HMI Vector Magnetic Field Data Products

Overview of Vector Field Pipeline

The vector pipeline overview paper is now published. Please cite as: Hoeksema et al., 2014, Solar Physics, doi:10.1007/s11207-014-0516-8.

The HMI instrument collects 4096*4096 filtergrams in various wavelengths and polarizations sufficient to determine the Stokes parameters, I, Q, U, & V, as rapidly as every 90 seconds. Currently a 135-second base sequence is collected using the HMI side camera. The sequence measures I+-Q, I+-U, I+-V at 6 wavelengths. Level1 filtergrams are interpolated in time and space, correcting for solar rotation, cosmic rays, distortion, and other instrumental effects. Every 12 minutes data collected over a 1215-second tapered interval are averaged to reduce noise to produce the Stokes IQUV parameters in a 720-second data series, hmi.S_720s

From many of the same filtergrams are determined the line-of-sight magnetic field (hmi.M_720shmi.M_720), continuum intensity (hmi.Ic_720s), line width (hmi.Lw_720s), and line depth (hmi.Ld_720s). The other HMI camera produces a time series of 45-second data. The pipeline produces near real time version of several of these data series.

To infer physical properties of the solar atmosphere where the spectral 617.3nm line is formed, an inversion is performed on the polarized light. The pipeline uses an inversion based on the Milne-Eddington approximations, called VFISV (Very Fast Inversion of the Stokes Vector). The Stokes vector is inverted to determine the prime magnetic field parameters: B-Total, Inclination angle relative to the line of sight, and 180-degree Azimuth, along with along with estimates of uncertainties from the fits. Several other plasma parameters employed in the model and co-variances are also determined. VFISV iteratively minimizes the difference between the observed and synthesized profiles. VFISV must fit for multiple physical parameters using observations of one-arc-second resolution obtained at only six wavelengths, consequently understanding the implications of the simple assumptions about the solar atmosphere and precise instrument characteristics are critical to interpretation of the data.

After inversion the field azimuth must be disambiguated because the absolute direction (+/- 180 degrees) cannot be uniquely determined from the Stokes parameters. The HMI pipeline uses a variant of Metcalf's Minimum Energy method to minimize a functional of the field divergence and the normal component of the electric current density. The minimum energy state may not be the best solution when noise dominates the signal (i.e. for the vast majority of HMI pixels that have average field strength below ~200 G); thus for weaker pixels we transition to methods that give preference to radial orientation, are most consistent with a potential field, or are sensitive to the direction of the field in surrounding pixels. Around active regions the neighboring-pixel acute-angle algorithm is our preferred choice. We are currently working to implement a pipeline version of the disambiguation for HARP regions and the full disk. Some notes about the noise mask to be used in the disambiguation are available at MagneticNoiseMask.

HMI Pipeline Inverted Vector Field Data

hmi.ME_720s_fd10 : Definitive, science-grade, inverted, not disambiguated HMI vector magnetic field data are available via JSOC. The polarimetric filtergram data are processed with the Milne-Eddington inversion code VFISV to produce full-disk vector magnetic field maps every 12 minutes (i.e. 720 seconds) starting 26 April 2012. Data products are available, on average, less than a week after observation. The inversion code assumes a fixed magnetic filling factor of unity. The fd10 version of the inversion code includes the addition of a regularization term to the merit function in order to bias the solution towards lower eta0 values. The aim of this modification is to reduce spatial discontinuities due to the presence of double minima in the chi-squared surface. Other 'fd10' update features focus on convergence criteria, speed optimization, and minor bug fixes.

We are retroactively processing selected data from 1 May, 2010 onwards. Currently the processed fd10 database includes at least 4 full-disk images per day. For some time steps the data are inverted only in HMI Active Region Patches (HARPDataSeries). The interval 1 Feb - 10 Mar, 2011 has full disk inverted data available every 720s. Check the Monthly Coverage Maps for current data availability.

