1. Creating The RHESSI Flare List J McTiernan 3-Oct-2006

2. Procedure: The flare list routine (hsi_flare_list::fill) is called for individual RHESSI orbits. It is run after the Observing Summary is found for one or more files are generated for a given orbit period. It uses data from the Observing Summary, in particular, the “corrected” count rate, the front to total count ratio, and the mod_variance (which is the variance in the count rate in 4 second intervals, divided by the variance expected from a Poisson distribution). These quantities are used to find intervals which may possibly be flare intervals. Next, hsi_locate_flare is used to verify if an interval is a flare; if we can find a position, then we say it’s a flare. Then, in each energy band, an image is attempted. If there is a discernable source, then the flare is ‘detected’ at that energy.

3. Common problem, often SAA and particles are tagged as “flares”

4. A more difficult problem is the fact that flares often occur right on top of a particle event.

5. Test Interval:17-Mar-2002 090100 to 103720 UT

6. Start with “corrected” count rate, 12 to 25 keV, and front-total counts ratio (later in mission, the program uses mod_variance)

7. Intervals are found with rate > background_rate+3*sigma_background_rate. Background is given by a 1 minute running average. Possible “flare” intervals have front_ratio (or mod_variance) > background+3*sigma.

8. Next, Check duration (default minimum of 8 seconds), Fraction of counts in 1 detector (if more than half of counts for the whole interval are in one detector, then the interval is thrown out.) Min peak count rate is 5.0 counts/sec/detector (this may be dropped)

9. Solar sources are determined by imaging. (This is new; previously front_ratio was used to determine whether the source was solar. This has become a problem as the rear detectors lose sensitivity. Mod_variance is unreliable because it is often large for particle events.) Hsi_locate_flare uses back-projection images to try to locate flares; full-sun images (128x128, 16 arcsec pixels) are made starting with detector 9, then 8, 7, 6, 5. If 3 images of 5 have the same maximum point, then the position is found. Hsi_ok_intv is used to find image intervals with good data, good SAS solutions, no att_state changes. Given these intervals for an event, the first try uses the interval which has the peak rate. If that fails then the intervals with the most counts are tried, a maximum of 3 tries. (Something else needs to be done during off-pointing.)

10. Sample output: % HSI_ANNSEC_PATTERN::PROCESS: Processing Annsec modulation patterns % Program caused arithmetic error: Floating illegal operand SPIN AXIS: 431.33763 -36.297018 -216.000 520.000 17-Mar-2002 09:25:44.000 17-Mar-2002 09:26:00.000 % HSI_ANNSEC_PATTERN::PROCESS: Processing Annsec modulation patterns -584.000 -376.000 17-Mar-2002 09:25:44.000 17-Mar-2002 09:26:00.000 % HSI_ANNSEC_PATTERN::PROCESS: Processing Annsec modulation patterns 136.000 904.000 % HSI_LOCATE_FLARE: Unsuccessful 17-Mar-2002 09:25:44.000 17-Mar-2002 09:26:00.000 % HSI_ANNSEC_PATTERN::PROCESS: Processing Annsec modulation patterns 648.000 760.000 % HSI_LOCATE_FLARE: Unsuccessful 17-Mar-2002 09:25:44.000 17-Mar-2002 09:26:00.000 % HSI_ANNSEC_PATTERN::PROCESS: Processing Annsec modulation patterns 104.000 952.000 % HSI_LOCATE_FLARE: Unsuccessful

11. Next Interval: % HSI_ANNSEC_PATTERN::PROCESS: Processing Annsec modulation patterns % Program caused arithmetic error: Floating illegal operand SPIN AXIS: 316.26622 -51.720544 72.0000 -168.000 17-Mar-2002 09:57:28.000 17-Mar-2002 09:58:24.000 % HSI_ANNSEC_PATTERN::PROCESS: Processing Annsec modulation patterns 40.0000 -152.000 17-Mar-2002 09:57:28.000 17-Mar-2002 09:58:24.000 % HSI_ANNSEC_PATTERN::PROCESS: Processing Annsec modulation patterns 40.0000 -168.000 % HSI_LOCATE_FLARE: Successful 40.0000 -168.000 Hsi_locate_flare is pretty reliable down to about 10 counts/second/detector for a 1 minute interval. Performance is spotty with fewer photons.

12. Next, use a smaller pixel size to zoom in, and get the position using hsi_map_evaluator: ANALYSIS DATE: Tue Oct 3 21:51:13 2006 INPUT PARAMETERS TIME_RANGE 17-Mar-2002 09:57:28.00 17-Mar- 2002 09:58:24.00 ENERGY_BAND 12.0 25.0 keV XYOFFSET 40.0 -168.0 PIXELS 128 x 3.0 arcsec MAP STATISTICS COUNTS IN MAP = 1810 FITTED PEAK = 634.836 PEAK LOCATION = 43.6526 -161.201 MAP AVG,SIGMA = 0.530991 96.6865 (PEAK-AVG)/SIGMA= 5.78225 SUBCOLLIMATOR NUMBER: 3f 4f 5f 6f 9f COUNTS: 272 327 534 351 326 : 0.1656 0.2652 0.1673 0.2713 0.2548 : 0.1701 0.2700 0.1658 0.2841 0.2396 0.9943 0.6299 1.0044 0.5628 0.7947 : 1.0043 0.6175 0.9488 0.5711 0.7612 : 0.8301 0.7747 0.9288 0.7609 0.7025

