Paper/Poster/Presentation Title

Maintaining the integrity of a complex model

Keywords

software process testing

Start Date

25-6-2018 2:00 PM

End Date

25-6-2018 3:20 PM

Abstract

Developing and testing complex models is difficult and time consuming so it is vital that once a model has reached an acceptable level it remains that way for the life of the model. This is particularly critical for models that undergo continuous maintenance and improvement. All too often, extensive effort is invested into model development and testing only to see the model performance drift over time, due to science and software enhancement. It would be good for the performance of a model to be visible to users, developers and the model’s governing body at any point in the life of the model.

The APSIM Initiative have developed a methodology to ensure this happens. This involves the use of tools such as version control, continuous integration, calculation of a broad range of model performance statistics, and web dashboards. Just as importantly, a process has been developed that links these tools to create a testing regime that all developers of APSIM must follow. It includes peer review of source code and science, performance metrics for all models freely available on the web and guiding principles to help developers commit new changes to models as painlessly as possible.

The extended abstract will provide more detail on the tools and processes that help maintain the integrity of a model.

Stream and Session

F5: New and Improved Methods in Agricultural Systems Modelling

Share

COinS
 
Jun 25th, 2:00 PM Jun 25th, 3:20 PM

Maintaining the integrity of a complex model

Developing and testing complex models is difficult and time consuming so it is vital that once a model has reached an acceptable level it remains that way for the life of the model. This is particularly critical for models that undergo continuous maintenance and improvement. All too often, extensive effort is invested into model development and testing only to see the model performance drift over time, due to science and software enhancement. It would be good for the performance of a model to be visible to users, developers and the model’s governing body at any point in the life of the model.

The APSIM Initiative have developed a methodology to ensure this happens. This involves the use of tools such as version control, continuous integration, calculation of a broad range of model performance statistics, and web dashboards. Just as importantly, a process has been developed that links these tools to create a testing regime that all developers of APSIM must follow. It includes peer review of source code and science, performance metrics for all models freely available on the web and guiding principles to help developers commit new changes to models as painlessly as possible.

The extended abstract will provide more detail on the tools and processes that help maintain the integrity of a model.