*, %, # - regressions significant at the levels of 0.05, 0.01, and 0.001, respectively, ## - mean from 27 Arctic stations, ### - mean air temperature for 65°N-85°N over the period 1951— 1986 (after Alekseev and Svyashchennikov 1991), & - Northern Hemisphere air temperature anomalies after Jones (1994)

Apart from the lines of regression, the curves of confidence have also been drawn for seasonal and annual T. of the Arctic and for BAFR, which is the only region characterised by a statistically significant change of T in the period examined (Figures 5.22 and 5.23). The coordinates of the lower and upper curves of confidence have been calculated using the following formula:

where:

y(i) - ordinates of curves of confidence +(upper) and -(lower),

T(i) - ordinates of the line of regression, ta - the value of Student's t statistic for the coefficient of confidence 0.95

and n-2 degrees of freedom,

Sr - averaged deviation from the line of regression, x - time, i - consecutive years, i = 1, 2, 3, ... 40, n - sample size

Curves of confidence mark off the confidence region of regression, encompassing with 95% probability the theoretical lines of regression. The courses of lines of regression and their intervals of confidence provide information about the nature oftrends and the dispersion of For seasons characterised by a higher variability of the upper and lower limits of intervals are further removed from the line of regression (Figures 5.22 and 5.23). Lines of regression, together with their curves of confidence, would only be useful in forecasting climate if the trends determined were statistically significant and if the variability of climatic conditions in the period for which the forecast is prepared were caused to the same degree and by the same factors. The last condition, however, cannot be guaranteed. Moreover, in the case of most of the data series analysed (Table 5.12, Figure 5.22), the slopes of their equations of regression can, within the confidence region determined, change within a wide range, including a change of sign (±). It is only in BAFR that a further decrease in T., and especially in spring, summer, and annual can be expected with quite a high probability in the near future, provided that the second condition is preserved (Figure 5.23).

Between 1961 and 1990, the course of trends changed significantly (Table 5.11, Figures 5.14 to 5.19) with positive trends clearly dominating in the Arctic. As mentioned above, this is connected mostly with the marked cooling observed in the Arctic and in the whole Northern Hemisphere in the 1960s. The trends of mean annual 7] of the Arctic were 0.15°C/10 years (TAj) and 0.34°C/10 years (T ).

1950 1960 1970 1960 1990

Figure 5.22. Lines of regression and curves of confidence at a = 0.95 for seasonal and annual Arctic r(27 stations) over the period 1951-1990.

1950 1960 1970 1960 1990

Figure 5.22. Lines of regression and curves of confidence at a = 0.95 for seasonal and annual Arctic r(27 stations) over the period 1951-1990.

A stronger and statistically significant trend in the zone 65-85°N is undoubtedly connected with the discernible warming which occurred in the continental Subarctic (Chapman & Walsh 1993). As follows from the data in Table 5.11 and Figures 5.14-5.19, the warming did not affect all regions of the Arctic. It was the strongest in ATLR (0.31°C/10 years) and in SIBR (0.32°C/ 10 years), where the trend was statistically significant at the level of 0.05. Mean annual in the Canadian Arctic did not manifest significant changes at that time. Similar results were obtained by Findlay et al. (1994b) for the region of the Canadian tundra.

BAFR, on the other hand, was characterised by a manifest and significant cooling (-0.50°C/10 years), similar to the long-term period between 1951 and 1990. Seasonal mean T. for particular regions, with the exception of CANR, behave similarly to annual temperatures. The greatest changes of occurred in the winter and spring (noted also by Chapman & Walsh 1993), whereas the greatest changes in CANR occurred in the summer and autumn. At that time an increase of in the warm half-year and its decrease in the cold half-year were observed in the Canadian Arctic (Table 5.11). It is worth adding that in the 30-year period under examination, a cooling affecting a large area of the Arctic was noted only in the autumn. Out of the 29 stations ana lysed, 15 were characterised by a negative trend. Areas characterised by a decrease in T. in the autumn were found in all the regions discussed; however, only CANR and BAFR were affected by this process in their entirety. Apart from the autumn, the greatest areas where decreased were noted in the winter. This is responsible for the fact that the trend of mean in the winter amounted only to 0.12°C/10 years and was smaller than its spring trend (0.30°C/10 years) and even its summer trend (0.13°C/10 years). It follows that the most significant warming in the Arctic between 1961 and 1990 occurred in the spring and summer and not in the winter and spring as Chapman and Walsh assert (1993). If, however, the Subarctic continental areas are taken into consideration, then indeed the greatest warming, in absolute values, occurred in the winter and spring, but a statistically significant trend at the level of 0.01 was noted only in the summer (0.28°C/10 years) (cf. Table 5.11).

Was this article helpful?

## Post a comment