This study develops a machine learning-based predictive model for identifying high-level clouds (HLCs). The model uses meteorological parameters as input features and is trained against human-recorded meteorological observations. A statistical analysis of the relationship between two independent methods of registering HLCs—lidar and meteorological observations—has been performed. Optimal thresholds for the total amount of cloud cover, at which meteorological data are consistent with lidar data, have been determined. The results demonstrate the promising performance of ML models in identifying the links between weather conditions and the probability of HLC detection, which is confirmed by ROC AUC (Area Under the Curve of the Receiver Operating Characteristic) values in the range of 0.87–0.88 for the presence and 0.77–0.78 for the absence of clouds, as well as balanced metrics Precision, Recall, and F1. The XGBoost (eXtreme Gradient Boosting) model proved to be the most robust, demonstrating the ability to effectively integrate data of various types for reliable prediction in various conditions.

During daylight laser polarization sensing of high-level clouds (HLCs), the lidar receiving system generates a signal caused by not only backscattered laser radiation, but also scattered solar radiation, the intensity and polarization of which depends on the Sun’s location. If a cloud contains spatially oriented ice particles, then it becomes anisotropic, that is, the coefficients of directional light scattering of such a cloud depend on the Sun’s zenith and azimuth angles. In this work, the possibility of using the effect of anisotropic scattering of solar radiation on the predictive ability of machine learning algorithms in solving the problem of predicting the HLC backscattering phase matrix (BSPM) was evaluated. The hypothesis that radiation scattered on HLCs has no effect on the BSPM elements of such clouds determined with a polarization lidar was tested. The operation of two algorithms for predicting the BSPM elements is evaluated. To train the first one, meteorological data were used as input parameters; for the second algorithm, the azimuthal and zenith angles of the Sun's position were added to the meteorological parameters. It is shown that there is no significant improvement in the predictive ability of the algorithm.