5.3 Atmospheric circulation indices and snow cover depletion
Our correlation analysis shows the AO and SH to be the circulation patterns most influential to patterns of snow cover depletion in the Upper Irtysh basin in winter and spring. The correlations between winter AO anomalies and DPSCD are significant below 1500-2000 m a.s.l.: notably, DPSCD shows higher correlation with the winter AO at higher elevation, while the opposite is true for the SH, and a similar inverse pattern is seen for DSCD.
Two winters stand out as highly anomalous in both the AO and SH record during the period of observation: 2007-2008 (positive) and 2009-2010 (negative), with matching winter temperature anomalies and patterns of subsequent spring snow cover depletion (Fig. 7). Winter 2009-2010 was in fact remarkably cold across Europe and central Asia. According to Cohen et al. (2010), the highly negative AO was initiated by the rapid advance of Siberian snow cover in October, followed by the strengthening of the SH, increased upward wave activity flux, stratospheric warming and downward propagation of height and wind anomalies to the surface. The link between autumn snow cover and phases of the AO has been reported in several studies, with feedback mechanisms including soil moisture and the albedo (Bojariu & Gimeno, 2003, Cohen & Entekhabi, 1999, Saito & Cohen, 2003); patterns of decline in Arctic sea ice have been linked to the AO (Liu et al., 2012). Compared to 2009-2010, the winter of 2007-2008 received little attention, although unusually mild temperatures were reported across Eurasia, linked with the positive phase of the NAO/AO (Ilkka et al., 2012), in addition to a La Niña event (Stevenson, 2016).
In our study, the winter AO AO is strongly negatively correlated with DPSCD and appears to influence spring snow cover retreat through the modulation of spring temperature anomalies. The lagged influence between the AO and spring snow cover has in fact been previously reported in the literature. Bamzai (2003) observed a correlation of 0.46 between JFM AO and AM snow cover melt date over Eurasia. Foster et al. (2013) found that the winter AO is negatively correlated with spring snowmelt and DSCD at latitudes between 50º and 60º N, while the opposite is true between 60º and 70º N. In Saito & Cohen (2003), evidence for reciprocal lagged effects was found, i.e. late winter/spring AO leading spring/summer snow and spring/ summer snow leading the following fall/early winter AO. In a subsequent study, Saito et al. (2004), found a consistent relationship between fall Eurasian snow cover and the winter AO between 1971 and 2000. During the 1980s, an association between winter AO and spring snow cover also emerged, as well as autocorrelation between winter and spring snow cover and winter/spring AO. This latter correlation was also reported by Foster et al. (2013) but found no confirmation in our study. While our analysis did not include fall snow cover, its possible direct influence on winter AO and spring snow cover retreat is appealing for long-range forecasts and should be discussed in future research, both in the Upper Irtysh and other Eurasian catchments.
Compared to the NAO, the winter AO exhibits a stronger connection with both winter/spring temperature anomalies and DPSCR/DSCD. While the similarity/difference between the AO and the NAO and the actual physical meaning of the AO are still a matter of debate (Báez et al., 2013), our study confirms the hypothesis of a weaker (Saito & Cohen, 2003) and more local effect of the NAO on snow cover variability, restricted to Europe and western Asia (Clark et al., 1999). The observed association between the spring SH, temperature anomalies and DPSCD, while confirming that the Siberian high is indeed a relevant atmospheric influence on the climate of Kazakhstan (Panagiotopoulos et al., 2005), also show that it holds a limited potential in hydrological forecasting, as no relevant lagged correlations were found in our study. In spite of evidence that a positive SH reinforces negative AO anomalies (Cohen & Entekhabi, 1999, Cohen et al., 2010), the interplay between the AO and the SH is not entirely understood (Huang et al., 2016), and whether one index can be predicted from the other still needs to be assessed.