gabrieles94/solarsae-sharp-keywords

SolarSAE SHARP keywords

Hourly keyword values of the Space-weather HMI Active Region Patches (SHARPs), series hmi.sharp_cea_720s (definitive, cylindrical equal-area), for every HARP on the disk, from 2010-05-01 to 2024-12-31: positions, NOAA region numbers, areas and the 16 SHARP space-weather parameters plus R_VALUE (Bobra et al. 2014). 1,072,654 rows, 6,887 HARPs, one parquet file per year.

Built for SolarSAE as the classical baseline for flare forecasting and as the calibration target for physical quantities computed from SDO images. Keywords only: no images or segments.

How it was made

• Source: JSOC, series hmi.sharp_cea_720s, through drms (0.9.0), keyword-only queries (no export request).
• Query: one recordset per month and year, all HARPs, hourly T_REC, e.g. hmi.sharp_cea_720s[][2010.01.01_00:00:00_TAI/744h@1h]; the exact queries of every year are in manifest.json. Keywords: HARPNUM, T_REC, NOAA_AR, NOAA_NUM, NOAA_ARS, LAT_FWT, LON_FWT, LAT_MIN, LAT_MAX, LON_MIN, LON_MAX, CRVAL1, CRVAL2, CRLN_OBS, CRLT_OBS, QUALITY, NPIX, SIZE, AREA_ACR, SIZE_ACR, NACR, USFLUX, MEANGAM, MEANGBT, MEANGBZ, MEANGBH, MEANJZD, TOTUSJZ, MEANALP, MEANJZH, TOTUSJH, ABSNJZH, SAVNCPP, MEANPOT, TOTPOT, MEANSHR, SHRGT45, R_VALUE.
• Retrieved: 2026-11-02 (UTC) by scripts/fetch_sharp_keywords.py in the SolarSAE repository, run by the GitHub Actions workflow sharp_keywords.yml.
• Each year's file is written once, its sha256 recorded in manifest.json; an interrupted run resumes from the manifest.

Files

The viewer shows one split, full: all years together (there is no train/test split; split by date in your own analysis).

Year	File	Rows	HARPs	Hours with records	First	Last	sha256
2010	sharp_cea_720s_keywords_2010.parquet	28,686	187	5,681	2010-05-01 00:00	2010-12-31 23:00	b8044ca66f1cbd8364fc508df475d04142fb274397ac60056a814c66d129e37a
2011	sharp_cea_720s_keywords_2011.parquet	85,470	519	8,576	2011-01-01 00:00	2011-12-31 23:00	06ac436c5e2e614f06866a5d601ade5a74e1201696d89148aa14374f93443b14
2012	sharp_cea_720s_keywords_2012.parquet	95,315	626	8,371	2012-01-01 00:00	2012-12-31 23:00	cafd484b9c35fdfba127f7872f34407f6fb81b423f04be0ffd926cf8f7dabf72
2013	sharp_cea_720s_keywords_2013.parquet	110,383	696	8,552	2013-01-01 00:00	2013-12-31 23:00	9253cb55e69b365c9c63e52336aa29f58306ba8c466f11eae9d5dc4c92e6054e
2014	sharp_cea_720s_keywords_2014.parquet	119,636	760	8,466	2014-01-01 00:00	2014-12-31 23:00	f89e89cb0bbef994fb36f12845011093fa27b4cd7856820e6e66f4fbd7a506c7
2015	sharp_cea_720s_keywords_2015.parquet	103,930	690	8,466	2015-01-01 00:00	2015-12-31 23:00	4eb44d1c07f2032cd2ccbfa8251ffd88237e8c7d8aa94cf5be241be8dd769163
2016	sharp_cea_720s_keywords_2016.parquet	54,275	412	8,410	2016-01-01 00:00	2016-12-31 23:00	7422db2be6b5036df56f76e149ef2985e127822ddbb43e774afd78251142347e
2017	sharp_cea_720s_keywords_2017.parquet	24,263	196	8,308	2017-01-01 00:00	2017-12-31 23:00	e51a83b5a48e48b7dfd732c2f48366a91d6b293369b9a9174ae58222369af1b2
2018	sharp_cea_720s_keywords_2018.parquet	11,525	84	6,415	2018-01-01 00:00	2018-12-31 23:00	94963e484d8727969c2cd433e8e51850f643073e2bf6dbc4b8226857003643c2
2019	sharp_cea_720s_keywords_2019.parquet	6,849	53	4,344	2019-01-01 00:00	2019-12-31 23:00	df48b466e898ece7ef77c7098ede35756df93b0f7dff35823c022ad7d0f9b324
2020	sharp_cea_720s_keywords_2020.parquet	13,476	89	6,016	2020-01-01 00:00	2020-12-31 23:00	df34348d3f813b0e64079630377460387fea773727993a57ea862a6629675982
2021	sharp_cea_720s_keywords_2021.parquet	42,327	286	8,355	2021-01-01 00:00	2021-12-31 23:00	88b32903cbfd76fe8908e0a7e4849c93328d494fe344cd8a595b0dfd329fb72f
2022	sharp_cea_720s_keywords_2022.parquet	98,282	629	8,439	2022-01-01 00:00	2022-12-31 23:00	011a1abbe589011fe0d356d3fa916464bde429bfb5f8eac77e01204ebcebd81c
2023	sharp_cea_720s_keywords_2023.parquet	140,477	876	8,447	2023-01-01 00:00	2023-12-31 23:00	78d4005655b2dd25769dcea24b333d91b17cf38cd24f8c2646f1fb1ff607d3af
2024	sharp_cea_720s_keywords_2024.parquet	137,760	902	8,453	2024-01-01 00:00	2024-12-31 23:00	dfc0924358359ada02625a5d378f40d905f14d0cfbad55d58dadcbbe5b7a58e7

