| |
 |
| Paper: |
Proof-of-concept Gridded Visibility Stacking Pipeline for Deep Spectral Line Interferometric Imaging |
| Volume: |
532, ASTRONOMICAL DATA ANALYSIS SOFTWARE AND SYSTEMS XXX |
| Page: |
341 |
| Authors: |
Rozgonyi, K.; Dodson, R.; Meyer, M.; Mitchell, D. A.; Roychowdhury, S.; Tobar, R. |
| Abstract: |
Deep Spectral Line imaging is one of the greatest challenges for the SKA and its pathfinders, as the raw visibility data volumes are too large to store long term. The Deep Investigations of Neutral Gas Origins (DINGO) survey is planning to observe for up to 2,500 hours per field for its deepest survey regions, likely spreading over multiple years. As the corresponding raw datasets would be >100PB and cannot be stored, the default pipeline will form daily images that are then averaged together to create the final deep image data products. However, to enable the mitigation of low-level systematic errors that might be undetectable on a daily basis and correlated across image pixels, compressed visibility storage is required. To meet this need we developed an alternate pipeline in which visibilities are stored and combined as a gridded data product. These grids are sparse, so can be stored efficiently. Gridding the data in this manner forms a product that is of the same scale as the image and applies the correct kernels, whilst maintaining the ability to flag, re-weight or even re-calibrate the data. Thus, this approach addresses the greatest risks of the default daily imaging strategy. We present our proof-of-concept pipeline, and we demonstrate that our pipeline introduces no significant systematics, but performs better, as compared to accumulating the daily images. |
|
|
|
|
|
