We selected the CaI 6122 Å line to derive the vector magnetograms and the photospheric Doppler maps. With an effective Landè factor of 1.75 this line is only moderately sensitive to Zeeman splitting but is relatively broad and better suited to the 160 mÅ bandpass of our Fabry-Perot etalon.

First, the raw images were corrected for the CCD dark current offsets and flat fielded to account for the pixel-to-pixel gain variations introduced by the detector electronics and optical system imperfections. To improve the signal to noise the images were resized to 512 × 512 pixels with a scale of 0.18 arcsecs per pixel, corresponding to a field of view of 92.2 × 92.2 arcsecs (66.3 × 66.3 Mm). The images were then carefully co-aligned to account for image motions caused by the residual telescope pointing errors and drifts.

The Stokes images Q/I, U/I, and V/I were obtained by calculating the normalized difference between measurements of opposite polarization states. For example:

The magnetic field components longitudinal () and transversal () to the line of sight (LOS) were calculated following Ronan et al. (1987):

The 180° ambiguity is not a straightforward problem to solve. Many techniques have been proposed (see Metcalf 1994 and references therein; Skumanich and Semel 1996). However, most of them work well only for active regions that do not deviate too much from a potential field configuration. Active region AR 8844 has several areas where the field structure appears to be highly complex and non-potential. Therefore, a totally automated algorithm would not be capable to properly resolve the 180° ambiguity for those areas. We adopted an approach that combines a potential field solution with an empiric estimation of the correct field orientation based on simple geometrical considerations.

We started by selecting one vector magnetogram out of the 56 recorded, which will be kept as reference, and we solved the 180° ambiguity for it. First, we compared the field vectors with a simple potential filed model derived by the longitudinal magnetograms. In this model we just considered the two main polarities in the active region and neglected the magnetic flux in-between. The field azimuths were chosen such that their angular difference with the model azimuths was the smallest. The resulting vector magnetogram was subsequently filtered with a "smoothing algorithm" that checks for maximum continuity between adjacent vectors to create a smooth azimuthal distribution of vectors across the entire region. The azimuth map was then visually inspected to correct for local geometrical inconsistencies. Finally, the vector field map was transformed from the LOS reference system to the local, heliographic reference system. In it, the zenith angle is expressed with respect to the local vertical and the azimuth with respect to the local N-S / E-W directions. The transformation into the heliographic reference system helps in revealing vectors that were incorrectly chosen due to the non-vertical viewing angle of the region (see Gary and Hagyard 1990 for a discussion on the effects of perspective). Therefore, the vector magnetogram was again visually inspected and corrected for local geometrical inconsistencies.

The rest of the vector magnetograms in the time sequence were corrected sequentially by comparison with the preceding vector magnetogram in the time sequence, starting with the reference vector magnetogram. Our assumption was that during the time interval between consecutive vector magnetograms the vector directions changed less than 90°. After this sequential comparison process, a few vector magnetograms still needed some manual correction to account for isolated local errors.

The results are two time sequences of 56 vector magnetograms corrected for the 180° ambiguity. One with the vectors expressed in the LOS reference system, and one in the heliographic reference system. The sequences cover a time span of approximately 4 hours, with an average time gap between consecutive vector magnetograms of about 3.5 minutes.

The photospheric Dopplergrams were obtained by calculating the difference between filtergrams recorded in the two wings of the CaI 6122 Å line of the vector polarimetric sequences.

We averaged all the images recorded in each wing during two consecutive vector polarimetric measurement sets:

To remove some residual contribution of the 5 minutes oscillation and to reduce image-to-image noise, we applied a space-time Fourier filter. The filter combined a high-pass filter, for the noise, with a bandpass filter, for the intensity oscillations.

We want to emphasize that the so derived Doppler velocity maps are just a rough estimation of the effective Doppler velocities. Only a line fit through multiple points across the spectral line can provide accurate velocity values.