Surface air temperature from meteorological data

Introduction go
References I go

1. Analysis of long-term air temperature observations in the Arctic go

1.1.Initial data go

1.3.1. Temperature changes from the longest observation series data go
1.3.2. Frequency structure of air temperature variability go

1.4.1. Estimates of temperature changes during the periods of warming and cooling go
1.4.3. Comparison of air temperature anomalies during the two warmest decades go

References II go

Introduction

Arctic climate is directly connected with climate of the rest of the Earth differing however from it in many respects. The main cause of this difference is related to a latitudinal decrease of the solar heat flux whose influence is attenuated due to internal thermodynamic processes occurring both in the entire climatic system and in its polar area. The poleward heat advection as a result of the atmosphere and ocean circulation makes the largest contribution to warming of arctic climate compared to climate that could be observed under the stationary atmosphere and ocean conditions. Due to these processes, Arctic climate “warms” with the increase of surface air temperature by several tens of degrees compared to climate in the absence of advection, which is much greater than the green house effect proper in the polar atmosphere /Alekseyev et al., 1991/.

The main heat advection to the Arctic occurs in the atmosphere. The portion of oceanic transfer on average for a year is estimated for the entire Northern polar area as 3-5% of the total advection in the atmosphere-ocean system /Budyko, 1969, Khrol, 1992/. The oceanic processes, however, also significantly influence the arctic climate formation, which is not restricted to the direct oceanic heat contribution to the heat balance of the Arctic.

The atmospheric component of the arctic climatic system is most exposed to the influence of the global energy-mass exchange oscillations related to instability of atmospheric circulation and its response to external forcing. Sensitivity of the arctic atmosphere and its dynamics to forcing from the Arctic Ocean and sea ice is restricted to local responses, as revealed from the numerical experiments /Kattsov et al., 1997/. Much of the interannual variability in the polar atmosphere is due to general atmosphere oscillations responsible for the heat transfer variations from warmer oceanic and lower latitudes to polar areas. These predominantly short-period oscillations generate a significant low-frequency weather noise in the climatic variability of atmospheric characteristics complicating a detection of signals of different origin /Alekseyev, Svyashchennikov, 1991/. The idea of increased atmospheric circulation and heat transfer to the Arctic as the main cause of arctic climate warming in the 1920s-1930s was substantiated by O.Yu.Viese /Viese, 1937, 1940/ and supported by other scientists /Dzerdzeyevsky, 1943, Vittels, 1946/. The development of the current Arctic warming is also related to the changes in the atmospheric circulation regime in the Arctic and over the adjoining latitudes of the Northern Hemisphere /Thompson and Wallace, 1998’ Johnson et al., 1999/.

At the background of strong natural oscillations of atmospheric characteristics of Arctic climate, its increased response to the man-made gaseous chemistry changes develops according to prediction of coupled atmosphere-ocean general circulation models. Although the models predict a very strong arctic climate response to the increased concentrations of carbon dioxide and other anthropogenic gases, an analysis of climatic data collected in the Arctic over the instrumental observation period, has not revealed up to now any unambiguous tendencies towards strong warming for the last twenty years. This discrepancy between the theoretical and empirical estimates can be eliminated if based both on improving the models and adequate allowance of polar processes and a detailed analysis of climatic archives using new ideas, methods, models and data. In this respect, a better insight to the climatic variability mechanisms connected with the arctic component of the climatic system is of importance. One of these mechanisms is transformation of the heat flux to the oceanic component of the arctic climatic system that includes summer processes of snow and sea ice melting, upper water layer heating and continental fresh water flow and liquid precipitation. In winter, this heat input to the atmosphere occurs predominantly due to the water-sea ice phase transition, which is compensated on average during the year by sea ice export outside the Arctic Ocean. This interseasonal “transfer” of oceanic heat serves as one of the main mechanisms of increased arctic atmosphere climate response in the winter to forcing, for example to the increased CO₂ content, which is also due to the inverse relation between the albedo and the solar radiation summer influx.

