Seasonal And Regional Characteristics Of Ice Extent Trends In The Twentieth Century

In the past, scientists focused mainly on comparatively short-term changes (2-3, 5-7 years) in ice extent, while in recent years there has been increasing interest in investigating long-term variability (10 years and more), which reflect changes in Earth's climate. Detailed information about these changes is contained in the monograph by Zakharov (1996).

Plotting twentieth century ice extent changes in the Arctic Seas reveals a gradual decrease in ice extent from the beginning to the end of the century. These changes can be expressed by a linear trend, whose parameter (inclination of the straight line) is derived by a least-squares procedure. Using the value of this parameter and the series length, one can analytically derive the variance value (measure of ice extent scattering) described by a linear trend. The formula1 for determining this variance is

^2 = ^ , where a] is the variance fraction, described by a linear trend; a is the trend parameter (km2/year); and n is the series length (years).

Table 2.1 presents statistical data for each of the seas discussed above and their combinations, allowing comprehensive estimation of linear trends that characterize ice extent changes in the study region during the twentieth century.

Table 2.1 shows that the most significant twentieth-century changes in Arctic Ocean ice extent mainly occurred as decreases and increases at the boundary with the Atlantic Ocean—as was noticed earlier by Zakharov (1978, 1997). The largest changes described by the linear trends are evident in the ice extent of the seas of the first group (Greenland and Barents) in April (547 thousand km2). Although similar changes in August are smaller (359 thousand km2), the total changes during this period in the Greenland, Barents, and Kara Seas ("Nordic region'') are comparable to the changes in April in the first two seas. The linear trend in the Laptev, East Siberian, and Chukchi Seas is an order of magnitude smaller. Thus, the most significant linear trend is in the ice extent of the Nordic seas. Its contribution both in April and August is typically greater than 30%. The share of the linear trend to ice extent variability in the seas located to the east of Severnaya Zemlya is only between 0 and 8%.

Figures 2.2 and 2.3 plot total ice extent changes in August for three "western" seas (Greenland, Barents, and Kara) and three "eastern" seas (Laptev, East Siberian, and Chukchi) in the twentieth century and their corresponding linear trends. The linear trend of ice extent for the western seas is much greater (4.3 times) compared to the eastern seas. One characteristic of the western seas is important: the decrease in

1 Equation 2.1 is obtained by a simple integration for the ice extent L variance o"2 = (1/n) J" i (L — L)2 dt by assuming that initial time moment t is located at the center of r n/2t Jn/2(

the series so that if L — at + b, we get L — b, (L — L)= at and o\ = (1/w) J"Wß(at)2 dt.

Table 2.1. Values of sea areas (S, thousand km2); average ice extent (L, thousand km2); ice extent changes for 100 years (1900-2000) (AL100); variance (ct2 x 106 km4), described by a linear trend; variance of ice extent series (ct2 x 106 km4); and their ratios (ct2/ct2)

Seas

S

L

AL100

AL100/L

AL100/S

a2

aj/a2

GS (IV)

1087

627

-303

0.48

0.28

10758

19304

0.56

BS (IV)

1388

857

-245

0.29

0.18

7645

23862

0.32

GS + BS (IV)

2475

1484

-547

0.37

0.22

24976

58714

0.43

GS (VIII)

1087

356

-95

0.27

0.09

759

6928

0.11

BS (VIII)

1388

201

-264

1.31

0.19

5802

17440

0.33

GS + BS (VIII)

2475

557

-359

0.64

0.15

10759

35517

0.30

KS (VIII)

830

444

-153

0.34

0.18

1959

23633

0.08

LS (VIII)

536

282

-38

0.13

0.07

122

9845

0.01

KS + LS (VIII)

1366

726

-192

0.26

0.14

3061

40184

0.08

ESS (VIII)

770

612

-37

0.06

0.05

113

11677

0.001

CS (VIII)

372

135

-45

0.33

0.12

171

2139

0.08

GS + BS + KS (VIII)

3305

1001

-505

0.50

0.15

21252

87166

0.24

LS + ESS + CS (VIII)

1678

1029

-120

0.12

0.07

1200

44158

0.03

Note: GS—Greenland Sea, BS—Barents Sea, KS—Kara Sea, LS—Laptev Sea, ESS—East Siberian Sea, CS— Chukchi Sea; IV, VIII—months of April and August

Note: GS—Greenland Sea, BS—Barents Sea, KS—Kara Sea, LS—Laptev Sea, ESS—East Siberian Sea, CS— Chukchi Sea; IV, VIII—months of April and August ice extent for the first half of the century was much more rapid than during the second half of the century. The formal calculation of parameters shows almost a sevenfold decrease in their value from the first half of the century to the second half.

Note that the linear trend parameter strongly depends on the time interval for which the trend is determined. However, as Gudkovich and Kovalev (2002b) show, in the presence of cyclic variability whose period is comparable to the time interval (see Sections 2.3-2.5), an assessment of the independent linear trend strongly depends not only on the series length but also on the choice of the starting point relative to the phase of cyclic variability. Ignoring this fact can lead to detection of a false trend or a strong distortion of the trend's value. The distortions are especially large in assessments of the trends for a time close to a cycle half-period. In such cases, they express a linear approximation of cyclic variability, which can be more strictly expressed by trigonometric functions (for example, a sinusoid). Even when we use sufficiently long time series, the linear trend can indicate both a unidirectional change in time and cyclic changes with periods exceeding the series length.

