State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Nanjing Hydraulic Research Institute, Nanjing, 210029, China.
State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing, 210008, China.
Environ Sci Pollut Res Int. 2021 Apr;28(13):16616-16632. doi: 10.1007/s11356-020-11544-y. Epub 2021 Jan 3.
Sediment resuspension is critical to the internal nutrient loading in aquatic systems. Turbidity is commonly used as an indicator for sediment resuspension and is proved to be highly correlated to wind speed in large shallow lakes. A field observation of wind speed and turbidity was conducted using a portable weather station and a YSI 6600V2-2, and an observation lasting for 39 days was evaluated in this study (the data points with wind speed > 4 m/s account for 75%). The daily average values (DA dataset) as well as daily maximum (MX dataset) and minimum values (MI dataset) were calculated from the instantaneous observations (IN dataset). Correlations in IN dataset were deduced based on machine learning methods and were compared to those obtained from DA, MI, and MX datasets. Furthermore, the correlation in IN dataset was analyzed by using two statistical methods, and from the view of statistical the turbidity is regarded as a variable. Results indicate that the correlations in IN datasets follow the exponential function or power function pattern with a critical wind speed of 6 m/s, Regression on IN dataset revealed that linear regression model had the best performance on predicting the turbidity in test dataset and no significant differences are observed between exponential function and power function pattern. Correlations in DA and MX datasets exhibit higher maximal information coefficient (MIC) than IN dataset and error of turbidity prediction introduced by using these correlations in IN dataset is within the tolerance level. Statistical analysis on the IN dataset shows that a strong relationship exists among the wind speed and expectation of turbidity with a MIC over 0.99, and follows the exponential function or the power function as well with a different critical wind speed of 4 m/s. Over 95% data points fall in the predicted intervals of turbidity for both methods, suggesting a high predicting accuracy.
泥沙再悬浮对水体内部营养负荷至关重要。浊度通常被用作泥沙再悬浮的指标，并且在大型浅水湖中被证明与风速高度相关。本研究使用便携式气象站和 YSI 6600V2-2 进行了风速和浊度的现场观测，评估了持续 39 天的数据（风速大于 4 m/s 的数据点占 75%）。从瞬时观测（IN 数据集）中计算出了日平均值（DA 数据集）以及日最大值（MX 数据集）和最小值（MI 数据集）。根据机器学习方法推断了 IN 数据集的相关性，并将其与从 DA、MI 和 MX 数据集获得的相关性进行了比较。此外，使用两种统计方法分析了 IN 数据集的相关性，从统计学角度来看，浊度被视为一个变量。结果表明，IN 数据集的相关性遵循指数函数或幂函数模式，临界风速为 6 m/s。对 IN 数据集的回归表明，线性回归模型在预测测试数据集的浊度方面表现最佳，并且在指数函数和幂函数模式之间没有观察到显著差异。DA 和 MX 数据集的相关性表现出比 IN 数据集更高的最大信息系数（MIC），并且在 IN 数据集使用这些相关性进行浊度预测时引入的误差在可接受范围内。对 IN 数据集的统计分析表明，风速和浊度期望之间存在很强的关系，MIC 超过 0.99，并且遵循指数函数或幂函数，临界风速为 4 m/s。两种方法的浊度预测区间内有超过 95%的数据点，表明预测精度较高。