The albedo of the arctic climatic system is most sensitive to the internal thermodynamic processes producing a strong inverse influence. The variable albedo component is connected with the sea ice cover extent in the Arctic Ocean changing under the influence of the summer heat flux. Some investigators assess the sea ice influence on the planetary albedo formation as about 50% and even less / Zilitinkevich, Monin, 1977/. Clouds comprise the remaining portion restricting the influence of the inverse albedo-sea ice relation.

Another variable parameter of the climatic system influencing the internal thermodynamic processes is the gaseous atmospheric composition subjected to changes due to the anthropogenic increase of CO₂ concentrations and other optically active trace gases. The effect of this factor has been most widely investigated in recent years. Most estimates of climate response to the increased CO₂ concentration in the atmosphere, obtained from numerical experiments and GCM of the atmosphere, coupled atmosphere-ocean models, as well as energy balance climate models predict significant climate warming. The mean global temperature increase is estimated between 1.5 to 4.5^o C. The distribution of warming by latitudes in the data of most model experiments is characterized by its increase towards high latitudes, especially in winter and late autumn. The cause for this simulated temperature increase in polar latitudes is primarily attributed to the influence of the inverse albedo relation with decreasing multiyear sea ice area in the Arctic Ocean.

In spite of sufficiently good agreement between the model estimates of climate response to the increased CO₂ concentrations, they cannot be considered as final due to incompleteness of models that do not take into account some important factors. In particular, incorporation of the ocean to a coupled model leads to a slower climate response, which depends on the thickness of the ocean layer considered in the model. Investigators also pay attention to the fact that models estimate the rate of global climate warming as 0.3^o C for 10 years, whereas the observed warming develops much slower. All this makes still important an objective of detecting and assessing manifestations of anthropogenic climate warming in the Arctic for two decades in the past at the background of earlier climate oscillations. Surface air temperature records at the arctic and sub-arctic meteorological stations where observations were carried out from the beginning of the 20^th century and earlier serve as a valuable information source of such oscillations.

This study aims at a comparative analysis of such records for identifying some specific features in climate changes throughout the 20^th century in the entire Arctic and its regions in different periods to assess similarity and difference of the thermal regime during the last decade and the preceding periods. The final objective is to clear up the validity of the suggestion about the green house warming manifestation in the Arctic for the last period.

Alekseyev G.V., Podgorny I. A., Khrol V. P., Svyashchennikov P.N. Climatic regime of Arctic on the boundary XX and XXI s., Part 1, 1991: L. Gidrometeoisdat, p. 4-29.

Alekseyev G.V., Svyashchennikov P.N. Natural variability of climate characteristics in northern polar region and northern hemisphere, 1991: Leningrad, Gidrometeoizdat, 159 p.

Budyko M. I. Polar ice and Climate, 1969: L. Gidrometeoisdat, 36 p.

Dzerdzeyevsky B. L. By the question of Arctic warming, 1943: Izv. AN. USSR, ser. Geophisic and geographic, ą 2, p. 60 - 69.

Khrol V.P. Atlas of the Energy Balance of the Northern Polar Area, 1992: St. Petersburg, Gidrometeoisdat, 72 p.

Kattsov V.M., Meleshko V.P., Alekseyev G.V., Matyugin V.A., Shneerov B.E., Gavrilina V.M. 1997: Sea-ice compactness effect on high-latitude atmospheric variability: Meteorology and hydrology, ą 4, p 43-54

Viese V.Yu. Climate of the sea Soviet Arctic, 1940: L-M., Glavsevmorput.

Vittels L.A. North Sea cyclones and Arctic warming, 1946: Meteorology and Hydrology, ą 5, p 32-40.

Zilitinkevich S.S., Monin A.S. Global atmosphere-ocean interaction, 1977: L., Gidrometeoisdat, 23 p.

1. Analysis of long-term air temperature observations in the Arctic

1.1.Initial data

The data archive on mean monthly surface air temperature measured at the 2 m height at meteorological stations north of 60^o N was used. The earliest observations began in 1813 with all observations encompassing the period through 1998. The database for climate monitoring in the Earth’s polar areas created at the AARI in 1991, served as a basis for the archive. The structure of the database and its description are presented in /Aleksandrov et al., 1995/. One of the main criteria for data entry to the archive was continuity of observations from their beginning through 1998.

