Abstract
• Centenary change of Bosten Lake water storage was reconstructed by machine learning. • The lake water storage underwent six dramatical fluctuations over the past century. • The annual salinity changes were affected mainly by vapor pressure and precipitation. • Lake salinities showed substantial spatial and intra-annual variability. In the context of accelerating climate changes and economic boom in the past decades, lakes have undergone drastic changes worldwide. Particularly in arid and semi-arid regions, those lakes were severely impacted and threatened due to their hydrologic sensitivity and ecological vulnerability. As the largest lake in Northwest China with arid climate, Bosten Lake provides precious water resources and ecosystem services to local communities. Although a large quantity of earlier efforts have been paid on Bosten Lake, there is still absence of tracking its changing trajectory at a long timescale (e.g., one century), which restricts the holistic understanding of the decadal periodic lake desiccations and their driving forces. This study employs a machine learning method to reconstruct the centenary covariations of water storage and salinity of Bosten Lake by integrating multi-source data. The results showed that, compared with the high stage of 6.76 × 10 9 m 3 in 1961, the lake water storage substantially dropped twice to 3.96 × 10 9 m 3 in 1987 and 4.67 × 10 9 m 3 in 2013. In recent years, the lake level rose rapidly and recovered back to the comparable stage of the 1960 s by 2020. Four metrics of accuracy evaluation employed in this study indicate the reliability of the XGBoost model, with the mean absolute error of 0.31 m, mean squared error of 0.37 m, r-square of 0.85, and adjusted r-square of 0.84. The centenary reconstruction results reveal that the lake salinity underwent six-phase fluctuations with the water level and storage changes during 1920–2020, with the highest value of 1.87 g/L in 1987 and the lowest value of 1.19 g/L in 2002. During the past century, the water salinity and storage of Bosten Lake were influenced chiefly by vapor pressure and precipitation, followed by wet day frequency, daily mean temperature, and potential evapotranspiration. Moreover, the uncertainty of the machine learning model was also explored and discussed. It could be mainly associated with the data accuracy of input climate variables and the ignorance of environmental impacts from the intense agricultural activities after the 1960 s. This study is expected to advance the scientific understanding of long-term change characteristics of Bosten Lake and to provide a technical reference of reconstructing centenary hydrologic and environmental trajectory for dryland lakes.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.