13. Next, check higher energy bands, if there are excess counts, then try images: ANALYSIS DATE: Tue Oct 3 22:07:43 2006 INPUT PARAMETERS TIME_RANGE 17-Mar-2002 10:12:42.00 17- Mar-2002 10:18:42.00 ENERGY_BAND 25.0 50.0 keV XYOFFSET -352.8 -245.0 PIXELS 128 x 3.0 arcsec MAP STATISTICS COUNTS IN MAP = 116254 FITTED PEAK = 25066.5 PEAK LOCATION = -337.171 -244.324 MAP AVG,SIGMA = 23.9669 2256.46 (PEAK-AVG)/SIGMA= 10.7852 SUBCOLLIMATOR NUMBER: 3f 4f 5f 6f 9f COUNTS: 19052 24924 24211 25185 22882 : 0.1619 0.2395 0.1786 0.2455 0.2247 : 0.1627 0.2406 0.1761 0.2454 0.2347 1.0464 0.7096 0.9179 0.6736 0.7633 : 1.0398 0.7146 0.9296 0.6978 0.7753 : 0.8414 0.7012 0.9287 0.7345 0.7155

14. Etc… If (peak-avg)/sigma > 3, and closer than 60 arcsec, keep it ANALYSIS DATE: Tue Oct 3 22:12:32 2006 INPUT PARAMETERS TIME_RANGE 17-Mar-2002 10:12:38.00 17- Mar-2002 10:18:38.00 ENERGY_BAND 50.0 100.0 keV XYOFFSET -352.8 -245.0 PIXELS 128 x 3.0 arcsec MAP STATISTICS COUNTS IN MAP = 25121 FITTED PEAK = 4133.12 PEAK LOCATION = -337.138 -243.736 MAP AVG,SIGMA = 2.21603 505.018 (PEAK-AVG)/SIGMA= 8.01909 SUBCOLLIMATOR NUMBER: 3f 4f 5f 6f 9f COUNTS: 4984 5197 5631 4721 4588 : 0.1803 0.2544 0.1800 0.2465 0.2253 : 0.1813 0.2554 0.1775 0.2464 0.2353 0.9553 0.6597 0.9115 0.6720 0.7627 : 0.9491 0.6646 0.9232 0.6955 0.7748 : 0.8414 0.7050 0.9303 0.7340 0.7165

15. At > 100 keV, this doesn’t work, so the hi_energy band is 50 – 100 keV ANALYSIS DATE: Tue Oct 3 22:17:03 2006 INPUT PARAMETERS TIME_RANGE 17-Mar-2002 10:12:42.00 17-Mar- 2002 10:18:42.00 ENERGY_BAND 100.0 300.0 keV XYOFFSET -352.8 -245.0 PIXELS 128 x 3.0 arcsec MAP STATISTICS COUNTS IN MAP = 17461 FITTED PEAK = 1259.76 PEAK LOCATION = -534.372 -338.434 MAP AVG,SIGMA = 2.19586 284.662 (PEAK-AVG)/SIGMA= 4.26195 SUBCOLLIMATOR NUMBER: 3f 4f 5f 6f 9f COUNTS: 3813 3024 4670 3059 2895 : 0.2071 0.2724 0.1855 0.2520 0.2295 : 0.2081 0.2735 0.1831 0.2519 0.2397 0.8402 0.6037 0.8963 0.6663 0.7617 : 0.8349 0.6080 0.9077 0.6902 0.7737 : 0.8414 0.7012 0.9287 0.7345 0.7155

16. Wait, try the rear detectors: There’s a peak in the corner, but thst’s too far away….. ANALYSIS DATE: Wed Oct 4 11:57:02 2006 INPUT PARAMETERS TIME_RANGE 17-Mar-2002 10:13:08.00 17- Mar-2002 10:18:42.00 ENERGY_BAND 100.0 300.0 keV XYOFFSET -352.8 -245.0 PIXELS 128 x 3.0 arcsec MAP STATISTICS COUNTS IN MAP = 60329 FITTED PEAK = 3512.24 PEAK LOCATION = -501.063 -391.409 MAP AVG,SIGMA = -2.46291 845.276 (PEAK-AVG)/SIGMA= 4.14072 SUBCOLLIMATOR NUMBER: 5r 6r 9r COUNTS: 18568 20972 20789 : 0.1858 0.2522 0.2293 : 0.1834 0.2519 0.2397 0.8939 0.6641 0.7630 : 0.9064 0.6903 0.7736 : 0.9836 0.9819 0.9832

17. What’s next? Given the time interval and position, fill the rest of the structure: Peak counts Total counts Flags: particles, SAA, eclipse, data gaps, decimation, att_state (hsi_flare_flag_fill) Filename Id_number: If the flare overlaps with a flare already in the flare_list, then it gets the same ID. (hsi_flare_id_check) New for this version: GOES Class and NOAA Active Region Finally, the flare_flag and non_solar_event flags are set in the observing summary, and we’re done.

18. Conclusions: based on a sample of 46 different orbits: see http://sprg.ssl.berkeley.edu/~jimm/temp_dbase/plots http://sprg.ssl.berkeley.edu/~jimm/temp_dbase/plots The new program works at least as well as the old one, but misses more small (cps/det <10) flares. Doesn’t flag the SAA or particle events as flares at all. Doesn’t do so well at high energy sources. Slower, from 1 to 30 minutes per orbit.