Regular meteorological observations in the sub-Atlantic sector of the Arctic and sub-Arctic began in the 18^th century. For the European part of the region, the longest and almost continuous series of instrumental meteorological observations are available from 1810 (for example, Arkhangelsk from 1813; Bergen, Petrozavodsk from 1816, etc.). In the Asian area, long continuous observation series are much fewer. The first polar meteorological station was opened in Norway (Vardo, 1829). In the Russian Arctic, first high-latitudinal stations appeared in its European area at the end of the last century (Maly Karmakuly, 1897; Kola, 1981). Polar stations in the Asian area appeared at the beginning of this century (Dikson, 1916; Kyusyur, 1918). Rapid development of a network of polar meteorological stations in the Russian Arctic was in the 1930s-1940s. Before 1930, only 16 meteorological stations operated there, whereas in 1940 their number was 66. The situation in respect of the stations with a long observation series in the Canadian Arctic is much worse. First meteorological stations north of the Polar Circle appeared in the late 1920s (Cambridge Bay, 1929).

The data archive quality control procedure included:

• checking the station number, coordinates, height and operation period using the station certificate (for Russian stations) and WMO catalogue for foreign stations;
• checking data dates for chronological succession;
• determination of observation gaps at each station. Each gap was filled with the absence indicator;
• checking for uniformity of the observation data. There are many causes for occurrence of non-uniformity in the observation series at the stations (replacement of instruments, change in their protection and changes in the station location, observation time and methods). Unfortunately, almost in all foreign editions insufficient attention has been paid to the problems of data series non-uniformity. Sometimes they do not contain data on the changes in station location and methods of mean daily temperature calculation. In unreliable cases, the quality and uniformity of the observation series was analyzed using mean monthly temperature differences at neighboring stations, as well as the Kolmogorov’s uniformity criterion /Aivazyan S.A. et al., 1983/;
• checking for spurious peaks. Mean monthly temperature values at each station were checked for 3s criterion. In cases where the value was beyond 3s , comprehensive checking was made that included a comparison of the values at nearby stations, as well as a comparison with the other meteorological parameters.

Note that there are few series with the absence of observation gaps. In the latter case, for solving some statistical problems (for example, calculation of spectra), connected with the requirement for series continuity, along with the original data variant, a variant of the data series with gaps filled in was created. Two approaches were applied: using multiyear norms and maps of anomalies. The former was used for the period with a small density of the meteorological station network mainly for the observation series up to 1920. In this case, the missing value was replaced by a 30-year norm including the gape date. In the second approach mainly for the observation series after 1920, the maps of mean monthly air temperature anomalies for the given month were plotted and the missing value was reconstructed by interpolation using the nearby stations. The locations of the stations with longest observational period are represented on fig.1

Fig.1. The locations of the stations with longest observational period

1.3. Features of air temperature temporal variability

1.3.1. Temperature change from the longest observation series data

First understanding of the character of surface air temperature oscillations in the Arctic throughout the 20^th century is provided by time series of its monthly means obtained from measurements at the stations with the longest observation period (Fig. 1.2). This Figure presents mean temperature changes in January-February and July-August at 9 stations with the longest observation series located over the entire Arctic. These months were selected as the most representative with respect to the features of the thermal regime changes in surface atmosphere both on average for a year and at extreme annual variations. An additional justification for such a choice is an analysis of air temperature departures within a year from its mean annual variations.

The connection of air temperature changes within a year was considered using estimates of the correlation linear relation between its changes in the individual months. It is known that at normal distribution with a 95% probability, the correlation coefficient will be meaningful if it is greater than the following threshold r-values at the N-number of pairs /Brooks C.E.P., Carruthers N., 1953/:

Table 1.1 presents matrices of air temperature correlation within a year at several long series stations located in different sectors of the Northern polar area. In its sub-Atlantic sector, the stations Upernavik (western sector) and Bergen (eastern sector) were selected. In the sub-Pacific sector in its western and eastern areas, these were Markovo and Fairbanks, respectively. Significant connection of air temperature changes within a year is observed in the sub-Atlantic sector of the Northern polar area. Here, significant (at the 95% significance level) relations of the temperature changes for two-three months are observed. Closer relations exist during the cold period when a succession of changes can be preserved for 3-4 months and more while the changes in January and February and in July and August are most intercorrelated. The connection within a year reaches 6-7 months in the western area of the sub-Atlantic sector, Baffin Sea (Upernavik station) being slightly less in the eastern area, the Norway Sea comprising 3-4 months.

