SHORT TERM PREDICTION OF POWER SOLAR FLARES FROM SSRT DATA

The algorithm for short term prediction of powerful solar flares, based on SSRT data, includes microwave emission characteristics in total and polarized flux, the structure of active region (AR) magnetic field at the photospheric level, and Ha patrol data.

Polarization Distribution of AR Microwave Emission

Unlike the familiar criterion reported by Tanaka and Enome (1975), our prediction technique takes into account physical characteristics of the generation of emission in the source region and during its propagation in the corona of the AR. For this purpose, the visible solar surface is divided into longitudinal zones, and the indicator of the preflare situation is taken to be the deviation of the observed polarization distribution from a 'normal' distribution (not a flare-hazardous one, but corresponding to quiet conditions of the AR) in the zone where the AR resides at a given time (Maksimov et al., 1990a). The criterion was tested through a retrospective prediction using the SSRT data obtained during the time interval 1982-1986. The analysis included 595 days of observation. We predicted flares of importance ³ 1N and quiet days with no flares of such importance. The predictability for the quiet days and for days with flares was 55% and 65%, respectively (the table of predictabilities may be found in a paper of Maksimov et al., 1990a). A significant number of the events with unsatisfactory predictability occurred near the boundaries between longitudinal zones. Therefore, a statistical study was made of the appearance (disappearance) of microwave emissions when the AR appeared from behind (disappeared beyond) the east (west) limb (the study covered 155 sunspot groups for the period 1988-1991; Maksimov and Bakunina, 1995), and of the polarization inversion phenomenon at the solar disk passage of the AR (the study covered 54 AR's for the period 1982-1988; Maksimov and Bakunina, 1991). This permitted us to obtain empirical formulas for determining the individual boundaries of longitudinal zones for each particular AR according to parameters of the sunspot group in the optical emission.

Microwave Emission Flux

An analysis of the relationship between the flare X-ray importance and the AR emission flux prior to the flare showed that solar flares occur in AR's with both low and high flux values. If, however, the flare occurs in the AR with the flux in excess of 20 sfu, it will most likely be a powerful one (Maksimov and Bakunina, 1996).

Flux-to-Area Ratio

The behavior of the ratio of the microwave emission flux to the sunspot area, F/S, represents quite well the contribution from the emission of a coronal condensation at different stages of AR development. By investigating 10 AR's, it was possible to identify with confidence AR's with a monotonic variation of F/S (4 AR's, 49 days of observation) and with a non-monotonic variation of F/S (6 AR's, 74 days of observation). Examples of such a behavior are given in Figure 1. The highest flare productivity corresponds to AR's which are characterized both by high flux values and by a non-monotonic character of flux-to-area ratio behavior (the table of the flare productivity for AR's with different values of Fand with a different behavior of F/S may be found in a paper of Maksimov and Bakunina, 1996).

Fig.1. The flux-to-area ratio behavior: a) monotonic (AR NOAA 6444, central meridian passage, January 14, 1991), b) non-monotonic (AR NOAA 6580, central meridian passage, April 13, 1991)

The Degree of Subflare - Microwave Burst Connection

The degree of subflare - microwave burst connection (N_b/N_f, where N_b is the number of microwave bursts, and N_f is the number of subflares) is typically below 30%. However, for some periods of AR lifetime the value of this parameter can exceed 70%. We investigated 12 AR's for 99 days of observation. A total of 295 subflares occurred during the time interval of observation, and 173 of them were accompanied by microwave bursts <N_b/N_f > » 0.6. During 37 days of this period when flares of importance ³ 1B were observed, of the 140 subflares recorded, 110 were accompanied by bursts (N_b/N_f = 0.79). On the quiet days, the value of this parameter was 0.24. One day before a powerful flare, it increased to 0.7 and still remained rather high the next day after the flare (Figure 2, Maksimov et al., 1990b).

Neutral Line Associated Sources (NLS)

NLS are weakly-polarized microwave sources which appear prior to large X-ray flares above the polarity inversion line of the AR magnetic field (Uralov et al., 1998a). They are indicators of powerful solar flare buildup. NLS are readily identified on radio images (radio maps) of AR's obtained either as part of a solar monitoring at the SSRT in the two-dimensional scanning mode (Uralov et al., 1998a) or on radio maps synthesized over several hours, based on monitoring data derived from scanning the Sun with one-dimensional resolution (Uralov et al., 1998b). A study of such radio images of more than 50 AR's showed that NLS occur in 2/3 of cases prior to a flare of importance X 1.0 or higher.

