介绍了一种估算大气光学湍流的方法--支持向量机。依据该方法用温度梯度、风速和相对湿度估算了2014年5月13日至2014年5月18日四川成都平原地区近地面连续6天的大气折射率结构常数C2n, 并将其与实际观测值进行了比较。结果表明两者吻合得较好, 能基本表现成都地区的光学湍流特征。与Wyngaard算法相比, 支持向量机得到的结果与观测值更加贴近, 其夜间估算的准确度高于Wyngaard算法。相关分析结果表明: 估算值与观测值的相关度在80%以上。平均绝对误差与平均相对误差等误差分析显示, 训练好的支持向量机能够基本准确地估算出成都地区的C2n, 但转换时刻的估算值与观测值仍有一定误差。该研究表明, 进一步提高估算精度需要改进模式在转换时刻的估算结果。

A method of estimating atmospheric optical turbulence is introduced which is called support vector machine (SVM). Based on this method, the gradient of temperature, wind speed, and relative humidity are used to calculate the refractive structure parameter C2n in Chengdu, Sichuan from 2014-5-13 to 2014-5-18, which is a continuous 6-day estimated result. The estimated results agree well with the measurement results, which can basically show the characteristics of optical turbulence in Chengdu area. Compared with results of Wyngaard algorithm, the results of SVM are more consistent with the measurements, and the estimation accuracy at night is higher than that of the Wyngaard algorithm. The relative analysis demonstrates that the correlation coefficient between the estimations and observations is over 80%. Besides, mean absolute error and mean relative error prove that after training, SVM can precisely estimate C2n in Chengdu. However, deviations are found between estimations and measurements during the transition time, which should be overcome to improve the performance of the model.