Anomalies of mean Jan-Feb surface air temperature (red line is 11-year running average, thick – linear trend)

Fig.1.2a

Anomalies of mean Jul-Aug surface air temperature (red line is 11-year running average, thick – linear trend)

Fig.1.2b

Table 1.1 Correlation coefficient matrices of mean monthly temperature

With eastward displacement from the Atlantic the connection within a year decreases and the succession of temperature changes from month-to-month decreases to 2-3 months near the Laptev Sea (East Siberia). No meaningful relations between the cold months of the year were observed in the sub-Pacific sector on Chukotka (Markovo station) and on Alaska (Fairbanks station). The relation of temperature changes at these stations is meaningfully correlated only in the warm months. At Fairbanks station, the relation of changes is recorded during the warm period for 3 to 4 months. Thus, data in Table 1.1 reveal significant differences in the character of closeness of relation between the air temperature changes in the sub-Atlantic and sub-Pacific sectors.

For assessing stability of relations in time within a year, the coefficients of pair correlation were calculated for the individual periods 1920-1959 and 1960-1998. In general, the correlation relations are stable by periods and variations observed are small. At the transfer from one period to another the relation sign between a pair of neighboring months was preserved in most cases with changes occurring only in the case of a non-meaningful statistical relation. During the second period (1960-1998) the correlation coefficients increased in January, March and July. In the other months, closeness of changes became less, especially in the sub-Atlantic sector.

Mean temperature changes in January-February (fig.1.2a) in the western sub-Atlantic Arctic (Upernavik, Angamassalik, Akureyry and Thorshovn) are characterized by its most significant increase from the early 1900s to the 1930s and a weaker increase during the last decade. The highest winter temperatures were recorded in the 1930s. At the stations located more eastward (Vardo, Salekhard and Kjusjur) the temperature increase in the 1920s-1940s was also noticeable but it was surpassed by the increase of the last decade. The absolute maximum winter temperature at the Kjusjur station was recorded in the 1930s whereas at two other stations in the last decade. The Markovo station located at the eastern margin of Eurasia demonstrates a different character of the winter temperature change without any trace of its increase in the 1920s-1940s at the background of a gradual rise during the entire observation period. The winter temperature changes at the Fairbanks station are close in structure to those at the western stations (for instance, Vardo) due to typical oscillations with a period of around 10-11 years. Estimates of a multi-year trend in general over the entire observation period at all stations yield a positive value varying from close to zero (Fairbanks) to almost 0.2^o C over a decade (Vardo).

The structure of summer air temperature changes (in July-August) (Fig. 1.2b) is both similar to and different from the structure of winter changes. First, the multiyear trend estimates over the entire observation period at their positive sign at all stations have significantly smaller values except for Fairbanks. At the western stations (from Upernavik to Thorshavn) the excess of summer warming in 1920-1940 over that in the last decade is even more pronounced than for winter. Both summer warmings are quite similar at the stations located eastward (from Vardo to Kjusjur). Of interest is a sufficiently high summer temperature in the 1840s-1850s at Vardo. Summer warming of the 1920s-1930s and during the last decade is quite evident at Markovo station, whereas at Fairbanks the summer temperature increases during the entire observation period.

