I. Data Processing and First Results

Transcription

1 ACTIVE-REGION MONITORING AND FLARE FORECASTING I. Data Processing and First Results PETER T. GALLAGHER, Y.-J. MOON and HAIMIN WANG Big Bear Solar Observatory, New Jersey Institute of Technology, Big Bear City, CA 92314, U.S.A. (Received 4 March 2002; accepted 23 May 2002) Abstract. This paper discusses a near real-time approach to solar active-region monitoring and flare prediction using the Big Bear Solar Observatory Active Region Monitor (ARM). Every hour, ARM reads, calibrates, and analyses a variety of data including: full-disk Hα images from the Global Hα Network; EUV, continuum, and magnetogram data from the Solar and Heliospheric Observatory (SOHO); and full-disk magnetograms from the Global Oscillation Network Group (GONG). For the first time, magnetic gradient maps derived from GONG longitudinal magnetograms are now available on-line and are found to be a useful diagnostic of flare activity. ARM also includes a variety of active-region properties from the National Oceanic and Atmospheric Administration s Space Environment Center, such as up-to-date active-region positions, GOES 5-min X-ray data, and flare-to-region identifications. Furthermore, we have developed a Flare Prediction System which estimates the probability for each region to produce C-, M-, or X-class flares based on nearly eight years of NOAA data from cycle 22. This, in addition to BBSO s daily solar activity reports, has proven a useful resource for activity forecasting. 1. Introduction Big Bear Solar Observatory (BBSO) has a long tradition of developing innovative instruments and interpreting data on solar activity and space weather. BBSO s lakeside location at an altitude of 2067 m, and its excellent seeing conditions (Goode et al., 2000), together with its sophisticated instrumentation and large team of experienced scientific and technical staff, make BBSO one of the world s most ideal sites for ground-based space weather research. This, in addition to the almost real-time availability of both ground- and space-based images of the photosphere, chromosphere, and corona, make scientific activity forecasting a timely and much needed resource for the solar physics and space weather communities. In collaboration with Kanzelhöhe Solar Observatory (KSO) in Austria and Yunnan Astronomical Observatory (YNAO) in China, BBSO has developed a highresolution, full-disk Global Hα Network (GHN) which provide 2 arc sec resolution images of the solar chromosphere with up to a 30-s cadence (Otruba, 1999; Steinegger et al., 2000). These images, when used in combination with longitudinal magnetograms and continuum images from the Michelson Doppler Imager Now at L-3 Communications Analytics Corporation, Laboratory for Astronomy and Solar Physics, NASA Goddard Space Flight Center, Greenbelt, MD 20771, U.S.A. Also at Korea Astronomy Observatory, Daejon , Rep. of Korea. Solar Physics 209: , Kluwer Academic Publishers. Printed in the Netherlands. CD ROM

2 172 PETER T. GALLAGHER, Y.J. MOON AND HAIMIN WANG Figure 1. Schematic of ARM showing the sources of data currently in use together with a possible future addition in dashed lines. (MDI) and EUV images from the Extreme-ultraviolet Imaging Telescope (EIT) onboard the Solar and Heliospheric Observatory (SOHO) further add to our ability to monitor active region evolution and make accurate predictions of solar activity (Gallagher et al., 2002). During solar maximum, flares, filament eruptions, and coronal mass ejections (CME), can result in conditions which influence the performance of space- and ground-based systems and which can endanger humans in space and the near-earth environment. This is particularly important today due to the increased number of commercial trans-polar aircraft flights and the increase in extra-vehicular astronaut activity required during the assembly of the International Space Station. This therefore prompts the need for real-time solar activity monitoring, not only to view current solar conditions, but also to make reliable, scientific forecasts of future solar activity. With this in mind, we have developed the BBSO ARM, which provides near real-time solar data and activity forecasting. This paper is organized as follows: in Section 2, the detailed design of ARM is discusses while preliminary results are given in Section 3. Section 4 then gives our conclusions and future prospects. 2. Active Region Monitor The Active Region Monitor (ARM) is a web-based solar activity monitoring tool which provides the most recent images from a variety of ground- and space-based solar observatories and can be found at ARM was developed within Research System s Interactive Data Language (IDL) using the current version of the SolarSoftWare (SSW; tree (Freeland and Handy, 1998). ARM was built on two key software developments, namely IDL sockets and map objects. IDL sockets provide a convenient method for reading

