Configuration

ERT requires some set-up. For example:

• Distributions must be assigned to priors.

• The forward model must be described so that ERT can run it.

• Observations must be provided.

• The analysis algorithms you want to use must be parameterized.

• If running on an HPC cluster, you will have to tell ERT how to interact with the queue system.

There are prescribed ways to provide all of this information, and ERT has strict formatting requirements for its configuration file. However, it also provides convenient syntax and structure for dealing with almost any scenario.

This chapter sets out everything you need to know to configure ERT.

• Data types
  • Scalar parameters with a template: GEN_KW
  • 3D field parameters: FIELD
  • 2D Surface parameters: SURFACE
  • Simulated data
  • General data: GEN_DATA
  • EnKF heritage
• Forward model
  • Configuring your own jobs
• Observations
  • General overview
  • SUMMARY_OBSERVATION keyword
  • GENERAL_OBSERVATION keyword
  • Matching GEN_OBS and GEN_DATA
  • HISTORY_OBSERVATION keyword
• List of keywords
  • Table of keywords
  • Commonly used keywords
  • Keywords controlling the simulations
  • Parameterization keywords
  • Analysis module
  • Keywords controlling the ES algorithm
  • Advanced keywords
  • Keywords related to running the forward model
  • Workflow hooks
  • Manipulating the Unix environment
• Queue system
  • Options that affect all queue systems
  • On Crash or Exit
  • LOCAL queue
  • LSF systems
  • TORQUE and PBS systems
  • Slurm systems
• Magic strings
  • List of keywords
  • Experiment constants
  • Experiment variables
  • Experiment dependent variables
• Site wide configuration