Seasonal and Regional Variation of Pan-Arctic Surface Air Temperature

Over the Instrumental Record

James E. Overland

1

Michael C. Spillane

2

Donald B. Percival

3

Muyin Wang

2

3

Box 355640

Revision submitted to Journal of Climate

teleconnectivity, yet there are systematic seasonal and regional differences. Analyses are based

on time/longitude plots of SAT anomalies and Principal Component Analysis (PCA). Using

monthly station data rather than gridded fields for this analysis highlights the importance of

considering record length in calculating reliable Arctic change estimates; for example, we

contrast PCA performed on 11 stations beginning in 1886, 20 stations beginning in 1912, and 45

stations beginning in 1936. While often there is a well-known interdecadal negative covariability

in winter between northern Europe and Baffin Bay, long-term changes in the remainder of the

Arctic are most evident in spring, with cool temperature anomalies before 1920 and Arctic-wide

warm temperatures in the 1990s. Summer anomalies are generally weaker than spring or winter

but tend to mirror spring conditions before 1920 and in recent decades. Temperature advection in

contrast with the recent Arctic-wide AO influence in the 1990s. The preponderance of evidence

supports the conclusion that warm SAT anomalies in spring for the recent decade are unique in

the instrumental record, both in having the greatest longitudinal extent and in their associated

patterns of warm air advection.

focus on changes in each season and in different regions of the Arctic, rather than concentrating

on annual zonal averages, because extensive averaging can often obscure the underlying physics.

Our analysis also provides a reevaluation of the instrumental temperature record in that we base

our methodology on station data rather than gridded analyses as in most earlier studies. This

approach avoids the possible introduction of artifacts due to gridding, which is particularly

important in understanding the results from Principal Component Analysis (PCA).

The most difficult issue in retrospective analyses of the Arctic is the range of starting

dates in the instrumental record. Although data coverage is far from complete, there is

considerable information from land areas of the Arctic dating from the early 20th century.

Przybylak (2000) notes that there is good Arctic coverage since the 1950s. Due to the strong

The fourth International Polar Year is scheduled for 2007 and will emphasize unresolved

issues of polar influence on climate variability. It is fitting to review the data that began with the

First Polar Year (1882–1883), which marked a transition from exploration to scientific study in

the Arctic. It is also fitting to update the analyses of the many authors in the 1920s–1940s who

noted the warm temperature anomalies of the period (Ahlmann 1948) and pioneered the concept

of high-latitude climate variability, in contrast to the prevailing uniformitarianism.

Recent studies show considerable change in the Arctic over the previous three decades in

both physical and biological indicators (Serreze et al. 2000; Overland et al. 2004). These

context of the period from the early 1800s to present, noting the warm temperature anomalies in

the mid 20th century as the end of a long period of rising temperatures. Recent analyses of the

mid-century warming suggest that internal atmospheric variability is important to its explanation

and that the regional dynamics were different compared to recent decades (Hanssen-Bauer and

Førland 1998; Bengtsson et al. 2003; Johannessen et al. 2003; Dickson, personal communication,

2003). As we shall show, it is also important to note the large seasonal and regional differences

in understanding temperature change on decadal time scales throughout the 20th century.

Two recent studies call into question whether changes in the recent period (1990s) are

unique compared to longitudinally averaged, historical temperature data; both papers note the

warm events in 1920–1950. Przybylak (2002) states that while 1991–2000 is the warmest decade

temperature records (Crowley 2000; Briffa and Osborn 2002) make the case for recent warming

relative to the previous two centuries based on external forcing driven by solar variations,

aerosols from volcanoes, and carbon dioxide. In particular, the cool period in the first half of the

1800s can in part be associated with major aerosol production, while the warm period in the first

half of the 20th century had almost no volcanic production (Robock 2000). Warming in the last

2

half of the 20th century is presented as an argument for the uniqueness of recent CO increases in

overcoming the increase in volcanic influence in the last 40 years. Thus we have two competing

usual formulation of PCA requires a complete temporal/spatial data matrix. Previous analyses

beginning in 1881 and 1892 (Kelly et al. 1982; Semenov and Bengtsson 2003) rely on gridded

fields where much of the data matrix for the early years is completed by a fill-in rule. We avoid

the need for gridding by conducting three PCAs with time series limited to those stations

beginning in or before 1886, 1912, and 1936. This approach requires that we fill only a few

temporal data gaps. We also limit our PCA to north of 64°N, remaining Arctic centric, while

earlier studies included considerably more stations southward to 60°N and 40°N.

