Zou Tianle, Yang Kun, Pan Meie, Zhu Yanhui, Zhang Yang, Su Danni
Faculty of Geography, Yunnan Normal University, Kunming 650500, China.
Faculty of Geography, Yunnan Normal University, Kunming 650500, China; The Engineering Research Centre of GIS Technology in Western China, Ministry of Education of China, Yunnan Normal University, Kunming.
Sci Total Environ. 2024 Sep 15;943:173618. doi: 10.1016/j.scitotenv.2024.173618. Epub 2024 Jun 7.
Turbidity is a crucial indicator of water quality. The European Commission's Copernicus Land Monitoring Service Platform provides free turbidity data for large lakes to monitor the global water quality of lakes. However, the data were missing from April 2012 to April 2016, severely limiting long-term analysis. Based on MODIS and turbidity data, Random Forest and XGBoost models are used to invert Tonle Sap Lake's turbidity. Random Forest outperformed the XGBoost model. Based on Random Forest model, missing data were filled in to construct long-term series data of Tonle Sap Lake turbidity (2004-2021). Trend, persistence and correlation analyses were conducted to reveal spatiotemporal characteristics and driving mechanism of turbidity. The results showed that: (1) spatially, the average annual, monthly, and seasonal turbidity was higher in the north but lower in the south, with regions of higher turbidity exhibiting more significant changes; (2) temporally, the annual turbidity mean was 53.99 NTU and showed an increasing trend. Monthly, turbidity values were higher from March to August and lower from September to February, with the highest and lowest recorded in June and November at 110.06 and 5.82 NTU, respectively. Seasonally, turbidity was higher in spring and summer compared to autumn and winter, with mean turbidity values of 84.16, 93.47, 15.33 and 23.21 NTU, respectively; (3) In terms of sustainability, the Hurst exponent for annual turbidity was 0.23, indicating a reverse trend in the near future; (4) Dam construction's impact on turbidity was not significant. Compared with natural factors (permanent wetlands, grasslands, lake surface water temperature, and remote sensing ecological index), human activities (barren, urban and built-up lands, croplands and population density) had a more significant impact on turbidity. Turbidity was highly correlated with croplands (r = 0.76), followed by population density (r = 0.71), and urban and built-up lands (r = 0.69).
浊度是水质的关键指标。欧盟哥白尼陆地监测服务平台提供大型湖泊的免费浊度数据，以监测全球湖泊水质。然而，2012年4月至2016年4月的数据缺失，严重限制了长期分析。基于中分辨率成像光谱仪（MODIS）和浊度数据，利用随机森林和极端梯度提升（XGBoost）模型反演洞里萨湖的浊度。随机森林模型的表现优于XGBoost模型。基于随机森林模型，填补缺失数据以构建洞里萨湖浊度的长期序列数据（2004 - 2021年）。进行趋势、持续性和相关性分析以揭示浊度的时空特征和驱动机制。结果表明：（1）在空间上，年、月和季节平均浊度北部较高而南部较低，浊度较高的区域变化更为显著；（2）在时间上，年平均浊度为53.99 NTU并呈上升趋势。月度上，3月至8月浊度值较高，9月至2月较低，最高和最低记录分别出现在6月和11月，分别为110.06 NTU和5.82 NTU。季节上，春季和夏季的浊度高于秋季和冬季，平均浊度值分别为84.16、93.47、15.33和23.21 NTU；（3）在可持续性方面，年浊度的赫斯特指数为0.23，表明近期呈反向趋势；（4）大坝建设对浊度的影响不显著。与自然因素（永久性湿地、草地、湖面水温以及遥感生态指数）相比，人类活动（裸地、城市和建成区土地、农田以及人口密度）对浊度的影响更为显著。浊度与农田高度相关（r = 0.76），其次是人口密度（r = 0.71）以及城市和建成区土地（r = 0.69）。
