Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userInterpreting LSTM Prediction on Solar Flare Eruption with Time-series Clustering
145
0
0.0
(
0
)
Authors
Hu Sun
Ask ChatGPT about the research
No Arabic abstract
We conduct a post hoc analysis of solar flare predictions made by a Long Short Term Memory (LSTM) model employing data in the form of Space-weather HMI Active Region Patches (SHARP) parameters calculated from data in proximity to the magnetic polarity inversion line where the flares originate. We train the the LSTM model for binary classification to provide a prediction score for the probability of M/X class flares to occur in next hour. We then develop a dimension-reduction technique to reduce the dimensions of SHARP parameter (LSTM inputs) and demonstrate the different patterns of SHARP parameters corresponding to the transition from low to high prediction score. Our work shows that a subset of SHARP parameters contain the key signals that strong solar flare eruptions are imminent. The dynamics of these parameters have a highly uniform trajectory for many events whose LSTM prediction scores for M/X class flares transition from very low to very high. The results demonstrate the existence of a few threshold values of SHARP parameters that when surpassed indicate a high probability of the eruption of a strong flare. Our method has distilled the knowledge of solar flare eruption learnt by deep learning model and provides a more interpretable approximation, which provides physical insight to processes driving solar flares.
rate research
Read More
We studied a circular-ribbon flare, SOL2014-12-17T04:51, with emphasis on its thermal evolution as determined by the Differential Emission Measure (DEM) inversion analysis of the extreme ultraviolet (EUV) images of the Atmospheric Imaging Assembly (AIA) instrument onboard the Solar Dynamics Observatory (SDO). Both temperature and emission measure start to rise much earlier than the flare, along with an eruption and formation of a hot halo over the fan structure. In the main flare phase, another set of ribbons forms inside the circular ribbon, and expands as expected for ribbons at the footpoints of a postflare arcade. An additional heating event further extends the decay phase, which is also characteristic of some eruptive flares. The basic magnetic configuration appears to be a fan-spine topology, rooted in a minority-polarity patch surrounded by majority-polarity flux. We suggest that reconnection at the null point begins well before the impulsive phase, when the null is distorted into a breakout current sheet, and that both flare and breakout reconnection are necessary in order to explain the subsequent local thermal evolution and the eruptive activities in this confined magnetic structure. Using local DEMs, we found a postflare temperature increase inside the fan surface, indicating that the so-called EUV late phase is due to continued heating in the flare loops.
Coronal disturbances associated with solar flares, such as H$alpha$ Moreton waves, X-ray waves, and extreme ultraviolet (EUV) coronal waves are discussed herein in relation to magnetohydrodynamics fast-mode waves or shocks in the corona. To understand the mechanism of coronal disturbances, full-disk solar observations with high spatial and temporal resolution over multiple wavelengths are of crucial importance. We observed a filament eruption, whose shape is like a dandelion, associated with the M1.6 flare that occurred on 2011 February 16 in the H$alpha$ images taken by the Flare Monitoring Telescope at Ica University, Peru. We derive the three-dimensional velocity field of the erupting filament. We also identify winking filaments that are located far from the flare site in the H$alpha$ images, whereas no Moreton wave is observed. By comparing the temporal evolution of the winking filaments with those of the coronal wave seen in the extreme ultraviolet images data taken by the Atmospheric Imaging Assembly on board the {it Solar Dynamics Observatory} and by the Extreme Ultraviolet Imager on board the {it Solar Terrestrial Relations Observatory-Ahead}, we confirm that the winking filaments were activated by the EUV coronal wave.
In this paper we present a topological magnetic field investigation of seven two-ribbon flares in sigmoidal active regions observed with Hinode, STEREO, and SDO. We first derive the 3D coronal magnetic field structure of all regions using marginally unstable 3D coronal magnetic field models created with the flux rope insertion method. The unstable models have been shown to be a good model of the flaring magnetic field configurations. Regions are selected based on their pre-flare configurations along with the appearance and observational coverage of flare ribbons, and the model is constrained using pre-flare features observed in extreme ultraviolet and X-ray passbands. We perform a topology analysis of the models by computing the squashing factor, Q, in order to determine the locations of prominent quasi-separatrix layers (QSLs). QSLs from these maps are compared to flare ribbons at their full extents. We show that in all cases the straight segments of the two J-shaped ribbons are matched very well by the flux-rope-related QSLs, and the matches to the hooked segments are less consistent but still good for most cases. In addition, we show that these QSLs overlay ridges in the electric current density maps. This study is the largest sample of regions with QSLs derived from 3D coronal magnetic field models, and it shows that the magnetofrictional modeling technique that we employ gives a very good representation of flaring regions, with the power to predict flare ribbon locations in the event of a flare following the time of the model.
We developed a solar flare prediction model using a deep neural network (DNN), named Deep Flare Net (DeFN). The model can calculate the probability of flares occurring in the following 24 h in each active region, which is used to determine the most likely maximum classes of flares via a binary classification (e.g., >=M class versus <M class or >=C class versus <C class). From 3x10^5 observation images taken during 2010-2015 by Solar Dynamic Observatory, we automatically detected sunspots and calculated 79 features for each region, to which flare occurrence labels of X-, M-, and C-class were attached. We adopted the features used in Nishizuka et al. (2017) and added some features for operational prediction: coronal hot brightening at 131 A (T>=10^7 K) and the X-ray and 131 A intensity data 1 and 2 h before an image. For operational evaluation, we divided the database into two for training and testing: the dataset in 2010-2014 for training and the one in 2015 for testing. The DeFN model consists of deep multilayer neural networks, formed by adapting skip connections and batch normalizations. To statistically predict flares, the DeFN model was trained to optimize the skill score, i.e., the true skill statistic (TSS). As a result, we succeeded in predicting flares with TSS=0.80 for >=M-class flares and TSS=0.63 for >=C-class flares. Note that in usual DNN models, the prediction process is a black box. However, in the DeFN model, the features are manually selected, and it is possible to analyze which features are effective for prediction after evaluation.
All three components of the current density are required to compute the heating rate due to free magnetic energy dissipation. Here we present a first test of a new model developed to determine if the times of increases in the resistive heating rate in active region (AR) photospheres are correlated with the subsequent occurrence of M and X flares in the corona. A data driven, 3 D, non-force-free magnetohydrodynamic model restricted to the near-photospheric region is used to compute time series of the complete current density and the resistive heating rate per unit volume $(Q(t))$ in each pixel in neutral line regions (NLRs) of 14 ARs. The model is driven by time series of the magnetic field ${bf B}$ measured by the Helioseismic & Magnetic Imager on the Solar Dynamics Observatory (SDO) satellite. Spurious Doppler periods due to SDO orbital motion are filtered out of the time series for ${bf B}$ in every AR pixel. For each AR, the cumulative distribution function (CDF) of the values of the NLR area integral $Q_i(t)$ of $Q(t)$ is found to be a scale invariant power law distribution essentially identical to the observed CDF for the total energy released in coronal flares. This suggests that coronal flares and the photospheric $Q_i$ are correlated, and powered by the same process. The model predicts spikes in $Q_i$ with values orders of magnitude above background values. These spikes are driven by spikes in the non-force free component of the current density. The times of these spikes are plausibly correlated with times of subsequent M or X flares a few hours to a few days later. The spikes occur on granulation scales, and may be signatures of heating in horizontal current sheets. It is also found that the times of relatively large values of the rate of change of the NLR unsigned magnetic flux are also plausibly correlated with the times of subsequent M and X flares, and spikes in $Q_i$.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