1.3.2. Frequency structure of air temperature variability

The time series spectra were calculated by means of self-correlation function expansion to Fourier series. The time series spectra of air temperature for January and July were calculated for all 28 long-series stations with the start of observations not later than 1920 In addition, mean spectra for all stations were obtained that allow general estimation of seasonal differences and similarities in the frequency variability structure. Mean spectra both in January and July have a significant “red” component indicating a considerable contribution to dispersion of long-period oscillations with periods of more than 24 years in July and more than 12 years in January. Apart from the difference due to a large range of long-period oscillations in January compared to July, mean spectra in general have a significant similarity. However, analyzing a frequency structure of temperature oscillations in different parts of the Arctic, four areas can be delineated with typical types of spectra: East Atlantic (Scandinavian peninsula and the Barents and Kara Seas), West Atlantic (East Greenland, Iceland and Jan-Mayen Island), North-American (Alaska and West Greenland) and Siberian (East Siberia and Chukotka). Averaged spectra within the areas delineated are presented in Figure 1.3a for January and in Figure 1.3b for July. In January, all areas are characterized by a slightly increased portion of high-frequency oscillations with a period of about 2.5 years. However, the frequency structure of January oscillations also has some

Mean spectral density of monthly mean air temperature variations in different parts of the Arctic

S a)

S b)

Period T_k=48/k years conditional frequency, k

Fig.1.3

Distribution of the air temperature long-term variability contribution

a) January, b)July

a)

b)

differences. The main contribution to the variability in the East Atlantic and North-American sectors is made by long-period oscillations with a period of more than 12 years. In the other areas, the main contribution belongs to the oscillations with a period between 4 to 12 years and the spectrum type is close to the “white” noise spectrum. The air temperature spectrum in July in the Siberian sector is similar to January while it is qualitatively different in the other areas. For example, the spectra in the North-American area in both months contain a significant portion of long-period oscillations, but this contribution in July is 1.5 times as great as a similar value in January. The difference is even more significant in the Atlantic sectors. Thus, in the East Atlantic sector in July, the contribution of oscillations is uniformly distributed for all frequencies and similar to January the dominance of long-period oscillations is absent whereas in the West Atlantic sector the seasonal changes are of a reverse character. Here, a significant long-period component is observed. Figures 1.4 provide some understanding of the distribution of contribution of long-period components for all 28 stations in January and July. The spatial non-uniformity of the temperature variability is also observed in the frequency structure of variability due to the seasonal differences as well. This is suggested by the spatial distribution of the contribution of long-period oscillations greater than 24 years in the air temperature dispersion in January (Fig. 1.4a) and in July (Fig. 1.4b). As can be seen from the Figures, the main contribution of long-period oscillations of more than 40% is observed in January in the area of the Scandinavian Peninsula while in July the maximum contribution of long-period oscillations is observed in southern Greenland and on Alaska.

1.4. Analisis of the structure of surface temperature variability

The preceding analysis has revealed a number of features in the air temperature variability over the entire observation period. These are primarily two periods of warming, the first in 1920-1939 and the most pronounced second period during the last decade. It is obvious that the character of warming differs from region-to-region and from season-to-season. What are these changes and is there any relation to the causes of both warming periods, in particular to the suggested anthropogenic origin of the last warming? The results of our answer to the first question are presented below, and namely, the quantitative comparisons of the thermal regime characteristics during different time intervals of the period under consideration.

1.4.1. Estimates of temperature changes during the periods of warming and cooling

We present here the features of spatial air temperature change distribution during the periods of the first and second warming and of cooling using linear regression coefficients B x ^oC/20 years of temperature change at 28 stations for the 1920-1939 and 1940-1959 periods and B x ^oC/39 years for the 1960-1998 period for each season of the year. The seasons are presented by averaged temperatures for January-February (winter), April-May (spring), July-August (summer) and October-November (autumn). The calculated coefficients of a linear air temperature regression by stations are plotted on the charts presented in Figures 1.9-1.12.

The air temperature changes are most pronounced in the winter season. Figure 1.9 shows the main warming centers in 1920-1939 in the northwest of Greenland and in the north of West Siberia. There is a definite similarity between the winter cooling of 1920-1939 and warming in 1960-1998. They however, differ in the intensity of the temperature rise. The warming value for 20 years in warming cores comprised around 4.5^o C in the first case and 3-3.5^o in the second. It can also be seen that the main core of warming in the Eurasian sector in 1960-1998 is displaced towards temperate latitudes. It is of interest that the winter cooling in 1940-1959 also began from the northwest of Greenland mainly encompassing the Canadian sector, north of West Siberia and the Arctic coast of the Eurasian sector. The temperature decrease core located in 1920-1939 in the northwest of the Canadian sector was also preserved in 1940-1959. The total temperature decrease in this core comprised around 10^o C.

