

40

__ • / Southern
Hemisphere __ 20
•,•-•.•/• _ (b)
1850 1875 1900 1925 1950 1975 1985
Date

Fig.4. (a) Numberof stations(histogram)and percentof globalarea locatedwithin 1200km of a station(heavyline).


(b) Percentof hemispheric
arealocatedwithin 1200km of a station.

1950-1980;this alternativemethod comparesunfavorablyto for t withavailableTn,and


ours with regard to both making use of the maximum
number of stations and defining the bias ST between Wl,n(t) : Wl,n.l(t) (3c)
stationsas accuratelyaspossible. rl,n(t): rl,n.l(t) (3d)
A complete description of our procedure for defining
large-areatemperature changeis as follows. We divide each for t withoutavailableTn. T represents
temperature
change,
of the 80 equal-area boxes of Figure 2 into a 10 by 10 t is time,andn identifies
the station.TLn(Ois an inter-
array of 100 equal-areasubboxes.(The east-westand north- mediate estimate of the temperature change based on
southdimensionsof a subboxare about 200 km; in the polar stations1 through n; these equationsare applied repeatedly
boxes the equal area requirement for the subboxescauses untilT•,N(t
) is obtained,
whereN is the totalnumberof
the polemost subboxesto be noticeably elongated in the stationswithin 1200km of the subboxcenter.Here dn is
north-southdirection.) For each subboxwe use all stations the distance of the nth station from the subbox center, and
located within 1200 km of the subbox center to define the
dn is usedto calculatethe weightWn by whichthe nth
temperature change of that subbox.The N stationswithin station temperaturechange is weighted. Wn decreases
1200 km are ordered from the one with the greatestnumber linearly from 1 at the subboxcenter to 0 at a distanceD,
of years of temperature record to the one with the least where we have taken D=1200 km as a representative
number. The temperature changesfor the first two stations direction-independentdistanceover which the temperature
are combined, then the third with the first two, and so on, changesexhibitstrongcorrelation.
using The temperature changesof the 100 subboxesare then
combinedto find the temperaturechangefor a box, in the
manner indicated by Figure 5 and (1)-(3). However, the
Wn = (D - dn)/D (2) subboxesare weighted equally, i.e., by area, except that
subboxes which have no station within 1200 km are ex-
W],n(
0: W],n.](
0 + Wn (3a)
cluded. The zonal mean temperature change for a latitude
+ W.(T.4T.)]/%,. band is obtained by combiningthe temperaturechangesof
13,350 HANSEN
ANDLEBEDEFF:
GLOBAL
TEMPERATURE
TRENDS

averagingthe station records to obtain annual mean data,


then applyingthe describedprocedure,and (2) averagingthe
temperaturechangesobtained with this procedurefor the
individual months. The differenc• in the results from the
two methodswere small, of the order of 10%, but we chose
T•(t) the latter method becauseit incorporatesall availabledata,
i.e., it usesthe recordsfor years in which data are missing
for 1 or more months.
If our data are employed at their highest (monthly)

I
resolution,it should be noted that, althoughthe seasonal

. pJ. cyclehas been removedto first order, the effect of changes


in the seasonalcycleare still present.Long-termchangesof
the seasonalcycle are not unexpectedand, indeed, we
illustratelater that somehave occurredin the past century.
Thus a spectral analysis of the long-term temperature
changesat monthlyresolutionshouldbe expectedto yield a
I I I I peak at 12 months.
One issuewith the bias method is how many years of
recordoverlapshouldbe requiredfor a stationrecordto be
Fig.$. Illustrationof how the temperaturerecordsof two stations combinedwith that of its neighbors.For example,it would
are combined.
The value•T is computed fromthe averagesof Tl(t) seeminappropriateto combinethe recordof a stationwhich
andT•.(t)overthe common yearsof record.The entirecurveT•.(t) had only 1 year in commonwith its nearby(within 1200km)
is thenmoveddownward by theamount•T, andTl(t) andTa(t) are
averaged.In general,Tl(t) and Ta(t) maycontaingaps,as may neighbors,becauselocal interannualfluctuationsare often
the commonperiodusedto compute•T. as large as the long-termchangeswhich we seek to def'me.
For the results we present, we used only station records
which had an overlap of 20 years or more with the combin-
ation of other stations within 1200 km. We tested other
the boxesin this sameway, with each box weightedby the choicesfor this overlap period and found little effect on
fraction of its subboxeswhich have a defined temperature the global and zonal results.Some effect could be seen on
change, i.e., the fraction of the box which has a station global maps of derived temperaturechange;a limit of $
within 1200 kin. Finally, the zonal temperature changesare years or less causedseveral unrealisticlocal hot spots or
combinedin the sameway to obtain hemisphericand global cold spots to appear, while a limit greater than 20 years
temperaturechange,with each latitude band weightedby the caused a significant reduction in the global area with
areawith a definedtemperaturechange. stationcoverage.
One potential disadvantageof the method we have We stressthat our procedure for defining temperature
describedfor combiningstation recordsis that the results, change is designed for obtaining the results for large
in principle, depend on the ordering of the stationrecords. regions,from the 1000 km scale to the global scale. For
However, we have tested the effect of other choices for some local studies it is better to start with the raw station
stationordering,for example,by beginningwith the station data, ff a local station exists, rather than to start with the
closest to the subbox center, rather than the station of temperaturechangewe have obtained for the local subbox
longest record. The differences between the results for (which is influencedby station data up to 1200 km away).
alternative choiceswere found to be very small, about 2 However,we includethe subboxtemperaturechangeon the
orders of magnitudelessthan the typicallong-termtemper- data tape which we make available,and we anticipatethat
ature trend.
it will be usefulfor manypurposes.For example,it provides
We have also tested alternativesto theseproceduresand an estimate of the small-scaletemperature changewhere
comparedthe error estimatesfor the alternatives,the error local stationsdo not exist (provided there is at least one
estimates being obtained as described in Section 5. For station within 1200 kin). Also, because of the variability
example,we tried weightingeach box by the box area and which exists on scales of a f•w hundred kilometers or more
each zone by the zone area, rather then weightingby the (see Figure 3), if a user is interested in area-average
area with a definedtemperaturechange.Overall temperature temperature change over scalesof at least a few hundred
changeswere similar with the different procedures,but the kilometers,it is probably better to use our subboxresults
procedureas we defined it previouslywas found to yield the rather than a local station. Finally, provisionof the subbox
smallest errors of the alternatives which were tested. We
results allows the user the possibility of averaging over
also tried alternatives to the 1200-kin limit defined earlier; largeregionsother than thosewhichwe havechosen.
althoughthe effectswere noticeableon geographicalmaps
4. DERIVED TEMPERATURE CHANGES
of temperature trends, there was no significanteffect on
zonal,hemispheric, or globaltemperaturechanges. We summarizehere some of the derived temperature
The time unit is the monthlymean in the stationrecords change characteristicswhich are significant,based on the
whichwe obtainfrom NCAR. We obtainmonthlytemperature error analysisin the following section. The global and
changesfor a given subbox by applying the previously hemispheric annual-meantemperaturechangesfor the period
describedprocedureindividuallyto each of the 12 months, since1880are shownin Figure6 and Table 1. Figure6 also
with the zero point for each month being its 1951-1980 includesthe 5-year smoothedtemperaturechangeand the
mean. We tried two methodsfor obtainingannual trends:(1) estimatederror(95% confidence limits,approximately
+_2a)at
HANSENAND LEBEDEFF:GLOBALTEMPEBATURE
']"RENDS 13,351

0.6

0.5

0.4

0.5

0.2

0.1

-0.1 0.6

-0.2 0.5

-0.3 0.4

-0.4 0.3

-0.5 0.2

0 o._..

-0.1<•
.... 0.2

0.4 -0.5

0.3' -0.4

0.2

0.1

-0.1

-0.2

-0.3
,i......•,......•i•......2{....•. Southern Hemisphere
•L..... •,............ i ...........................................................................
-0.4

-0.5 i i I i i I I I
1880 1900 1920 1940 1960 1980

Date
Fig.6. Global and hemispheric surfaceair temperaturechangeestimatedfrom meteorological stationrecords.The
northernhemispherescaleis on the right. The 5-year runningmean is the linear averagefor the 5 yearscenteredon
the plottedyear. The uncertaintybars (95% confidence limits) are basedon the error analysisin section5; the inner
barsrefer to the 5-yearmeanand the outer bars to the annualmean.

several dates based on the error analysisin the following We includethe 5-year runningmean in a number of our
section.The smaller bar refers to the uncertaintyin the 5- graphsbecauseit providesa simplesmoothingwhichhelps
year mean and the larger bar refers to the uncertaintyin clarify long-termchange.However,we emphasizethat the
the annualmean temperature.Note that the stationdistri- resultingcurveis not appropriatefor studyof "cycles,"
such
butionsfor 1900,1930and 1960are givenin Figure 1. as those appearingin the southernhemisphererecordin
by about0.5øC Figure 6. For studiesof periodicitiesone may employthe
The smoothedglobaltemperatureincreases
unsmoothed data with appropriate filtering techniques
between 1880 and 1940, decreasesby about 0.2øC between
[Mitchellet al., 1966].
1940 and 1965, and increasesby about 0.3øCbetween1965
and 1980. The northern hemispheretemperature changeis The global surfaceair temperaturein 1981 reacheda
rather similar to the global change,increasingby 0.6øC warmer level than obtained in any previous time in the
between 1880 and 1940, decreasingby 0.3øCbetween 1940 periodof instrumentalrecord.The 1981maximumexceeded
and 1970,and increasingby 0.3øCbetween1970 and 1980. the maximumof 1940 by about 0.2øC,which is larger than
The southernhemispheretemperaturechangeis noisier,but, the estimatederror. In the northernhemisphere1981is also
especiallyif averagedover a 10-yearinterval or longer, it the warmestyear on record, as already noted by Joneset
shows a more steady increase in temperature, with a al. [1982]. The 1981 peak in the northern hemisphereis
warmingof about 0.6øCbetween1880 and 1980.The largest 0.1øCgreater than any previousyear in the record,but the
rate of warmingis between1965and 1980. 5-yearsmoothedtemperaturein 1981is only aboutthe same
13,352 HANSEN
ANDLEBEDEFF:
GLOBAL
TEMPEI•TLIRE
TRENDS

TABLE 1. SurfaceAir TemperatureChangefor the GlobeandSpecifiedRegions.


,

Zone Box

Year Globe NH SH 1 2 3 4 5 6 7 8 6 7 9 10 15 16

1880 -0.40 -0.52 -0.10 -1.06 -0.72 -0.32 -0.53 -0.13 -0.08 0.23 .... 1.69 -0.75 -0.41 -0.97 -0.92 0.36
1881 -0.37 -0.41 -0.25 -1.19 -0.69 -0.13 -0.29 -0.19 -0.26 0.07 ßßß -0.77 -0.47 -1.20 -0.74 -0.34 0.25
1882 -0.43 -0.47 -0.33 -1.71 -0.39 -0.32 -0.39 -0.35 -0.28 -0.11 ßßß -0.56 -1.30 0.37 -0.85 -0.42 0.20
1883 -0.47 -0.55 -0.25 -0.26 -0.86 -0.49 -0.43 -0.18 -0.30 0.17 .... 1.77 -1.92 -0.29 -0.81 -0.78 -0.20
1884 -0.72 -0.78 -0.55 -1.35 -0.90 -0.68 -0.65 -0.33 -0.54 -0.77 .... 1.51 -2.09 -0.23 -1.03 -0.86 -0.06
1885 -0.54 -0.61 -0.34 -1.83 -0.74 -0.50 -0.17 -0.14 -0.35 -0.41 ßß ß -0.27 -1.44 -0.36 -1.06 -0.48 -0.70
1886 -0.47 -0.51 -0.34 -1.55 -0.53 -0.45 -0.25 -0.20 -0.34 -0.38 ßßß -0.79 -0.95 -0.18 -1.07 -0.75 -0.68
1887 -0.54 -0.59 -0.43 -1.88 -0.60 -0.31 -0.53 -0.53 -0.35 -0.10 .... 1.74 -1.44 -0.27 0.21 -0.46 -0.29
1888 -0.39 -0.46 -0.22 -1.50 -0.69 -0.28 -0.15 0.09 -0.18 -0.65 .... 0.43 -1.24 -1.28 -0.10 -0.51 -0.80
1889 -0.19 -0.24 -0.07 -0.84 -0.29 -0.25 0.00 0.30 -0.18 -0.22 ß ßß 0.86 -0.11 -0.46 -1.03 -0.17 -0.12

