PUNCH data overview

PUNCH is an imaging mission and most data from the mission are images. Because the corona and solar wind are very faint compared to the various backgrounds in the data, high photometric precision and many steps of processing are required. This means finding PUNCH data can be complicated.

There are four standard levels of processing for PUNCH data, numbered 0 through 3, plus two ancillary “levels” that are also produced but are branches from the primary science data flow. Each has a different collection of data products. Each data product is identified with a three-character code, that indicates the general type of product and (where relevant) which spacecraft produced the data.

PUNCH is a polarimeter, so image data arrive in both unpolarized and polarized forms. PUNCH represents polarization in two major ways: first, as the quasi-Stokes parameters B, ^⟂ pB, and ^⟂ pB’; and second, as polarization triplets in the “M,Z,P” tri-polarizer system which uses three (real or virtual) polarizer channels at M(inus) 60 degrees, Z(ero) degrees, and P(lus) 60 degrees relative to a reference angle in the instrument or image plane. Both systems are described by DeForest et al. 2022.

The four levels of processing are:

Level 0

These are data direct from the PUNCH cameras, assembled into FITS files and merged with metadata from the spacecraft. The data are typically square-root coded and losslessly compressed on board the spacecraft; L0 images have been decompressed into their original square-root coded form, but the coding is preserved. The square-root coding is tuned to match the effective digital step size to the photon noise level, across the dynamic range of the image. To reconstruct direct camera values that are roughly proportional to photometric intensity, you can examine the “ISSQRT” field in the header. If “ISSQRT” has a nonzero value, then the radiance B of a particular pixel is given by

B = P * P / SCALE

where B is an approximation of the original value (in digitizer number units) from the camera, P is the value of a particular pixel, and SCALE is the “SCALE” field in the header.

Raw camera pixel values are digitized at 16 bits. NFI pixel values are sums of several camera frames, and can therefore have values greater then 2 ¹⁶.

Level 0 files have two-letter codes with a spacecraft number appended. WFI1, WFI2, and WFI3 are assigned the numbers 1-3, and NFI is assigned the number 4. Polarized images have the codes “PMn”, “PZn”, and “PPn” where “n” is the spacecraft number. The M, Z, and P refer to physical polarizer angles relative to camera-frame horizontal, though these are not calibrated at this level. Clear images (taken with no polarizer in the beam) have the codes “CRn”. These sixteen codes (four image types and four spacecraft) form the bulk of the L0 dataset.

Level 1

These are photometrically calibrated, conditioned data from the PUNCH cameras, maintained as separate data files from each spacecraft. The Level 1 data have similar naming scheme to Level 0, with the same product codes. The data are destreaked, despiked, flat-fielded, PSF-corrected, and stray-light subtracted; and have WCS fields with precise alignments derived from the in-image starfield. The images are floating-point values in mean-solar-brightness units.

Level 2

These are rectified NFI images and full-constellation mosaic trefoil images, resampled to output projection coordinates (gnomonic projection for NFI and azimuthal-equidistant projection for the mosaics). Polarization is given in the virtual-polarizer M,Z,P system. The images are intended to be photometric, astrometrically regularized images of the celestial sphere with all light sources included.

The product codes are “PTM” for polarized trefoil mosaics, “CTM” for clear trefoil mosaics, “PNN” for polarized NFI images, and “CNN” for clear NFI images.

Level 3

Two intermediary product codes describe data products just before the final L3 products, “PIM” for polarized intermediate mosaics, and “CIM” for clear intermediate mosaics. These products have F-corona subtraction, but no starfield removal.

Level 3 products are background-subtracted and are intended to be usable directly as coronal and solar-wind images. The basic product codes are similar to the Level 2 codes, but include long-term average products (“PAM”, “PAN”, “CAM”, and “CAN”) that fill in the entire circular FOV by averaging across 32 minutes of PUNCH data acquisition. In addition, a wind-flow product “VAM” describes derived solar-wind motion.

Level Q: QuickPUNCH

These images are intended to be useful for space weather forecasting. They use a time-asymmetric background modeling scheme that permits low-latency production. They’re produced for NOAA’s space weather forecasting infrastructure, but are also made available to other users.

NFI images include a large and very dynamic stray light component. We are working on a technique to handle this. For now, we are using a PCA-based approach to fit and remove this stray light, and the F corona is also removed by it. We fit about 50 PCA components to a group of about 500 good images (i.e. not outlier images—no corrupted regions, no heavy saturation). The fitted components are then used to transform and un-transform an image to be subtracted. This reconstructed image, composed of only the leading PCA components, tends to contain the stray light and F corona but little of the K corona, and we subtract it out. The image to be subtracted is excluded from the set of images the PCA is fit to, to reduce the likelihood that dynamic K corona structure is contained in the PCA components. To reduce the impact of planets and the Moon, which can leave prominent artifacts in the leading PCA components, we divide the image plane into quarters. Each quarter is fit and filtered separately, with the fitting only using images that do not contain a planet or the Moon in that quarter. This quartering approach, combined with outlier image rejection, tends to reduce an initial set of 1000 images to about 500 that are actually fit for any one quarter of the image.

Because this PCA approach removes the F corona from the NFI images but we are not yet subtracting the F corona from WFI images, NFI is currently excluded from the CTM mosaics, but this will be changed in the future.

Level L: QuickLook

These images are intended for viewing and are published in JPEG2000 (.jp2) format.