[Paper Review] Differential Emission Measure Evolution as a Precursor of Solar Flares

C. Gontikakis, I. Kontogiannis+5 more|arXiv (Cornell University)|Nov 12, 2020

Solar and Space Plasma Dynamics31 references4 citations

TL;DR

This study investigates the differential emission measure (DEM) evolution in solar active regions as a predictor of major solar flares (M- and X-class). Using time-series data from SDO/AIA and GAIA-DEM maps, it finds that positive temporal derivatives of emission measure ($d\text{EM}/dt$) and maximum DEM temperature ($dT_{\text{max}}/dt$) serve as credible short-term precursors, with $d\text{EM}/dt$ showing higher predictive power than unsigned magnetic flux for flare forecasts within 2–12 hours.

ABSTRACT

We analyse the temporal evolution of the Differential Emission Measure (DEM) of solar active regions and explore its usage in solar flare prediction. The DEM maps are provided by the Gaussian Atmospheric Imaging Assembly (GAIA-DEM) archive, calculated assuming a Gaussian dependence of the DEM on the logarithmic temperature. We analyse time-series of sixteen solar active regions and a statistically significant sample of 9454 point-in-time observations corresponding to hundreds of regions observed during solar cycle 24. The time-series analysis shows that the temporal derivatives of the Emission Measure dEM/dt and the maximum DEM temperature dTmax/dt frequently exhibit high positive values a few hours before M- and X-class flares, indicating that flaring regions become brighter and hotter as the flare onset approaches. From the point-in-time observations we compute the conditional probabilities of flare occurrences using the distributions of positive values of the dEM/dt, and dTmax/dt and compare them with corresponding flaring probabilities of the total unsigned magnetic flux, a conventionally used, standard flare predictor. For C-class flares, conditional probabilities have lower or similar values with the ones derived for the unsigned magnetic flux, for 24 and 12 hours forecast windows. For M- and X-class flares, these probabilities are higher than those of the unsigned flux for higher parameter values. Shorter forecast windows improve the conditional probabilities of dEM/dt, and dTmax/dt in comparison to those of the unsigned magnetic flux. We conclude that flare forerunner events such as preflare heating or small flare activity prior to major flares reflect on the temporal evolution of EM and Tmax. Of these two, the temporal derivative of the EM could conceivably be used as a credible precursor, or short-term predictor, of an imminent flare.

Motivation & Objective

• To evaluate the temporal evolution of the differential emission measure (DEM) in solar active regions as a precursor to major flares.
• To assess whether DEM-derived parameters such as $d\text{EM}/dt$ and $dT_{\text{max}}/dt$ can serve as reliable short-term predictors of solar flares.
• To compare the predictive performance of DEM evolution metrics against the conventional unsigned magnetic flux predictor.
• To determine the statistical significance and forecast utility of positive DEM derivative values in predicting imminent flares.

Proposed method

• Utilized full-disk DEM maps from the GAIA-DEM archive, derived from SDO/AIA EUV observations using a Gaussian fit to the logarithmic temperature dependence of the DEM.
• Analyzed time-series data from 16 active regions and a statistically significant sample of 9,454 point-in-time observations across solar cycle 24.
• Computed temporal derivatives of DEM-related parameters ($d\text{EM}/dt$, $dT_{\text{max}}/dt$, $d\sigma/dt$) using Savitzky-Golay filtering with a 3rd-degree polynomial to smooth noise and enhance signal detection.
• Focussed on positive derivative values ($^{+}$) to isolate heating phases preceding flares, excluding flare onset periods.
• Calculated conditional flare probabilities for different forecast windows (2, 6, 12, 24 hours) based on the distribution of $d\text{EM}/dt^{+}$ and $dT_{\text{max}}/dt^{+}$.
• Compared the predictive power of DEM parameters against the standard unsigned magnetic flux predictor using receiver operating characteristic (ROC) analysis and conditional probability metrics.

Experimental results

Research questions

• RQ1 Can the temporal evolution of the differential emission measure (DEM) serve as a reliable precursor to major solar flares (M- and X-class)?
• RQ2 Do the positive temporal derivatives of $d\text{EM}/dt$ and $dT_{\text{max}}/dt$ show enhanced values prior to flare onset, indicating pre-flare heating?
• RQ3 How does the predictive performance of $d\text{EM}/dt^{+}$ compare to that of the unsigned magnetic flux for short-term flare forecasting (2–12 hours)?
• RQ4 Is the DEM-based predictor more effective than magnetic flux for C-class, M-class, and X-class flares across different forecast windows?
• RQ5 To what extent do DEM parameters reflect pre-flare heating or small-flare activity preceding major flares?

Key findings

• Positive temporal derivatives of emission measure ($d\text{EM}/dt^{+}$) and maximum DEM temperature ($dT_{\text{max}}/dt^{+}$) frequently exhibit high values a few hours before M- and X-class flares, indicating pre-flare heating and plasma brightening.
• For M- and X-class flares, conditional probabilities derived from $d\text{EM}/dt^{+}$ exceed those from unsigned magnetic flux when forecast windows are shortened to 2–12 hours and threshold values exceed 0.3.
• $d\text{EM}/dt^{+}$ yields higher conditional flare probabilities than $dT_{\text{max}}/dt^{+}$, indicating that EM is more sensitive to pre-flare heating than temperature maximum.
• The $d\sigma/dt^{+}$ parameter showed high overlap in distributions and was excluded from further analysis due to poor discrimination between flaring and non-flaring cases.
• The DEM-based predictor outperforms the standard unsigned magnetic flux predictor for short-term forecasting (2–12 hours), particularly for major flares.
• The results suggest that pre-flare heating and small-flare activity are reflected in the DEM evolution, with $d\text{EM}/dt^{+}$ emerging as a credible short-term precursor for imminent flares.