Introduction to the JWST Pipeline


The JWST Science Calibration Pipeline processes data from all JWST instruments and observing modes by applying various science corrections sequentially, producing both fully-calibrated individual exposures and high-level data products (mosaics, extracted spectra, etc.). The pipeline is written in Python, hosted open-source on Github, and can be run either via command line interface (strun) or via the Python interface.

The full end-to-end ‘pipeline’ (from raw data to high-level data products) is comprised of three seperate pipeline stages that are run individually to produce output products at different calibration levels:

Stage 1:

Detector-level corrections and ramp fitting for individual exposures.

Stage 2:

Instrument-mode calibrations for individual exposures.

Stage 3:

Combining data from multiple exposures within an observation

As such, the term ‘pipeline’ may refer to a single pipeline stage or to the full three-stage series.

Because JWST has many different instruments and observing modes, there are several different pipeline modules available for each stage. There is one single pipeline for Stage 1 - corrections are applied nearly universally for all instruments and modes. There are two pipeline modules for Stage 2: one for imaging and one for spectroscopic modes. Stage 3 is divided into five separate modules for imaging, spectroscopic, coronagraphic, Aperture Masking Interferometry (AMI), and Time Series Observation (TSO) modes. Details of all the available pipeline modules can be found at Pipeline Modules.

Each pipeline stage consists of a series of sequential steps (e.g, saturation correction, ramp fitting). Each full pipeline stage and every individual step has a unique module name (i.e Detector1Pipeline, or DarkCurrentCorrection). Steps can also be run individually on data to apply a single correction. The output of each pipeline stage is the input to the next, and within a pipeline stage the output of each step is the input to the next.

The pipeline relies on three components to direct processing: input data, step parameters, and reference files. The inputs to the pipeline modules are individual exposures (fits files) or associations of multiple exposures (asn.json files). The parameters for each pipeline step are determined hierarchically from the parameter defaults, parameter reference files, and any specified overrides at run time. Finally, reference files provide data for each calibration step that is specific to the dataset being processed. These files may depend on things like instrument, observing mode, and date. In both the command line and Python interface, a pipeline or step module may be configured before running. Reference files can be overridden from those chosen by CRDS, steps in a pipeline can be skipped, step parameters can be changed, and the output and intermediate output products can be controlled.

A pipeline (or individual step) outputs corrected data either by writing an output file on disk or returning an in-memory datamodel object. The output file suffix (i.e cal.fits, rate.fits) depends on level of calibration - each full pipeline stage as well as each individual step have a unique file suffix so that outputs may be obtained at any level of calibration. Other pipeline outputs include photometry catalogs and alignment catalogs (at stage 3).

Overview of Pipeline Code

The following is a brief overview of how the pipeline code in jwst is organized.

Pipeline and Step Classes

The JWST pipeline is organized into two main classes - pipeline classes and step classes. Pipelines are made up of sequential step classes chained together, the output of one step being piped to the next, but both pipelines and steps are represented as objects that can be configured and run on input data.

Detector1Pipeline  # an example of a pipeline class
DarkCurrentStep    # an example of a step class

Each pipeline or step has a unique module name, which is the identifier used to invoke the correct pipeline/step when using either the Python or the Command Line Interface.

Package Structure

Within the jwst repository, there are separate modules for each pipeline step. There is also a pipeline module, where the pipeline classes, consisting of step classes called in sequence, are defined.

        # contains AssignWcsStep
        # contains DarkCurrent Step
        # contains Detector1Pipeline
        # contains Image2Pipeline


The jwst package has several dependencies (see the pyproject.toml file in the top-level directory of jwst for a full list). Some notable dependencies include:


ASDF, the Advanced Scientific Data Format is the file format the JWST uses to encode world coordinate system (WCS) information.


GWCS, Generalized World Coordinate System - is an generalized alternative to FITS WCS which makes use of astropy models to describle the translation between detector and sky coordinates. In JWST data, WCS information is encoded in an ASDF extension in the FITS file that contains GWCS object. In contrast, FITS WCS is limited because it stores the WCS transformation as header keywords, which is not sufficient to describe many of the transformations JWST uses.


STPIPE contains base classes for pipeline and step, and command line tools that are shared between the JWST and Nancy Grace Roman Telescope (Roman) pipelines.


The stcal package contains step code that is common to both JWST and the Roman telescope, to avoid redundancy. All step classes for the JWST pipeline are still defined in jwst, but some of the underlying code for these steps lives in stcal if the algorithm is shared by Roman (for example, ramp fitting, saturation).