Figure 2.2. Linear trends in the changes in total ice extent in the Greenland, Barents, and Kara Seas for the period 1900-2003 (August). The inset shows the corresponding equations for linear trends: 1) 1900-2003, 2) 1900-1969, 3) 1945-2003.
Figure 2.3. Linear trends in changes in total ice extent for the Laptev, East Siberian, and Chukchi Seas for the period 1900-2003 (August). The inset shows the equation for linear trends.

Errors in estimating the trend will be minimal if the trend parameter is calculated for a time interval between the neighboring maxima and minima of the most energy-intensive cyclic variability. For western region ice extent, 1900-1969 and 1945-2003 can be assumed to be such time intervals. Nevertheless, in this case the values of the corresponding linear trend parameters (—7.06 and —1.63) differ by more than four times.

Changes in the ice extent trend of the seas of the North European Basin in April for periods lasting for decades exhibited the same features; however, the differences in the trend values for the first and second halves of the century were much smaller. The trend of the ice extent in the eastern seas (Figure 2.3) is quite small and is not significant, as will be shown below. Its small increase in the second half of the century is determined exclusively by a large negative anomaly (—3.8c), noted in 1990, and therefore it should be considered random in the study of climatic change.

A slower rate of ice extent decrease with time in the Nordic Seas was detected by Vinje (2000). By analyzing the ice extent changes in these seas for the period 18641998, he concluded that the total ice extent decrease in April in this region for 135 years, expressed by a nonlinear regression equation, was 0.79 • 106 km2, or 33% of its initial value. Further, the rate of the ice extent decrease gradually deceased from 8 • 103 km2/year in 1880 to 3 • 103 km2/year in 1980. So, about 50% of the total sea ice extent decrease was observed during the four decades of the nineteenth century preceding the "period of Arctic warming."

What is the significance of these trends? We calculated the significance of the linear trends by means of standard procedures (Rozhkov, 2001; Stuart and Ord, 1994). Table 2.2 presents the estimates and confidence intervals at 95% significance level of linear ice extent trends in the Eurasian Arctic Seas. The calculations were conducted in general for the entire observation period (1900-2003), and also for two time intervals noted above for the western region seas.

Analysis of Table 2.2 shows that a negative trend in ice extent for August during the period 1900-2003 was observed for the Greenland, Barents, Kara, and Chukchi Seas as well as for the whole Eurasian Arctic. At the same time, a positive trend is not excluded but rather has a 95% probability for the Laptev and East Siberian Seas. Hence, the estimates of negative trends in ice extent for these seas are unreliable, similar to the total ice extent of the eastern region seas.

To determine the reliability of the linear trends in the western seas for the first and second halves of the twentieth century, the observation series were subdivided into two overlapping time intervals. The corresponding values are given in the lower rows of Table 2.2. These data indicate that the calculated linear trend coefficient for the first time interval is reliable, but it appears to be unreliable for the second time interval.

The seasonal and intra-secular changes in the linear trend of ice extent noted above suggest that this phenomenon should be studied in greater detail. The available observational data for the Barents Sea in the second half of the twentieth century allow us to calculate the values of the linear trend in ice extent of this sea for each month as listed in Table 2.3. As the table shows, the ice extent decrease in the Barents Sea in the second half of the twentieth century was observed only during the spring-

Table 2.2. Estimates of the linear trend coefficients (y — ax+ b) of Eurasian Arctic Seas ice extent and their confidence intervals at 95% significance level

Sea (region)

Month

Linear trend coefficient

Estimate

Lower bound

Upper bound

Observation period 1900-2003

Greenland

IV

-3.241

-4.007

-2.475

Barents

IV

-2.447

-2.905

-1.090

Greenland, Barents

IV

-6.482

-8.013

-4.951

Greenland

VIII

-0.954

-1.463

-0.446

Barents

VIII

-2.638

-3.324

-1.952

Kara

VIII

-1.533

-2.488

-0.578

Laptev

VIII

-0.381

-1.023

+0.261

East Siberian

VIII

-0.368

-1.068

+0.332

Chukchi

VIII

-0.453

-0.743

-0.163

Greenland, Barents, Kara

VIII

-5.126

-6.763

-3.488

Laptev, East Siberian and Chukchi

VIII

-1.202

-2.488

+0.084

Eurasian Arctic

VIII

-6.330

-8.367

-4.293

Observation period 1900-1969

Greenland, Barents, Kara

VIII

-7.064

-10.393

-3.735

Observation period 1945-2003

Greenland, Barents, Kara

VIII

-1.633

-5.154

+1.887

Table 2.3. Seasonal linear trend changes in ice extent for the Barents Sea for 1945-2000, thousand km2/year

Months

I

II

III

IV

V

VI

VII

VIII

IX

X

XI

XII

+3.50

+2.45

-0.41

-1.65

-1.37

-0.99

-0.80

-0.96

-1.01

+3.56

+3.11

+0.31

summer period (from March to September). During the autumn-winter period (October to February), when there is intense formation of young ice, a significant positive linear trend occurs in the variability of ice extent (Table 2.3).

Buzin (2006) found a similar trend for the Barents Sea and its northeastern sector for 1928-2003. It should be noted that the linear change in mean annual ice extent was close to zero.

Ponomarev et al. (2003, 2005) detected similar behavior in seasonal changes in twentieth-century climatic trends of surface air temperature in the middle and temperate latitudes of northeast Asia. Significant warming in winter was accompanied by noticeable cooling in summer. So, the seasonal changes in the trends of this parameter were opposite to those observed in the Barents Sea. The causes of these anomalies require further investigation.

Was this article helpful?

0 0

Post a comment