3 ACTIVE-REGION MONITORING AND FLARE FORECASTING, I 173 (or writing) data between remote machines without actually having to transfer data via ftp before reading takes place. This vastly increases the speed and reliability of accessing and reading data on remote archives. IDL object maps were developed by Dominic Zarro (L-3 Communications Analytics Corp./GSFC) and are simply structures containing two-dimensional image data with accompanying information such as the Cartesian coordinates of center of the image, the pixel scale, the reference time, etc. This therefore allows typical processing tasks such as roll-correction, stretching, translation, solar rotation-compensation, and image co-alignment to be performed. Further information on sockets and map objects can be found at the SSW web-site. In addition to providing images from the photosphere, chromosphere, and corona, ARM includes tables of active-region data from NOAA, including up-to-date active region positions. Furthermore, a Flare Prediction System (FPS) has been included, which provides a table of region names together with flaring probabilities. Further detail on these, and other features of ARM are discussed below FULL-DISK IMAGES ARM displays full-disk images from GONG, GHN, magnetograms, and continuum and EUV images from MDI and EIT. On the hour, ARM connects to the BBSO data archive at ftp://ftp.bbso.njit.edu and locates the most recent full-disk Hα image. When located, this file is then transferred via ftp to the local machine on which ARM is running. The archived data selected has already been flat-fielded using an improved version of the Kuhn Lin method (Kuhn, Lin, and Loranz, 1991; Denker et al., 1999) and limb darkening corrected using the method described in Denker et al. (1999). See Figure 1 for a graphical summary of the software. EIT (Delaboudinière et al., 1995) provides full-disk EUV images in narrow bands centered on He II (30.4 nm), Fe IX/X (17.1 nm), Fe XII (19.5 nm) and FeXV (28.4 nm) which have formation temperatures of K, K, K, and K, respectively. The instrument provides a 45 arc min 45 arc min field of view and 2.6 arc sec pixels together with a maximum fulldisk frame-rate of about 2 min in high cadence mode. On the hour, the ARM software opens a client-side TCP/IP Internet socket and reads the most recent Fe XII (19.5 nm) and Fe XV (28.4 nm) FITS file from the SOHO archive, and then automatically background subtracts, de-grids, and flat-fields the data. MDI (Scherrer et al., 1995) records high-resolution magnetograms, Dopplergrams, and line-depth images of the photosphere in the Ni I nm absorption line. The instrument images the Sun onto a pixel CCD camera through a series of increasingly narrow spectral filters. The final elements, a pair of tunable Michelson interferometers, enables MDI to record full-disk filtergrams with a spatial resolution of 2 arc sec per pixel. On the hour, the ARM software also transfers the most recent continuum image and magnetogram from the SOHO archive. The

4 174 PETER T. GALLAGHER, Y.J. MOON AND HAIMIN WANG Figure 2. Full-disk Hα image from BBSO on 21 January :02:43 UT as displayed on ARM. The Stonyhurst grid-lines are drawn at 10 heliographic degree intervals, while the NOAA active region numbers are also indicated. This figure can be seen in color on the accompanying CD-ROM. continuum image is then flat-fielded and corrected for limb darkening, while the magnetogram is scaled appropriately. Full-disk longitudinal magnetograms are also sourced from four of the six sites of the Global Oscillation Network Group (GONG; Harvey et al., 1996). Namely, data are obtained from: Big Bear, California; Mauna Loa, Hawaii; Learmonth, Western Australia; Cerro Tololo, Chile. The longitudinal magnetograms are in turn used to derive magnetic field gradient maps. Photospheric field gradients have long been known to be related to flaring activity in active regions (Zhang et al., 1994) although there has been little recent observational or theoretical developments. We focus our analysis on the magnitude of the gradient of the line-of-sight field which can quite simply be calculated using