The next section discusses the available data series. This is followed by a visual and

semiquantitative (PCA) inspection and interpretation of the nearly complete instrumental record

of Arctic SAT.

(WMO) location index, with records from adjacent locations supplementing that from the

primary site. A secondary source of monthly mean data is the World Monthly Surface

Climatology (WMSC) dataset (http://dss.ucar.edu/catalogs/ranges/range560.html). The selected

time series were cross checked with the data set from Polyakov et al. (2002).

b. Time series selection

It is important to define a southern border of the Arctic for climatological study. There

are few stations north of 80°N, but a major increase in the number of stations as one extends the

multimeteorological element approach (Przybylak 2000). The definition is more complex if one

considers the southern limits of hydrologic river basins, which flow into the Arctic. There is no

firm selection criteria.

For our purpose, we propose a practical southern limit of 64°N. Adding the many stations

south of this latitude bias results to the subarctic region. On the other hand, 64°N provides a

reasonable climatological limit for stations that lie north of the Arctic front in winter and provide

sufficient geographic and temporal station coverage. For example, we include Fairbanks, Alaska

but not Anchorage. We also include some interior stations in northern Canada and Russia that

have long instrumental records. In maps of recent temperature trends (Chapman and Walsh 1993)

there are north-south orientations to anomalies, which suggest these relatively lower latitude

stations in the Baffin Bay region.

To minimize data gaps and maximize the number of useable long time series, data were

combined as follows:

• stations were limited to north of 64°N in the GHCN database,

• each record was initially populated with data from the primary WMO location,

• data gaps were successively filled using information from the supplementary adjacent

location, if available, in the order they appear in the GHCN Version-2 file,

values were eliminated, and

• stations which begin after 1936 were not used, with a few exceptions where regional

coverage is sparse, such as the Canadian Arctic.

The zonal coverage of the identified stations is not uniform. In data rich areas, such as

Scandinavia and the western Russian Federation, some series were removed; the stations retained

were those deemed optimal in terms of duration of coverage, continuity of data, and

representativeness. The result is a set of 59 stations which will be used for visual inspection.

They are plotted in Fig. 1 and listed in Table 1 by longitude. An overall measure of

completeness for each station is given by the percent of monthly observations present in the

available period for that location.

For visual presentation of time series and principal components, we apply a five-year

running average to suppress interannual fluctuations. We justify the five-year smoothing as an

approach for investigating century-long decadal change. Decadal records often shift in response

to changes in the frequency of extreme events. For example, the number of winters with cold

stratospheric temperature anomalies increased in the 1990s relative to the 1980s. Changes in the

frequency of volcanic events can affect SAT on decadal time scales. There are natural changes in

storminess from year-to-year. There is also the potential for a high-latitude influence from the

Combining Arctic data into conventional seasonal three-month averages obscures the patterns we

seek to understand. Thus we group monthly data into quasi-seasonal series based on their

similarity. We define Arctic “winter” as December–January, “spring” as April, “summer” as

July–August and “fall” as October. Other months weakly resemble these patterns or can be

considered transitional. We discuss the consequences of this approach in section 3c.

We conduct a PCA for comparison with visual inspection of the time series. Two

important issues are gap filling of the data and the variable length of the records. With regard to

the few remaining data gaps, we use both a Monte Carlo approach, where missing data are filled

by sampling from a statistical model and repeating the PCA multiple times, and an estimation

procedure for the PCs (Davis 1976). Both approaches gave similar results and indicate that the

3. Visual inspection and PCA of instrumental temperature time series

As discussed below, visual inspection suggests strong spatial correlation of temperature

anomalies within different segments of the Arctic at decadal scales. We define six sectors in the

text corresponding to northern Europe, Siberia, Beringia, northwestern Canada, Baffin Bay, and

east Greenland; these regions are located in Fig. 1. The northern Europe sector extends eastward

from Scandinavia to Archangelsk and includes the islands of the northeastern Atlantic. Beringia

represents the region on both sides of Bering Strait. The inland location of Fairbanks groups

Przybylak (2003), except for his combining east Greenland with northern Europe, which from

visual inspection of the station time series is not unreasonable. The longest records are from

northern Europe, Baffin Bay, and eastern Greenland; there are several sites in other regions

which begin before 1900.

a. Winter

We begin the analysis by inspecting time/longitude plots of temperature anomalies

relative to 1961–1990 means for December–January (Fig. 2). The time/longitude plots for

individual months are available at www.unaami.noaa.gov. The northern Europe sector (Stns.

