Analysing ALMA data with CASA

The radio astronomical data analysis package CASA was selected to be the designated tool for observers to analyse the data from the Atacama Large mm/sub-mm Array (ALMA) which is under construction and has recently started taking its first science data (Cycle 0). CASA is a large package which is being developed by NRAO with major contributions from ESO and NAOJ. Generally, all radio data from interferometers and single dish observatories can be analysed with CASA, but the development focuses presently on the needs of the new observatories EVLA and ALMA. This article describes the main features of CASA and the typical analysis steps for ALMA data.

💡 Research Summary

The paper presents a comprehensive overview of the Common Astronomy Software Applications (CASA) package as the designated analysis environment for early science data from the Atacama Large Millimeter/sub‑millimeter Array (ALMA). Developed primarily by the National Radio Astronomy Observatory (NRAO) with substantial contributions from the European Southern Observatory (ESO) and the National Astronomical Observatory of Japan (NAOJ), CASA is a large, modular software suite written in C++ with a Python front‑end, designed to handle data from interferometers and single‑dish telescopes alike. The authors first motivate the need for a unified, modern data‑reduction platform as ALMA moves from construction into Cycle 0 operations, emphasizing that the sheer volume and complexity of ALMA data demand a robust, extensible tool chain.

CASA’s architecture is built around a “tool‑task” paradigm. Low‑level tools expose the underlying data structures—most importantly the Measurement Set (MS) format—while high‑level tasks combine these tools into user‑friendly commands. Examples include importasdm for converting ALMA’s native ASDM format into an MS, listobs for summarising observation metadata, and flagdata for masking radio‑frequency interference (RFI) or antenna failures. This separation allows developers to extend functionality without breaking existing scripts, and it enables astronomers to write reproducible Python pipelines.

The paper then walks through the typical ALMA reduction workflow, which can be divided into four stages: (1) initial data inspection and flagging, (2) calibration, (3) imaging, and (4) spectral‑line analysis. In the flagging stage, flagdata is used interactively or automatically to excise corrupted visibilities; setjy establishes an absolute flux scale based on standard calibrators. Calibration proceeds with gaincal (time‑ and frequency‑dependent complex gains), bandpass (frequency response), and fluxscale (bootstrapping the absolute scale). Although ALMA supplies a default pipeline script, users retain full control via applycal, allowing custom solutions for problematic datasets.

Imaging is performed primarily with tclean, a highly configurable implementation of the CLEAN algorithm. tclean supports multi‑scale deconvolution, multi‑frequency synthesis (MFS) for wide‑band continuum imaging, and automatic masking (auto-multithresh). Parallel processing options enable the handling of terabyte‑scale datasets within reasonable wall‑clock times. Post‑imaging analysis uses tasks such as immath, imstat, and imfit to compute statistics, perform arithmetic on images, and fit source models.

For spectral‑line work, the authors highlight uvcontsub, which subtracts continuum emission directly in the uv‑plane, and specfit, which fits Gaussian or Voigt profiles to extracted spectra. Interactive visualization tools like viewer and plotms provide channel‑by‑channel inspection, facilitating quality assessment and scientific interpretation.

A notable strength of CASA is its scriptability. The standard ALMA pipeline is delivered as a Python script (pipeline.py) that can be edited, re‑run, or incorporated into Jupyter notebooks. This flexibility encourages reproducible science and allows observers to tailor the reduction to specific scientific goals, such as deep field imaging or high‑dynamic‑range spectral line studies.

Finally, the paper outlines future development directions: GPU‑accelerated imaging kernels, machine‑learning‑based RFI detection, and extensions to support the upcoming Square Kilometre Array (SKA). The authors stress the importance of community involvement through mailing lists, a public GitHub repository, and continuously updated documentation and tutorials.

In conclusion, CASA provides a unified, end‑to‑end environment for ALMA data reduction, covering everything from raw visibility flagging to high‑fidelity imaging and spectral analysis. By detailing the standard workflow and offering concrete script examples, the paper serves as a practical guide for ALMA Cycle 0 users and establishes a foundation for the increasingly sophisticated data challenges that will accompany the full operation of ALMA.