Multi-model means from 1920-2099 for continuous ice-free period (IFP), last retreat day (LRD) and first advance day (FAD) using either a delta-shift bias correction (BC) or weighted mean (WM). Three regions are used: Hudson Bay (including James Bay, but excluding Hudson Strait and Foxe Basin), western Hudson Bay, and southern Hudson Bay. The latter two regions are commonly used for dividing the polar bear population in Hudson Bay. (e.g., https://polarbearagreement.org/polar-bear-biology/population-distribution-and-trends)
The bias correction is conducted by comparing the observational average of sea ice phenology from the 1979-2021 passive microwave record to the years in each model for which the temperature anomaly relative to 1850-1900 is in the same range as recorded by the Berkeley Earth surface temperature dataset. For each model simulation, the difference in average IFP, LRD, and FAD for those years sharing the same temperature anomaly range is considered the bias, and this bias is subtracted from all years of data before calculating a multi-model mean from the first ensemble member (replicate) of each model.
The weighted mean has two factors. First, a model simulation receives more weight if it shows better performance (relative to the passive microwave record for sea ice phenology and the Berkeley Earth surface temperature record) for that same period 1979-2021. Second, model simulations that show more independence from other simulations also receive more weight. In this way, it is possible to use multiple simulations (up to 7) from each participating model. The bias-corrected mean uses 20 simulations total, whereas the weighted mean uses 49.
| Colonne | Type | Intitulé | Description | Units | Media type | Result Value Type | Statistic Applied |
|---|---|---|---|---|---|---|---|
| Region | text | ||||||
| Year | text | ||||||
| IFP-BC | text | ||||||
| LRD-BC | text | ||||||
| FAD-BC | text | ||||||
| IFP-WM | text | ||||||
| LRD-WM | text | ||||||
| FAD-WM | text | ||||||
| taglobal | text | ||||||
| tareg | text |