USING MACHINE LEARNING IN STUDYING THE IMPACT OF SOLAR ACTIVITY ON EARTH

Solar activity can significantly impact the Earth’s atmosphere through variations in electromagnetic radiation and solar energetic particles (SEPs). While the investigation of connections between electromagnetic radiation and Earth’s atmosphere has a long history, recently, there has been a growing interest in research on the rapid impact of SEPs on the weather and climate of the Earth. Entering the atmosphere, fast SEPs originating from a coronal mass ejection and coronal holes can increase ionization rates and affect the atmosphere’s processes. Although the penetration of SEPs into the magnetosphere in the Earth’s polar area is recognized, how exactly they spread and disperse over air masses and move toward the Earth’s surface is still unclear. The recent availability of a large amount of satellite data monitoring the processes between the Sun and the Earth and the development of machine learning (ML) algorithms have made it possible to process massive amounts of data and to find precise patterns and trends that are not recognizable using traditional techniques based on known physical laws. Several studies using different ML techniques have been conducted at the Geographical Institute “Jovan Cvijic” (GIJC) SASA to investigate the functional dependence between factors characterizing solar activity and the occurrence of forest fires, hurricanes, and precipitation-induced floods. The underlying premise of these investigations is that Earth’s atmosphere is affected by any magnetic field energy and particles ejected from the Sun’s geoeffective position. Due to the high risk to people and their property, as well as indications that floods will become more frequent in the future, the investigation of a possible link between solar activity parameters and precipitation-induced floods is of particular interest. In GIJC, floods in the United Kingdom and across Europe have been investigated. Studies used classification ML modeling and took into account the time delay between outbreak SEPs and the effect on the atmosphere. It has been shown that classification ML algorithms are valuable tools for establishing the nonlinear relationship between SEPs and precipitation-induced floods, and they can explain the appearance of the precipitation that may cause floods up to nine days in advance with accuracy between 81% and 91%.