6 Conclusion
In this study, we explored multi-annual (2000-2017) snow cover variability in the upper course of the Irtysh, a large transboundary river basin of Central Asia, which is significantly impacted by spring snowmelt floods. Given the scarcity of direct runoff observations in the region, we analysed snow covered area as a means to understand the seasonal and inter-annual variability of snow hydrology, to provide valuable information in the context of flood forecasting. The analysis was carried out using 8-day MODIS snow cover composites, both within the five major tributary basins and the Upper Irtysh as a whole. Cloud filtered data were used to investigate: 1) the timing of snow cover disappearance, using the day of year when snow cover disappears from different 500 m elevation ranges within each basin (DSCD), 2) peak snow cover depletion, the day of year of maximum areal snow cover depletion in a basin (DPSCD) and 3) peak snow cover depletion rate in each basin in each year (PSCDR expressed in km2d-1). In addition, we employed ERA-Interim reanalysis data, local weather stations and atmospheric circulation indices to investigate potential meteorological drivers of the observed snow cover depletion patterns.
The analysis of snow cover variability in the Upper Irtysh basin points to large inter-annual and inter-basin differences, for both DSCD and DPSCD, with the largest, low-lying basin showing the earliest dates and providing the largest snowmelt contribution to runoff. DPSCD generally occurs between mid-March and April, depending on the basin, while DSCD occurs later during spring. No clear trends over the 2000-2017 study period were identified in any basin for either variable. In contrast, PSCDR, which peaked at over 5900 km2day-1 in 2017, displays a weak increasing trend for the Upper Irtysh basin over the same period. In view of their relevance for flood hazards, the existence of longer-term trends, as well as the causes of the increase in PSCDR, should be addressed in further research.
Multi-year fluctuations in peak snow cover depletion and snow cover disappearance are controlled by spring temperature anomalies and regional atmospheric patterns detected by weather stations and reanalysis data in the Upper Irtysh basin. The atmospheric patterns that best correlate with temperatures and snow cover variables are the winter Arctic Oscillation and spring Siberian High indices. Among the two, the winter Arctic Oscillation shows the strongest correlation with DPSCD, especially in the larger, low-lying basins (-0.71), while lower correlations are seen in the smaller mountain basins, where topographic variability likely plays an important role.
The ability of the winter Arctic Oscillation index to predict a relatively late or early timing of the spring peak snow cover depletion rate in parts of Eurasia makes it a potentially useful tool for long-term forecasts, and for use in early-warning flood prevention schemes. However, the lack of empirical runoff data in the Upper Irtysh basin complicates validation of the feasibility and effectiveness of this approach. Despite the lack of validation, our approach is based on readily available data and can therefore be applied to similar catchments where snowmelt is the dominant source of runoff, providing meaningful insights into the interactions between large scale snow cover depletion patterns and streamflow, for informing water management strategies. A denser network of ground based observations or different remote sensing approaches could be employed to further assess the influence of the Arctic Oscillation in modulating snow cover depletion at the catchment scale.