CASDA

a repository for CASDA-related tools

I've developed this notebook (based on code obtained from Dr. Laura Driessen, who got it from Dr. Minh Huynh) to download publicly available CASDA data around a given object's coordinates (or a .csv of objects/coordinates). View Notebook

All you need to do is have an OPAL login, add your source details, (pray that CASDA is working), and it will:

1. Download the component catalogues into a newly created, named folder (you can play around with the code and download other data products too)
2. Match components to within a given sky separation (default 5")
3. Produce a lightcurve from all public ASKAP data
4. Save the lightcurve data points to a .csv.

A couple of small things to keep in mind with this data:

• The matches are to Selavy component source catalogues (within 5") of the object coordinates and Selavy has a 5sigma detection threshold.
• There was a recent bug fix due to a change in the CASDA file naming convention. It is possible that some files are missed by my search if they are unconventionally named.
• In some cases CASDA may store multiple versions of a single observation -- especially from pilot programs -- if the data was reprocessed. This may produce, for example, two detections on the same day with slightly different fluxes.
• The columns in each .csv should be self-explanatory (feel free to reach out if not) with the exception of flux_err_quad which is a more conservation flux uncertainty I defined for the lightcurves as: $$((flux_{err}^2) + (rms^2) + (0.06*flux)^2)^{0.5}$$

Updates:

October 8th 2025:

We overhauled casda_download() to optimise the process for speed.

• Chunked batch staging (10x faster than individual staging); chunk_size variable defined within function
• Parallel downloads for simultaneous file retrieval (3-5x faster); max_workers variable exposed to user
• Vectorized coordinate matching in data_filter() (2-5x faster)
• Dictionary lookups instead of DataFrame filtering (10-100x faster for lookups) Overall: 20-100x speedup depending on your specific use case. For benchmarking purposes you can set optimized=False.

November 13th 2025:

Added a CASDA_Cutout notebook to get image cutouts:

• Similarly optimised with batch staging and parallel downloads
• Currently available only on dev branch
• Single source processing has been tested
• Bulk downloads is untested and not recommended unless you want to fill up all of your storage