Space distribution of parameters of rectilinear trend in a winter season Bx ^îŃ /20 of years for periods 1920-1939 (a) and 1940-1959 (b) and Bx ^îŃ /39 of years for period 1960-1998 (c)

Fig.1.9

Fig.1.10

Space distribution of parameters of rectilinear trend in a summer season Bx ^îŃ /20 of years for periods 1920-1939 (a) and 1940-1959 (b) and Bx ^îŃ /39 of years for period 1960-1998 (c)

Fig.1.11

Space distribution of parameters of rectilinear trend in an autumnal season Bx ^îŃ /20 of years for periods 1920-1939 (a) and 1940-1959 (b) and Bx ^îŃ /39 of years for period 1960-1998 (c)

Fig.1.12

Spring warming was also intense in 1920-1939 (fig.1.10). The main core of this warming was also located near Greenland in the western North Atlantic. The value of this warming comprised here more than 3^o C. The temperature decrease core was located above Scandinavia and northern European Russia. There is much common in the distribution of warming and cooling cores between the spring warming in 1920-1939 and 1960-1998. However, similar to the case of winter warming, the intensity of the latter warming is less. The increase for the last period comprised around 2^o C in the main core located on Alaska. In addition, like in the winter, the area where the spring temperature increase was observed at this time was greater. The spring cooling during the 1940-1959 period did not span the entire polar area. The main cooling was observed in the Asian sector and in the western Canadian sector. A pronounced temperature decrease core was situated in West Siberia. The decrease here comprised around 4^o C. Above Scandinavia and the northern area of the North-European Basin the temperature increase at this time comprised more than 1^o C.

The summer is also characterized by the periods of warming and a period of cooling between them (fig.1.11). This season there is much less similarity between both warming periods compared to the previous seasons. The 1920-1939 warming mainly developed in the sub-Atlantic sector and in East Siberia. In 1960-1998, the main core of warming was located on Alaska. Large intensity of the temperature increase in 1920-1939 was also similar to the previous seasons. During the period of cooling the temperature decrease was predominantly observed in the Canadian and the sub-Atlantic sectors. A pronounced temperature increase core was located in West Siberia.

For autumn, the spatial distribution of the temperature increase and decrease cores in different periods is characterized by a pronounced tendency for a general temperature decrease from the beginning of the century to present (fig. 1.12). Due to this, the autumn warming is pronounced relative to other periods only in 1920-1939 when a temperature increase was recorded over much of the territory. In 1960-1998, the temperature increase is less pronounced. These two periods of warming are characterized by the location of the permanent warming area in the Asian sector. Another typical feature is persistence of the cooling area in the Canadian sector throughout the entire study period.

In summary of the seasonal distribution of the warming and cooling cores, the following main feature can be distinguished. The 1920-1939 warming was more intense than in 1960-1998. In the former case, the pronounced warming cores were located in high latitudes whereas in the latter case they were displaced southward. In general, warming was more pronounced in the winter and spring seasons.

1.4.2. Comparison between summer and winter warming

The normalized average anomalies for 20 year periods were calculated from air temperature data at 28 stations. Based on air temperature changes during the century, four periods were selected. These were 1920-1930, 1940-1959, 1960-1979 and 1980-1998 taken in such a way as to reflect the effect of warming in the 1930s and the 1980s-1990s, as well as the periods of cooling separating them. Normalizing of anomalies was as follows. The air temperature anomaly presenting the difference between the air temperature for the selected month of each year and mean value for the 1920-1998 period was divided by the root mean square deviation determined for the same period. This allows a comparative analysis of the summer and winter air temperature change characterized by a significantly different variability. After normalizing, the air temperature anomalies were averaged within each 20-year period. The character of the summer and winter air temperature change was investigated on the basis of January and July data. Table 1.2 presents the results of a comparative analysis of individual 20-year periods.

