Algorithm for validation of thickness of thin sea ice
Validation vs. data derived from SMOS
SMOS quality screening
Thickness data derived from SMOS observations are not generally reliable, so these data are accompanied by error estimates. In order to filter out regions with unreliable thickness data, we take the advice from the data provider into account. From their web page we are informed that
Data with an uncertainty >1m or with a ratio between retrieved and maximum retrievable sea ice thickness near 100% must not be used.
Denoting the thickness by H and its uncertainty by dH, we have implemented the following requirement in the pre-processing of thickness data:
-
Keep data wherever
- dH < 1m
and - dH/H < 0.75
In order to perform a proper validation, the mask resulting from the data screening outlined above, as well as regions without observational coverage, is overlaid on the model results prior to all computation of validation metrics. Note that during the melting season, when no observational product is available, no validation is performed.
Thickness categories, thin sea ice
In order to summarize the comparison between observations and model results, we provide results in the form of contingency tables for selected sea ice thickness intervals. Their definitions are made so that the intervals spanned are smilar in range, and that the highest category will only contain few observations (to indicate how many model values are outside the observational range).
| Class | Thickness range |
|---|---|
| Open ocean | <0.01m |
| Category 1 | 0.01 - 0.3m |
| Category 2 | 0.3 - 0.6m |
| Category 3 | 0.6 - 1m |
| Category 4 | >1m |
Validation vs. data derived from CryoSat-2
CryoSat quality screening
Thickness data derived from CryoSat observations are generally reliable, with data typically to within few decimeters from in situ observations; see this overview for details and references.
The data product distributed by CPOM also has a quantification of the standard deviation of the data. In line with the strategy adopted for quality screening of the SMOS data, we have adopted the following:
-
Keep data wherever
- H_std/H < 0.75
In practice, this screening removes little data from the CPOM product. In order to perform a proper validation, the mask resulting from the data screening outlined above, as well as regions without observational coverage, is overlaid on the model results prior to all computation of validation metrics. Note that during the season when the presence of meltponds leads to a halt in the production of the observational product, no validation is performed.
Thickness categories, thick sea ice
The thickness categories used for validation of thick sea ice is tabulated below. Category limits are set based on a rough estimate of the distribution of observational data from CryoSat.
| Class | Thickness range |
|---|---|
| Open ocean | <0.1m |
| Category 1 | 0.1 - 1m |
| Category 2 | 1 - 1.5m |
| Category 3 | 1.5 - 2m |
| Category 4 | 2 - 2.5m |
| Category 5 | 2.5 - 3m |
| Category 6 | >3m |