Workflow: A series of orchestrated and repeatable processes defined for a modeling system (usually means supercomputer jobs at NCEP) supported by systematic organization of resources. (Example: Global Workflow) This concept could both mean the workflow of a particular modeling system in general, or an instantiated workflow of that system.

CROW: Acronym of “Configuration manager of Research and Operational Workflows”. A Python 3.6 based toolbox designated to configure and generate workflows for weather industry.

Global Workflow: A comprehensive atmospheric modeling system consisting of numerical forecast and data assimilation providing weather guidance for the entire surface of the earth.

CROW-based workflow: A workflow that is configured and generated by CROW.

Workflow manager: A software that orchestrates workflow in a controlled environment, usually a supercomputer system. (Example: Rocoto, ecFlow)

Workflow Cycle (“Cycle” there-after): Execution of a series of orchestrated activities of all or a subset of a workflow. Usually defined based on the model hour.

Workflow Task (“Task” there-after): A certain functional activity(ies) performed within a workflow. Often defined by a series of scripts, an application, and a resource allocation.

Workflow Job (“Job” there-after): An instantiation of workflow task within a workflow at a given cycle.

Legacy configuration system: Existing configuration system for global workflow. A manual approach that requires users to edit configuration on a job-by-job basis.

Experiment directory: A directory created during the configuration process, storing configuration settings of the entire workflow.

Configuration Suite: A python object created during the execution of CROW defining all configuration settings of a workflow.

Workflow Suite: A file or file system generated by CROW defining every job within a workflow, the dependencies between them, and resource specification. This is the file(s) passing to the workflow manager (Rocoto, ecFlow)

COMROT/ ROTDIR: rotating directory. Subdirectories are organized in $RUN.$HOUR convention (ie: gdas.00, gfs.06 ). Model input and output, and restart files for certain hours will be stored here.

RUNDIR/ DATAROOT: run-time directory. Subdirectories are created for each job. All input file, restart file, and executables will be soft-linked to the corresponding subdirectory under this place. Namelists and temporary outputs will be generated here. When a job has finished successfully, outputs will be copied to ROTDIR.