5 ( B ) 2 B = + x ACTIVE-REGION MONITORING AND FLARE FORECASTING, I 175 ( ) B 2. (1) y The calibrated data from GONG, GHN and EIT are then over-laid with Stonyhurst grids using SSW mapping software, while active regions are identified and labeled using positions from NOAA. An example of a full-disk Hα image in given in Figure 2. A10arcmin 10 arc min field centered on each active region is also extracted from each full-disk image and written to the ARM web pages. A zoom view of NOAA 9672 from 24 October 2001 is given in Figure 3. This therefore allows users to view both full-disk images together with zoomed views of each region ACTIVE-REGION TABLES Active-region properties, such as area in millionths of the solar hemisphere, McIntosh and Mount Wilson classifications, and total number of constituent spots, are obtained from the Solar Region Summary issued by the NOAA Space Environment Center at 00:30 UT on a daily basis. The region positions, obtained from this report, are then differentially rotated to the current time and are tabulated as both heliographic and heliocentric positions. ARM is thus the only source of active region positions which are valid to the nearest hour FLARE PREDICTION SYSTEM Similarly to forecasters at NOAA s Space Environment Center, we have developed a flare prediction system based on the McIntosh active-region classification scheme (McIntosh, 1990). The McIntosh system describes the fundamental properties using three parameters: Z gives the Modified Zürich Class; p describes the penumbra of the largest spot; c gives the sunspot distribution. There have been several attempts in the past to calculate flare rates from active region flaring probabilities using both the Mount Wilson (Sammis, Tang, and Zirin, 2000) and the McIntosh classification schemes (McIntosh, 1990; Bornmann, Kalmbach, and Kulhanek, 1994; Bornmann and Shaw, 1994). Our system uses almost eight years of SEC flare occurrence and active region McIntosh classification data from November 1988 to June 1996, thus sampling the majority of activity in cycle 22. Following Kildahl (1980), and average flare rate is calculated based on tables of daily average flare rates. For example, between November 1988 and June 1996 there were 302 regions with a McIntosh classification Eai. As this class produced 172 C-class events, 62 M-class events, and 2 X-class events, a flare rate can be estimated by dividing the number of events by the number of occurrence of this class i.e., the average daily flare rate of M-class flares for an Eai region is 62/302 or 0.21 flares per day. As for many random phenomena in nature, the frequency of occurrence of solar flares follows Poisson statistics, and so knowing the average flare rate allows us to

6 176 PETER T. GALLAGHER, Y.J. MOON AND HAIMIN WANG Figure 3. Four automatically calibrated and aligned 10 arc min 10 arc min images centered on NOAA 9672 and taken directly from ARM. (Note: due to the loss of Yohkoh in December 2001, EIT 19.5 nm images are now used in place of SXT.) This figure can be seen in color on the accompanying CD-ROM. calculate the probability for a region of a particular class to produce one or more C-, M-, or X-class event for a given 24-hour period. Recently, Wheatland et al. (2000) found that the waiting-time distribution of X- ray flares is consistent with a time-dependent Poisson distribution using 25 years of GOES data. Moon et al. (2001) also showed that the waiting-time distribution for X-ray flares from 1989 to 1991 is well represented by a time-dependent Poisson function with time-varying mean flaring rates and the period most suitable for a constant mean flaring rate is about 3 days. In addition, they found that the waitingtime distribution in six flare productive active regions follows a constant Poisson distribution. Analyzing solar flares that occurred in individual active regions from

7 ACTIVE-REGION MONITORING AND FLARE FORECASTING, I to 1999, Wheatland (2001) supported the results of individual active regions by Moon et al. (2001) and showed that a time-dependent Poisson process is a good model for the observed waiting-time distributions. As far as daily solar flare forecasting is concerned, these results imply that the assumption of a constant Poisson process is reasonable. The probability of observing N events per unit time interval is given by the basic Poisson distribution: P µ (N) = µn exp( µ), (2) N! where µ is the average number of events expected in each time interval. The probability of observing one or more events in a one day (24-hour) period can then be estimated via P µ (N 1) = 1 P µ (N = 0), (3) where P(N = 0) is the probability of no events occurring. Explicitly, the probability can be given by P µ (N 1) = 1 exp( µ). (4) Thus, for the Eai region described above with an average daily flare rate of 0.21 flares per day, the probability for one or more M-class events to occur in a given 24-hour period is P µ=0.21 (N 1) = 1 exp( 0.21) (5) = This methodology can also be extended to each of the three parameters used in the McIntosh classification scheme. In Figure 4 the daily flare rate for each of the three McIntosh classification parameters was used to calculate the probabilities for M- and X-class activity. This is discussed further in Section SOLAR ACTIVITY REPORTS AND WARNINGS Once per day, BBSO issues a solar activity report providing information on all the active regions on the Sun, including the probability for flare events. In addition, solar activity warnings are posted on the ARM web pages and distributed immediately to an list. These warnings have replaced BEARALERTS (Zirin and Marquette, 1991) as of 24 November Requests for subscription to the activity warning distribution list can be sent to activity@bbso.njit.edu.

8 178 PETER T. GALLAGHER, Y.J. MOON AND HAIMIN WANG Figure 4. Derived 24-hour active-region flare probabilities for each of the three McIntosh classification parameters using Poisson statistics. 3. Recent Results ARM has been in operation for just under a year and has proven crucial to solar activity monitoring and forecasting, and has also produced several new and interesting results. In this section, we briefly detail some of the recent results, leaving more detailed analysis to follow-on publications.

