Full-Disk Disambiguated Vector Magnetic Field Observations from HMI
The hmi.B_720s Data Series
HMI full-disk disambiguated vector field observations available in the data series hmi.B_720s have been computed routinely every 12 minutes since 1 May, 2010. (See the coverage maps). The full-disk disambiguated field is available only as a definitive data product, thus the data become available several days after the observation is made.
The hmi.B_720s data series mostly inherits the quantities from the VFISV ME inversion (hmi.ME_720s_fd10), with added keywords and segments (images) related to disambiguation. For a more complete description of the HMI vector magnetic field data series, please refer to the HMI Vector Field Pipeline paper (Hoeksema et al., 2014). The VFISV ME inversion paper (Centeno et al., 2014) describes the VFISV inversion based on the HMI Stokes parameters. The additional keywords in hmi.B_720s contain information for the disambiguation module parameters. The two newly added full-disk images give the results of the disambiguation in disambig and information about its quality in conf_disambig. Both near-real-time and definitive disambiguated vector magnetic field observations from HMI are available for the entire mission in automatically identified Space-weather HMI Active Region Patches (SHARPs), as described in the SHARP paper (Bobra et al., 2014).
Algorithm and Disambiguation Solution
The full-disk disambiguation differs from the SHARP version in that only the strong-field pixels are subject to the minimum-energy algorithm, in order to speed up the processing. Certain other parameters have also been adjusted to find a balance between quality and speed.
The information indicating the applied algorithm is contained in the conf_disambig image, where a value of "90" indicates the pixel is above the noise threshold, "60" indicates pixels peripheral to the strong field, "50s" are weak-field pixels, and '0' is off the disk. The "90" and "60" pixels are "annealed" using the minimum-energy algorithm, and the "60" pixels are subsequently smoothed. The weaker "50" pixels are not annealed; three different algorithms are used instead: potential-acute; random, and radial-acute.
For the potential-acute method, a potential field is extrapolated from the vertical field component; the azimuth that makes the field vector more similar to the potential field vector (i.e. a larger dot product) is selected. For the radial-acute method, the azimuth that makes the field more nearly radial is selected. For the random solution, whether the azimuth should be flipped is decided randomly. Both the potential and radial methods suffer from systematic errors, but the random method discards any available information about the disambiguation in weak-field regions. In most cases, the random method is preferred.
The result from the disambiguation determination is a 3-bit mask, where a 1 indicates that 180 degrees should be added to the azimuth at that pixel. The three results are stored in the disambig image. Only the lowest three bits have meaning. For weak-field pixels (conf_disambig = 50), the three bits correspond to solutions from potential-acute (lower bit), random (middle), and radial-acute (higher) algorithm, respectively. That is, 011 means 0 from radial acute and 1 from random and 1 from potential-acute. For stronger-field pixels (conf_disambig = 90 or 60) the three bits are all 1's or all 0's.
Note that the azimuth image in hmi.B_720s is NOT DISAMBIGUATED and has a value between 0 and 180. To include the disambiguation solution, the user must pick one of the three bits in the disambig image. Where a disambig pixel bit has value=1, 180 degrees should be added the azimuth. The randomly assigned disambiguation is safest because the radial-acute and potential-acute methods can show systematic large-scale patterns in weak regions.
Using the Data
Generally speaking, one needs to download at least four images (segments) to study the vector data: field, inclination, azimuth, and disambig. The field is the total field strength; inclination is the angle of the field relative to the line of sight; azimuth is the ccw angle between 'up' on the image and the transverse field before disambiguation; and the selected bit in disambig indicates whether the azimuth should be increased by 180 degrees.
We provide two sample SSWIDL modules for data processing: hmi_disambig.pro and hmi_b2ptr.pro. The first combines disambig and azimuth; the second generates Br, Btheta, Bphi, i.e. the field vector components projected onto spherical-coordinate unit vectors using other geometry keywords. Headers provide an explanation of usage and examples.
As with any inversion, the results at some pixels are erroneous, so users should also pay careful attention to the indicators of quality, uncertainty, and confidence that are provided in otherhmi.B_720s map segments. There are also systematic errors associated with the daily orbit of the SDO spacecraft. See the HMI Vector Field Pipeline paper and the HMI Observables paper for discussion.
The full-disk azimuthal disambiguation solutions have been used to produce the two definitive SHARP data series since 15 January, 2014: hmi.sharp_720s and hmi.sharp_cea_720s . They can be distinguished from earlier versions by looking at the AMBPATCH keyword (0 for full disk disambiguation, 1 for the earlier method where all pixels were annealed and the weaker pixels smoothed). The default algorithm for the weaker-field pixels is radial-acute in both SHARP series. NRT SHARP data are still disambiguated only in the rectangular HMI AR patches and all pixels are annealed.
Due to time constraints, some higher-noise non-optimal-quality data are not disambiguated as of now. This averages 1-2 frames per day, typically when one or more filtergrams are missing due to calibrations. The QUALITY keyword indicates when data are not optimal.