Columns

Column	Units	Meaning
HARPNUM	-	HMI Active Region Patch number
T_REC	TAI string	record time, e.g. 2014.01.07_16:00:00_TAI
NOAA_AR	-	NOAA active region number closest to the HARP; 0 if none
NOAA_NUM	-	number of NOAA regions in the HARP
NOAA_ARS	-	comma-separated NOAA regions in the HARP; 'MISSING' if none
LAT_FWT	deg	flux-weighted centroid latitude (Stonyhurst)
LON_FWT	deg	flux-weighted centroid longitude (Stonyhurst, from disk centre)
LAT_MIN	deg	minimum latitude of the HARP
LAT_MAX	deg	maximum latitude of the HARP
LON_MIN	deg	minimum Stonyhurst longitude of the HARP
LON_MAX	deg	maximum Stonyhurst longitude of the HARP
CRVAL1	deg	Carrington longitude of the CEA patch centre
CRVAL2	deg	latitude of the CEA patch centre
CRLN_OBS	deg	Carrington longitude of the observer (disk centre)
CRLT_OBS	deg	Carrington latitude of the observer (B0)
QUALITY	bit flags	HMI quality word of the record (0 = no flag set)
NPIX	pixels	number of pixels within the patch
SIZE	micro-hemispheres	projected area of the patch on the image
AREA_ACR	micro-hemispheres	de-projected area of active pixels on the sphere
SIZE_ACR	micro-hemispheres	projected area of active pixels on the image
NACR	pixels	number of active pixels in the patch
USFLUX	Mx	total unsigned flux
MEANGAM	deg	mean angle of the field from radial
MEANGBT	G/pixel	mean gradient of the total field (see caveats: G/pixel, not G/Mm)
MEANGBZ	G/pixel	mean gradient of the vertical field (see caveats)
MEANGBH	G/pixel	mean gradient of the horizontal field (see caveats)
MEANJZD	mA/m^2	mean vertical current density
TOTUSJZ	A	total unsigned vertical current
MEANALP	1/Mm	mean characteristic twist parameter alpha
MEANJZH	G^2/m	mean current helicity (B_z contribution)
TOTUSJH	G^2/m	total unsigned current helicity
ABSNJZH	G^2/m	absolute value of the net current helicity
SAVNCPP	A	sum of the modulus of the net current per polarity
MEANPOT	erg/cm^3	mean photospheric magnetic free energy density
TOTPOT	erg/cm	total photospheric magnetic free energy density
MEANSHR	deg	mean shear angle
SHRGT45	%	fraction of area with shear angle > 45 deg
R_VALUE	log10(Mx)	Schrijver R: log10 of the unsigned flux near strong-field PILs
time_tai	datetime (TAI)	T_REC parsed (naive datetime, TAI scale); added by SolarSAE

