Data processing pipelines

All images are processed by the open-source LSST Science Pipelines.

Prompt processing

Every night as new ~30 second exposures are obtained, prompt processing and alert production run automatically.

Within 60 seconds, the following processing is done:

• instrument signature removal (image reduction)
• image characterization (e.g., point-spread function)
• astrometric and photometric image calibration
• template generation and image differencing
• difference-image source detection and measurement
• association of difference-image sources with known static and moving objects
• (re-)evaluation of time-variability parameters for static objects
• alert production and transmission to brokers
• forced photometry at the locations of known time-domain objects

Within 24 hours, the Prompt Products Database (PPDB) is updated with the same data as were released in the alerts. Difference-image source linking is run to discover new moving objects. Detections of moving objects are ingested by the Minor Planets Center and orbital parameters are refined.

After 80 hours, the promptly-processed images and difference images become available to users.

Data Release processing

Every year the entire set of LSST raw images obtained to date are reprocessed with the most up-to-date version of the LSST Science Pipelines. This annual Data Release processing is often referred to as DRP.

Annual DRP includes almost all of the steps listed above under Prompt processing, except alert production and transmission. However, since the 60-second and 24-hour limits do not apply to the annual processing, the underlying algorithms can be different (i.e., more resource-intensive algorithms can be used).

The following processing is done on annual timescales only:

• image stacking to create deep coadds
• deep coadd image characterization (e.g., point-spread function)
• source detection, deblending, and measurement in deep coadds
• forced photometry in individual images for all objects detected in the deep coadds
• photometric redshift evaluation for objects in deep coadds
• the generation of survey property maps

Unlike the products of Prompt processing, the DRP data products do not change after release.

Special processing

The above Prompt and Data Release processing applies to all standard images that are obtained as part of the LSST's Wide Fast Deep main survey.

For non-standard or "special programs", such as alternative survey strategies or regions, special processing will be done (e.g., the creation of very deep stacks in the deep drilling fields).

Go to the technical documentation for special programs processing.

User processing

As the LSST Science Pipelines are open-source, anyone may use them to re-process LSST data to create custom data products.

For example:

• sub-threshold source detection (<5-sigma)
• custom-depth coadded images (e.g., monthly)
• alternative sky background subtraction methods

Additional resources

In-depth descriptions of the processing done with the LSST Science Pipelines are provided in the data release documentation that accompanies each data release.

Questions?