1890 -0.40 -0.40 -0.39 -1.18 -0.47 -0.17 -0.43 -0.60 -0.28 0.08 ßß ß -0.48 -0.93 -0.04 -0.94 -0.17 0.04
1891 -0.44 -0.45 -0.43 -1.27 -0.48 -0.40 -0.26 -0.55 -0.30 -0.17 .... 0.10 -0.38 -0.34 -1.27 -0.78 -0.35
1892 -0.44 -0.49 -0.32 -1.22 -0.66 -0.35 -0.32 -0.30 -0.35 0.03 ßßß -0.49 -0.14 -0.54 -0.85 -0.84 -0.84
1893 -0.49 -0.52 -0.43 -0.65 -0.46 -0.56 -0.49 -0.65 -0.32 0.41 .... 1.51 -0.77 -0.80 0.00 -0.74 -0.77
1894 -0.38 -0.42 -0.30 -0.79 -0.43 -0.37 -0.38 -0.46 -0.22 0.39 ßßß 0.03 -0.89 0.01 -0.75 -0.51 -0.01
1895 -0.41 -0.48 -0.26 -0.71 -0.69 -0.46 -0.25 -0.30 -0.23 0.08 .... 0.39 -0.14 -0.44 -0.67 -0.96 -0.57
1896 -0.27 -0.37 -0.08 -1.04 -0.66 -0.25 -0.06 -0.22 0.15 0.12 .... 0.68 -0.92 -0.30 -0.83 -0.07 -0.15
1897 -0.18 -0.23 -0.07 -0.37 -0.45 -0.31 0.06 0.04 -0.02 -0.29 .... 0.27 43.85 0.12 -0.85 -0.48 -0.01
1898 -0.38 -0.41 43.33 -1.36 43.49 -0.22 -0.29 -0.23 -0.32 43.40 ß ßß 0.05 43.45 0.18 -0.99 -0.71 0.08
1899 -0.22 -0.22 -0.24 -0.97 -0.20 -0.09 -0.18 -0.32 -0.03 -0.24 ßß ß -0.83 43.57 43.09 0.49 -0.77 -0.22

1900 -0.03 -0.08 0.07 43.69 -0.11 -0.05 0.09 0.14 0.19 -0.27 --- 0.58 0.01 -0.15 -0.44 0.13 0.12
1901 -0.09 -0.09 43.07 -0.58 0.10 -0.22 0.00 -0.06 43.05 0.10 .-. 0.45 0.12 0.00 -0.15 -0.22 -0.66
1902 -0.28 -0.39 43.09 -1.67 -0.49 -0.25 -0.07 0.08 -0.03 -0.68 ß ß. 0.05 0.05 -1.20 -0.95 43.44 -0.41
1903 -0.36 -0.40 -0.29 -0.44 -0.28 -0.59 -0.30 -0.17 43.29 -0.46 .... 0.25 -0.75 0.36 -0.41 43.84 43.56
1904 -0.49 -0.48 -0.52 -0.47 -0.43 -0.47 -0.55 -0.63 -0.24 -0.41 .... 0.35 -1.36 -0.31 0.18 -0.40 -1.07
1905 -0.25 -0.29 -0.19 -0.06 -0.06 -0.59 -0.27 -0.01 43.28 -0.40 ... 0.58 -0.76 0.04 -0.19 -0.75 -0.56
1906 -0.17 -0.17 -0.17 -0.28 0.03 -0.39 -0.12 -0.07 -0.11 -0.41 ß. ß 0.63 -0.44 0.35 0.14 43.46 -0.18
1907 -0.45 43.50 -0.37 -0.41 43.71 -0.54 -0.34 -0.37 -0.28 -0.25 . .. -0.64 -1.55 -0.68 -1.10 -0.32 -0.56
1908 -0.32 -0.37 -0.23 -0.28 -0.41 43.42 -0.33 -0.25 -0.25 0.07 -.- 0.39 -0.17 -0.67 -1.36 -0.31 0.02
1909 -0.33 -0.39 -0.22 -0.77 -0.45 -0.38 -0.25 -0.20 -0.21 -0.06 .... 1.02 0.06 -0.49 0.13 -0.67 -0.33

1910 -0.32 -0.40 -0.19 -0.79 -0.15 -0.45 43.43 -0.25 -0.13 0.09 .-. 0.42 0.02 0.46 -0.39 -0.03 43.53
1911 -0.29 -0.30 -0.28 -0.21 -0.27 -0.41 -0.25 -0.29 -0.23 43.09 ßß . 43.63 43.43 0.17 -0.56 -0.30 0.08
1912 -0.32 -0.45 -0.08 -0.86 -0.66 -0.50 -0.13 0.09 43.07 -0.39 ..ß 0.08 -1.03 -0.60 -1.01 -1.17 -0.58
1913 -0.25 -0.35 -0.08 -0.69 -0.27 -0.45 -0.24 -0.03 0.06 43.32 - ß. -0.13 -0.57 0.31 0.19 -0.73 0.07
1914 -0.05 -0.13 0.08 -0.74 -0.03 -0.18 0.02 0.12 0.16 0.01 - ßß 0.32 -1.35 0.27 0.16 43.25 43.62
1915 -0.01 -0.06 0.09 -0.97 -0.10 -0.07 0.24 0.28 0.11 -0.29 ... 0.85 0.21 43.54 0.49 43.62 -0.32
1916 -0.26 -0.34 -0.11 -0.73 -0.43 -0.26 -0.25 -0.06 -0.04 -0.08 .... 1.05 -0.12 0.16 43.44 -0.65 -0.48
1917 -0.48 -0.60 -0.26 -1.53 -0.54 -0.58 43.39 -0.38 -0.28 0.55 .... 1.00 -0.99 43.57 0.11 -0.89 -1.27
1918 -0.37 -0.43 -0.27 -1.25 -0.29 -0.37 -0.33 -0.34 -0.13 -0.07 -.. 0.35 -1.29 0.12 -0.52 -0.30 -0.39
1919 -0.20 -0.29 -0.03 -0.86 -0.53 -0.24 0.00 0.00 0.01 0.01 -.. -0.25 -0.43 -0.70 -0.92 -0.64 -0.03

1920 -0.15 -0.18 -0.11 0.07 -0.12 -0.30 -0.21 -0.04 -0.04 -0.29 . ß- 0.01 -0.56 0.52 -0.31 -0.67 -0.62
1921 -0.08 -0.08 -0.09 -0.16 0.17 -0.05 -0.25 -0.03 0.00 -0.22 . ß. 0.44 -0.33 0.49 0.72 0.54 0.55
1922 -0.14 -0.19 -0.07 -0.15 -0.35 -0.08 -0.18 -0.02 -0.07 0.05 .... 0.11 -0.71 -0.48 0.69 -0.21 0.04
1923 -0.13 -0.10 -0.19 0.44 0.06 -0.22 -0.24 -0.11 -0.12 -0.42 ..ß 0.58 -0.77 0.05 0.89 43.41 -0.36
1924 -0.12 -0.12 -0.11 0.20 -0.13 -0.25 -0.09 0.05 -0.26 -0.10 -.. -0.23 -0.45 -0.22 0.50 -0.76 -0.71
1925 -0.10 -0.04 -0.21 -0.17 0.41 -0.14 -0.22 -0.11 -0.18 -0.42 ..ß 0.34 -0.54 0.47 1.16 0.05 0.03
1926 0.13 0.17 0.06 0.52 0.46 -0.11 0.12 0.20 0.12 -0.39 -.- 0.87 -0.91 0.15 0.04 -0.05 -0.55
1927 -0.01 0.01 -0.04 0.01 -0.05 -0.01 0.05 0.09 0.10 -0.61 .... 0.67 0.00 -0.06 -0.22 0.02 0.20
1928 0.06 0.11 -0.01 0.77 0.07 -0.12 0.12 0.14 0.08 -0.42 ... 0.86 -0.02 -0.14 -0.51 -0.12 43.20
1929 -0.17 -0.16 -0.18 0.40 -0.29 -0.31 43.11 -0.12 -0.09 -0.33 .... 0.39 43.54 43.64 -0.92 -0.65 43.15

1930 -0.01 0.09 -0.19 0.66 0.19 -0.04 -0.04 -0.11 0.07 -0.72 . ß. 0.55 0.29 0.79 -0.12 -0.12 0.13
1931 0.09 0.16 -0.04 0.70 0.01 0.05 0.19 0.12 -0.12 -0.26 ... 1.61 1.01 -0.39 -0.29 0.57 0.63
1932 0.05 0.10 -0.03 0.51 0.27 -0.05 -0.03 0.04 0.09 -0.27 ß ß. -0.19 0.19 0.41 0.84 -0.32 0.37
1933 -0.16 -0.23 -0.04 -0.07 -0.47 -0.25 -0.12 0.03 0.03 -0.11 -.. -0.60 -0.56 -0.58 -1.12 0.38 0.49
1934 0.05 0.07 0.01 0.87 0.36 -0.08 -0.25 0.04 0.07 0.07 ... 0.98 -0.77 1.16 -0.32 1.12 -0.09
1935 -0.02 0.00 -0.05 0.51 0.08 -0.10 -0.13 0.11 -0.03 -0.51 .... 0.32 -0.44 0.31 0.03 -0.03 -0.23
1936 0.04 0.06 0.01 0.47 0.19 -0.13 0.00 0.08 -0.01 -0.02 .... 0.59 -0.27 0.74 0.46 0.12 -0.03
1937 0.17 0.25 0.04 1.40 0.21 0.13 0.04 0.14 -0.02 -0.14 ... 0.04 0.47 0.84 -0.38 -0.22 0.14
1938 0.19 0.29 0.01 1.49 0.52 0.17 -0.12 0.05 0.13 -0.14 ß. . 1.10 0.12 1.10 0.30 0.47 0.51
1939 0.05 0..?7 0 14 0.91 0.32 0.14 -0.14 0.00 -0.03 -0.75 ... 0.57 -0.37 0.52 0.44 0.59 0.32

1940 0.15 0.16 0.13 0.90 0.15 -0.01 0.06 0.20 0.09 0.16 -.- 0.95 0.18 -0.96 0.09 0.13 -0.74
1941 0.13 0.13 0.14 0.06 -0.13 0.21 0.24 0.31 0.05 -0.11 ... 1.10 0.28 -1.20 -0.97 0.25 0.24
1942 0.09 0.07 0.12 0.41 0.01 0.04 0.03 0.24 0.14 -0.07 -.. 0.63 0.36 -1.09 -0.35 0.02 0.10
HANSENANDLEBEDEFF.'
GLOBALTEMPERATURE
TRENDS 13,353

TABLE 1. (continued)

Zone Box

Year Globe NH SH 1 2 3 4 5 6 7 8 6 7 9 10 15 16

1943 0.04 0.08 -0.03 1.13 0.37 -0.05 -0.32 -0.05 -0.05 0.37 ßßß 0.49 -0.58 0.71 0.31 0.30 -0.21
1944 0.11 0.16 0.04 1.07 0.47 -0.11 -0.14 0.07 0.15 -0.04 ß ßß 1.12 0.22 0.46 0.98 -0.13 0.12
1945 -0.03 -0.04 -0.02 0.49 -0.05 -0.13 -0.13 0.10 0.14 -0.45 .... 0.17 0.00 0.04 -0.92 0.00 0.09
1946 0.03 0.08 -0.05 0.08 -0.03 0.21 0.03 0.05 -0.02 -0.32 ß ßß 0.16 -0.04 0.31 -0.04 0.53 0.52
1947 0.15 0.18 0.11 0.93 -0.03 0.15 0.11 0.22 0.05 0.05 ßß ß 0.30 0.61 0.12 -0.11 0.06 0.07
1948 0.04 0.10 -0.06 0.27 0.32 0.04 -0.05 0.10 0.01 -0.44 .... 0.12 0.04 0.58 0.75 -0.11 0.24
1949 -0.02 0.04 -0.10 0.40 0.15 0.02 -0.12 0.08 0.01 -0.49 .... 0.12 0.01 0.88 -0.12 -0.06 0.83

1950 -0.13 -0.13 -0.14 0.45 -0.22 -0.09 -0.28 -0.10 0.08 -0.44 .... 1.33 -0.54 0.43 -0.38 0.13 -0.03
1951 0.02 0.07 -0.05 0.14 0.08 0.07 0.03 0.05 -0.16 0.04 .... 1.07 0.47 0.45 0.31 -0.23 0.19
1952 0.07 0.09 0.05 0.22 -0.08 0.16 0.09 0.20 -0.04 0.00 ß ßß 0.53 1.06 -0.08 -0.61 0.15 0.48
1953 0.20 0.33 0.03 1.01 0.44 0.22 0.14 0.15 0.06 -0.08 ß- ß 1.05 0.91 0.41 0.23 0.64 0.81
1954 -0.03 0.02 -0.10 0.81 -0.22 0.02 -0.07 -0.04 -0.02 -0.20 ßßß 0.16 0.25 -0.25 -0.93 0.84 0.43
1955 -0.07 -0.09 -0.04 -0.31 -0.05 0.14 -0.22 -0.16 -0.14 0.22 .... 0.81 0.90 -0.29 0.25 -0.13 0.20
1956 -0.19 -0.29 -0.05 -0.07 -0.57 -0.23 -0.25 -0.24 -0.13 0.39 .... 0.43 -0.30 -1.28 -0.68 0.27 0.06
1957 0.09 0.03 0.17 0.09 0.26 -0.12 0.01 0.14 0.10 0.22 0.43 0.05 0.00 0.61 0.35 0.12 0.35
1958 0.11 0.16 0.04 -0.16 0.20 0.17 0.24 0.24 0.09 -0.22 -0.53 0.99 0.57 -0.09 -0.31 0.18 -0.42
1959 0.06 0.13 -0.02 0.55 0.17 0.05 0.06 0.15 0.06 -0.41 -0.46 -0.11 -0.09 0.53 -0.32 0.13 0.30