Caveats

• Time scale. T_REC and time_tai are TAI. TAI - UTC was 34-37 s over 2010-2024 (irrelevant at hourly cadence, but convert before joining with UTC catalogues).
• NOAA numbers. HARPs without a NOAA region have NOAA_AR = 0 and NOAA_ARS = 'MISSING' (48.9% of the rows; the two disagree in 396 rows). A HARP can contain several NOAA regions (NOAA_ARS) and a NOAA region can be split over HARPs.
• Empty records. 17.8% of the rows have no pixels (NPIX, positions and intensive parameters null, extensive parameters 0): the HARP exists at that time but its patch holds no pixels. Most of these rows are on the disk (median distance of the patch centre from the central meridian 42 deg; only 9% beyond the limb). Filter with NPIX.notna().
• QUALITY is the HMI quality word (unsigned 32 bits); 13.7% of the rows have a non-zero value. Records per bit set: bit 5 (0x00000020): 1,469, bit 7 (0x00000080): 37,494, bit 10 (0x00000400): 56,483, bit 11 (0x00000800): 3,113, bit 12 (0x00001000): 10,047, bit 13 (0x00002000): 65, bit 14 (0x00004000): 62, bit 15 (0x00008000): 3, bit 16 (0x00010000): 102,933. The SHARP authors' tutorial selects QUALITY < 65536 (no bit >= 16 set; github.com/mbobra/SHARPs, plot_swx_d3.ipynb); in this dataset that removes 9.6% of the non-empty records. Check the bit meanings in the JSOC definition (jsoc.stanford.edu/doc/data/hmi/Quality_Bits/QUALITY.txt) before choosing a selection.
• Gradient units. MEANGBT, MEANGBZ and MEANGBH are in gauss per CEA pixel, not G/Mm as the JSOC keyword description says (SHARP issue tracker by the SHARP authors, https://github.com/mbobra/SHARPs/blob/master/SHARP_Issue_Tracker.md); multiply by the pixel size in Mm (0.03 deg of heliographic angle, ~0.36 Mm) to convert.
• Near the limb (|LON_FWT| > ~60-70 deg) projection effects and noise make the parameters unreliable; most studies restrict to |longitude| <= 70 deg.
• Disambiguation. SHARP vector quantities use the radial-acute solution for low-confidence (weak-field) pixels (Hoeksema et al. 2014); the parameters are computed over pixels above the noise thresholds of Bobra et al. (2014).
• Series start. SHARP records begin on 2010-05-01.

