Statistical Assessment of Photospheric Magnetic Features in Imminent Solar Flare Predictions

Abstract

In this study we use the ordinal logistic regression method to establish a prediction model, which estimates the probability for each solar active region to produce X-, M-, or Cclass flares during the next 1-day time period. The three predictive parameters are (1) the total unsigned magnetic flux T flux , which is a measure of an active region's size, (2) the length of the strong-gradient neutral line L gnl , which describes the global nonpotentiality of an active region, and (3) the total magnetic dissipation E diss , which is another proxy of an active region's nonpotentiality. These parameters are all derived from SOHO MDI magnetograms. The ordinal response variable is the different level of solar flare magnitude. By analyzing 174 active regions, L gnl is proven to be the most powerful predictor, if only one predictor is chosen. Compared with the current prediction methods used by the Solar Monitor at the Solar Data Analysis Center (SDAC) and NOAA's Space Weather Prediction Center (SWPC), the ordinal logistic model using L gnl , T flux , and E diss as predictors demonstrated its automatic functionality, simplicity, and fairly high prediction accuracy. To our knowledge, this is the first time the ordinal logistic regression model has been used in solar physics to predict solar flares.

Image patch analysis of sunspots and active regions

“…Typically, a set of scalar properties is derived from line of sight (LOS) or vector magnetogram and analyzed in a supervised classification context to derive which combination of properties is predictive of increased flaring activity (Leka & Barnes 2004;Guo et al 2006;Barnes et al 2007;Georgoulis & Rust 2007;Schrijver 2007;Falconer et al 2008;Song et al 2009;Huang et al 2010;Yu et al 2010;Lee et al 2012;Ahmed et al 2013;Bobra & Couvidat 2015). Examples of scalar properties include: sunspot area, total unsigned magnetic flux, flux imbalance, neutral line length, maximum gradients along the neutral line, or other proxies for magnetic connectivity within ARs.…”

“…First, we determine regions of high magnetic flux of each polarity using an absolute threshold at 50 Gauss. Second, we compute for each pixel the distance to the closest high flux region in each polarity using the Fast Marching method (Sethian 1995). Once the two distance fields (one for each polarity) are calculated, the neutral line can be obtained by finding the pixels that lie on or are close to the zero level-set of the difference of these two distance fields.…”

“…2. From this two-valued array, we retrieve the distances from the zero level-set using a fast marching method (Sethian 1995) implemented in the Python SciKits' module ''scikit-fmm''. 3.…”

Solar flare forecasting using sunspot-groups classification and photospheric magnetic parameters

“…Experimental results show that photospheric parameters is indeed can be used a precursor for solar flares forecasting. In this study, aiming to improve the solar flare forecasting performance in our previous study (Song et al 2009) ,we use both sunspot-groups classification and photoshperic magnetic parameters. We view the solar flare forecasting as a classification problem in machine learning field.…”

“…However, people sometimes prefer to get a probability instead of a binary label, just like what people get from daily weather reports. In our previous studies by Song et al (2009) and Yuan et al (2010), we have shown that logistic regression, which is a statistical learning method for probability estimation, can be used for flare forecasting. In this paper, the solar flare forecasting is regarded as a classification problem in machine learning field, i.e., flaring population vs. non-flaring population.…”