Skip to content
Simulation Modelling Integration Framework
Branch: master
Clone or download
Latest commit 61a855e May 7, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
ci Attempt reverting CI changes Mar 22, 2019
src/smif Merge branch 'master' into develop May 7, 2019
.appveyor.yml Drop conda-forge build from Appveyor Apr 11, 2019
.checkignore Exclude test files from QC analysis Nov 16, 2016
.environment.yml Add mypy to dev packages in environment.yml, ignore cache Mar 6, 2019
.gitignore Add mypy to dev packages in environment.yml, ignore cache Mar 6, 2019
.pre-commit-config.yaml Refactor smif around decisions and interventions Jul 16, 2018
.readthedocs.yml Bump numpy for RTD Oct 30, 2018
.travis.yml Configure database fixture setup/teardown, in 'test_smif' test db Nov 29, 2018
CHANGELOG.rst Update CHANGELOG May 7, 2019
CITATION.cff Update changelog and version before minor release Jan 29, 2019 Create Sep 20, 2017
CONTRIBUTING.rst Expand on contributing notes Dec 13, 2018
LICENSE.txt Added Roald as an author Feb 12, 2018
README.rst Fix whitespace May 7, 2019
mypy.ini Add mypy config file Mar 15, 2019
requirements.txt Add database connection dependencies Nov 29, 2018
setup.cfg Make spatial libraries optional (pip install smif[spatial]) Sep 10, 2018



Simulation Modelling Integration Framework

Travis CI build status Appveyor CI Build status Code Coverage PyPI package conda-forge package Archive Journal of Open Research Software paper Documentation Status

smif is a framework for handling the creation, management and running of system-of-systems models.

A system-of-systems model is a collection of system simulation models that are coupled through dependencies on data produced by each other.

smif provides a user with the ability to

  • create system-of-systems models
    • add simulation models to a system-of-systems model
    • create dependencies between models by linking model inputs and outputs
    • pick from a library of data adapters which perform common data conversions across dependencies
    • create user-defined data adapters for more special cases
    • add scenario data sources and link those to model inputs within a system-of-systems
  • add a simulation model to a library of models
    • write a simulation model wrapper which allows smif to run the model
    • define multi-dimensional model inputs, outputs and parameters and appropriate metadata
  • run system-of-systems models
    • link concrete scenario data sets to a system-of-systems model
    • define one or more decision modules that operate across the system-of-systems
    • define a narrative to parameterise the contained models
    • persist intermediate data for each model output, and write results to a data store for subsequent analysis

In summary, the framework facilitates the hard coupling of complex systems models into a system-of-systems.

Should I use smif?

There are number of practical limits imposed by the implementation of smif. These are a result of a conscious design decision that stems from the requirements of coupling the infrastructure system models to create the next generation National Infrastructure System Model (NISMOD2).

The discussion below may help you determine whether smif is an appropriate tool for you.

  • smif is not a scheduler, but has been designed to make performing system-of-systems analyses with a scheduler easier
  • Geographical extent is expected to be defined explicitly by a vector geometry
    • smif is not optimised for models which simulate on a grid, though they can be accomodated
    • smif is designed for models that read and write spatial data defined over irregular grids or polygons using any spatial format readable by fiona
  • Inputs and outputs are exchanged at the ‘planning timestep’ resolution
    • smif makes a distinction between simulation of operation, which happens at a model-defined timestep resolution, and application of planning decisions which happens at a timestep which is synchronised between all models
    • smif is not focussed on tight coupling between models which need to exchange data at every simulation timestep (running in lockstep)
    • smif does accomodate individual models with different spatial and temporal (and other dimensional) resolutions, by providing data adaptors to convert from one resolution to another
  • smif has been designed to support the coupling of bottom-up, engineering simulation models built to simulate the operation of a given infrastructure system
    • smif provides a mechanism for passing information from the system-of-systems level (at planning timesteps scale) to the contained models
    • smif is appropriate for coupling large complex models that exchange resources and information at relatively course timesteps
  • smif is not appropriate for
    • discrete event system simulation models (e.g. queuing systems)
    • dynamical system models (e.g. predator/prey)
    • equilibrium models without explicit timesteps (e.g. Land-Use Transport Interaction)
    • for simulating 100s of small actor-scale entities within a system-level environment

Installation and Configuration

smif is written in Python (Python>=3.5) and has a number of dependencies. See requirements.txt for a full list.

Using conda

The recommended installation method is to use conda, which handles packages and virtual environments, along with the conda-forge channel which has a host of pre-built libraries and packages.

Create a conda environment:

conda create --name smif_env python=3.6

Activate it (run each time you switch projects):

conda activate smif_env

Add the conda-forge channel, which has smif available:

conda config --add channels conda-forge

Finally install smif:

conda install smif

Installing smif with other methods

Once the dependencies are installed on your system, a normal installation of smif can be achieved using pip on the command line:

pip install smif

Versions under development can be installed from github using pip too:

pip install git+

To install from the source code in development mode:

git clone
cd smif
python develop

Spatial libraries

smif optionally depends on fiona and shapely, which depend on the GDAL and GEOS libraries. These add support for reading and writing common spatial file formats and for spatial data conversions.

If not using conda, on Mac or Linux these can be installed with your OS package manager:

# On debian/Ubuntu:
apt-get install gdal-bin libspatialindex-dev libgeos-dev

# or on Mac
brew install gdal
brew install spatialindex
brew install geos

Then to install the python packages, run:

pip install smif[spatial]

Running smif from the command line

Follow the getting started guide to help set up the necessary configuration.

To set up an sample project in the current directory, run:

$ smif setup

To list available model runs:

$ smif list

To start the smif app, a user-interface that helps to display, create and edit a configuration, run:

$ smif app

To run a system-of-systems model run:

$ smif run demo_model_run
Model run complete

By default, results will be stored in a results directory, grouped by model run and simulation model.

To see all options and flags:

$ smif --help
usage: smif [-h] [-V] {setup,list,run} ...

Command line tools for smif

positional arguments:
{setup,list,app,run}  available commands
    setup               Setup the project folder
    list                List available model runs
    app                 Open smif app
    run                 Run a model

optional arguments:
-h, --help        show this help message and exit
-V, --version     show the current version of smif


If you use smif for research, please cite the software directly:

Here's an example BibTeX entry:

      author       = {Will Usher and Tom Russell and Roald Schoenmakers and Craig Robson},
      title        = {nismod/smif vX.Y.Z},
      month        = Aug,
      year         = 2018,
      doi          = {10.5281/zenodo.1309336},
      url          = {}

Please also cite the software description paper:

  • Will Usher and Tom Russell. (2019) A Software Framework for the Integration of Infrastructure Simulation Models. Journal of Open Research Software, 7: 16 DOI:

Here's an example BibTeX entry:

      author       = {Will Usher and Tom Russell},
      title        = {A Software Framework for the Integration of Infrastructure Simulation Models},
      journal      = {Journal of Open Research Software},
      volume       = {7},
      number       = {16},
      pages        = {1--5},
      month        = May,
      year         = {2019},
      doi          = {10.5334/jors.265},
      url          = {}

A word from our sponsors

smif was written and developed at the Environmental Change Institute, University of Oxford within the EPSRC sponsored MISTRAL programme, as part of the Infrastructure Transition Research Consortium.

You can’t perform that action at this time.