1–9) shows evidence of an interdecadal signal with alternating cold and warm anomaly periods

Maximum warm anomalies in the late 1930s and early 1940s occur several years later than the

mid-1930s maximum in northern Europe. After the 1970s the tendency for Siberia is to be out of

phase with northern Europe, with a warm period in the 1980s and a cool period in the late 1990s

similar to 1890–1910. Far Eastern Siberia and North America generally stayed cool until the mid

1970s, with Beringia (Stns. 27–36) turning cold again starting in the late 1980s and NW Canada

(Stns. 37–44) remaining warm.

The Baffin Bay region (Stns. 45–53) was especially cold during 1865–1910 when Europe

was warm. It was mostly in phase with Europe from the late 1910s to 1948, with a cold followed

strong longitudinal Arctic-wide covariability. Particularly strong are the warm anomalies from

the mid 1930s to the early 1950s. This pattern is arguably repeated in the 1980/1990s but with

more temporal variability at individual stations. East Greenland eastward through northern

Europe (Stns. 54–7) was generally cool before 1930.

We have combined two months with similar anomaly patterns for winter (December–

January) and summer (July–August) and have used one month to represent spring and fall

transitions. The following is a brief description of how the other six months relate to those

shown; monthly plots are available at www.unaami.noaa.gov. Our four seasonal figures 2, 8, 12,

13 are in general representative of the variations for all months. February is similar to

December/January except for a warming in north America in recent years. March is similar to

4. Discussion

a. Comparison with previous PCAs

All previous PCAs that we are aware of have been based on gridded fields of SAT. Walsh

(1977) conducted a PCA analysis north of 60°N for 1955–1975 from temperature anomalies for

all individual months. His first mode is the N. Atlantic seasaw and his second mode has a central

Siberia and northern N. America in phase, which echos our winter pattern C. Kelley et al. (1982)

uses annual data north of 60°N for 1881–1980. Their first mode is an in phase behavior between

positive after 1970. Their second mode is a positive-negative hemispheric pattern with a high

frequency character and potential ENSO influence. Their third mode shows a mid-century

1920–1955 warm period with Baffin Bay and northern Europe in phase.

All analyses include the winter seasaw pattern as an early detected mode. Our winter

pattern A time series stay interdecadal in character during the recent decade, while the Semenov

and Bengtsson (2003) first PC shows an upward trend during the 1980–1990s. Perhaps their

compositing of March–April, which does have this trend, with winter months contributes to this

result. Our second winter pattern B and first spring pattern A show mid-century warming events

in the 1920–1930s and 1940–1950s, respectively. The Semenov and Bengtsson (2003) mid-

century third PC also seems to composite these two seasonal events. The Kelly et al. (1982)

a general warming over the instrumental record (Figs. 2, 8, 12, 13). For comparison to the 19th

century several references are of note in addition to the Stockholm record. Gervais and

MacDonald (2001) state that trees on the Kola peninsula show suppressed growth for a twenty-

year period after the 1809 eruption. Lee et al. (2000) reports cold springs in Finland before 1900.

It is likely that the Arctic was cold in certain decades in the 19th century with a warming rebound

in the first half of the 20th century. That these trends are often in non-winter months is consistent

with papers that suggest a reduction in volcanic influence from the mid 1800s through the 1950s.

However, the upward trend of SAT has continued during recent decades despite an increase in

several authors make a case for winter warming in Eurasia from an increase in volcanic eruptions

(Robock and Mao 1992; Graf et al. 1993; Stenchikov et al. 2002). Robock and Mao (1992) note

high latitude winter warming from the 12 largest volcanoes since 1883. The physical argument is

that the radiational effects are large at low latitude, producing primarily a dynamical response in

mid and high latitudes due to increased latitudinal temperature gradients. Thus the winter

warming of the European sector in the 1930s, which influences the pan-Arctic annual average

temperature anomalies, cannot be clearly attributed to the lack of vulcanism. The winter warming

in Europe in the 1930s–early 1940s appears to be more of a high latitude internal variability

event following a warm phase of the North Atlantic seasaw in the 1920s.