Gaps

• 65 intervals longer than 24 h contain no record. 57 of them are bounded by hours with fewer than 3 HARPs, almost all in 2017-2020 (solar minimum): the disk had no HARP, so no record exists; these are not missing data.
• 8 gaps are bounded by an active disk (>= 3 HARPs on both sides), so records are missing (an active disk does not empty for a day). Each line gives the SDO coverage of those days and the outcome of the per-day re-query:
  • 2012-02-16 23:00 -> 2012-02-18 00:00 (24 h without records; 11 / 13 HARPs before / after); SuryaBench has 120 files on 02-17; day re-query returned 0 records: missing from the JSOC series
  • 2012-02-22 23:00 -> 2012-02-25 00:00 (48 h without records; 8 / 13 HARPs before / after); SuryaBench has 111 and 120 files on 02-23 and 02-24; day re-queries returned 0 records: missing from the JSOC series
  • 2012-02-26 23:00 -> 2012-02-28 00:00 (24 h without records; 12 / 11 HARPs before / after); SuryaBench has 120 files on 02-27; day re-query returned 0 records: missing from the JSOC series
  • 2012-05-04 23:00 -> 2012-05-06 00:00 (24 h without records; 13 / 12 HARPs before / after); SuryaBench has 120 files on 05-05; day re-query returned 0 records: missing from the JSOC series
  • 2012-06-12 23:00 -> 2012-06-14 00:00 (24 h without records; 12 / 9 HARPs before / after); SuryaBench has 120 files on 06-13; day re-query returned 0 records: missing from the JSOC series
  • 2012-08-12 23:00 -> 2012-08-14 00:00 (24 h without records; 5 / 5 HARPs before / after); SuryaBench has 120 files on 08-13; day re-query returned 0 records: missing from the JSOC series
  • 2013-01-08 23:00 -> 2013-01-10 00:00 (24 h without records; 20 / 16 HARPs before / after); SuryaBench has 119 files on 01-09; day re-query returned 0 records: missing from the JSOC series
  • 2022-11-24 01:00 -> 2022-11-25 20:00 (42 h without records; 8 / 10 HARPs before / after); SDO data gap: SuryaBench has only 8 and 20 files on 11-24 and 11-25; day re-queries returned only records already present
• Days re-queried one by one because of a gap: 10 (2 returned records).

References

• Bobra, M. G., Sun, X., Hoeksema, J. T., et al. 2014, The Helioseismic and Magnetic Imager (HMI) Vector Magnetic Field Pipeline: SHARPs - Space-Weather HMI Active Region Patches, Sol. Phys. 289, 3549, doi:10.1007/s11207-014-0529-3.
• Hoeksema, J. T., Liu, Y., Hayashi, K., et al. 2014, The Helioseismic and Magnetic Imager (HMI) Vector Magnetic Field Pipeline: Overview and Performance, Sol. Phys. 289, 3483, doi:10.1007/s11207-014-0516-8.
• Data: courtesy of NASA/SDO and the HMI science team; JSOC, Stanford University.

License

This compilation (the parquet files, manifest.json and this card) is released under CC0 1.0. The underlying SDO/HMI data are courtesy of NASA/SDO and the HMI science team (JSOC, Stanford University) and follow the SDO data-use policy.

Please cite

CC0 imposes no obligation, but if you use this dataset please cite it together with the two papers that describe the SHARP data:

@misc{solarsae_sharp_keywords,
  title        = {SolarSAE SHARP keywords: hourly hmi.sharp\_cea\_720s keywords,
                  2010-2024},
  author       = {{SolarSAE contributors}},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/datasets/gabrieles94/solarsae-sharp-keywords}},
  note         = {Code: SolarSAE (to be released with the associated publication)}
}
@article{bobra2014sharps,
  author  = {Bobra, M. G. and Sun, X. and Hoeksema, J. T. and Turmon, M. and Liu, Y. and
             Hayashi, K. and Barnes, G. and Leka, K. D.},
  title   = {The Helioseismic and Magnetic Imager (HMI) Vector Magnetic Field Pipeline:
             SHARPs -- Space-Weather HMI Active Region Patches},
  journal = {Solar Physics},
  volume  = {289},
  pages   = {3549--3578},
  year    = {2014},
  doi     = {10.1007/s11207-014-0529-3}
}
@article{hoeksema2014vector,
  author  = {Hoeksema, J. T. and others},
  title   = {The Helioseismic and Magnetic Imager (HMI) Vector Magnetic Field Pipeline:
             Overview and Performance},
  journal = {Solar Physics},
  volume  = {289},
  pages   = {3483},
  year    = {2014},
  doi     = {10.1007/s11207-014-0516-8}
}

Links

• Code: SolarSAE (to be released with the associated publication)
• JSOC SHARP documentation: http://jsoc.stanford.edu/doc/data/hmi/sharp/sharp.htm