9 ACTIVE-REGION MONITORING AND FLARE FORECASTING, I 179 Figure 5. The left panel gives a GONG+ magnetogram taken at 22:03:16 UT on 9 January 2002, while the derived gradient of the longitudinal magnetic field is given at right. This figure can be seen in color on the accompanying CD-ROM FLARE PREDICTION In Figure 4 we show the region flaring probabilities for each of the three McIntosh parameters, Zpc. The upper panel compares the probabilities for events of M- and X-class to occur in the seven Modified Zürich Classes Z. It can clearly be seen that there is a systematic increase in the probability for regions to flare from class A through F, while regions of class H show a low flaring probability. This is not unexpected as region with classes D, E, and F represent well-developed bipolar regions with numerous spots, while regions of class H usually represent the final, quiescent state of region development. The middle panel of the Figure 4 compares the flaring probabilities associated with the second McIntosh parameter, p, which describes the penumbra of the largest spot. Although the trend is not as well behaved here, spots of class s through k certainly show a higher probability than the other classes here, with regions with large asymmetric sunspot penumbras (class k) having the largest flaring probability. The bottom panel of Figure 4 gives the third parameter of the McIntosh scheme, c, which describes the interior sunspot distribution of the group. Here we can see that regions with larger, more compact distributions of spots (i and c) and more likely to flare FLARE-ASSOCIATED MAGNETIC GRADIENTS There have been several attempts to derive properties of the photospheric magnetic field associated with flaring activity such as shear, electric current, and field gradient. The left-hand panel in Figure 5 shows a arc min image extracted from a full-disk GONG magnetogram obtained on 9 January 2002 in which the data has been scaled to ±250 G to highlight weak features which would otherwise

10 180 PETER T. GALLAGHER, Y.J. MOON AND HAIMIN WANG Figure 6. The top-left panel gives a GONG magnetogram taken at 22:03:16 UT on 9 January 2002, while the top-right panel gives the gradient of the longitudinal magnetic field. A three-day GOES X-ray plot from 7 January to 9 January 2002 is given at the bottom. This figure can be seen in color on the accompanying CD-ROM. have been difficult, if not impossible, to observe. Also labeled in the image are two simple active regions: NOAA 9772 is an α region and NOAA 9775, a β. The right-hand panel of the same figure gives the gradient of the magnetic field derived using Equation (1). As expected, the solitary spot of NOAA 9772 shows the largest gradients ( 0.12 G km 1 ) due to the significant magnetic flux difference between the spot center and its boundaries. NOAA 9775, on the other hand, consists of nu-

11 ACTIVE-REGION MONITORING AND FLARE FORECASTING, I 181 merous smaller spots, each being identifiable in the gradient map as small, isolated double-peaks with gradients of 0.07 G km 1. In contrast to these simple regions, the top-left panel of Figure 6 shows a arc min magnetogram centered on the large complex region NOAA This data was again extracted from a full-disk GONG magnetogram taken on 9 January 2002 and is scaled to ± 250 G. This region first became visible on the Sun on 4 January 2002 and developed rapidly from a simple α spot to a complex region consisting of some twenty constituent spots on 9 January 2002 when the data analyzed here were obtained. On this date, the region was classified as a βγδ region with a distinct inversion line within the region s interior. The topright panel shows the gradient of the magnetic field (in G km 1 ) derived from this magnetogram. It is clear there are strong magnetic gradients of up to 0.22 G km 1 visible across the inversion line of the delta, which only became visible on this day: magnetograms from preceding and subsequent days displayed gradients of less than 0.12 G km 1. Indeed, the majority of active regions typically show gradients of less than 0.12 G km 1, with only the most flare productive regions showing larger, more distinct gradients. The bottom panel of Figure 6 shows the GOES X-ray activity for 7 to 9 January 2002 showing the increase in flaring activity from the end of the seventh to when the magnetic gradient measurements presented here were obtained. Figure 7 gives a further example of an active region (NOAA 9830) which was both flare prolific and showed distinct gradients in the longitudinal magnetic field. The top row shows a arc min MDI continuum and corresponding magnetogram when NOAA 9830 was almost on the central meridian. On this date (19 February 2002) NOAA 9830 was classified as a βγδ region. During this region s passage across the disk it produced numerous C- and M-class events and, from the bottom panel of Figure 7, showed distinct gradients along the magnetic inversion line of the δ. Although the field gradient magnitude is a little less ( 0.07 G km 1 ) than the previous example, it again straddles the inversion line of the δ, a fact not too surprising considering the right-hand-side of Equation (1). That is, fields in close proximity (i.e., with small x and/or y) which have large, oppositely polarized magnetic fluxes (i.e., large B) will result in large field gradients. Further examples of the relationship between magnetic flux gradients and flare productivity and their physical interpretation will be discussed in a follow-on paper which is currently in preparation. Finally, it should be noted that GONG is a line-of-sight instrument and so longitudinal field strengths, and hence gradient maps, are particularly dependent on their position on the disk. We have therefore only selected active regions close to the central meridian for this work.