1960 0.01 0.11 -0.10 0.39 0.04 0.11 0.07 -0.03 0.06 -0.07 -0.89 0.30 0.74 0.30 -0.88 -0.18 -0.14
1961 0.08 0.07 0.08 -0.12 0.37 0.16 -0.12 0.01 0.20 -0.06 0.27 0.31 0.33 0.86 0.84 -0.11 0.00
1962 0.02 0.10 -0.07 0.45 0.31 0.05 -0.07 -0.08 0.03 0.20 -0.69 0.08 -0.37 -0.22 1.46 0.13 -0.22
1963 0.02 0.10 -0.08 -0.11 0.32 0.08 0.06 -0.03 -0.05 -0.35 -0.01 0.64 -0.29 -0.60 0.68 0.33 -0.35
1964 -0.27 -0.24 -0.31 -0.47 -0.32 -0.20 -0.14 -0.17 -0.27 -0.47 -0.73 -0.35 -0.44 -0.13 -0.33 -0.22 0.00
1965 -0.18 -0.20 -0.15 -0.31 -0.31 -0.15 -0.15 -0.12 -0.12 -0.21 -0.31 -0.52 -0.55 -0.65 0.28 -0.06 43.22
1966 -0.09 -0.08 -0.10 -0.90 -0.23 0.10 0.11 0.00 -0.27 -0.29 0.06 -0.51 0.60 0.21 -0.44 -0.26 43.31
1967 -0.02 0.01 -0.06 0.39 0.28 -0.15 -0.15 -0.14 0.00 -0.08 0.11 0.01 -0.47 0.53 0.98 -0.04 43.37
1968 -0.13 -0.12 -0.14 -0.44 -0.01 -0.13 -0.09 -0.26 -0.16 0.22 -0.13 0.07 0.16 -0.11 -0.68 43.40 -0.39
1969 0.02 -0.03 0.09 0.00 -0.53 -0.10 0.32 0.19 0.01 -0.22 0.20 -0.09 0.47 -0.74 -1.89 -0.09 -0.22

1970 0.03 -0.02 0.09 -0.25 -0.10 0.01 0.07 0.02 0.09 0.03 0.36 -0.18 -0.25 -0.17 -0.02 -0.20 -0.14
1971 -0.12 -0.15 -0.08 -0.05 -0.04 -0.10 -0.29 -0.24 -0.07 0.15 0.22 -0.22 -0.10 0.21 0.49 -0.28 0.04
1972 -0.08 -0.26 0.14 -0.34 -0.65 -0.28 0.03 0.14 0.07 -0.12 0.63 -1.22 -2.08 0.33 -0.65 -0.16 -0.28
1973 0.17 0.17 0.17 0.13 0.30 0.06 0.19 0.25 0.15 -0.01 0.13 0.43 0.51 0.19 0.66 -0.10 0.43
1974 -0.09 -0.16 0.00 -0.20 -0.19 -0.10 -0.18 -0.28 -0.02 0.31 0.70 -0.17 -0.91 0.72 -0.13 0.15 0.14
1975 43.04 -0.03 -0.05 0.09 0.31 -0.05 -0.28 -0.22 0.02 0.09 0.28 -0.34 -0.14 1.10 1.16 -0.44 0.39
1976 -0.24 -0.27 -0.20 -0.11 -0.38 -0.27 -0.24 -0.22 -0.21 -0.16 -0.21 0.49 -0.74 -0.54 -0.76 43.28 -0.39
1977 0.16 0.15 0.17 0.14 0.27 0.12 0.11 0.16 0.13 0.19 0.29 0.78 0.64 0.14 0.41 0.36 0.02
1978 0.09 0.07 0.12 -0.12 0.14 0.04 0.11 0.06 0.12 0.31 0.08 -0.18 -0.93 -0.63 0.32 -0.16 43.36
1979 0.12 0.12 0.13 -0.64 0.14 0.22 0.26 0.21 0.15 0.24 -0.40 -0.25 0.18 -0.14 0.22 -0.46 -0.06

1980 0.27 0.21 0.33 0.21 0.08 0.12 0.36 0.27 0.32 0.33 0.59 0.59 -0.12 -0.67 -0.04 0.52 0.03
1981 0.42 0.51 0.31 1.26 0.93 0.25 0.21 0.02 0.24 0.54 1.63 1.79 1.18 0.23 1.63 0.76 0.13
1982 0.02 0.02 0.03 -0.26 0.08 -0.08 0.15 0.03 0.14 0.22 -0.55 -0.82 43.88 0.40 0.84 43.13 0.11
1983 0.30 0.37 0.21 0.13 0.74 0.10 0.40 0.38 0.09 0.15 -0.09 0.52 -0.15 0.84 1.96 0.10 0.18
1984 0.09 0.08 0.10 0.26 0.04 -0.05 0.14 0.03 -0.05 0.18 0.85 0.44 -0.18 0.19 -0.30 0.04 0.11
1985 0.05 -0.03 0.14 0.21 -0.26 -0.04 0.03 0.02 0.20 0.40 0.71 -0.37 -0.14 -0.96 -0.29 -0.29 0.04

as in 1940.In the southernhemisphere,1980is warmerthan degree; the more recent of these studies also found a
any previousyear in the record. We have included pro- warming during the past decade.The only exceptionto the
visionalresultsfor 1985 in Figure6, basedon a limited generalagreementis Yamamotoand Hoshiai[1980],who
numberof stations. obtaineda hemispherictemperaturechangewith a shape
Other investigatorsstudyingtemperaturechangewith the similar to that found in the other studiesbut with the
meteorological
stationdata have limited their analysesto magnitude
of temperature
variationsonlyabouthalf aslarge
the northernhemisphere(an exception,the analysisrecently as in the other studies.Yamamoto and Hoshiai used the
published
by Joneset al. [1986b,c], is discussed
in the final Gand#•[1963]methodof analysis,
which,in effect,assumes
section6). Our resultsfor that regionare in general that areaswithoutstationcoveragehad no temperature
agreementwith most other studies,for example,I45'llett change.The other studies,in effect, assumethat the areas
[1950],Mitchell [1961], Budyko [1969], Vinnikovet al. withoutstationcoverage had the samechangeas the area
[1980],andJoneset al. [1982].All of thesestudiesfounda with stations.It wouldbe difficultto provea priori which
hemisphericwarmingof the orderof 0.5øCbetween1880and of these assumptions is better, especiallysincethere is a
1930-1940,and a subsequent coolingof a few tenthsof a high correlationof data-richareaswith the continentsand
13,354 HANSEN
ANDLEBEDEFF.'
GLOBAL
TEMPERATURE
TRENDS

1.8

1.5

1.2
1.2j T ............ ••,.•5Year
J__•_j__.j
.... Running
Mean
,•___•___•__•
....
•__•___•
.... 'i
•_+_•_•P 0.9
0.9
0.6 •.... L-f.•-l:'•i,-'.• • -4---½--•
....•....P--•
....L-C-;-;
....L-i....•....L--;-;--L--;•--;: 0.6

0.3•

-0.3 .....
64-90øN ---T--T--T--T--T
....[--•---•--T---:-•--• -0.3

0.6
-0.6 ' '•.......
- , ,,,,',, •---• t--t--t---•---l--•
....
•-t--t---t--•---•---•
....
,;---•---•f•',
:',j-•f---•
....
•- 0.9

-•.• ,, u ,,,,, •l---i-


....
- -r-f-•--• ....F--T
....'F-?--f---•
....h!--: ....•....'•-ff--f-f--ht ....

-1.5 ..... " 44_64ON.


......
J....
i....
[....
,--•---•---•--
•.. • ' :•
.. liF4-+---•
• ••,i....
4....
•--;--•---•-•
....
•....
•....
•--•---•-½---•-•
!• i 44- 64øS'
....
•....
•,....
O.6
.-. 0.3 ......
,, ,,_•_

! : : ' ' n 94ON ' '


ß
-•---:---,-,
....
:....
:---t--*.---•---,.--•
....
•?---,---:---:--•-•
....
--.
0.3
,...,.
-•,,"'i•--i-•"•-d
--
-0.5 -- .... -..........
, • • , ij ,, • ', ', ! i , • ! J ! !
-0.6
1880 1900 1920 1940 1960 1980
Date

Fig.7. Surfaceair temperaturechangefor the eightlatitudezonesof Figure2. Graphicaldetailsasin Figure6.

data-free areas with the oceans. However, based on the the high northern latitude zones. For most zones it is more
error estimateswhich we obtain for our own results in accurateto say that there was little trend for the period
section5, we concludethat the Gandin methodyieldsa less 1940-1970.There is significantwarmingin mostzonesduring
realistictrend. the past15 years.
Temperaturechangesfor the eight latitude zones are The zonal temperaturechangeis shownmore compactly
shownin Figure 7. One aspectwhich they demonstrate, andwith higherlatitudinalresolutionby the colorpresenta-
especially in the northern hemisphere, is high-latitude tion in Plate 1. Annual mean results are shownin Plate la
enhancementof long term change.The magnitudeof the and the 5 year smoothedtemperaturechangein Plate lb.
1880-1940warmingis progressively lessfrom the northpolar Blues and greens represent negative 6Ts for a given
region to the equatorialregions;this warmingis significant latitude, while yellows and reds are positive, where
in all zones except Zone 7 (44ø-64øS) and Zone 8 6T = 0 is the 1951-1980mean. Plate 1 strikinglyportrays
(64ø-90øS),where observationsare inadequate for that the high-latitudeenhancementof temperaturechangesand
period. The 1940-1965 cooling is only dearly apparentfor the warming maxima of 1940 and 1980. Note that the
HANSENAND LEBEDEFF:GLOBALTEMPERATURETRENDS 13,355

TEMPERATURE CHANGE (øC) ANNUAL MEAN


90

6O 3.0
1.4
0.7
3O
0.3
0.1
0
-0.1
-0.3

-0.7

-I.4
-3.0

.. !

-90
1880 1895 1910 1925 1940 1955 1970 1985

TEMPERATURE CHANGE (øC) 5-YEAR RUNNING MEAN


90 .:.

..

3.0
6or 1.4

,of 0.7
0.3

0.1

o 0

-0.1
-0.3
-30 -0.7

-I.4
-3.0
-60

-90
1880 895 1910 1925 1940 1955 97O 1985
(b)
Plate1. (a) Annualand(b) 5 yearsmoothed zonalmeansurfaceair temperature
change in thepastcentm•y.
At each
latitudethe temperature
zeropointis definedasthe 1951-1980
mean.Bluesandgreens representnegativevaluesand
yellowsandredsarepositivevalues.
A nonlineartemperature
scaleis usedsothattherearecomparable areasfor each
color.Regionswithoutdataarewhite.
13,356 HANSEN
ANDLEBEDEFF:
GLOBALTEMPERATURE
TRENDS

TEMPERATURE
9O
• -,•.

, .

.... .:..•,...
• .. --,,,r'•
,•,•i,
% ....
......

60' ---,' • 5.0


.

• :.
3.0
: :.
,..
..

2.0
3O
.... :. -.....
.•.. -•: ...... . 1.0
. ...

..
0.5
0.0
-0.5

-I.0
-3O
-2:.0
-3.0

-6O -5.0

-90 •
-180 -I 20 -6O 0 60 120 180
(•)

TEMPERATURE TRENDS 1880-1940


9O

6O 5.0
3.0
.... 2..0
3O
1.0
0.5
0.0
-0.5
-I.0
-:;'.0
-3.0

-60 • -5.0

-90 '
-180 -120 -60 0 60 120 180
(b)

Plate2. Temperaturetrendsinferred for the four time intervals(a) 1880-1985,(b) 1880-1940,(c) 1940-1965,and
(d) 1965-1985,from the linearbestfit to all availabletemperatures
for eachinterval.The indicatedtemperaturetrend
is the productof the derivedslopeand the indicatedtime interval.In Plate 2a, resultsare shownonly for those
subboxes whichhad a definedtemperatureby 1925or earlier;in 2b, 2c and 2d, the minimumrecordlengthsfor a subbox
are 25, 20, and 15 years,respectively.
HANSENAND LEBEDEFF:G•.OB^•.TEMPERATURE
TRENDS 13,357

TEMPERATURE TRENDS (øC) 1940-1965


90 !

60 5.0
3.0

2.0
30
1.0
0.5

0 0.0
-0.5

-I.0

-2.0

-3.0

-60 • -5.0

-90 '
-180 -120 -60 0 60 120 180
(c)

TEMPERATURE TRENDS (øC) 1965-1985


90,' •,................. .

