Manual visual ship-based observations are often the only data source for sea-ice information in an otherwise data sparse region.
The data offered here have been quality checked and standardized. Standardized means that the observations of different ship cruises are combined such that observations of the same parameter occur in the same column. Missing columns are filled. The values used as missing value vary from parameter to parameter, depending on whether it is a floating point or integer variable and depending on whether the variable can have negative values.
Observations are usually conducted by voluntary observers with varying skills / expertise. This limits the accuracy of the observations.
Of the data included the total sea-ice concentration, given in tenth, can be regarded as the most accurate parameter; its accuracy is 5 to 10%. Accuracy is worst in the inter-mediate sea-ice concentration range from ~30% through ~70%. Partial concentrations of ice types of different thickness are possibly less accurate - especially when fractions of quite similar ice types such as thin first-year ice and medium-thick first-year ice are reported.
Next accurate are observations of sea-ice thickness and snow depth (see below though). Accuracies of the observations of other fractions, such as melt-pond fraction or fraction of ridged ice are ~10%. Ridge height observations are possibly the most difficult to evaluate and have a rather unknown accuracy.
Note that most ships tend to follow easy-to-navigate sea-ice conditions. Because of this sea-ice concentrations might be biased low during summer; the same applies all-year round for sea-ice thickness, snow depth, fraction of ridged ice and possibly also ridge height (see Worby, et al., J. Geophys. Res., 113, C06S91, 2008 and Kern, et al., The Cryosphere, 13, 3261-3307, DOI: 10.5194/tc-13-3261-2019, 2019).
I recommend to consider the quality of this data set as Level 2b (Geophysical Variable - basic quality control).