While there is a general minimum in temperature anomalies in the Arctic during the

to the 1800s.

Temperature anomalies in the Arctic, at least in fall through spring, are primarily driven

through temperature advection. This is documented for the Scandinavian sector in winter (Fu et

al. 1999) and the remaining Arctic in spring (Overland et al. 2002). The positive phase of the AO

in the winter provides warming and cooling in the Atlantic sector but not a strong temperature

advection signal in the remainder of the Arctic (Rigor et al. 2002). Considerable advective

warming does occur in the remainder of the Arctic in the spring when the polar vortex breaks

down. The strong dynamic (wind) control in the Arctic cool season suggests care should be taken

circulation through changes in latitudinal temperature gradients (Stenchikov et al. 2002). It is

also possible that sea ice processes, change in land cover such as the increase in shrubs (Sturm et

al. 2001), and ozone chemistry (Newman et al. 2001) promote the persistence of spring and

summer circulation anomalies on decadal scales. Such a decadal feedback process could have

occurred in the late 1930s (Bengtsson et al. 2003).

5. Conclusions

There are considerable differences in decadal SAT variability in the Arctic across seasons

and regions. These differences are apparent in both the visual and semi-quantitative (PCA)

investigation of the data. Much previous work centers on annual, six month, and Arctic-wide

pattern, and then extrapolating to earlier time periods by projecting onto fewer stations to

determine the PCs, is questionable.

In this study we have used PCA to track the major features shown in time/longitude plots

of SAT. For example, the nearly Arctic-wide simultaneous events of the 1910s and 1980s are

resolved by our winter pattern C. Our PCs differ and support previous multi-month composite

analyses. Our winter pattern A emphasizes a continuing NAO interdecadal oscillation and a

separate longer duration warming event is represented by pattern B. Our spring pattern A

emphasizes warming in the 1940s/1950s and 1990s. However, based on the preponderance of the

Acknowledgments. We appreciate discussions with K. Wood and N. Bond on aspects of this

paper. We thank NSF through the SEARCH Project Office for support of this study. The paper

is also a contribution to SEARCH through the NOAA Arctic Research Office. JISAO

contribution number 997 under NOAA Cooperative Agreement NA17RJ1232.

Climate, 15, 1969–1978.

Bengtsson, L., V. A. Semenov, and O. Johannessen, 2003: The early century warming in the

Arctic—a possible mechanism. Report 345 Max Planck Institut fur Meteorologie, ISSN

0937-1060, 31 pp.

Briffa, K. R., and T. J. Osborn, 2002: Blowing hot and cold. Science, 295, 2227–2228.

Chapman, W. L., and J. E. Walsh, 1993: Recent variations of sea ice and air temperature in high

latitudes. Bull. Am. Meteorol. Soc., 74, 33–47.

Crowley, T. J., 2000: Causes of climate change over the past 1000 years. Science, 289, 270–277.

d’Arrigo, R. D., and G. C. Jacoby, 1993: Secular trends in high northern latitude temperature

reconstructions based on tree rings. Clim. Change, 25, 163–177.

Climatol., 19, 581–606.

Gervais, B. R., and G. M. MacDonald, 2001: Tree-ring and summer-temperature response to

volcanic aerosol forcing at the northern tree-line, Kola Peninsula, Russia. The Holocene,

11, 499–505.

Graf, H. F., I. Kirchner, A. Robock, and I. Schult, 1993: Pinatubo eruption winter climate effects:

model versus observation. Clim. Dyn., 9, 81–93.

IPCC Report, Houghton, J. T. and Coauthors, 2001: Climate Change 2001: The scientific basis.

Cambridge Press, 881 pp.

Johannessen, O. M., and Coauthors, 2003: Arctic climate change—observed and modeled

temperature and sea ice. Nansen Center Technical Report, 26 pp.

Kelly, D. M., and Coauthors, 1982: Variations in surface air temperatures: Part 2, Arctic regions

1881–1980. Mon. Weather Rev., 110, 71–83.

Lee, S. E., M. C. Press, and J. A. Lee, 2000: Observed climate variations during the last 100

years in Lapland, Northern Finland. Int. J. Climatol., 20, 329–346.

