5.2 Variability and trends in snow cover depletion
In this study, we found no significant trend but strong fluctuations in DSCD and DPSCD in all sub-basins of the Upper Irtysh between 2000 and 2017. Both variables appear strongly negatively associated with spring temperature anomalies at elevations below 1500 m a.s.l. (see Table 4), which represent the largest contributing areas to runoff for most basins (see Table 2), with the highest correlation in the largest sub-basin of Kara Ertis. Above these altitudes, temperature variations are likely less relevant, and other topographic features such as aspect. shading, snow redistribution by avalanching or wind may dominate over the influences of regional climatic variability. Fluctuations in DSCD agree with the findings of Dietz et al. (2011) in their study of snow cover in central Asia using MODIS data, with matching years of relevant snow cover disappearance anomalies (e.g. 2008, 2010). High multiannual variability in Kazakhstan snow cover during the melting season was also found by Mashtayeva et al. (2016) although snow depth was the variable investigated. Considering longer periods of observation, Dai & Che (2014) found a significant decrease in late spring snow depth and snow cover duration (-1 day/year-1) in the Chinese Altay between 1987 and 2011, while in the Ob basin, Yang et al. (2003) identified an upward trend in spring snow cover between 1966 and 1999, caused by increased precipitation over Western Siberia. Long-term temperature records between 1966 and 2015 from southern Altay show warming trends in winter and spring, with the greatest rise in spring (1.15°C per decade at Fuyun, Zhang et al., 2018). Warming in average annual temperatures were also reported by Hu et al. (2014) using several reanalysis datasets between 1979 and 2011. If the warming trend continues as projected, a shift towards earlier snow cover disappearance and peak snow cover depletion would be expected to occur in the Upper Irtysh and across Eurasia. The existence of longer-term trends for the study area might be investigated using AVHRR data, albeit with lower spatial accuracy (Dietz et al., 2014).
In this study, peak snow cover depletion rates were not significantly correlated with the timing (DPSCD) in any sub-basin. PSCDR are also uncorrelated with the onset of melting conditions at weather stations, including rates and magnitudes of temperature increase at crossing of the 0º C isotherm. A significant correlation (0.50) was found between PSCDR and snow cover area on DOY 57 in late winter (see Table 1), which can represent a proxy for maximum SCA for a given year. Late winter snow cover explains 25% of the variance in PSCDR, but it shows no significant trends between 2000 and 2017, thus it does not appear to be the cause of the observed increase in PSCDR (Fig. 5), which remains unresolved. Among the sub-basins, Narym and Kurchum show the highest multi-annual variability; large peaks are also seen in Bukhtarma (2001 and 2010) and Uba (2000 and 2011) basins, with potential implications for local flood risks and water availability for reservoir operation.