Prediction of Solar Flares and Background Fluxes of X-Ray Radiation According to Synoptic Ground-Based Observations Usin
- PDF / 751,755 Bytes
- 6 Pages / 612 x 792 pts (letter) Page_size
- 96 Downloads / 159 Views
iction of Solar Flares and Background Fluxes of X-Ray Radiation According to Synoptic Ground-Based Observations Using Machine-Learning Models A. G. Tlatova, *, E. A. Illarionovb, c, I. A. Berezina, and A. D. Shramkoa aKislovodsk
Mountain Astronomical Station, Russian Academy of Sciences, Kislovodsk, 357700 Russia b Moscow State University, Moscow, 119234 Russia c Moscow Center of Fundamental and Applied Mathematics, Moscow, 119234 Russia *e-mail: [email protected] Received March 1, 2020; revised April 24, 2020; accepted May 29, 2020
Abstract—The paper presents machine-learning models for predicting powerful solar flares and background X-ray fluxes in the range of 1–8 Å. To predict solar flares for the next day, information was used on the current level of solar activity obtained from ground-based synoptic observations, such as characteristics of sunspots and radio fluxes at wavelengths of 10.7 and 5 cm, as well as the level of the background flux and the number of solar flares of the current day obtained from the GOES satellite. To predict the background fluxes of X-ray radiation, only data from ground-based telescopes were used. The high efficiency of the forecast for the next day is shown. The neural network was trained on data available since 2002. DOI: 10.1134/S0010952520060106
INTRODUCTION At present, forecasting space weather (SW) caused by solar activity is a prerequisite for the successful implementation of space programs, air traffic at high latitudes, and other tasks, including special-purpose tasks. To effectively predict space weather parameters, forecasts of solar flares and the level of X-ray radiation are used. This paper presents machine-learning models that solve these problems. In this paper, we present models for forecasting solar flares based on daily synoptic observations of solar activity in the optical and radio bands. The first multivariate model is based on machine learning. The model makes it possible to forecast the number and intensity of solar flares in one or two days with high reliability. The second is a model for predicting the background flow X-ray radiation from ground-based observations based on machine-learning models. The models are based on data that include data from “classical” optical telescopes for observing the photosphere, chromosphere, and corona of the Sun. FORECASTING METHODS AND BASELINE DATA The mechanism of solar flares is one of the main unsolved problems in solar physics. The processes of energy accumulation and triggering of flares are
caused by the appearance of a magnetic field flux in the photosphere (for example, [1, 2]). The shape and complexity of sunspots in white light emission have been classified according to the growth rate of sunspots [3]. It is known empirically that larger spots with a large number of nuclei and a more complex magnetic flux structure tend to cause larger flares (see, for example, [4–9], as well as repeated flares in the same active regions (ARs)) [10]. One area in the development of forecasting methods is the study of the characteri
Data Loading...