. • '• '•i:' 5.0


. • 3.0
--• •o 0

".. 1.0
0.5
0.0
.......-0.5
-I.0
-30
-2.0
" -3.0

': - 5.0

-90
-180 -60 0 60 120 180

(d)

Plate 2. (continued)
13,358 HANSENANDLEBEDEFF:GLOBALTEMPERATURE
TRENDS

warming centered in 1940 was most intense in high northern Finally, with regard to Plate 2, we note that Jones and
latitudes,while the recent warming is more global in extent. Kelly [1983]presentedmaps of hnear temperaturetrendsfor
The geographicaldistribution of temperature trends in the several intervals within the period 1917-1980. Their results
past centuryis illustratedin Plate 2a. The resultsshownin for periods similar to those which we chose, specifically,
Plate 2 (and Plate 3a) are the linear trends multiplied by their results for 1940-1964 and 1965-1980, are in good
the indicated time interval. Stationsemployedin construct- generalagreementwith our results.
ing Plate 2a were restricted to those with records which We presentexamplesof the data at monthlyresolutionin
extend at least as far back as 1925. The estimatedtemper- Plate 3a, which illustratesthat the surface air warming in
ature trends can be unreliable in fringe areas where only the past century has been greatest in the winter months,
one or two stations contribute to the trend, as shown in especiallyat high latitudes. Plate 3b, which presents the
section 5; such regions with potentially anomalousresults standard deviation about the 5-year smoothedtemperature
includecentralAfrica, Mongolia, and manyoceanlocations. change, shows that the natural variability on short time
It is apparent that the warming of the past centuryis by scaleshas a qualitativelysimilar geographicaland seasonal
no meansgloballyuniform. The greatestwarming occursin pattern, the variability increasingwith increasinglatitude
the Arctic, but note that the area of that region is and front summer to winter. Plate 3c shows the confidence
exaggeratedby the map projection. Areas of greater than limits [Ostle, 1963], for the trends in Plate 3a. It is
average warnting occur at lower latitudes in west and apparent that the trends are highly significant in most
central Asia, central and eastern Canada, Greenland, Alaska, cases,especiallyat low and middle latitudes. The lesser
and partsof SouthAmerica. significanceat high latitudes is a result of the greater
In Plates 2b, 2c, and 2d we have broken up the past variabilitythere, the reducedarea for a given incrementof
centuryinto the periods of warming and coolingsuggested latitude, and shorter records in the southern hemisphere
by the global mean temperaturechangein Figure 6: the highlatitudes.
1880-1940warntingperiod,the 1940-1965coolingperiod,and Plates 3a-3c suggestthat there has been a measurable
the 1965-1985warming period. In 1880-1940 there is a changein the seasonalcycleof temperatureduring the past
strongwarnting trend at high northern latitudes,especially century,in addition to a general warming. We explicitly
Alaska, central Canada, Greenland, the northern coastlineof illustrate this in Plate 3d, which shows the difference
Asia, and that portion of the Arctic Ocean where a between the winter and summertemperaturechanges,as a
temperaturetrend could be inferred. At lower latitudes function of latitude and time. On the basis of error
there is warmingof more than 0.5øCin most of the United estimatesobtained in section 5, it is apparent that there
Statesand China and in parts of SouthAmerica and Africa. has been a significantreduction in the amplitude of the
The few local areaswhere it appearsthat coolingmay have seasonalcycle in the past century, but little net change
occurred are mainly in the southern hemisphere,but the between 1940 and 1985.
stationcoveragethereis poor duringthat period. Finally, Figure 8 and Tables 1 and 2 provide examplesof
The period 1940-1965 (Plate 2c) is marked by strong temperaturechange and variability at higher temporal and
cooling at high northern latitudes, especially Alaska, spatialresolution.Figure 8 showsseasonalmean temperature
northern Canada, northwesternGreenland, and the northern changefor the globe, box 9 (most of Europe) and box 15
coastline of ^o:• ^, •...... •,:,...•,•o ,-h,•..,• :....,- ....... •:.... twc•c-•,emicu uHi[cu otcttcbialtUnorthern•v•cx,cu/,the latter
in China and Africa. A substantialarea of warming stretches re•ons being defined in Figure 2. The variability(standard
from western Europe acrosscentral Asia almost to Lake deviation)of the seasonalmean temperature(averagefor
Baykalat latitudesabout40ø-60øN. the four seasons)is 0.2øC for the globe, 1.0øCfor box 9,
The period 1965-1985 (Plate 2d) is marked by rapid and 0.6øCfor box 15. The temperaturein these regionshas
warming over Alaska, northwestCanada, and the northern little trend during the period 1950-1985,thoughthe 5-year
half of Asia. Substantial cooling occurs in the southern runningmean global temperaturein the 1980sis about one
Greenlandregion.In comparingPlates2d and 2b, note that standarddeviation above the mean. Although the area of a
the recent period is only one-third as long as the earlier box is more than 6 million squarekilometers,interseasonal
period; if the temperaturetrendswere graphedas a rate the variations at the box resolution are as much as several
resultsfor the recentperiodwould appearmore pronounced. degrees Celsius. At high latitudes, such as box 9, the
Note in Plate 2 that the geographicalpatterns of the extremevariationsare mostlyin the winter, as can be seen
warnting and the locations of the areas of greater than in Figure 8 since the winter points fall on the lines
average warming are not of the nature that would be separatingsuccessiveyears. Table 1 includes the annual
expectedfor anthropogenicheat island effects.We present temperaturechangefor six individualboxes(6, 7, 9, 10, 15,
additional evidence in section 5 indicating that the global and 16) and Table 2 gives seasonaltemperature changefor
warming trend of the past centuryis not due to anthropo- three regions (boxes 6 and 7, Canada; boxes 9 and 10,
genicheat islandeffects. Europe and western Asia; boxes 15 and 16, United States
Examination of Plate 2 reveals several places where the andnorthernMexico).
1880-1985 temperature trend is not the sum of the trends We have presentedonly a few examplesfrom our data set
for the three subperiods.This is becausethe resultsin Plate for surface air temperature change.Clearly, more informa-
2 (and Plate 3a) are the linear trends for the indicated tion couldbe extractedfrom analysesof the data'stemporal
intervals. One characteristic of the linear trends is that a and spatial characteristics,but such analysesare beyond the
strong change at the end of a long record, such as the scope of this paper. A documentedcomputer tape of the
warming in the early 1980s, is not reflected as fully as it derived temperature changes, which we intend to keep
would be if the temperature change were obtained by updated in the future, is available from the authors or Roy
differencingthe initial and final temperatures. Jenne at NCAR. The tape contains annual, seasonal,and
HANSENAND LEBEDEFF:GLOBALTEMPERATURETRENDS 13,359
(• o o o o

i.,,i._l

z
z

b_l

".

_J
<•
z

ILl

Z
LI

o o o o o o 0 0 0 0 0 0
ro f,,,.o 0'• 0'• •O 1,0 rO •O
i i i i i

i ,

I.

i,i

0 0 0 0 o o
{,D 0'•
i i i
13,360 HANSEN
ANDLEBEDEFF:
GLOBAL
TEMP•,RATURE
TRENDS

' ' • i i i i i i i i i ..... • , i i • .... !,


1.0 ---'-•-•---':--•--"--•---•'
I I I I I I I I...."----: I I....:I...."....:I....L__.J.,'
....••___.J.,'___Z___',___',___',___:
-'- I- ,'I- I....I:....I:....I:....I:....;:....t:....t:.....'-----,*
: :_j_ : :
.... -r----r--i---
Global
0.5)--4--J--$-4-•---i--J----i
....•....';....i....i....i---•....';....;....;---=,---•--$---'+---i
....i....i....•....i....i....L---'•
....i--.'.-•,-L--¾--•--L-I
o -[-,,z,"•..f••,•-r"'"
'•---i.:-''
",.' '" '•. '•'", '•''•
"", :•••,•'-:;..•
-0.5
[---f--,+---t--+--½-r-•
....!....i....t....i....'?
....',,,
....'?
....F+--•---f---+---+--•
....•....!....i....i....'!....)....',,-
....'?
....'Wr---F+--+-•-'
--T---T---T-T-•....?--•....:I....:.
....tI.........tI....:---•....,---•....r---=--,-
.......•,........•,---:
....•....,:
....t.....tI....tI....h-'?---h--h-+,---i-' 3.0
I I I I I I I I I I .....
, , Box 9 (Euro e) . , 1' , , , ,
___.__.___.___.___,___,
.... .l.... .'___.'
.... ;........ L................. ' .L . .......... 2.5
I---i
....
+---+
...................
•---4
....4---4
........................................
.......4.........................................
•----4
•.........
....4....i'.---4
!....
i....i....F-•7--T-T---T-
....•....i....•....i--i....i--•t----L--$
.... 2.0
•-?'
......
•--+--!
....
i.......
i....
! i---•---i
....
i---i
....
!....
'•....
,•-•---'+---,•---!
....
•....
r-t---•--4
....
i....
':....
1-4....
k-4--,•--•---+'
.... 1.5
,'' : ,'• •. , •' :' ,,
,, f.....: ,,, ,, . :: ,: •:: •,, :,, :.,,,
,, ,,, : : .•, •, t,
•, ,:, ' ...... : : : ..•....:
,L-•---•---.'---z--4---•
....,'i---:-" ,-,:---L-,:
....:__?.
....•....L _L,..:•___•__:___•
....•....•L__:i--z:_',__]
....•....i....i....[....i....L•',_! __L_ 1.0

+-e•
:4-' '"--'+,.4-'-
•_ '---4 _•<•i....½-- ::-•
....•,--'-- ,....';--•',•
....',• :,,',:,,;,•.
....•-,-•-,• ; •',•
..... +- :.......
• •;-,•
i:,:, --H•---e•-;
,:, ,:, :......
I,, ;•....)--•
• :..... .e>--t+-Ir•,
: I: ,.,,, ....
'u:,•. •!'•,,',. •i•,, O.5 (•
',' l',J
) ',l ' •:I' "='",
:' "' I
,I"
•,•' :•:,,•',•!
• ', 'li• ' i I'i•:,,
" ,, •,,':,,,•, ":•,•,
'•"',,I,
,•'i?:'-'",...-,
, •,,,
', ,',,,'•,',
',..•!'' '•, ,!,
•',, II I, I f•?,•.
I,', ', ,,,: I',
•I ' ,,I,:,I o

:t • ' ," •,', ,: '" ,, ,, ' , ' , ' , ,,•...... .. ,' ,,


o t-
....
L__•!_•i___ji•L..."___Jj
....
j:,i_j_.
iz,'
......[;.iiiL:j_.,'_L_iL_
1__L•j_i•___L_ii___i.,,__•
.............' .....
dL__•'
....... -o.5 <1
: :'*: ::: :;;::/',:: ::: : ,I,• :z,,:::t: :',ii: : ) ::, :i; i i i !:l i i i l',:',lri',:} | i
-I.0

--•--½---•----'.•---q:---'ii--•
....•....•....•....:....';---47,--4....',--4 --i----½-4
-4--4....4....4....4....4....,,'
....k---,',•
....',•
....•....•....,L__4___•
....

....-•---+,'----•--+--•
....F-4....4---j--4....;....;--:,'•--4
....•-4---•---,•---•
....J--•---4
....4....J---4
....4....•....;----•
....•....•---J-
....•----•--4
.... -2.0

-2.5
....
+--+--+---,"---i
....
: • • • ,, •....
i-!
• •....
!•....
•---i
• •....
!---•---i
• ' ....
',--',•
•....
',-+
•....
, =--•
....
••.........
•!....
:i.........
,....
•....
::....
=--i....i[---7,'"
.... .......
-3.0
....•---777,"---i
....•,....]....i....i....i....i....i........
!....i....•....!-7--r--T--i
....•....i....!....i....•....!....r....!....•....!....•---:,,'---T---,
....
-3.5

2.0 ....+,'--+---+--+---f---•
....•---4....4--4---+--4....•- •- -L-• ---• • .....J----L ' 4 --•--•---i "-4....L....•....•__-+
....•___+
__+---
BOX15 (West-Central U.S.,NorthernMexico)
1.5 ---&---.-,--
.---.--i ...................½
: • ', ......
....•- 4- -4.... • ...............................................................
', ', ', '•= ! • ,• T •'-
:--! .........
•, : ] • • i ' ; ," F • -r' T -' "-
, .f! ..........
I.O ....T--T---•,--•--E....]....F--]....V -T- -T....; --,,.H,,•
....t....'r- 7- -* -+- -+ --v--•t-d --¾-*,'
....l-....... ,•....
, •,
., , •,'' ,.... : ,,, ? •,',, i '
........ ' I ,, ', •' } •,,,,•,•
i T ;i',,'!',•',i : I• ! i : : : -i:i:',! ! i i ', ' i i ii: : i;',i '$ •! ! $;',q',:',i
0.5 .--•,---;---:•-;-•-i•---4---;4'.---b-C:---..'--•,•
' • •'? , , • " • ....
....•-4';.•;i'.
- •--:•
,
....,,'• 4 ..... •--J---•::--•---J-i--4--
• ,
$,".• , , ,
• ,• i:.... : l', :l',.',l I
',ii :i!•
..... '•I :i ;tiiiii
":','":"tl ',:/: , ::, I
:••11•
',•ii i ' l , , I , ,# : ,, , r,,
O ....i ",i :!11:ii i ',ii',i'":":
......
. ",.' _:.,L?•_•_'

-0.5 '":'"'!i'½'"'"' .....


......""'....
ß
',,:,:
-t---t---!---i
....
i....
i---li
....
!-'I
?-;T
-I .O __•___L__L__L__L___J
....
':'i.... .... ....
....i....J....•....L__
i....
?--?l--'
-I
i....i....!....•....•....,:....i__Z_:___L___J___J___•
....J....•....J...... i___!ii___i
....!....L_2.........

