|
|
Use of machine learning methods for satellite data processing
Doležalová, Anežka ; Pišoft, Petr (advisor) ; Piskala Gvoždíková, Blanka (referee)
Meteorological features, such as clouds, can be observed in 2 ways - from ground stations and, more recently in the last decades, by distance methods. The fact that the two views mentioned are different makes it not easy to compare them. In our work, we use MSG data, such as level 1.5 data and NWC SAF product cloud type, and cloud determination at ground stations by observers from CHMI. Based on these data, we build ML models that best determine from the input satellite data the categories that would be observed by the ground-based observer. It is necessary to select the appropriate model type for this task, set the most suitable parameters and also deal with the imbalanced representation of the different meteorological phenomena and associated cloud categories.
|
guest ::
