All the ALMA Cycles 2, 3 and 4 MOUS (with project ID starting with "2013", "2015", "2016") that can be processed with the ALMA imaging pipeline and that have not already been processed during the QA2 stage will be processed with the ARI-L procedures. It is expected that, by the end of the project (in 2022), more than 70 % of them will have an associated ARI-L product folder ingested into the archive.

Currently, the latest version of the imaging pipeline cannot be applied to full Stokes, VLBI, total power, and Solar observational data; MOUS belonging to any of these classes will not have an ARI-L product folder associated. We also do not work with Total Power data.

Only datasets that are stored in the archive and accessible for public download (i.e. not in QA3 or under proprietary restriction during the duration of the project) with the QA2 'passed' or 'semi-passed' flag can be properly calibrated and are included in our project lists.

For each MOUS that is processed with the ARI-L procedures, the ARI-L products include the calibrated measurement set and the ARI-L imaging folder. See the following questions for details.

Each MOUS to be processed with ARI-L is retrieved from the ALMA Science Archive.

The information needed for the processing (e.g. the CASA version needed for calibration) is retrieved automatically from the downloaded data. Then, for the execution blocks (EBs) of each MOUS, calibration is performed with the proper CASA version using the prescriptions of the calibration scripts included in the downloaded folder. Assuming that the QA2 calibration script generates correct measurement sets, no change or verification is applied to the calibration stage.

The QA2-produced imaging scripts included in the script folder are not used except for extracting meaningful additional flagging commands. This may happen in manual imaging when the QA2 analyst realized that, at the end of the calibration, a portion of the data (i.e. some channels, a time range, an antenna, a baseline, etc.) are still misbehaving with respect to all other data, so the analyst decided to flag the misbehaving data after the calibration but before imaging. Such meaningful commands are extracted and applied in the ARI-L procedure after the calibration and before the imaging.

The most recent version of the pipeline for imaging is then applied to the measurement sets to generate images for each MOUS, combining all the execution blocks belonging to it.

A README file is automatically generated to include a summary of the processed data.

A Quality Assurance step dedicated to the ARI-L product is performed. If passed, the imaging folder is sent to ESO and JAO for ingestion into the archive as external products, while the calibrated measurement sets are stored in the IA2 repository.

The ARI-L products have a quality that is virtually always comparable to the currently archived products where they overlap. However, the ARI-L products include complete cubes for all the sources in the datasets, while often the QA2 manually imaged products show only a small portion of the images of the representative target.

All the ARI-L products will be quality checked before ingesting them in the archive. The ARI-QA procedure consists of three layers:

• A check will be performed to ensure that the ARI-L code run was performed correctly. In case of failures, ARI-L products (or portion of it) will not be ingested into the ASA.

• The weblog of the imaging pipeline products will be reviewed to verify that the pipeline has properly generated good quality images according to its heuristics (i.e. all the stages have been executed with a score larger than 0.9). Any discrepancy will be analyzed and, if not justified, the relative ARI-L products will not be ingested into the ASA.

• The portion of the data corresponding to what is available as QA2 products will be extracted and smoothed to the same resolutions, and then the rms noise, fluxes, and dynamic ranges will be compared between the products. Differences should be evaluated on the basis of the processes used during QA2, but in case differences are not justified and the reprocessed data have worse rms noise or dynamic range by more than 30% in all the comparisons or the imaged structure is clearly wrong, the ARI-L products will not be ingested into the ASA.

The ARI-QA only evaluate the quality of the produced images assuming that the calibration part has already been evaluated during the QA2 process. For this reason, we can compare our imaging products with the existing QA2 image.

In cases where only some of the images of an MOUS can pass the ARI-QA, even after accurate evaluations and attempts to recover the missing ones, the quality assurance process might consider to ingest into the ASA only the passed ones.

In case any feature is noticed in the products as a consequence of the ARI-QA, a note is added in the README file enclosed in the ARI-L product folder.

The ALMA QA2 aims to verify that the resolutions and sensitivity requested by the PI are reached. The QA2 analysts are requested to generate images developing dedicated scripts or using the imaging pipeline of at least a portion of the dataset, enough to verify that the requested conditions can be obtained.

ARI-L images have been through a distinct and less manually intensive QA compared to the products generated in the ALMA QA2 process. Image characteristics resulting from these distinct processes are typically comparable but can vary, particularly in cases where manual adjustments were required for QA2 (e.g., self-calibration or optimization of the continuum channel selection). The purpose of the ARI-QA process is to verify that the products are complete ans suitable for general archive user cases.

The ARI-QA is not intended to be a repetition of QA2 and does not change the QA2 outcome for an MOUS. It will only evaluate the quality of the produced images and, for each MOUS, will rely on the calibration as it is in the archived scripts. For this reason, once all the codes have successfully produced an image product, any quality comparison will be done with the existing QA2 images and without referring to the PI requests.

All the calibrated measurement sets produced during the ARI-L processing will be stored by the Italian node of the European ALMA Regional Centre, exploiting the capabilities offered by the Italian center for Astronomical Archives. They will be stored for the whole duration of the ARI-L project and for a minimum of 3 years after the project conclusion.

It is possible to request the calibrated MS via e-mail to help-desk@alma.inaf.it

In the ALMA Science Archive, projects with ARI-L products appear in the search table results with the flag 'ari-l' in the 'Collections' column .

Single FITS images, README and weblog files can be previewed and downloaded exploiting ALMA Archive preview system.

The ARI-L image are ingested into the ASA and distributed as externally-generated products and can be downloaded as "External Products" and open with the CARTA viewer.

The targets and calibrators images

For each source in an MOUS, including the calibrators, the ARI-L code produces:

• an image of the aggregated continuum

• an mfs image of each spectral window

• a cube at the native spectral resolution for each spectral window

All the images are produced with Briggs robust parameter equal to 0.5 unless the imaging pipeline indicates a different approach. The imaging pipeline can define automatically the most suitable weighting scheme if the beam values requested by the PI is available, which is not usually the case in the ALMA data from Cycles 1-4. Hence, there may be a few cases in which the robust parameter is different from 0.5, for datasets observed in later cycles. In all the cases the ROBUST keyword is stored in the header of the produced images.

In cases where only some of the images of an MOUS can pass the ARI-QA, even after accurate evaluations and attempts to recover the missing ones, the quality assurance process may ingest only the passed ones into the ASA.

The README file

All the product folders for each MOUS include an automatically generated README file including:

• a synthetic description of the purposes of the ARI-L products

• the description of the MOUS as taken from the archived QA folder README notes on the ARI-L process (including at least the CASA version used for processing)

• a table describing the content of the EBs included in the MOUS (one per line): sequential number of the EB, EB name, time range in MJD of the observations, array (7m, 12m), number of antennas, frequency range for each SPW in GHz, spectral resolution in MHz for each SPW, range of angular scales covered in arcsec, first, second and third quartile of the uv range in metres

• a table summarizing the properties for each source (target or calibrator; one per column) included in the MOUS: RA and DEC in deg, the sequential number of EB as in the previous table in which the source is included, the number of pointings (different from 1 in case of mosaics), largest angular scale and resolution of the observation (in arcsec), intent of the observation of the source

A general description of the project and its collaboration and the update of its status and all the documents can be accessed through the ALMA Science portal .

The project rationale, workflow and science applications are in the ALMAmemo \#614, and in the general ARI-L document, both available through the ALMA Science portal, or on the project website..

The use of ALMA data elaborated by ARI-L data must be acknowledged using the standard ALMA acknowledgement statement.

It is suggested and would be appreciated that the publication Massardi et al. (2021) is cited when ARI-L data are used.