-I.5
....
....
'"
....
'"
....
':....
....
'---:
....
:::
....
....
.... ....
........
....
'7....
i....
i---h:
....
........
!---+---+--- ', I • ', •

-2.0
--L-4-•---•---i--i
....
L-i....
i....
i....
i---i
....
t....
i....
i....
i---i---•---J
....
$--!-4
....
i....
i....
i....
i....
i....
i....
f....
[........
-L--L--+-$-•
1950 1955 1960 1965 1970 1975 1980 1985
Dmte
Fig. 8. Temperature changeat seasonalresolution;the solid curveis the 21 point (approximately5-year) runningmean.
Box 9 containsmost of Europe and Box 15 containswest-centralUnited States and northern Mexico (see Figure 2).
Each seasonis 3 calendar months, for examplefrom December 1 through February 28 or 29, computedas the simple
mean of the ATs for the 3 months.

monthlytemperaturedeviationsfrom the long-termmeansat temporalcoverageprovidedby the finite number of meteor-


global, hemisphericzonal, box, and subboxspatial resolu- ologicalstations.Indeed, the coveragein the southernhem-
tions. The tape includes a small program to interpolate isphereis sufficientlysparse(Figure 1) that the data from
resultsto a 1ø by 1ø (latitude by longitude)grid; for most that hemisphere have been dismissedin most previous
regionssucha resolutionis higher than the data meaning- studiesof the station data. Thus prior to our earlier paper
fully provides,but the high resolutionallows the user to [Hansen et al., 1981], there were no published southern
constructreadily the resultsat arbitrarycoarserresolutions. hemisphereor global surfaceair temperaturechanges.More
The user shouldbe aware of basic limitationsand probable recentlyFoRandet al. [1984] have analyzedmarine tempera-
errors in the data set, which are partially clarified in the ture changesin both henrispheres and Joneset al. [1986b,c]
followingsection. have studied southern hemisphereand global temperature
changesbasedon land-basedstationsand marine data.
5. ERROR ESTIMATES
We obtain a quantitative estimate of the error due to
The greatestsourceof error or uncertaintyin the derived imperfectspatial and temporal coveragewith the help of a
temperature changesis due to the incompletespatial and 100-yearrun of a general circulationmodel (GCM). The
HANSENAND LEBEDEFF:GLOS^L TEMPERATURETRENDS 13,361

TABLE2. Seasonal
SurfaceAir Temperature
Changefor Specified
Regions.

Box 6 + Box 7 • Box 9 + Box 10 Box 15 + Box 16

Year D-J-F M-A-M J-J-A S-O-N D-J-F M-A-M J-J-A S-O-N D-J-F M-A-M J-J-A S-O-N

1880 -1.35 -1.73 -0.46 -1.67 -2.09 -0.50 0.10 -0.76 1.40 0.27 -0.36 -1.81
1881 -1.62 0.39 -0.75 -1.82 -1.09 -0.63 -0.45 -1.58 -1.32 -0.24 0.17 0.03
1882 -0.15 - 1.90 -0.46 -0.80 0.86 0.08 -0.23 -1.24 1.38 -0.26 -0.41 -0.34
1883 -3.07 -1.38 -0.63 -1.75 -2.15 -0.99 -0.36 -0.14 -0.80 -0.81 -0.19 -0.53
1884 -3.24 -0.99 -0.98 -1.28 1.21 -2.02 -1.22 -0.76 -0.23 -0.58 -0.61 0.03
1885 -3.33 -0.53 -0.88 -0.46 -0.12 -0.63 -0.61 -1.10 -1.27 -0.48 -0.40 -0.60
1886 -0.93 -0.33 -0.17 -1.07 -0.64 -1.44 -0.58 -0.53 -0.79 -0.43 -0.34 -0.80
1887 -3.85 -0.62 -0.82 -1.19 0.95 -0.41 -0.20 0.34 -0.55 0.15 -0.35 -0.73
1888 -1.95 -1.36 -0.39 -0.58 -0.88 ' -0.22 -0.11 -0.32 -1.06 -0.88 -0.52 -0.74
1889 0.81 1.72 -0.18 -0.23 -1.76 -0.50 -0.03 -1.03 -0.12 0.32 -0.75 -1.01

1890 -2.21 -0.95 -0.25 0.19 -0.56 -0.24 0.28 -1.20 2.16 -0.46 -0.56 -0.38
1891 -O.43 -0.23 -0.61 -0.19 -2.30 -0.47 -0.28 -1.47 0.33 -1.15 -1.03 -0.60
1892 0.74 -0.56 -0.36 -0.06 -0.07 -1.30 0.32 -0.44 0.12 -1.28 -0.79 -0.65
1893 -1.90 -1.73 0.06 -0.86 -3.88 0.34 0.05 0.76 -1.28 -0.85 -0.59 -0.94
1894 -1.94 -0.25 0.38 -0.65 1.10 -0.31 -0.36 -1.10 -0.25 0.22 -0.76 -0.24
1895 -0.17 0.55 -0.55 -0.89 -2.33 -0.53 -0.05 0.53 -1.26 -0.30 -0.75 -0.51
1896 0.26 -0.98 -0.09 -2.36 -1.14 -1.21 0.18 0.15 0.18 -0.03 -0.16 -0.74
1897 -0.24 -0.69 -0.40 -0.71 -1.37 -0.18 0.37 -0.39 0.09 -0.21 -0.58 0.14
1898 -0.13 0.11 0.12 -0.87 -0.38 -2.56 -0.06 -0.15 0.24 -0.23 -0.31 -0.61
1899 -1.57 -1.52 -0.50 0.42 2.25 -0.40 -0.42 1.50 -1.51 -0.59 -0.41 0.13

1900 0.59 0.76 0.01 -0.43 -2.57 -0.69 0.21 0.22 -0.02 -0.26 -0.14 0.62
1901 0.61 0.64 -0.10 -0.03 0.15 0.76 0.40 -0.75 -0.15 -0.74 0.08 -0.46
1902 1.64 0.61 -0.88 -0.16 0.62 -1.75 -0.24 -2.25 -0.90 0.05 -0.89 0.02
1903 -0.52 -0.78 -1.06 -0.40 0.63 -0.51 -0.48 -0.55 -0.54 -0.08 -1.27 -0.76
1904 -1.74 -1.02 -0.54 0.08 0.50 -0.84 -0.35 -0.12 -1.16 -0.45 -1.19 -0.30
1905 -1.14 0.77 -0.33 -0.17 0.25 -1.36 -0.16 0.72 -1.79 0.03 -0.67 -0.25
1906 1.15 -0.20 0.28 0.07 0.15 1.32 0.38 -0.48 0.29 -0.88 -0.68 -0.32
1907 -2.34 -2.16 -0.93 -0.09 -0.81 -0.48 -0.42 -0.94 0.56 -0.79 -1.06 -0.59
1908 1.34 -0.55 -0.17 0.59 -1.01 -1.57 -0.47 -1.33 0.28 0.38 -0.81 -0.34
1909 -1.48 -0.58 0.09 0.10 -0.86 -1.33 -0.23 0.95 0.79 -0.88 -0.58 -0.18

1910 -0.67 1.64 -0.11 -0.26 1.49 0.09 -0.22 -1.19 -1.51 0.74 -0.76 -0.20
1911 -1.47 0.22 0.01 -1.18 -0.43 -0.69 0.25 0.22 0.19 -0.05 -0.44 -0.53
1912 -0.31 -0.59 -0.93 -0.11 -0.23 -0.59 -0.66 -1.73 -1.20 -0.83 -1.14 -0.41
1913 -0.77 -0.56 -0.40 -0.04 0.16 0.05 -0.58 0.45 0.07 -0.47 -0.56 -0.29
1914 -0.15 -0.06 -0.56 0.26 2.95 0.14 -0.36 -1.18 0.04 -0.52 -0.49 0.03
1915 0.03 1.30 -0.76 0.15 1.37 -0.08 0.22 -1.01 -0.58 -1.06 -1.19 0.26
1916 -1.03 -0.14 0.08 -0.27 1.07 -0.59 -0.62 -0.27 0.07 -0.40 -0.76 -0.83
1917 -1.81 -1.06 -0.10 -0.12 -1.24 -0.91 0.00 1.20 -0.77 -1.56 -0.70 -0.95
1918 -2.42 -0.37 -0.89 -0.03 0.43 -0.88 -0.30 0.43 -1.65 0.24 -0.37 -0.54
1919 1.53 -0.11 0.02 -1.59 -1.53 -1.15 -0.54 -0.53 0.27 -0.32 -0.47 -0.18

1920 -1.32 -0.98 -0.01 0.03 0.09 2.16 0.49 -1.68 -0.73 -1.02 -0.98 -0.32
1921 0.79 0.12 0.23 -0.62 0.06 1.45 0.72 -0.51 0.93 0.66 -0.28 0.54
1922 -1.36 0.43 -0.18 0.28 -1.04 0.84 0.32 -0.48 0.04 -0.07 -0.36 0.13
1923 -1.56 -1.33 -0.34 1.22 0.65 -0.62 0.29 1.85 0.36 -0.86 -0.76 -0.57
1924 1.05 0.13 -0.54 0.15 -0.47 -0.12 0.05 0.96 0.18 -1.24 -0.47 -0.30
1925 -1.85 0.53 -0.11 -0.96 2.86 0.07 0.25 0.30 0.07 0.62 -0.17 -0.75
1926 2.11 0.00 -0.41 -0.89 0.70 -0.15 -0.64 0.79 0.26 -0.83 -0.48 -0.16
1927 0.11 -0.17 -0.32 -0.30 -0.50 -0.32 0.92 0.30 0.88 0.15 -1.06 0.73
1928 -0.06 0.11 -0.22 0.06 -0.70 -1.63 -0.38 0.52 -0.02 -0.33 -0.57 -0.24
1929 -0.11 -0.12 -0.12 -0.13 -3.29 -0.92 0.49 1.24 -0.74 0.30 -0.51 -0.66

1930 -0.96 0.47 0.66 0.17 -0.40 0.45 0.47 0.54 0.59 0.00 -0.06 -0.17
1931 3.21 0.64 0.77 0.99 -2.07 -0.54 0.78 0.00 0.61 -0.55 0.29 1.27
1932 0.44 0.16 0.51 -0.27 0.68 -0.12 0.43 0.97 1.84 -0.52 0.02 -0.49
1933 -0.37 -0.10 0.31 -0.97 -1.20 -0.64 0.08 -0.04 0.43 -0.02 0.12 0.50
1934 -1.41 0.32 -0.19 0.38 0.04 -0.19 -0.21 0.78 0.81 0.68 0.49 0.63
1935 -O.75 -0.73 0.24 -0.97 0.97 -0.26 0.25 0.19 0.39 -0.14 -0.07 -0.26
1936 -1.63 0.39 0.70 -0.64 0.13 -0.51 0.71 0.70 -1.59 0.59 0.59 -0.16
1937 -1.18 0.96 0.84 0.29 1.26 -0.36 0.63 1.03 0.33 -0.34 0.38 -0.21
1938 -0.16 0.45 0.58 0.89 0.17 0.97 0.70 1.58 0.78 0.61 0.05 0.28
1939 0.00 -0.54 0.10 0.02 -0.11 0.04 0.75 -0.43 0.61 0.37 0.11 0.40

1940 2.32 0.72 0.00 0.17 -1.99 -0.17 0.73 0.40 -0.46 -0.35 -0.14 -0.21
1941 1.61 1.18 0.33 -0.34 -1.17 -1.22 -0.23 -1.28 0.70 -0.14 -0.19 0.61
1942 1.31 1.43 0.00 0.09 -2.81 -1.73 0.10 0.49 0.21 0.28 -0.01 0.28
1943 -O.72 -1.25 -0.37 1.15 0.11 1.21 0.35 0.28 0.51 -0.14 0.40 -0.43
13,362 HANSENAND LEBEDEFF:GLOBALTEMPERATURE
TRENDS

TABLE 2. (continued)

Box 6 + Box 7 Box 9 + Box 10 Box 15 + Box 16

Year D-J-F M-A-M J-J-A S-O-N D-J-F M-A-M J-J-A S-O-N D-J-F M-A-M J-J-A S-O-N

1944 1.68 0.20 0.12 0.85 2.46 1.02 -0.04 0.26 0.34 -0.25 0.02 0.02
1945 0.99 0.18 -0.44 -0.60 -1.59 -0.21 0.40 -0.07 -0.07 0.59 -0.44 0.28
1946 -0.58 1.18 -0.27 -0.26 0.77 0.14 0.08 -0.32 -0.09 1.11 -0.31 0.31
1947 0.60 -0.42 -0.01 0.91 -2.58 0.80 0.05 0.99 0.64 -0.38 -0.07 0.61
1948 0.12 -1.34 0.45 1.32 1.30 0.68 0.24 0.65 -0.53 0.25 0.02 0.21
1949 -1.49 0.51 0.20 0.51 1.60 -0.04 -0.08 0.07 0.49 0.37 0.27 0.43

1950 -2.48 -0.91 -0.52 -0.51 -1.30 1.08 -0.04 -0.12 1.40 -0.56 -0.62 0.23
1951 0.43 0.40 -0.14 -0.65 -0.95 0.78 0.52 0.20 0.37 -0.24 -0.07 -0.24
1952 -0.72 1.16 0.30 0.54 2.33 -1.39 0.50 -1.22 1.10 -0.18 0.65 -0.23
1953 2.20 0.95 0.14 1.36 -0.32 0.93 0.82 -0.91 1.35 0.34 0.43 0.65
1954 0.42 -0.85 0.12 0.82 -2.45 -0.89 0.48 0.81 1.43 0.12 0.35 0.66
1955 1.19 -0.24 0.69 -0.40 0.93 -1.11 0.37 0.63 -0.12 0.61 0.06 -0.09
1956 -0.03 -0.90 -0.33 -0.17 -2.87 -0.89 -0.28 -0.90 0.03 -0.21 -0.01 0.10
1957 -1.72 0.45 -0.19 0.46 1.18 -0.45 0.41 0.41 1.29 -0.05 0.13 -0.44
1958 2.27 1.56 -0.04 0.18 1.80 -1.59 -0.43 0.14 0.19 -0.22 0.00 0.35
1959 -0.84 -0.11 0.09 -0.82 0.90 0.64 0.47 -1.08 -0.23 0.29 0.30 -0.09

