Snow depth reconstruction over last century: Trend and distribution in the Tianshan Mountains, China

Clarifying the long-term variations in snow depth is important for hydrological, meteorological and ecological implications. However, the short period of snow depth records in the mountains has made the assessment of regional long-term variation of snow depth difficult. Based on reanalysis datasets from 1901 to 2014 as well as observational data from 1961 to 2014, this study applied the artificial neural networks (ANN) method to reconstruct the historical snow depths in the Tianshan Mountains of China from 1901 to 1960. The variations in monthly snow depth were analyzed during the periods of 1901-1960, 1961-2014, and 1901-2014 as historical, observed, and overall periods, respectively. The results indicated that the reconstructed snow depths captured the long-term variation and spatial distribution in the study area. For temporal scale, increases in snow depth were detected in the southern and eastern Tianshan Mountains during all three periods. The trends in snow depth indicated an increase in the western, northern and overall Tianshan Mountains during the periods of 1901-1960 and 1961-2014, but showed a decrease during the overall period of 1901-2014. The difference in variation of snow depth trends in different temporal scales indicates that the time scale of snow depth increase is decadal rather than centennial. For spatial scale, higher values of snow depth occurred in the western and northern Tianshan Mountains, while lower ones appeared in the southern and eastern Tianshan Mountains. In addition, the increasing trend in the reconstructed snow depth was more profound with increased elevation in the Tianshan Mountains during 1901-1960, with the smallest increasing rate at the elevation of 1000 m-1500 m. Although some uncertainties exist within the reconstruction, this work proposed a method for developing historical snow depth for observation-limited areas, which provided additional data for hydrological and ecological simulation. The results also allow us to achieve a better understanding of regional climate change.