How does it work ? : IcePac Regional Model (12.5km)
BASE DATA : OSI-SAF (Available from : http://osisaf.met.no/p/ice/index.html#conc-reproc)
1. Data is gathered for all days of the year, starting in october 1978 through the most up to date available dataset (Actually 2015).
2. Data is clipped to the spatial extent of the study area.
3. Time series of ice concentrations per day of the year (DOY) are gathered for every pixel of the study area.
4. Gaps are, when present (1-2 years max), filled with values generated from linear interpolation.
5. Datasets of mean values per week according to a stable (365 days - no leap) calendar are built.
6. Trend detection is run on weekly mean ice concentration time series using the Mann-Kendall test.
7. Trend is removed on all datasets using a Logit link linear regression.
8. Beta distribution is fitted on all detrended time series and model parameters (alpha and beta) are estimated using least squared estimator.
9. Model is exploited on every pixel for different probabilities (PDF) and concentrations (CDF) at a 5% pace [0-95%].
Note on calendar
The calendar used for the model is the 365 days - no leap calendar. Therefore, Feb. 29th is ignored in all calculations.