1960 2.26 -0.39 0.38 0.49 -0.36 -1.29 -0.18 -0.75 0.16 -0.59 0.11 0.53
1961 0.64 -0.12 0.78 0.03 2.38 1.40 0.07 0.17 -0.26 -0.20 -0.15 -0.03
1962 -1.05 0.05 -0.40 0.66 2.11 0.97 -0.42 0.24 -0.03 0.03 -0.24 0.14
1963 0.23 -0.15 0.02 1.09 -0.77 -0.44 0.10 1.35 -0.94 0.63 -0.04 1.00
1964 1.01 -1.18 -0.45 -0.44 -0.74 -1.27 0.07 0.11 -0.99 0.10 -0.24 -0.18
1965 -1.41 -0.36 -0.54 -0.94 0.53 -0.29 -0.31 -0.62 0.04 -0.46 -0.57 0.11
1966 0.15 0.18 0.12 -0.26 0.20 0.34 0.27 0.26 -0.40 -0.15 -0.03 -0.12
1967 -0.17 -1.86 0.43 0.50 -1.49 1.83 0.12 1.38 0.10 -0.09 -0.49 -0.45
1968 0.45 0.61 -0.86 0.84 -0.04 1.09 -0.47 -1.06 -0.60 -0.18 -0.32 -0.08
1969 0.11 -0.21 -0.22 -0.10 -4.36 -1.47 -0.38 0.55 -0.36 -0.39 0.12 -0.31

1970 0.88 -0.37 0.46 -0.11 -0.53 0.03 -0.34 0.15 -0.49 -0.30 0.12 -0.19
1971 -1.02 0.24 -0.01 0.14 0.27 -0.68 -0.18 0.96 0.00 -0.69 -0.27 0.28
1972 -2.85 -0.94 -0.66 -1.65 -0.42 0.02 0.32 -0.35 0.51 0.31 -0.57 -0.73
1973 -1.20 1.03 0.50 -0.03 1.31 0.86 0.09 -0.56 -0.20 0.05 0.07 0.55
1974 -0.75 -1.34 0.03 -0.52 0.49 0.59 -0.06 0.26 0.66 0.70 -0.23 -0.55
1975 0.12 -0.62 0.41 0.05 2.45 1.72 0.34 0.10 0.59 -0.55 -0.15 0.02
1976 -0.17 0.17 0.00 -0.33 0.35 -0.32 -0.12 -1.73 0.75 0.20 -0.41 -1.43
1977 0.70 1.84 -0.15 0.34 -1.16 1.48 0.05 0.47 -1.36 0.86 0.32 0.43
1978 -0.25 -0.47 -0.47 -1.34 0.44 0.16 -1.01 0.50 -1.29 0.08 0.27 0.16
1979 -1.79 0.35 0.11 0.27 -1.41 -0.03 -0.26 0.27 -1.73 0.06 -0.05 0.22

1980 1.30 1.14 -0.10 -0.03 -0.24 -0.74 -0.57 0.22 0.44 0.12 0.92 -0.05
1981 1.93 1.56 0.26 0.87 2.24 0.16 0.82 0.73 0.38 0.46 0.76 0.10
1982 -0.96 -0.94 -0.51 -0.46 0.45 0.51 -0.08 0.99 -0.42 0.16 -0.18 0.01
1983 0.87 -0.03 0.53 0.54 3.82 1.12 0.50 0.31 1.09 -0.58 0.57 0.71
1984 -0.24 0.59 0.55 -0.80 1.80 0.21 -0.14 -0.22 -0.51 -0.30 0.10 -0.24
1985 -0.57 0.36 -0.13 -1.45 -2.69 -0.44 -0.28 -0.22 -0.48 0.99 -0.06 0.09

GCM is modelII, described by Hansenet al. [1983].In the for the rawinsondestations.Our problem is somewhat
100-yearrun the oceantemperaturewas computed,but different,sincewe are seekingthe error in the long-term
hor/zontalocean heat transportswere fixed (varying surfaceair temperaturechangedue to spatialgapsin
geographically and seasonally, but identicalfrom year to meteorological stations,and we are mainlyinterestedin the
year)as described by Hansenet al. [1984].The oceanmixed error in large-areaaverageresults,such as the global
layerdepthalsovariedgeographically andseasonally, andno surfaceair temperaturechange.
heat exchange occurredbetweenthe mixedlayer and the Our procedure is to usethe surfaceair temperatureat all
deeper ocean.This 100-yearrun will be describedin more grid points in the 100-yearrun to define a "perfect"
detail elsewhere, since it serves as the control run for temperaturedata set, i.e., one with completegeographical
severaltransient CO2/tracegasclimateexperiments. andtemporal coverage. We thensample thatdatasetat only
We note that Oo•,t[1978]previously usedoutputfrom a the pointswhere stationsexistedat a given time, and
GCM to estimateerrorsin meteorological variablesdue to examine howwella givenstationdistribution canreproduce
spatialgapsin observing stations,specifically
the errorsin the full 100-yeartemperaturevariation. In order for this
atmosphericwinds and temperaturesfor the rawinsonde test to provide an accuratemeasureof the error, the
station distribution. He found rms errors of about 0.5ø to model'sspatialand temporalvariabilitymustbe similarto
iøC for the temperature distributionin the free atmosphere,that in the real world. Therefore we first examine the
HANSENAND LEBEDEFF:GLOBALTEMPERATURETRENDS 13,363

0.4

Global Surface Air Temperature-I00 Year GCM Control Run


0.:5

0.2
ii i II ii •..
ß' * A •nuc•l Mean
O.I
m $ ml I I i m

w i

-0.1
. ::• , . ,.'• ß • , .

-0.2 0.2

O.I

-0.1

-0.2

-0.:5

-0.4
0 I0 20 :50 40 50 60 70 80 90 I00

Time (year)

Fig. 9. Global annual mean (left scale) and seasonalmean (right scale)surfaceair temperaturechangein a 100-year
GCM run with fixed climate forcing. The "seasonal"AT is computedas the simplemean of the three ATs for the
relevantmonths(e.g., December-January-February).The AT for a givenmonth is the differencebetweenthat month's
temperatureand the 100-yearmean temperaturefor that month.

model'svariability and compareit to available estimatesfor 30-year running mean, with a result very similar to that
the real world, the latter, of course,only being well known illustrated in Plate 4b.
in areasof densestationcoverage. Plate 4 suggeststhat the model's variability of annual
The global annual-mean and seasonal-meansurface air mean temperature is realistic to first order. The standard
temperaturechangesin the 100-yearGCM run are shownin deviation of annual mean temperature is typically
Figure 9. Evidently the interannual variability and the 0.25ø-0.5øCat low latitudes, increasingto about iøC in
longer term fluctuationsare of the same order of magnitude polar latitudes.At mid-latitudesthe variabilityis greater in
as in the estimate from observations(Figure 6), but the midcontinentsthan in coastalregions.
GCM has somewhat less global variability. A smaller The interannualvariability of seasonalmean temperature
variability in the model is not surprisingsince the ocean is illustrated in Plate 5 for the model and for observations.
heat transportsare specifiedto be identical each year, thus The magnitude and global distribution of the model's
excludingphenomenasuchas E1 Nifio, and sinceatmospheric variability are fairly realistic, for example as regards the
trace gas composition,aerosols,solar irradiance and other change in variability from low latitudes to high latitudes,
external forcing are fixed, thus excludingany long-term from ocean to continent, and from summer to winter. There
trend in the GCM temperature. It is relevant to note that are discrepanciesin the locations of some of the centersof
the model exhibitsunforcedtemperaturechangeson decadal high variabilityin northernhemispherewinter. The largest
scales as well as on shorter time scales. Examination of the
discrepancyin the model is the overestimateof variability
mechanismsof the model'sinternal variability is beyondthe in continentalareasduring summer,with the model yielding
scopeof this paper; a discussionof this subjectis given by a variability about twice as large as in the observations.
Lorenz[1985]. However, the nature of this discrepancyis such as to make
A crucial model variability, for the purposeof examining it more difficult for a given station distributionto repro-
the error due to incompletespatial coverageof stations,is duce the model's variability, i.e., this model characteristic
the geographicaldistributionof interannual variability. The should tend to cause us to overestimate the error due to
standarddeviation of the annual mean surface air tempera- incompletestationcoverage.
ture is shown in Plate 4 for both observations and the On the other hand, the variability of surfaceair tempera-
GCM. Plate 4a shows the observed standard deviation for ture over the ocean may be too small in the model, because
1950-1980, a period with good station coverageover the the oceanhorizontalheat transportsare identicaleachyear.
continents, and Plate 4b shows the standard deviation about The modeled and observedsurfaceair temperaturevariabil-
the 100-year mean for the GCM. We also computedthe ities over the oceanappearto be in good agreement(Plates
GCM's variability as the mean standard deviation about the 4 and 5), but it shouldbe rememberedthat the observations
13,364 HANSENANDLEBEDEFF:
GLOBAL
TEMPERATURE
TRENDS

STANDARD DEVIATION (øC), ANNUAL 1951-1980


90 "--"iF r

60 1.6
1.4

1.2
30
1.0

0.8

0.6

0.4
0.3
-3O
0.2
0.1

-60 . 0.0

-90
-180 -120 -60 0 60 20

STANDARD DEVIATION (øC), ANNUAL GCM


90

60- 1.6
1.4
'" 1.2
30 •-
1.0

0.8
0.6
0.4

0.3
-30 •
0.2
0.1

-60 - 0.0

-90
-180 - 120 - 60 0 60 120 180
(b)
Plate4. Interannualvariabilityof annual mean surfaceair temperature,specificallythe standarddeviationof the annual
mean surfaceair temperatureabout the long term mean. (a) Standarddeviationabout the 30-yearmean for our analysis
of observationsof meteorologicalstationsfor the period January1, 1951, throughDecember31, 1980. (b) Standard
deviationabout the meanfor the 100-yearrun of our GCM.
HANSENAND LEBEDEFF:GLOBALTEMPERATURETRENDS 13,365

• T

- o o

0 0 0 0 0 o o o o o O,
i i
I I

.,-• •

.--.

o 0
.=. 0 •

=L

o 0

o 0

c) 0 0 0 0 0 O,

i !
13,366 HANSENANDLEBEDEFF:GLOBALTEMPERA•RE TRENDS

0.3.............................................................................
i ...........
i.............
i................
•:
..............................
•/i----i--!,
•i...........
ø'•/..............
..... •,...... • ....•,,•---',-
]----):,,'•
-N--!--• _ ] Global
..............
1880's
Stations
if' ! • •', • • , ' ', ' !

0.I .....
•,.......
•;-•--•i-T
---!
........
T ......
t,-i
...........
•.....
•..........
r...............
",' • • ' • •i', • •, [ '

O.
I I............
•...........
• ..........
•....... '•..... • .........
•...........
•- --•?---'•
......
•/-
....•-•-?
......
-o.,l
/
.............
j
............
i
............

.......
',__I!• .....
• •
_:
• , 1930 s Sations i
........... 0.2

.....• ....... '• .... .___t --,......... •____•


............ • .............. ,,.............

___• __• . •....... •..................... •....... • --

, , ,
.j, • : •. • •'. .
-0.1

o.•.............
• •.......
0.3 ............
•i.......
, •! -•
,....
•-........
I ,'................
•, •............
•- --• - i__-
Southern Hemisphere,___
-0.3
0.2

.... ?: •T r 1880'sStations
I :

-o.3 .... • •-• ?-- • ?-- -•-- •-- :.... •- o.•

, •
•j•• , Southern Hemisphere, o.3
• , .
--,-:.... • T---n'•-: •[• •:-•---1930
s Stations 0.2
--: ) -• , -• •..... •.......... '•....... 0.1

•,• • ....... •_ • ....... J_,•


...... •

C_•• •',• ,• i',•',•,


?,,',,<•
, • ,,•,,

I I I I I I I I " -0.3
0 10 20 30 40 50 60 70 80 90 00
Year
Ei•. lo. Global and southernhemispher•t•mpetatuf• chanEein the ]00-year GCM tun basedon all the mod•l •rid
points(solid li•) and as •stimat•d from •quations(])=(3), usin• only t•mpefatufesfrom locationswhet• stations
dutin• the indicateddecade(dashedlin•).

are based on continent and island locations: interannual the model, they are beyondthe scopeof this paper. At this
variability of surface air temperature due to variable ocean time we can only say that the available comparisonssuggest
heat transports, such as the IE1 Nifio phenomenon, that the model's temperature variability on large spatial
are probably muted in these observationsand they are scalesis comparablein magnitude to the variability in the
absentin the model.This model deficiencyprobablycauses real world.

some underestimateof interannual temperature variability Our first error estimate was obtained as follows. For each

and perhaps also an underestimate of interdecadal decade(1880s, 1890s, etc.) we determinedthe locationsof
variability. those stationswhich reported temperature records for at
A more comprehensive comparisonof the model'sspace- least 5 years during the decade. We used the GCM-cal-
time variability with observationsis desirableand relevant culated temperatures at only the station locations for a
to the issue of how good is the error estimate obtained given decadein an attempt to reproducethe entire 100-year
from the model. Although we plan to pursue such com- globaltemperaturerecord of the GCM. An exampleis shown
parisons,as an aide to developmentof the next version of in Figure 10 for the 1880's and 1930's stations. The
HANSENANDLEBEDEFF:
GLOBALTEMPERATURE
TRENDS 13,367

9O

6O

$0

-30

-60

-90
- 180 - 120 -60 0 60 120 180
Fig. 11. Circlesmark the southernhenrispherestationlocations.Each SouthernHenrispherestationis connectedto a
northernhemispherestationwith similarlongitude,reflectedlatitude,and lengthof temperaturerecord.

deviationsin individual years from the "true" GM tempera- 0.04øC in the 1960s. These error estimates are illustrated for
ture trend (i.e., the trend based on all grid points) were several decades in Figures 6 and 7, where the bars are
found to be normally distributed.The standard deviation a _+1.96a,correspondingto 95% confidencelimits, based on
for the global mean temperature decreasesfrom about the GCM's variability. Since the GCM and real-world
0.07øC in the 1880s to about 0.02øC in the 1960s and for temperature variabilities are not identical and since there
the southernhemisphere,from about 0.13øCin the 1880sto are other sources of error, the bars only represent a

0.2

0I '. I ', • I ', ', ,, : .,i


.... :I'.•',•
--:?••,
: : : •'•,'
, , , , , , ,-1: --•' 1',,•,',•,,,,:',:
.....

o
........
.......
.......
............
- o.• ......................
........
.......
:::::::::::::::::::::::::::::
]......................................

