Estimation of Surface Layer Optical Turbulence using K-Nearest Neighbor Model in Northwest China

This paper presents an estimate of surface layer optical turbulence in Northwest China's Gansu Province using K-Nearest Neighbor (KNN) model with conventional meteorological parameters such as temperature, relative humidity, wind speed, etc. In the KNN model, the Euclidean distance is used as a metric, the mean function is used as a label calculation method, and the optimal value of the number of neighbor samples k is determined by the grid search method. The estimated and measured results were compared. The results shows that the overall trend of the estimated and measured values is relatively consistent. The mean absolute percentage error of the model is 2.48% and the mean absolute error is 0.343. Under clear weather conditions, the model performs well in all-day variations, as well as in the morning and twilight changes. However, for rainy days, the model estimations are on the high side, and the model accuracy needs to be further improved.