4.1 Multiannual snow cover variability
While there are no notably strong trends in mean DSCD in any of the elevation intervals, there appear to be large inter-annual fluctuations at the lower elevations, especially in the larger Kara Ertis basin (Fig. 3d), while very little variability is seen above 3500 m a s.l.. The most consistent patterns across the basins are the late DSCD events in 2010, best seen in Kara Ertis basin under 2000 m a.s.l., and in 2013, mostly seen in the smaller basins above 2000 m a.s.l.. Other notable features include the early DSCD in 2003, 2008 and 2012 above 2500 m a.s.l. in the smaller basins. 2008 also shows a large shift towards earlier DSCD in Kara Ertis basin between 500 and 1000 m a.s.l.. An isolated spike towards later DSCD is present in 2005 in Uba basin (see Fig. 3c), indicating snow did not completely melt during the observation period. At the lower elevations (< 1000 m a.s.l.), snow disappears on average before DOY 105 (mid-April, see Table 1), in the smaller basins and much earlier (DOY 73, third week of March) in Kara Ertis basin. At the upper elevations (3500-4000 m a.s.l.), it either disappears at the end of June in Kara Ertis basin, or does not melt at all during the observation period (until the end of July) in Bukhtarma basin (see Fig. 3b). Between 2500 and 3000 m a.s.l., snow disappears on average in May in Kara Ertis basin and early June in the smaller basins, with the exception of Uba basin in 2005.
In comparison with DSCD, larger inter-annual variability can be seen at all elevation ranges for DPSCD (Fig. 4). However, the patterns of DPSCD for entire basins seem to closely match those of their 500-1000 m elevation range, with the exception of Kurchum basin (Fig. 4a) where the basin-wide values lie between those of two elevation ranges, due to the different hypsometry of this basin (see Fig. 2). As with DSCD, late DPSCD can be seen in 2010 in Bukhtarma, Kara Ertis and Narym (see Fig. 4b,d,e) basins at most elevation ranges, while 2008 shows earlier DPSCD mainly at the lower elevations of Kara Ertis (< 2000 m a.s.l.) and Bukhtarma (< 1000 m a.s.l.) basins. The different elevation ranges generally show little cross-correlation, with several contrasting values of early/late peak snow depletion, e.g. between 2500-3000 m and 3500-4000 m in Kara Ertis basin and 1500-2000 m and 2500-3000 m in Uba basin (see Fig. 4c). In the smaller basins, DPSCD generally occurs in April and on average, peak snow cover depletion occurs in early April in Bukhtarma and Uba (DOY 97) basins, and during the first week of April (DOY 89) in Narym basin. In the whole Upper Irtysh basin, DPSCD is synchronous with snow depletion in Kara Ertis basin except for 2010, when it occurred a week later.
PSCDR itself is largely controlled by snow retreat between 500 and 1500 m a.s.l., with smaller contributions from the other elevation ranges, and is up to 8 times larger in Kara Ertis basin, due to its greater size. For this reason, we here show the basin-wide values, not individual elevation ranges, together with the global value for the Upper Irtysh catchment (see Fig. 5). PSCDR for Upper Irtysh closely resembles the Kara Ertis basin series, with the exception of 2010, when there was a large drop in the Kara Ertis basin PSCDR. PSCDR maxima occurred in 2011 and 2017 with almost 5800 and over 5900 km2 day-1, respectively, with the lowest amount in 2007 (3160 km2day-1), and an average of over 4500 km2 day-1. We evaluated trends for each basin and the whole Upper Irtysh using both ordinary least-squares linear regression and the Mann-Kendall test. A significant (p < 0.05) positive trend of +67 km2day-1 per year was identified for the Upper Irtysh basin. However, the trend is only significant at the 90% confidence level when the Mann-Kendall test is used. No other significant trend is evident for individual basins.