This result was obtained, unavoidably, by using crude optical data. It is not inconceivable that if less powerful flares are accounted for, the correlation of NLS with flare buildup will be higher. This is supported by the detection of NLS by the experienced observer at the SSRT using only one-dimensional solar scans in the monitoring process. This fact, combined with analyses of AR radio images, demonstrated that the SSRT data alone are sufficient to detect NLS. Of course, NLS are associated with places of storage and quasi-stationary release of energy in the corona of AR's. NLS brightness temperatures in the cases treated here were in the range 300-3000 thousand degrees.

Fig. 2. Character of variation of the degree of subflare - microwave burst connection prior to, during (time 0) and after the flare (region with 25% - 75% of values, median values, maximum and minimum values).

ALGORITHM AND PREDICTABILITY

An algorithm for short term prediction of powerful solar flares was generated on the basis of our identified signatures of the preflare stage of AR development in microwave emission. The algorithm is essentially as follows. Optical data are used to measure the heliolatitude and heliolongitude of the AR, its extent and area, the slope of the axis of a sunspot group with respect to the equator, and the magnitude and polarity of the magnetic field in sunspots. The individual (for each AR) boundaries of longitudinal zones are calculated from these data using empirical formulas. The zones boundaries, AR positions, and one-dimensional solar scans in total and polarized microwave emissions are then plotted on the heliographic grid. This step is followed by an identification of AR's in optical and microwave emissions and a comparison of the observed polarization distribution with a normal distribution in those zones where the observed AR's are located. For each AR, the flux F and the area S are determined, and F/S and N_b/N_f are calculated. If the AR is away from boundaries of longitudinal zones, then the weight factor 2 is assigned to the flare-potential polarization distribution. If an abrupt change in F/S is observed, the value of the microwave emission flux F is higher than 20 sfu, and the value N_b/N_f exceeds 0.5, then the weight factor 1 is assigned to each of these indicators. Otherwise the weight factors are set equal to 0. If the AR lies near the boundaries of longitudinal zones (± 24 hours from the calculated position of the boundary), then the value 1 is also assigned to the flare-potential polarization distribution. If the sum of numerical factors exceeds the value 3 (in both cases), it is assumed that a flare of importance ³ M1.0 would occur in the AR within the subsequent 24 hours. The prediction is performed using a computer program. Input data are renewed every 2 hours. Intermediate information (position of the longitudinal zone boundaries, temporal behavior of F and F/S) is displayed in graphics windows.

The algorithm was tested through a retrospective prediction for the period from 11 to 25 December 1990 (the polarization distribution and the ratio N_b/N_f were taken into account). During this period, 7 AR's crossed the solar disk, in which 20 flares of importance ³ M1.0 and 2 flares of importance ³ X1.0 were recorded. The forecast in terms of flares scored 60%, and no false alarms occurred. Thirty-eight cases of no flares were predicted for all AR's during the above-indicated period, with 79% predictability; flares were recorded in 8 cases (21%) with "no" prediction.

For the periods from 1 to 5 October and from 15 to 22 December 1992, real-time predictions were carried out. Data on AR's and magnetic fields were received from the Sayan observatory by teletype. For lack of communication with the Baikal observatory, the parameter N_b/N_f was not used. In October, 6 AR's were observed to cross the solar disk. A no-flare prediction was made for all days, and no flares were in fact recorded. For December, 7 quiet days and a flare for 1 day were predicted. No flares occured in actual fact.

CONCLUSION

At the moment there is no way of claiming that the prediction algorithm has passed a reliable, statistically sound test. However, some optimism derives from the fact that the four indicators share a common physical reason: a substantial increase of the role of microwave emissions from the corona of the AR at the preflare stage of its development. An important distinctive feature of this algorithm is the possibility of forecasting flares in near-limb regions (both on the visible and averted side) of the solar disk, which is beyond the reach of predictive techniques based on optical monitoring data.