An interesting observation is that the weakest mean negative trends occurred in the winter and summer and, what is more, in the winter they encompassed the smallest area. It must be added, though, that in this season the greatest diversification of the magnitude of trends was also noted (Table 5.11, Figure 5.21). They underwent evident cooling only in ATLR (especially in the area of the Barents Sea, where the trend was less than -0.6°C/10 years) and in BAFR, especially in its southern part (ca. -0.4°C/10 years). At that time the greatest positive upward tendency of winter occurred in Alaska (0.6°C/10 years), In the summer, negative trends were noted in the area encompassing large parts of SIBR, CANR, the region of the Arctic Ocean (IARCR) and the whole area of BAFR. The area where the downward tendency in was noted was slightly bigger in the spring than in the summer. It affected the whole area of BAFR and most areas of CANR, PACR, and IARCR. The remaining area of the Arctic was characterised predominantly by an increase in T. in the period under examination (Figure 5.21).

Between 1951 and 1990 all seasonal and annual air temperatures in the Northern Hemisphere (TNff) were characterised by slight positive trends. In the case of the series also incorporating the Sea-Surface Temperature (SST), the trends are statistically insignificant (except for the spring). More evident, on the other hand, is the warming of continents (Table 5.11). The mean temperature of the 200-metre surface layer of water along a profile through the Barents Sea is characterised by a negative trend which is slightly weaker than the mean The greatest cooling of water occurred in the winter, which is in conformance with the behaviour of T in ATLR. These facts correlate very well with the sea-ice cover of the Barents Sea, which in the winter showed the greatest statistically significant increase (3.28%/10 years).

Equations of the regressions of winter, summer, and annual T. have been formulated for 9 selected stations representing particular climatic regions and sub-regions, for all climatic regions of the Arctic, and for the Arctic and the Northern Hemisphere (Table 5.12). The following formula has been used:

In order to estimate statistical significance of the results obtained, the lower and upper limits of confidence intervals of the linear coefficients of regression were stated, assuming the confidence level a = 0.95. The calculations were made using the formula published by Kozuchowski (1985):

t - value of Student's t statistic for n - 2 degrees of freedom and coefficient of confidence

T = ax + b where: Sr - average deviation from line of regression,

Additionally, the standard error of the dependent variable T and statistic t have been calculated in order to determine the probability of the occurrence of a null hypothesis that there is no trend in the series:

where: aT - standard deviation of T variable r,. - correlation coefficient of the variables x and T

In order to explain the share of the linear trend in the general variability of T, appropriate calculations have been made, the results of which are presented in Table 5.12. An analysis of Tables 5.11 and 5.12 shows that for the majority of the series analysed, T. is characterised by statistically insignificant trends. In the winter significant positive trends occurred only in stations Barrow and Coppermine; in the spring significant negative trend has been calculated only for BAFR (-0.32°C/10 years). The magnitude of 40-year changes of T. in this region lies, with a 95% probability, within the range from-2.4 °C to -0.2°C. A significant downward tendency for the whole Arctic (Taj) was observed only in the autumn (-0.18°C/10 years), and for the series from the area between 65-85°N (TA2) the significant increase of T. turned out to be that for the summer (0.19°C/10 years) (cf. Table 5.11). As far as the area of the Northern Hemisphere is concerned, as has already been mentioned, only T. over the continents was characterised by a significant increase at that time. The remaining coefficients of linear regression are so small (Table 5.12) that there is no basis for considering them different from zero.

The share of linear trends in the general variability of T. is, in most cases, very slight and approaching zero (Table 5.12). Only the statistically significant trends explain 10-20% of its variance.

Was this article helpful?

## Post a comment