Time Series Clustering and Influencing Factors Analysis on Qinghai-Tibet Plateau Lake Area Change

Hao He,Weide Li,Min Qian,Shujuan Hu
DOI: https://doi.org/10.1007/s10666-023-09913-1
2024-01-01
Environmental Modeling & Assessment
Abstract:Qinghai-Tibet Plateau has a special geographical location and large lake groups making it an important research subject of global climate change and regional response. Taking its 149 typical lakes as the research subject, using the relevant features of lake area data from 1985 to 2020, the variation of the lake area, the chronological division of climate change, and the impact of climate features on the lake area growth rate are analyzed. The main methods and results include the following: (1) By using the time series clustering algorithm to mine the shape similarity of lake area change, three patterns of area change, C1, C2 and C3 modes, which show a slow growth trend, a strong fluctuation, and a significant growth trend, respectively, are summarized. (2) When the optimal segmentation point index of the binary sequence is proposed and combined with the clustering results of the lake regional climate to find the year of the optimal segmentation point for all lakes, the regional climate segmentation points of lakes are concentrated from 1995 to 2005. (3) When researching the lakes in C3 mode, using the XGBoost model as the regression model to extract the feature importance affecting lake growth rate, the effect of precipitation on lake area is the highest, followed by surface temperature. When a double machine learning model is established to conduct causal analysis to obtain an unbiased estimation of the lake area growth rate influenced climate features, the effect of precipitation is the highest, followed by evaporation. The methods used and the results show a new perspective in understanding climate change and lake response.
What problem does this paper attempt to address?