Moberg, A., H. Bergström, J. R. Krigsman, and O. Svanered, 2002: Daily air temperature and

pressure series for Stockholm (1756–1998). Clim. Change, 53, 171–212.

variability. Part II: Development of 1901–96 monthly grids of terrestrial surface

temperature. J. Climate, 13, 2217–2238.

Newman, P. A., E. R. Nash, and J. E. Rosenfield, 2001: What controls the temperature of the

Arctic stratosphere in spring? J. Geophys. Res., 106, 19 999–20 010.

Omstedt, A., and D. Chen, 2001: Influence of atmospheric circulation on the maximum ice

extent in the Baltic Sea. J. Geophys. Res., 106, 4493–4500.

Overland, J. E., M. Wang, and N. A. Bond, 2002: Recent temperature changes in the western

Arctic during spring. J. Climate, 15, 1702–1716.

wavelets. J. Amer. Stat. Assoc., 92, 868–880.

Polyakov, I. V., and Coauthors, 2002: Observationally based assessment of polar amplification

of global warming. Geophys. Res. Lett., 29, doi:10.1029/2001GL011111.

Przybylak, R., 2000: Temporal and spatial variation of surface air temperature over the

instrumental observations in the Arctic. Int. J. Climatol., 20, 587–614.

—, 2002: Variability of Air Temperature and Atmospheric Precipitation in the Arctic. Kluwer,

330 pp.

—, 2003: The Climate of the Arctic. Kluwer, 270 pp.

Rigor, I. G., R. L. Colony, and S. Martin, 2000: Variations in surface air temperature

observations in the Arctic 1979–1997. J. Climate, 13, 896–914.

northern North Atlantic in the 1920s. J. Climate Appl. Meteorol., 24, 1303–1310.

—, and E. Mosley-Thompson, 1995: Atlantic cyclones and the mild Siberian winters of the

1980s. Geophys. Res. Lett., 22, 799–802.

—, and M. J. McHugh, 2002: On the separability of the North Atlantic Oscillation and the Arctic

Oscillation. Clim. Dyn., 19, 599–608.

Semenov, V. A., and L. Bengtsson, 2003: Modes of the wintertime Arctic temperature

variability. Geophys. Res. Lett., 30, 10.1029/2003GL017171.

framework. Geophys. Res. Lett., 27, 2569–2572.

Shindell, D., 2003: Whither Arctic climate. Science, 299, 215–216.

Stenchikov, G., and Coauthors, 2002: Arctic Oscillation response to the 1991 Mount Pinatubo

eruption: effects of the volcanic aerosols and ozone depletion. J. Geophys. Res., 107,

doi:10.1029/2002JD002090.

Stott, P. A., and Coauthors, 2001: Attribution of twentieth century temperature change to natural

and anthropogenic causes. Clim. Dyn., 17, 1–21.

Sturm, M., C. Racine, and K. Tape, 2001: Increasing shrub abundance in the Arctic. Nature, 411,

546–547.

Thompson, D. W. J., and J. M. Wallace, 1998: The Arctic Oscillation signature in the wintertime

fluctuations. Mon. Weather Rev., 105, 1527–1535.

Greenland. Smaller station numbers in italics denote stations used in the time/longitude plots but

excluded from the PCA analysis due to short records or geographic proximity.

2. Time/longitude plot of five-year averaged surface air temperature (SAT) anomalies for winter

(December–January) relative to the 1961–1990 mean for each station. Note the evidence of an

NAO interdecadal seesaw response in northern Europe, warm Siberian temperatures in the

1940s, and more hemispheric warming in the 1980s. Temperatures were generally cold before

1920 outside of northern Europe.

3. Winter pattern A is represented by the PCA analysis with records beginning in 1886, 1912,

4. Polar stereographic plots of SAT and sea level pressure anomalies for December–January

1933, 1936, and 1983–1984. The years correspond to the month of January. SAT data are from

Climate Research Unit CRU T2.0 and the SLP data are from the Trenberth set at NCAR.

5. Winter pattern B based on EOF mode 2 for data beginning in 1886 and EOF mode 3 for data

beginning in 1912. All patterns show an in phase behavior between Baffin Bay and the northern

Europe.

1750 1800 1850 1900 1950 2000

year

    



 

   



 

   