Comparison and evaluation of existing detection methods and correction methods for annual to monthly data.
- Creation of software to test the different methods: adaptation of existing software, creation of new software if necessary, implementation of evaluation tools (statistical or graphical).
- Test runs over the benchmark-dataset: real series, simulated time series.
- Evaluation of test results: performance over real Series (comparison with metadata, spatial coherency), performance over simulated time series (in terms of level, power, robustness).
- Ranking of the detection methods: partial rankings (accounting for different Climate - Variables, network densities, etc), overall ranking.
- Redaction of practical recommendations for detection.
- Dissemination of the results: uploading to website, scientific paper in refereed journal.
- Evaluation of the results of D.1: selection of methods to be considered.
- Considerations over detection methods: should any detection be performed using daily data?
- Creation of software to test the different methods: adaptation of existing software, creation of new software, selection/creation of evaluation tools (statistical, graphical).
- Testing over the Benchmark Dataset of existing correction methods/strategies: interpolation of monthly factors, distribution-oriented methods, approaches based on two-phase regression.
- Development of a new correction method: evaluation of tests results, determination of strengths and weaknesses of the existing methods, theoretical definition of the new method, implementation: software suite, handbooks, testing, debugging, etc.
- Dissemination of the new method: upload to the Action’s web site the software suite and handbooks, scientific paper in refereed journal, dissemination in meetings, preparation of pre-formatted seminars for future dissemination.
COST Benchmark Workshop Zurich
Sept. 14, 2010.
Participants: Victor Venema, Petr Štepánek, Gregor Vertacnik, Tamás Szentimrey, Jose A Guijarro, Barbara Chimani, Renate Kocen, Michele Brunetti, Lars Andresen.