While satellite observations of sea ice extent and concentration are available from 1979, long-term high quality (daily and high spatial resolution) observations of sea ice thickness remain limited as a result of few satellite and in situ observations. Reconstructions using numerous observational sources show a 65% decline in annual mean sea ice thickness in the central Arctic since the 1970s (Lindsay and Schweiger , 2015). Existing observations of sea ice thickness can differ through spatial and temporal coverage, measurement uncertainties, and methods of estimation.

In the last decade, satellite-based altimetry has improved our spatial and temporal coverage of sea ice thickness. Prior to these satellite observations, spare sea ice thickness data are available from submarine upward looking sonar and moorings. Limited data is even available from 1947 in the Canadian Arctic Archipelago. More recently, ERS-1/ERS-2 (1993-2001), ICESat (2003-2009) and now CryoSat-2 (2010-2017) provide nearly pan-Arctic observations of sea ice thickness. These altimetry-based estimates of sea ice thickness can be derived by measuring the height above the water level (freeboard), using snow and sea ice densities, estimating the snow depth on top of the ice, and assuming hydrostatic equilibrium. Satellites estimates of sea ice thickness can also be derived from brightness temperature, such as in the Soil Moisture and Ocean Salinity (SMOS) Mission. Meanwhile, estimates of sea ice thickness from upward-looking sonar and mooring data measure ice drift before converting to sea ice thickness. This can also induce uncertainties and errors through changing bottom roughness.

Through these various observational data sets and methodological assumptions, sea ice thickness remains one of the more poorly observed variables in the Arctic. Sources of error and uncertainty include: freeboard measurements, snow depth, and density estimates for sea ice and snow (Zygmuntowska et al., 2014). Most satellite sea ice thickness data is not available during the melt season due to the formation of melt ponds.

As a result of the limited temporal and spatial estimates of sea ice thickness, ice-ocean models with data assimilation are also a useful tool in providing sea ice thickness “reanalysis.” The Pan-Arctic Ice-Ocean Modeling and Assimilation System (PIOMAS) is a coupled ice and ocean model with sea ice thickness data available over the satellite era (from 1979) (Zhang and Rothrock , 2003). PIOMAS has been widely validated against sea ice thickness data sets (such as ICESat), and its uncertainties are addressed in Schweiger et al.(2011). PIOMAS has the capabilities of assimilating sea surface temperatures, sea ice concentration, and sea ice velocity data. The model is driven by atmospheric surface forcings from daily mean NCEP-NCAR (R1) reanalysis.

The following list of data sets and links provide a starting point for using sea ice thickness data. Particularly, the “Unified Sea Ice Thickness Climate Data Record” is a comprehensive archive of various sea ice thickness observational data sets, which are available at a 50 km resolution from 1947 to 2017 (Lindsay, 2010). The archive identifies each data set type along with any associated uncertainty statistics and additional concise documentation.

The use of these data sets requires an understanding of the documentation and potential sources of error. For example, estimates of snow depth on the top of sea ice often use a climatology from Warren et al. (1999), which may not reflect the actual snow depth at the time of sea ice thickness estimation. Comparisons and validations between many of these sea ice thickness products can be found in recent studies (e.g., Stroeve et al., 2014; Wanget al., 2016).##