, ,,,,.,/i"\:,,,, ,,•..,.'l ,;
, .... .,;.•,
.............

o O.• ........
4.......
;i,.......
4--4---A---•--",-+
.......
+-4....t,--•,-"- ---
t-- i il ',, i ',i i ,,, ",,• :: i i i i 'i i i 'r i i
i ;: • .... ! :! i•' ! ! i J i i ! i ! i i i
<• - 0.4 .......
4.....
• --+•--,•--->
.......
:'•
......
>-•-•.......
>.......
•.......
•........
'•.......
•'........
•.......
•'........
'•.......
•.......
>.......
•.......
>.......
•........
i ,,• :: i• %: • ', i i i ', " .....
', , ', • ', ',' ,'',i : • ', ', ', : ', I, ,, ,, • : •
0.5 .......f ½-•-•
.......•.......• ...... ;--'•-•-•.......•.......• ..... •.......4........•.......!........'r.......!........'r.......•.......½.......f.......+.......*........
all NH stations
-0.6 ..................
,•/
,,
............. NH stations similar
-0.7 .--,
....;'.....................................................................................to SH coverage --+
........

- o. 8 --:t-fi
...................................................................................................................................................................
',•l

-0.9 '
•880 •900 •920 •940 •960 •980
Date
Fig. 12. Five-yearrunningmean of the annualnorthernhemispheretemperaturechange,as estimatedfrom all northern
henrispherestations(solid CUlWe)
and as estimatedusingonly thosestationswhichyield a coverageequivalentto that in
the southernhenrisphere(seeFigure 11).
13,368 HANSEN
ANDLEBEDEFF:
GLOBALTEMFERATURE
TRENDS

0.5

0.4

0.:5

0.2
]---:
....;.....
,exc,
lud,es
cities
!.......
i--'! ' /
o 0

....
---,:':•,-
........ -:.....
:r- , ',
, ,
- -4...... P....... 4....... 'r....
• ,,
-0.2
,,f ,½ ' • i
-0.:5 ..................
'7 Northern Hemisphere.................. 0.:::5

-0.4
,,
', , !!
..........

• •
,.' ,
i ',
-0.5
---•
..... • .......
-0.6 ---?
................
i....................
,
+
' ,
, ?3......
':.......
• .....• ...... i--- ' -0. I
' •"•,. i , ', i •
" i' i, ,,, -0.2
i
I),,,, , ,,
,
,'
_ __•__ _
-0.3

0.2 -0.4
Southern Hemisphere
',
-0.5
,

o -0.6

<:•
-o.i
-0.2

-o.:5

-0.4
! ' ' i " ' ' ',
-0.5 , i "
,
' I
188o 1900 1920 t940 1960 198,0
Date

Fig. 13. Comparison


of 5-yearrunningmeantemperaturechangederivedwith andwithouturbanstationsincluded.

nominal measure of the expected error in the temperature southern hemisphere error estimate obtained in the GCM
change. studies described earlier. This result is not surprising,
One potential difficulty with basing an error estimate on becausethe southernhemispherestation distributionshould
the GCM's 100 year control run is that, unlike the observa- be less adequate in the northern hemisphere,which has
tions for the period 1880-1985,the GCM temperatureshad greatervariability;also no northern hemisphereanalogcould
no long-term trend. Thus we performed the same calcula- be found for several of the southern hemispherestations.
tionsusingthe temperatures
from the transient
CO2/trace We infer that the error estimate from the second method is
gas climate experiment(ScenarioA) of Hansenet al. [1986, roughly consistentwith the value obtained from the GCM.
1987]. The resulting error estimateswere practically the The "error bars" which we display in our figures are
sameasthoseobtainedfrom the 100-yearcontrolrun. ___1.96a,
based on the GCM a; although formally a "95%
We obtained an independentestimate of the error in the confidence"limit, we only regard it as an approximate
southernhemispheretemperature changeby estimatingthe measureof the uncertainty.
northernhemispheretemperaturechangeusingonly a subset We concludethat the principalfeaturesin the global and
of the northern hemispherestations, specificallya subset hemispherictemperaturechangesare real, in the sensethat
having spatial and temporal coverage in the northern they are not artifacts due to poor spatial coverage of
hemisphereequivalent to the coverageby all the southern stations.The long-termglobal trendsillustratedin Figure 6,
hemispherestations.The northern hemispherestationswere i.e., the 1880-1940 warming, 1940-1965 cooling, and
chosenby reflecting the locationsof each southernhemi- 1965-1985 warming, are much larger than the estimated
sphere station about the equator and finding the nearest errors. Also, shorter-term features such as the 1961-1964
available station with similar record length. As shown in cooling and 1981-1984 cooling are much greater than the
Figure 11, it was possible to preserve the clustering of error, as are many of the single-year changes.It is not
station locationsto closelymimic the southernhemisphere clear, though,whether 1980, 1981, or 1983 was the warmest
station distribution. The northern hemispheretemperature year in the globalrecord.
change obtained with this subset of stations is compared We also conclude that the uncertainty in the southern
with the temperaturechangebased on all of the stationsin hemispheretemperaturesis large in the period 1880-1900,
Figure 12. The standarddeviationbetweenthe two curvesin and the southernhemispheretemperaturetrend up to about
Figure 12 is about a = 0.1øC, somewhatlarger than the 1930 is too small to be meaningful.However, the results
HANSENANDLEBEDEFF:GLOBALTEMPERATURE
TRENDS 13,369

1.0
._ •,.-,- GIobal '
0.5 ............
•........
•_____.•..
.......
]!_•...•
............................................
•..........
•,..........
'•.......................
L........
,...........

!--'".
t-!-;•--t
,
....
i?•,
•':
........
ii--i--.•
.....
", ':
;•
.......
,,•
..........
!---,•-
.....
•--•
,• '
......... -
' • : ' ' ,• ;': ? :' ' ' il

-0.5
•,,• :•,:' ' '•,,
••,?,6::,'':
, :,,i! !i', i'•....
-1.0

-I.5
,, ',',-',,, .....i
,,----i-•
...........
i---"
J'-i
..........
i-i.......
':.......
-•.............
i.........
:..................
i.........
•..................
..........
+.........
..........
.........
+.........
'?
......•.......!.........•
.......:..
.......................
•..........
:......•...........
•..................
.+
......
.....................
!......
i......
i!........
...........
....? ' ........
.... .........,.o
......
'r,_•_•_•
•i:', :"',......
ß •!:
',':•---,:.•
' i......
•-
' !',--i-'•--::
::'•".........ß •--•---!,•-•-¾,'---•'
': Ji '•--?
is", ,....•i 0' 5
' a'
•;,,
....: ',:,• , ,, •, ,.,' ,,
• ::• ,,,,
' ,•,,,, ,•• •,•',.,,
•',•,,'t,.
.,',•,, '., ,,,
.,' ,•,,
'... ,'..,

.................................... 1.0

ii--i---
,•---?-
.....
..'--
-':•---'½½
--•
........
i--?i...........
!...........
:.........
!................
'•........
ß......... 1.5
2.0 • ' _:__]__
i ?_'.• '• ; __•_ ' • -2.0
1.5
!:! Box 16(Eastern United States)
1.0

0.5

-0.5

-1.0

-I.5

- 2.0
1760

Fig. 14. Estimatedtemperaturechangein the period 1760-1985for the globe,box 9, and box 16 (see Figure2). The
error barson the globaltemperaturechangeare 95% confidence
limitsderivedas described in the discussion
of Figures
6 and 10.

show that there was a real warming trend in the southern this assumption is supported by empirical studies, e.g.,
hemisphereduring the past century,and the sharpwarming Mitchell [1953].We usedTable E of Davis [1969]to identify
trend which has occurredsince1965 is much larger than the populationcentersexceeding100,000people. Elimination of
uncertainty. all stations within these population centers reduced the
An additonalissueor uncertaintyabout the derived global numberof stationsby aboutone third.
temperature change is the following: How much of the Removal of the city data reducedthe magnitudeof the
change is a result of the growth of urban heat island globaland hemisphericwarmings,as illustratedin Figure 13.
effects? There is abundant evidence that the growth or For example, the global temperature change in the past
developmentof urban areas is a significantcontributor to century was reduced from 0.7ø to 0.6øC, where these
local temperature trends [Mitchell, 1953; Landsberg,1981; numbers represent the difference between the mean
Cayanand Douglas,1984;Karl, 1985;Kukla et al., 1986].We 1980-1985temperatureand the mean1880-1885temperature.
obtainedan estimateof the magnitudeof urban influenceon We subjectivelyestimatethat completecorrectionfor urban
the global temperature change of the past century by heat island effects should not reduce the global warming in
eliminatingfrom the data set all stationsassociatedwith the past century, defined as the temperature difference
populationcenterswhich had more than 100,000people in between1980-1985and 1880-1885,to lessthan about 0.5øC.
1970. The usefulnessof the test is based on the assumption As mentionedalready,the nature of the observedtemper-
that even though the urban heat island effect existsfor all ature trends, especiallythe geographicaldistributionof the
city sizes, the effect generally increaseswith population; warming, also provides strong evidence that the global
13,370 HANSENANDLEBEDEFF:
GLOBALTEMPERATURE
TRENDS

0.5

0.4

0.3
thispaper .... 4..... • ....... •--

.............Jonese! al. (1986) i : i Globol


0.2

0.1 ,

-0.1

-0.2

-0.3 0.5
,

-0.4 0.4
• • Northern Hemisphere
-0.5 •........ •...... •---- ' --• ...... 4....... 4-....... •....... 4-....... 0.3
,

-0.6 0.2

-0.7
o

0 v

-0.1
<:•
-0.2

i
-0.3

0.4 -0.4

0.3 -0.5

0.2 ß "Southern
k......i.... Remisphere
i i

0.1

-0.1

-0.2

-0.3

-0.4

-0.5

-0.6
1880 1900 1920 1940 1960 1980
Date
Fig. 15. Comparisonof our resultsfor annualglobaland hemispherictemperaturechangewith the resultsreportedby
Joneset al. [1986c].

temperaturechangeis not a figment due to urban heat designedto obtainmaximumtrend informationfrom sparse
island effects.This evidenceand the quantitativetest just data, it is apparent from the magnitudeof the uncertainty
describedlead us to concludethat the global warming of bars that the global change prior to 1880 is not very
the past centuryis a real climatetrend, even thoughit meaningful.What we can say is that the data provide no
does contain a significantcontributiondue to urban heat evidencethat 1880was unusuallycold, and thus no evidence
island effects.More detailed and comprehensive studiesof that much of the subsequentwarming representsa return to
urban influenceare warranted;perhapsthe data set we have levelsprevailinga few years earlier. The data do suggest
developed,includingthe seasonalvariation of trends,can that the 1884 temperatureis lower than that in 1880-1881,
contribute to such studies. an effect that some have associatedwith expectedcooling
Finally, we considerthe contentionof Ellsaesser et al. from a largevolcaniceruption(Krakatoa)in 1883.However,
[1986]that much of the warmingof the past centuryis an the 1984 temperatureis lower than that in 1980-1981,and
artifact due to most analysesbeginningnear a minimumin there was a large volcaniceruption (El Chichon)in 1982.
the temperature record. Specifically,they suggestthat Thus we also do not find evidenceof prejudicein the com-
25-50% of the approximate0.5øC temperaturerise in the parisonof t-he1880-1885meanwith the 1980-1985mean.
past centuryis due to the fact that severalanalysesbegan We concludethat the evidence from the meteorological
in 1880or 1881"nearthe apparenttemperatureminimumof stationrecordssupportsthe estimateof a globalwarmingof
1883." approximately 0.5øCin the past century,an estimatewhich
Our estimated"global"temperaturechangefor the period has been used in many studiesof climate sensitivity.The
1760-1985is shownin Figure 14, alongwith the temperature reason for beginning temperature analysesin 1880 is
changeof the two regions(Europe and the United States) apparentfrom Figure 4, especiallythe curvefor the station
where the records extend over two centuries.Although our coveragein the northern hemisphere,and also from the
procedure for estimating global temperature change is error barsin Figure 14.
HANSENAND LEBEDEFF:GLOBALTEMPERATURETRENDS 13,371

0.8 -- i• Thispaper
i............
Angell
0.6 --•" -•-i.....
+.....and Korshover !
i........ : ................. '-''-i-'i---j
• • ! '---* ....
i.....
i.....
!..... • ..................

