Electrically charged gases on the Sun’s surface generate zones of strong magnetic fields. Sun’s gases are constantly moving, tangling, stretching, and twisting the magnetic fields. This motion generates enormous activity on the Sun’s surface. At times, the Sun’s surface becomes tremendously active. At other times, the atmosphere is a little more subdued. Solar flares are the most violent events that occur in our solar system. A solar flare is a powerful blast of radiation caused by the magnetic energy released by sunspots. Flares occur when the Sun’s highly entangled magnetic fields change from poloidal to toroidal. As with a rubber band, tangled magnetic fields release energy when they snap. Solar flares are commonly visible because to the broad spectrum of light they emit. Solar flares must be regularly studied since their intensity provides insight into the sun’s activity, which has a direct impact on Earth. X-rays and optical light are commonly used to monitor flares. We employed the light curve of solar data in the X-ray region to identify solar flares autonomously in this manner. To recognise various solar flare properties automatically, we combined a statistical model with a few machine learning techniques.
- Introduction
- Solar Flares feature extraction
- Classification of Flares
- Use of Fit Parameters
- Limitations of our method
- Conclusion
- References
Solar flares occur when magnetic energy accumulates in the solar atmosphere and must be released.
There are these three steps involved in a typical flare:- In the first stage—the precursor stage—the release of magnetic energy is triggered. This stage will be used to detect soft x-ray emission.
- Electrons and protons are accelerated to energies significantly during the second stage—the impulsive stage. This step results in the emission of X-rays, radio waves, and gamma rays. In this technique, we are using only the X-rays data to detect the flares.
- The third and final stage is called decay. Soft X-rays are identified during this stage due to their progressive accumulation and decay.
Solar flares go through several stages, and there is no accurate way to forecast their strength or duration. Each of these stages might last anywhere from a few seconds and an hour. Solar flare statistics can be generated by detection of sudden impulsive increases of the soft X-ray intensity in light curves. Here, we develop an automated flare detection algorithm that allows us to analyze the time series data of light curve.
Since the Sun is an active ball of gases, there is always some energy releasing from the surface of it. But, there is a very sudden change in flux when a solar flare takes place. This energy change remains almost constant in the absence of the solar flare. We have used this feature to detect solar flares
In order to determine the characteristic features of solar flares, we will require the rise time i.e.; the time when the flare started taking place, the peak time i.e.; the time when flux from the flare is at peak, the end time i.e; time when flux from the flare becomes zero etc. Here are the steps we have involved to extract these features:
- First of all, we are re-binning the time steps to ∆t = 60s by the median value during each bin. Thus, the daily record has the number of nbin equal to 86400/60 = 1440 data points.
- We have taken the median of fluxes during quiescent time periods to find a noise level of fnoise ≈ 4.7 × 10−8Wm−2 and define a corresponding threshold level of fthresh = fnoise .
- The smoothing of the re-binned light curve was accomplished by the use of the boxcar technique with nsm = 10 time bins.
- We now identify all of the smoothed light curve’s local peaks and minima (consecutively in daily intervals). The flux maximum times ti correspond to probable flare peak time, while the flux minimum times ti−1 and ti+1 correspond to flare start and finish time, respectively.
- The light curve is composed of several flux shifts. Flares are characterised by the following characteristics:
- The flare begins at a flux minimum time ts, where a preflare background fBG is generated from the median flux of time interval [ts − ∆tBG,ts]. In this case, we used a time interval of ∆tBG = 300 seconds to calculate background flux, which translates to nbin = 300/60 = 5 bins. Now, here a problem arises that there may be the case when the particular flare is already preceded by some flares especially in case when solar activity is high. So, here, taking median will give the median of fluxes of preceding flare instead of background, so to solve this issue, we put a condition that if the median of fluxes is more than five times of preceding background flux, then we take background flux of this flare as thatof previous background flux (without updating the background of this flare).
- The flare ends at the first subsequent flux minimum time te, when the flux drops below the level fBG, now here too, in case when some other flare is happening when a flare is ending, then the flux level will not touch the background, so, in this case we extrapolate the decay of this flare and rise of subsequent flare to find the ending and rising time of this and the next flare respectively.
-
The peak flux value fp occurs at time tp during the intermediate time interval ts < tp < te. Now, to find the flare peaks, our algorithm searches for consecutive increases of four points such that count rate of 4th point is greater than or equal to 1.04 times that of first point. Then, we search for three consecutive decreasing points after above fourth point and then the global maximum between those points is used to calculate the peak flux of that flare.
-
The background-subtracted peak flux is F = fp − fBG.
For the duration flare event, we extrapolate the curve at the beginning and at the end of a flare to the fBG level in case it is preceded or followed by some other flare. Now, as we know that the observed burst has a fast rise and slow decay shape. The fast rise follows the linear form while the slow decay shape follows the power-law function (f (x) = at−k). We have fitted the function according to the given data and found out the parameters of these two forms for every flare and then we found out the duration of rising and decay times by extrapolating the curves to fBG level.
-
We classify the solar flares on the basis of the peak flux value as following:
- A class: When peak flux value ranges in the order of - 10−8W/m2 or less
- B class: When peak flux value ranges in the order of 10−7W/m2
- C class: When peak flux value ranges in the order of 10−6W/m2
- M class: When peak flux value ranges in the order of 10−5W/m2
- X class: When peak flux value ranges in the order of 10−4W/m2 or more.
-
We have also tried to further sub classify these classes with an additional digit, like a C3-class flare has a peak flux value of 3 × 10−6W/m2.
-
The use of peak flux values can serve as a warning; for example, if it exceeds a specific level, we can trigger the satellite’s protection mechanism or something similar in order to avoid damage to the satellite’s electronics.
-
We used the A, B, C, M, and X classification criteria because they provide a direct indication of the strength of the solar flare, both qualitatively and quantitatively, and also allow for easy comparison of the strengths of solar flares, for example, an X3 class flare is five times stronger than an M6 class flare.
- If a solar flare lasts shorter than seven minutes, it is unlikely to be spotted, as we first re binned to 60 seconds and then searched for four growing and three declining spots, as previously explained. Using the box car average method did result in a tiny reduction in our peak flux value. Time complexity is high for high time scale data. Detailed explanation of equation is more complex in real than it is in proposed in the mail.
- Solar flares are the most violent events that occur in our solar system. We constantly study flares because their magnitude provides insight into the sun’s activity, which has a direct impact on the Earth. We developed an automated identification method for solar flares using this technique by evaluating light curves. By flattening background energy variations, we refined the quality data for the flares and determined the fit parameters such as rise time, peak flux, and decay time for each flare by extrapolating the entire curve using a Gaussian equation for the rising half and using a power-law equation for the decay half curve. Additionally, we categorised each flare according to its peak flux.
- Gryciuk et al., Solar Physics, 292, 77, 2017
- Aschwanden and Freeland, Astrophysical Journal, 754, 112, 2012
- Markus J. Aschwanden1 and Samuel L. Freeland1
- M. Gryciuk1,2 · M. Siarkowski1 · J. Sylwester1 · S. Gburek1 · P. Podgorski1 · A. Kepa1 · B. Sylwester1 · T. Mrozek1,2