One goal of the Frontier Fields Implementation Team is to make the best possible high-level reduction products available to the community as quickly as possible after the data are taken. This will enable the entire community to start working with high-quality data at the same time. To this end, we have been working on several reduction strategies that are optimized for the Frontier Fields project.
One of these is a limited “self-calibration” procedure, which involves solving for the darks using the science data itself. This does a better job identifying and removing the many marginal warm pixels that can be CTE-trailed out of existence in the no-background internal darks. The self-cal procedure will also identify and flag the CTE trails behind bright cosmic rays. These are a major source of streakiness in drizzled images. (Often the CRs are identified and removed, but when the trails fall below the clipping threshold, they make it through to infect the drizzle product with faint vertical streaks.)
We are also working on streamlining the reduction procedure so that these reduction products can be assembled as quickly as possible after the data are taken. Image alignment is one of the major bottlenecks in the stacking procedure, since it often requires human intervention to identify galaxies or point sources that are suitable for tweak-reg analysis. We have been developing an automated procedure that goes through the individual flt exposures and identifies potential sharp objects that should have good astrometric information. The procedure then collates these detections, constructs an accurate template for each one, and uses the templates to measure a consistent position for each object in each individual exposure. These positions are inter-related to determine which galaxies provide the best handles on the inter-image registration. The entire procedure is iterated, and in the end it yields a refined template for each galaxy and a raw position for it in each exposure. This basic information serves as input to tweak-reg, so that the WCS headers of each exposure can be exquisitely aligned with the reference frame.
We have been using the 108 half-orbit F606W exposures of the UDF to do this development. Figure 1 shows a portion of the UDF stack, with the best galaxies identified in green, the intermediate-quality ones identified in yellow, and the low-quality ones identified in red. There are many galaxies that, for whatever reason, did not even get identified as possible astrometric handles.
Figure 2 shows an assortment of galaxy templates. These are x4 supersampled relative to the image pixels. In total, we found over 400 usable galaxies within the ACS UDF field.
Finally, Figure 3 shows the improvement in astrometric quality between when we use crude centroid-based positions for galaxies (left panel) and when we use template-fit positions (right panel). By using tailor-made templates, we can improve the astrometric precision for all galaxies from about 0.2 pixel to better than 0.07 pixel, and by properly weighting the galaxies, we can arrive at a net registration that is better than 0.01 pixel.
We hope to make these tools available to users eventually, but at the moment they are being optimized for the particulars of the Frontier Fields data set, wherein we have between 20 and 80 exposures through a given filter of the same field at the same orientation. Stay tuned!
Jay Anderson, Frontier Fields team member