Using the satellite images for the territory of Ukraine
The purpose of this article is to identify the wind direction and speed using the images from geostationary satellites and through application of two-dimensional wind vectors, the magnitude and direction of which corresponds to the speed and direction of cloud masses, on a satellite image. The results may be used for making a short-term forecast of dangerous weather events within the territory of Ukraine. To make the technique work, it is necessary to select cloud areas on a satellite image using the threshold method. Then, based on the brightness temperature distribution between two tracking modules (parts of an image based on which two consecutive satellite images are compared), the maximum correlation coefficient for infrared brightness temperature is to be determined. The coefficient corresponds to the movement of cloud masses and sets the beginning and end of the wind direction vector. To determine the optimum application of the technique for the territory of Ukraine, the analysis of accuracy of tracking modules of different sizes was also performed. The analysis revealed that the accuracy of determining the wind vector direction depends on the tracking module size: the larger it is, the more accurate is the direction vector found, but given that the time interval between images is 15 minutes, the optimum algorithm to be used in Ukraine is the one with 5x5 pixel tracking module. The technique performance was also compared with the data of ICON and GFS forecast models. The results of the applied algorithm showed that the direction of air masses was more reliable than the data retrieved from the above-mentioned forecast models, because the algorithm analyzes the real-time movement of air masses while the forecast models assess the formation and movement of air masses in advance (with an interval of several hours up to dozens of hours). Numerical wind speed forecast of ICON and GFS models is more accurate, because the algorithm determines the wind speed based on the movement of cloud masses on satellite images whereas the forecast models consider several factors (pressure fields, development and subsequent evolution of cyclones, anticyclones, geographical characteristics etc.) which makes them more realistic.