Borrowing terminology used in software testing, we will from now on consistently use the terms alpha, beta and release candidate testing to describe the significant stages of the pre-operational testing of the IFS. They are:
- Alpha testing: Initial testing by Research Department to establish the content and performance of the next IFS cycle for implementation in Operations. During this testing stage regular modifications and updates of the experiments are quite common and these changes may alter the model results. Users will not have access to these preliminary datasets.
- Beta testing: Once a cycle has passed this initial stage it will be passed to the Forecast Department for further tests to evaluate the potential impact on Operations. At this stage a technical test data set will be made available to users to test the technical impact on user's applications and tools. This testing stage is usually prone to minor updates, starts and restarts and users should not rely on the availability of these test data sets for their own tests before the next stage of testing has been reached. During this phase there may be a few changes that may alter the model results but these will be fully documented. Users will have access to part of the data generated during this phase.
- Release candidate testing: One month for normal changes, and three months before the planned implementation of a high-impact change (e.g. resolution upgrades) the test system will be frozen and the test suite will run close to real-time. There will be no changes that alter the model results. A full set of product services (e.g. dissemination of test data, ecCharts) will be offered during the whole period. Users will have access to all data generated during this phase. At this stage we will also declare which of the beta test data sets can be used to acquire longer time series of test data.
These terms apply to the IFS direct model output. Post processed forecast products and charts are gradually incorporated during the beta testing phase.