Round Robin (Algorithm Comparison)
The Round Robin exercise for SST CCI was fully launched on 22 September 2011.
In this part of the SST CCI project, methods for estimating SST from clear-sky brightness temperature observations from space were objectively compared. On the basis of these comparisons, the best algorithm for implementation in the SST CCI project was chosen.
The exercise therefore involved both the project team and external participants and was open until 31 January 2012 (deadline for submission of results by participants).
The manager of the SST_CCI round-robin algorithm selection exercise was Gary Corlett (firstname.lastname@example.org). Gary was in charge of:
- information about accessing the multi-sensor matchup data set provided as the common data set for the exercise
- the document explaining the data contents
- the Round Robin Protocol, which sets out the procedures for involvement
An important aspect of the exercise were the steps taken to ensure an objective comparison of algorithms. Most importantly, subsets of the provided information were ear-marked for particular uses: as training data (for algorithm development, including and empirical tuning), testing data (for internal testing of results by participants) and selection data (reserved for calculation of selection metrics, not used in algorithm development). This concept is shown schematically below. For more information on the design of the exercise, see the presentation from the GHRSST 12 meeting in Edinburgh, June 2011.
Lastly, we encouraged participants to look regularly at the FAQ for information about issues raised by other participants about the RRDP.