Table 1.2

Comparison of mean normalized anomalies (N.A.) of the surface air temperature for 20-year periods at 28 Arctic stations for 1920-1998

As can be seen from data in the Table 1.2, there is a significant difference in the temperature regime between the 20-year periods. Thus, for example, the 1920-1939 and 1980-1998 periods of warming were characterized by a large number of positive anomalies that were observed over a selected station network compared to the other periods. Positive anomalies were observed during the 1920-1939 period at 20 and 18 stations in January and July, respectively. The number of positive anomalies during the 1980-1998 period comprised 18 and 19, respectively. Note also the fact that even during the warmest periods no positive air temperature anomalies were observed at all stations. In this case, only the increased number of positive anomalies can be stated. These periods are also characterized by a much greater maximum positive anomaly compared to the other periods that comprised +0.5 in January and +0.8 in July, the anomaly value during both periods in July being almost twice as large as in January. However, the most significant negative anomalies reaching 0.7 were also recorded during these periods, being observed only in July, as in January the value of negative anomalies was approximately the same for all four periods. The 1920-1939 and 1980-1998 periods of warming differed in that during the last two decades at most stations the July anomalies are greater than in January, whereas the early period of warming was characterized by prevailing by value January anomalies over July anomalies at most stations. This feature indicates a different character of air temperature change during the compared periods of warming, in particular, during the last two decades the summer temperature increase relatively predominated.

1.4.3. Comparison of air temperature anomalies during the two warmest decades

As mentioned above, the air temperature changes in the Arctic are distinguished by the spatial non-uniformity. This non-uniformity is manifested in particular, in the time of occurrence of the maximum mean air temperatures over a decade during the 1920-1930 period of warming. This is indicated by the calculated running 11-year mean temperatures for each of 28 stations. Their analysis reveals that in January the maximum temperatures were recorded over a wide time interval, the middle of the warmest 11-year periods falling to the time between 1926 to 1939. For most stations, predominantly in the East-Atlantic sector the 1929 to 1939 period was the warmest. In July, the time interval of the onset of the air temperature maximum is even wider from 1926 to 1949 with the warmest period for most of the stations in the East Atlantic and the Siberian sectors falling to 1932-1942. These two periods were selected as the warmest winter and summer decades of the first warming period. The second current decade presents the 1988-1998 period. The spatial distribution of air temperature changes during these periods is illustrated by the maps of differences between mean air temperature for 1929-1939 and 1988-1998 (January) and 1932-1942, 1988-1998 (July) and mean multiyear temperature calculated over the entire 1920-1998 period (Fig. 1.13). As follows from the analysis of the Figures, no general warming is observed in January in 1929-1939 in the Arctic. In Canada and East Asia in contrast to the other regions, cooling is observed, being especially significant in Canada. During the 1988-1998 period, one can speak about warming in the individual regions. Three centers of warming (Scandinavia, East Siberia and Canada) and three centers of cooling (Greenland,

Anomalies mean surface air temperature for warmest decade in January (1929-1939, a) in July (1932-1942, c) and for decade of 1988-1998 (January, b; July, d)

Fig. 1.13

Chukotka and Alaska and the area of the Barents and Kara Seas) are identified. During the periods compared, in the area of Greenland and in the Barents and Kara Seas there are temperature change tendencies in different directions. Finally during the last decade, cooling is observed in these regions. No uniaxial tendencies towards warming are observed in July during the study periods over the entire territory of the Arctic. In those regions where positive anomalies were observed during the 1932-1942 period, negative anomalies of air temperature were recorded during the last decade. However while in 1988-1998 there was warming in January and cooling in July in Scandinavia, it was cooling in January and warming in July on Alaska.

A comparison between the temperature anomalies of the warmest decade during the 1920-1940 period of warming and the last 1988-1998 decade was extended to all seasons of the year. In this case, the 1927-1937 period was selected as the warmest decade.

Seasonal temperatures were calculated for two months, and namely January-February, April-May, July-August and October-November. The temperature over the 1961-1990 period was assumed as a norm.