Be aware that reported magnetic field values in very strong umbrae and in other regions with complicated Stokes profiles are sometimes erroneous in HMI vector data. The dynamic range of the HMI magnetic measurements extends to ~3500 Gauss. Strong magnetic fields leads to a complete splitting of the spectral line (Zeeman saturation). In our case, values stronger than 3500 Gauss shift components of the spectral line outside of HMI's observed spectral range. High Doppler velocities can also shift the spectral line partially outside of the HMI spectral range. When either situation or a combination occurs, the inversion code does not have enough information to constrain the field strength, and occasionally it hits the preset upper limit of 5000G. It is also good to bear in mind that the VFISV code relies on certain simplifying assumptions about the solar atmosphere that can lead to incorrect inversion results. Only a very, very small fraction of the pixels exhibits such problems, many of which are easily seen 'by eye'. These and other minor issues will be described in detail at ReleaseNotes3.

hmi.ME_720s_fd10_nrt : Near real time fd10 inversion data are also available. Currently the VFISV fd10 nrt data are computed only in HARP regions.

Space Weather HMI Active Region Patches

The SHARP paper is now published. Please cite as: Bobra et al., 2014, Solar Physics, doi:10.1007/s11207-014-0529-3

Early Special Release Disambiguated Vector Magnetic Field Data

ReleaseNotes2 : The first release of disambiguated full vector magnetic field data for science analysis took place in November, 2011. That release included 5-days of 12-minute vector magnetograms of AR 11158, the X-class flare producing region from February 12-16, 2011. The processing for that interval was not completely standard because we compute longer and with an earlier version of the VFISV code and use an earlier version of the disambiguation code. A skeleton of the VectorPaper "Measuring the Vector Magnetic Field of AR 11158 with HMI" describing the processing is available. Presently the data for this a few other regions are available for science analysis; however we consider this a 'beta' release and look forward to receiving your feedback. See ReleaseNotes and ReleaseNotes2 for more information.

The method of computation (particularly the disambiguation) may differ for quick-look and final data products, for full-disk and selected patches, and for strong and weak field regions. Routines are available to export the data in useful heliographic coordinates, either in rectilinear {Bx, By, Bz} for patches or for full-disk observations in spherical coordinates, either {Br, Btheta, Bphi} or {Br, inclination, and azimuth}. Contact an HMI team member for assistance. Data can also be exported after remapping to standard synoptic or synchronic map coordinates. Averaging vector field maps in time and/or space can also be accomplished. A module will exist to merge the various methods.

Future Availability

These data will be combined to produce Carrington charts, synoptic maps, and synchronic frames of the vector field components. A long-term estimate of the polar field will be available for applications requiring global coverage.

From the photospheric data various products and models will be computed, including local and global harmonic coefficients, global potential field - source surface models, global and local magnetostatic models (including NLFFF on patches), global and local MHD models of various types.

Methods for tracking surface features will include local correlation tracking, ILCT, and DAVE4M. Interfaces to other models will be provided as practical.

List of VFISV FD10 Inversion Parameters

A description of the VFISV data products is given at ME Data Segments.

The VFISV FD10 Inversion provides the following 25 full-disk quantities

Field - |B| - total flux

Field_err

Inclination (gamma) to the line of sight

inclination_err

Azimuth (phi) relative to 'up' on the CCD

azimuth_err

Vlos_mag - Magnetic line-of sight velocity

vlos_err

Alpha_mag - Magnetic Filling Factor Set to 1

alpha_err

Dop_width

Eta_0

Damping

Src_continuum - Intensity

Src_grad - Intensity Gradient

Correlation Coefficients

field_inclination_err

field_az_err

inclin_azimuth_err

field_alpha_err

inclination_alpha

azimuth_alpha_err

Confidence, Quality, and VFISV Processing Info

Chisq - Chi Squared Value of VFISV Inversion

Conv_flag - Convergence Flag for VFISV

Qual_map - Bit Map Indicators of Data Quality

Confid_map - Judgement of quality of inversion

Older Documentation

A summary of OlderVectorDocumentation is available.