Elevation-dependent trends in peak snowpack amount and timing illustrate emerging snow water resource changes: a case study from the Kashkadarya River, Uzbekistan

Mountainous regions contribute disproportionately to streamflow, particularly in arid regions such as Central Asia, and may be more susceptible to climate change with implications for downstream water resource development. The Kashkadarya is a regionally important river located in the Republic of Uzbekistan and within the Amu Darya watershed. The hydrology of the Kashkadarya is dominated by snowmelt generated from the headwaters in the western Pamir-Alai Mountains. The watershed has an elevation range of 404 m to 4,332 m and is data-scarce, particularly at high elevations, with only three of eight weather stations in the watershed above 2,000 m and no weather stations above 2,700 m.  This investigation presents a case study to understand trends and predictability in snow-water resources in the Kashkadarya watershed above Qarshi, Uzbekistan (11,344 km²). We developed a high-resolution (100 m), long-term (1950–2023, 73 water years) snow water equivalent (SWE) dataset using a physics-based snow model (SnowModel) forced with the ERA5-Land meteorology reanalysis. To improve the SnowModel simulation, local station-derived air temperature and precipitation lapse rates were used with a spatial precipitation correction grid. The spatial precipitation correction grid was generated by comparing snow persistence, the long-term average of percent snow covered days from January 1 – July 1, from an initial SnowModel simulation to observed MODIS cloud-gap-filled snow persistence. These modifications in the model improved mean Kling–Gupta Efficiency (KGE) of simulated and observed SWE time series at the three high-elevation weather stations from 0.44 (default model configuration) to 0.64 (model configuration with local lapse rates and precipitation grid correction). Watershed wide nonparametric Mann–Kendall trends in annual peak SWE amount and timing were not present; however, some decreasing mean peak SWE trends were present in 200 m elevation bands between 1,100–1,500 m (mean Sen’s slope = -0.35 cm/decade, mean p-value < 0.05). Peak SWE timing trends illustrates broader changes in the watershed with earlier mean day of water year of peak SWE from 1,300–2,900 m and 3,100–3,500 m (mean Sen’s slope = -1.62 days/decade, mean p-value < 0.05). To understand the predictability of the mountain snowpack in the watershed, first of the month mean SWE values for each elevation band will be compared to teleconnection indices (e.g., the Pacific Decadal Oscillation) and other variables (e.g., preceding precipitation and air temperature) as well as streamflow measurements in the watershed. These results suggest that while the volume of snow water resources remain stable in the high elevations of the watershed, the timing of snowmelt is shifting in the mid- to high-elevation portions of the watershed, portending changes to the melt dynamics in the most hydrologically productive areas of the watershed.