:•,:!i i '• ! , • : ob . :, •
0.41
0.2 ••i•.....
•-•.....
••.....
...... •-•
_.t;._/_I,
"
_.•,• G Ii•al • •-•.•
•.....
': •', i '•
•-•...... •', ••-'-•
....•? •:•---•,
• • •
! :,,i,,,,

':
' •-•--t--
¾-•.....
I" :• •, • :• •',

o "-
0.2
..... i ",- ........

o,6--T
• :......
•'?f-'
• .• '....... Northern
Hemisphere
-"

, ,j f ,i ',•
•:',•
', ,i,• : , t ....
, . : ' , ', .,'v,
• ,•'i:
•,,,i ] •'--

---•--•??•....•....T-, : • ', , • i--•-i-f-'r :•,-•-0.2


---_:___:....: --:---•---L:-•:'-_',d:
..... •_...... i__J_ , , _;....L__,'_
....

o.6 , ' , : • '• ,


0.4•i,?t
•, ',•: Southern
Hemisphere
;........
0.2

0.2

0.4

1958 1960 1965 1970 1975 1980 1985


Date

Fig.16. Comparison of our resultsfor seasonal surfaceair temperature


changewith resultsprovidedby Angelland
Korshovef,The arbitraryzero-pointis set as the 1951-1980meanfor our data;we set the zero-pointfor Angelland
Korshover's
databymakingthemeanthesameasfor ourdatafor theperiodof recordin common (1958-1985).

6. COMPARISONS WITH OTHER DATA SETS the temperaturechangethey find is slightlyflatter than
We have shown that the network of meteorological that whichwe find for that period. The fact that thesetwo
stationswhich measuresurfaceair temperatureis sufficient analysesof surfaceair temperaturechange,which differ
changefor both the northern substantially
to yieldreliabletemperature in their methods,are in goodoverallagreement
and southern hemispheres,despite the fact that most lendsadditionalcredibilityto the grossconclusions.
J. K Angell[privatecommunication,
stations are located on the continents. The estimated errors 1987]kindlyprovided
smallto implythat long-term,decadaland us with updatedtemperaturesbasedon analysisof radio-
are sufficiently
interannualtemperature changesobtainedfrom the station sondedata as describedby Angell and Korshover[1983].
data are meaningful. Their analysisis basedon 63 radiosondestations(38 in the
Joneset al. [1986c]have recentlyestimatedglobalnear- northernhemisphereand 25 in the southernhemisphere)
surfacetemperaturechangeby combiningmeteorologicalwhich have wide geographical distribution. The great
station data with marine data. The marine data contain a advantageof their data is that it includes records of
numberof systematic errors,for whichJoneset al. devel- temperature changein the upper air, through the tropo-
opedan empirical correction usingnearbymeteorological sphere and lower stratosphere.Thus it is free of any
station data. Their global and hemisphericresults are significantinfluenceof the urban heat island effect, and
generallyin good agreement with ours, as shownin more importantly,it providesinformationon the altitude
Figure15. One difference is thattheydescribe theirresults dependence of anyobserved temperaturechange.
for the globaltemperature changeas beingflat from 1940 Our global and hemisphericseasonalmean surfaceair
to 1970,while we note a temperaturedeclinein the period temperature changes are compared with thoseof Angelland
1940-1965.This is partly a matter of perception,although Korshoverin Figure 16. The standarddeviationbetweenthe
13,372 HANSENANDLEBEDEFF:GLOBALTEMPERATURE
TRE•DS

two results is 0.09ø, 0.09ø, and 0.17øC for the globe, Projections
of globalclimatechange,
in Effectsof Changes
in
northern
hemisphere,
andsouthern
hemisphere,
respectively.Stratospheric
Ozone andGlobal
Climate,
edited
byJ. G.Titus,
379 pp., U.S. EnvironmentalProtectionAgency, Washington,D.
The corresponding1-a error estimatesare 0.06ø, 0.07ø and C., 1986.
0.09øCfor the radiosonde
stations
and 0.03
ø, 0.03
ø, and Hansen,
J.,A. Lacis,
D. Rind,
G.Russell,
I. Fung,
andS.Lebedeff,
0.05øCfor the (1970s)distributionof meteorologicalstations Evidence for future warming: How large and when, in The
used in our analysis, where the error estimates were Greenhouse Effect: Climate Changeand ForestManagementin the
United States, edited by W. E. Shands and J. S. Hoffman,
obtained as describedin section5 and refer only to the
ConservationFoundation,Washington,D.C., 1987.
error due to incompletespatial coverageby the stations. Jenne,R., Data setsfor meteorological research,NCAR-TN/JA-111,
These results suggestthat most of the difference between 194pp., Nat. Cent. for Atmos. Res., Boulder, Colo., 1975.
the two temperaturerecordsis due to the incompletespatial Jones,P. D., and P.M. Kelly, The spatialand temporalcharacter-
coverageof stations.The error in surface air temperature isticsof northern hemispheresurfaceair temperaturevariations,
J. Climatol., 3, 243-252, 1983.
change is only about twice as large for the radiosonde
Jones,P. D., S.C. B. Raper, R. S. Bradley,H. F. Diaz, P.M. Kelly,
stations as for the meteorological stations, even though
and T. M. L. Wigley, Northern hemispheresurfacetemperature
these are more than an order of magnitude more of the variations:1851-1984,J. Clim.•lppl. Meteorol.,25, 161-179,1986a.
meteorologicalstations (Figure 4). Of course the main Jones, P. D., S.C. B. Raper, and T. M. L. Wigley, Southern
advantageof the radiosondestationsis the informationthey hemisphere surface air temperature variations: 1851-1984,
provideon changesin the upperair. J. Clim.Appl. Meteorol.,25, 1213-1230,1986b.
Jones,P. D., T. M. L. Wigley,and P. B. Wright, Global temperature
Acknowledgments. We appreciate useful commentson the first variations between 1861 and 1984. Nature, 322, 430-434, 1986c.
draft of this paper, which were provided by Tim Barnett, Phil Jones,P. D., T. M. L. Wigley and P.M. Kelly, Variationsin surface
Jones,Peter Stone, and the JGR referees.Tom Wigley and Jim air temperatures: 1. Northern hemisphere, 1881-1980. Mon.
Angell kindly provided the data which allowed the comparisonsin WeatherRev., 110, 59-70, 1982.
Figures15 and 16, respectively.We thank JeffreyJonasand Patrice Karl, T. R., Perspectiveon climatechangein North Americaduring
Palmer for helping produce the color figures, Jose Mendoza for the twentiethcentury,Phys.Geogr.,6, 207-229, 1985.
draftingother figures,CarolynPaurowskifor typingthe manuscript, Kukla, G., J. Gavin, and T. R. Karl, Urban warming,J. Clim. •lppl.
and Elizabeth Devine for assistancewith the desktop typesetting. Meteorol.,25, 1265-1270, 1986.
We defined our approach for combiningstation data, equations Landsberg,H. E., The Urban Climate,275 pp., Academic,Orlando,
Fla., 1981.
(1)-(3), in the late 1970s and subsequentlyreceived help from
several students in implementing and checking the data; we Lorenz, E. N., The index cycleis alive and well, Proceedingof the
particularlynote the help of Dan Johnsonand C3mthiaJones,who Namias Symposium,edited by J. Roads, 98 pp., ScrippsInst.
worked on the data as participants in the Summer Institute on Oceanogr.,La Jolla, Calif., 1985.
Planetsand Climate in 1979 and 1985, respectively.This work has Mitchell, J. M., On the causesof instrumentallyobservedsecular
been supported by the NASA Climate Program, managed by temperaturetrends,J. Meteorol.,10, 244-261, 1953.
R. Schiffer,and by EPA grant R812962-01-0. Mitchell, J. M., Recent secularchangesof globaltemperature,•lnn.
N.Y. Acad. Sci., 95,235-250, 1961.
REFERENCES Mitchell, J. M., B. Dzerdzeevskii, H. Flohn, W. L. Hofmeyr,
H. H .Lamb, K. N. Rao and C. C. Wallen, Climate Change,WMO
Angell,J. K., and J. Korshover,Global temperaturevariationsin the Tech. Note 79, 79 pp., World Meterol. Organ.,Geneva,1966.
troposphereand stratosphere,1958-1982,Mon. WeatherRev., 111, Oort, A. H., Adequacy of the rawinsonde network for global
901-921, 1983.
circulationstudiestested through numericalmodel output, Mon.
Barnett, T. P., Long-term trends in surfacetemperatureover the WeatherRev., 106, 174-195, 1978.
oceans,Mon. WeatherRev., 112, 303-312, 1984.
Ostle, B., 1963: Statistics in Research, 585 pp., Iowa State
Bradley, R. S., P.M. Kelly, P. D. Jones, C. M. Goodess,and UniversityPress,Ames, 1963.
H. F. Diaz, A climatic data bank for northern hemisphereland Paltridge,G., and S. Woodruff, Changesin global surfacetemper-
areas, 1851-1980, DOE Tech. Rep. 017 (DOE/EV/10739-2), ature from 1880 to 1977 derived from historical records of sea
335 pp., Dep. of Energy,Washington,D.C., 1985. surfacetemperature,Mon. WeatherRev., 109, 2427-2434, 1981.
Budyko, M. I., The effect of solar radiation variations on the Spangler,W. M. L., and R. L. Jenne,World monthlysurfacestation
climate of the Earth, Tellus, 21,611-619, 1969. climatology.Computer data tape documentation,14 pp., Natl.
Cayan, D. R., and A. V. Douglas, Urban influences on surface Cent. for Atmos. Res., Boulder, Colo., 1980.
temperaturesin the southwesternUnited States during recent Vinnikov, K. Ya., G. V. Gruza, V. F. Zakharov, A. A. Kirillov, N. P.
decades,J. Clim.Appl. Meteorol.,23, 1520-1530,1984. Kovyneva,and E. Ya. Ran'kova,Current climaticchangesin the
Davis, K., World Urbanization 1950-1970, Vol. I. Basic Data for northernhemisphere,(in Russian)Meteorol.Gidrol.,6, 5-17, 1980.
Cities, Countries,and Regions,Population Monogr. Ser., no. 4, Wigley, T. M. L., J. Angell, and P. D. Jones,Analysisof the
Universityof California,Berkeley,1969. temperaturerecord, in Detectingthe Climatic Effectsof Increas-
Ellsaesser,H. W., M. C. MacCracken, J. J. Walton, and S. L. Grotch, ing CarbonDioxide, DOE/ER-0235, 198 pp., Dep. of Energy,
Global climatic trends as revealed by the recorded data, Rev. Washington,D.C., (availableNatl. Tech. Inf. Serv.,Springfield,
Geophys.,24, 745-792, 1986. Va.), 1985.
Folland, C. K., D. E. Parker, and F. E. Kates, Worldwide marine
Willett, H. C., Temperaturetrends in the past century.Centen.
temperaturefluctuations1956-1981,Nature, 310, 670-673, 1984. Proc.,R. Meteorol.Soc., 195-206, 1950.
Gandin, L. S., ObjectiveAnalysisof MeteorologicalFields,242 pp., Yamamoto, R., and M. Hoshiai, Fluctuations of the northern
Israel Programfor ScientificTranslations,Jerusalem,1963. hemispheremeansurfaceair temperatureduringrecent100years,
Hansen, J., D. Johnson,A. Lacis, S. Lebedeff, P. Lee, D. Rind, and
estimated by optimum interpolation,J. Meteorol. Soc. Jn., 58,
G. Russell, Climatic impact of increasingatmosphericcarbon 187-193, 1980.
dioxide, Science,213, 957-966, 1981.
Hansen, J., G. Russell, D. Rind, P. Stone, A. Lacis, S. Lebedeff,
R. Ruedy, and L. Travis, Efficient three-dimensionalglobal J. Hansen, NASA Goddard Space Flight Center, Institute for
models for climate studies: Models I and II, Mon. Weather Rev., SpaceStudies,2880 Broadway,New York, New York 10025.
111, 609-662, 1983. S. Lebedeff, Sigma Data ServicesCorporation, 2880 Broadway,
Hansen, J., A. Lacis, D. Rind, G. Russell, P. Stone, I. Fung, New York, N.Y. 10025.
R. Ruedy and J. Lerner, Climate sensitivity:Analysisof feedback
mechanisms, in Climate Processes and Climate Sensitivity,
Geophys.Monogr. Ser., Vol. 29, edited by J. E. Hansen and
T. Takahashi,AGU, Washington,D.C., 1984. (ReceivedMay 7, 1987;
Hansen, J., A. Lacis, D. Rind, G. Russell, I. Fung, P. Ashcraft, revisedJune 26, 1987;
S. Lebedeff, R. Ruedy, and P. Stone, The greenhouseeffect: acceptedJuly 1, 1987.)

You might also like