12 182 PETER T. GALLAGHER, Y.J. MOON AND HAIMIN WANG Figure 7. Top row: MDI continuum and magnetogram images of NOAA 9830 which produced numerous C- and M-class events during its traversal of the solar disk. Bottom row: magnetic gradient map derived from a GONG magnetogram. This figure can be seen in color on the accompanying CD-ROM. 4. Conclusions Monitoring solar activity with a global network of full-disk Hα telescopes, in conjunction with longitudinal magnetograms, EUV and continuum images from SOHO, put us in an advantageous position for detecting and predicting changes in solar activity. For the first time, this suite of ground- and space-based instruments allow us to characterize active regions from the photosphere, through the chromosphere, to the corona. Future automatic detection of flares, filament eruptions, and other events (Gao, Wang, and Zhou, 2002) will also allow us to rapidly respond with additional observations such as high-cadence vector magnetograms, narrow-band filtergrams, and almost diffraction limited speckle images.

13 ACTIVE-REGION MONITORING AND FLARE FORECASTING, I 183 Acknowledgements This work was supported by NASA grants NAG5-9682, NAG5-9738, ONR N , NSF grants ATM , ATM , ATM , and an AFOSR MURI grant. YJM also acknowledges grant NRL M J of the Korean Government while PTG acknowledges SESDA contract NAS while at GSFC. EUV, continuum, and magnetogram data are courtesy of the EIT and MDI consortia. SOHO is a project of international cooperation between ESA and NASA. This work utilizes data obtained by the GONG project, managed by the National Solar Observatory, which is operated by AURA, Inc. under a cooperative agreement with the National Science Foundation. References Bornmann, P. L. and Shaw, D.: 1994, Solar Phys. 150, 127. Bornmann, P. L., Kalmbach, D., and Kulhanek, D.: 1994, Solar Phys. 150, 147. Delaboudinière, J.-P. et al.: 1995, Solar Phys. 162, 291. Denker, C., Johannesson, A., Marquette, W., Goode, P. R., Wang, H., and Zirin, H.: 1999, Solar Phys. 184, 87. Freeland, S. L. and Handy, B. N.: 1998, Solar Phys., 182, 497. Gao, J., Wang, H., and Zhou, M.: 2002, Solar Phys. 205, 93. Gallagher, P. T. et al.: 2002, Ann. Geophys., submitted. Goode, P. R., Wang, H., Marquette, W. H., and Denker, C.: 2000, Solar Phys. 195, 421. Harvey, J. W. et al.: 1996, Science 272, Kildahl, K. J. N.: 1980, in R. F. Donnelly (ed.), Solar-Terrestrial Predictions Proceedings 3, 166. Kuhn, J. R., Lin, H., and Loranz, D.: 1991, Publ. Astron. Soc. Pacific 103, McIntosh, P. S.: 1990, Solar Phys. 125, 251. Moon, Y.-J., Choe, G. S., Yun, H. S., and Park, Y. D.: 2001, J. Geophys. Res. Space Phys. 106, Otruba, W.: 1999, in B. Schmieder, A. Hofmann, and J. Staude (eds.), Third Advances in Solar Physics Euroconference: Magnetic Fields and Oscillations, ASP Conference Series 184, 314. Sammis, I., Tang, F., and Zirin, H.: 2000, Astrophys. J. 540, L583. Scherrer, P. H. et al.: 1995, Solar Phys. 162, 129. Steinegger, M. et al.: 2000, ESA Special Publ. 463, 617. Wheatland, M. S.: 2000, Astrophys. J. 536, L109. Wheatland, M. S.: 2001, Solar Phys. 203, 87. Zhang, H., Ai, G., Yan, X., Li, W., and Liu, Y: 1994, Astrophys. J. 423, 828. Zirin, H. and Marquette, W.: 1991, Solar Phys. 131, 149.