Fig. 1.14 presents the fields of average air temperature anomalies over a season for two periods under consideration. In 1927-1937, the main core of winter air temperature anomalies was located near the Yamal-Taimyr region. In 1988-1998, the main core of warming was also located in the Eurasian sector of the Arctic. The difference consisted of the displacement of the warming core in the second case to temperate latitudes. In the summer season, the fields of anomalies differ only in the location of negative anomaly areas, which in 1927-1937 were in the sub-Pacific sector and in 1988-1998 in the sub-Atlantic sector.

The distribution of the fields of anomalies during the transient seasons was most interesting. In spring, an extensive core of negative air temperature anomalies was located above the Northern polar area during both periods. The location of the main core of negative anomalies near the New-Siberian Islands (Laptev and East-Siberian Seas) was also the same during both periods. In autumn, warming was most pronounced in 1927-1937 and less pronounced in 1988-1998.

Thus, a comparison of the temperature anomaly distribution during two periods under consideration revealed that warming in the first period was most pronounced in winter, summer and autumn. Winter, summer and partly autumn were the warmest during the last decade. A striking coincidence of anomalies in spring was observed in both periods. Intensity of anomalies during the winter, spring and summer seasons was approximately the same whereas in autumn it was greater in the 1927-1937 period.

Anomalies of mean surface air temperature for warmest decade (mean annual temperature)

of 1927-1937 (a) and for decade of 1988-1998 (b) in April-May

Fig.1.14

Aleksandrov Ye.I., Dementyev A.A., 1995: Surface meteorological database of polar regions and its use. In: Formation of the database on sea ice and hydrometeorology. St. Petersburg: Gidrometeoizdat, p. 67-75.

Aivazyan S.A., Yenyukov I.S., Meshalkin L.D., 1983: Applied statistics. Reference edition. Moscow:Finansy i statistika,. p. 369-375.

Meditch J.S.,1973: A survey of data smuothing for linear and nonlinear dynamic systems. Automatica, vol. 9, ą 2, p. 151ľ 162.

Kolmogorov A.N.,1941: Interpretation and extrapolation of the stationary variate process, Izv. AN USSR, ser. Mathem., vol. 5. ą 1, p. 15ľ 28.

Kalman R.E., Bucy R.S. ,1961: New results in the linear fitration theory prediction, Proceedings of the Inst. Ingin. And Mech. Ser. D , ą 1, p. 3ľ 32.

Goldman S., 1957: Theory of the information, M., Inost. lit., 446 p.

Brooks C.E.P., Carruthers N., 1953: Handbook of statistical methods in meteorology.-London,.416 p.

Parsen E., 1962: On estimation of a probability density and mode. – Ann. Math. Statist., Vol. 33, p. 1065-1076.

Conclusions

The main scientific results of these research are as follow:

Trends and features of the winter and summer temperature change at different time in different sites of the Arctic were estimated. Trends are positive at all sites with long-term observation in the Arctic started before 1900. Trends for 1920-1998 are negative in the Greenland, Canada and the Nordic Seas region.

Significant correlation between the mean temperature change in winter in six uniform Arctic areas with respect to the temperature regime was established. Negative correlation between temperature changes in the Canadian-Greenland and Eurasian areas (r=-42) and in the Nordic Seas and Chukotka/Alaska (r= -0.24) is stronger for the oscillations with 6-7 year periods.

Based on spectral analysis, four types of spectra were revealed in the frequency structure of temperature oscillations ranging between “red to white noise”;

The air temperature increases on average in the Arctic from data of 28 stations, this increase being especially pronounced after 1966-1967.

A comparison of the characteristics of warming in the Arctic in 1920-1940 and during the last 1989-1998 decade was performed. The location of the main warming cores in time and space was established during both periods that alternate with the relative cooling areas.

In 1920-1940, warming extended to the sub-Atlantic area of the Arctic from West Greenland to the Laptev Sea. It was stronger east of zero meridian in winter and west of it in summer. In 1989-1998, warming was stronger in the American-Asian sector of the Arctic and relatively stronger in summer. It was found that in spring (April-May) the temperature over much of the Arctic decreased during both periods.