IPCC Fourth Assessment Report: Climate Change 2007
Climate Change 2007: Working Group I: The Physical Science Basis

4.4.2 Sea Ice Extent and Concentration Data Sources and Time Periods Covered

The most complete record of sea ice extent is provided by passive microwave data from satellites that are available since the early 1970s. Prior to that, aircraft, ship and coastal observations are available at certain times and in certain locations. Portions of the North Atlantic are unique in having ship observations extending well back into the 19th century. Far fewer historic data exist from the SH, with one notable exception being the record of annual landfast ice duration from the sub-antarctic South Orkney islands starting in 1903 (Murphy et al., 1995).

Estimation of sea ice properties from passive microwave emission requires an algorithm to convert observed radiance into ice concentration (and type). Several such algorithms are available (e.g., Steffen et al., 1992) and their accuracy has been evaluated using high-resolution satellite and aircraft imagery (e.g., Cavalieri, 1992; Kwok, 2002) and operational ice charts (e.g., Agnew and Howell, 2003). The accuracy of satellite-derived ice concentration is usually 5% or better, although errors of 10 to 20% can occur during the melt season. The accuracy of the ice edge (relevant to estimating ice extent) is largely determined by the spatial resolution of the satellite radiometer, and is of the order of 25 km (recently launched instruments provide improved resolution of about 12.5 km). Summer concentration errors do lead to a bias in estimated ice-covered area in both the NH and SH warm seasons (Agnew and Howell, 2003; Worby and Comiso, 2004). This is an important consideration when comparing the satellite period with older proxy records of ice extent.

Distinguishing between first-year and multi-year ice from passive microwave data is more difficult, although algorithms are improving (e.g., Johannessen et al., 1999). However, the summer minimum ice extent, which is by definition the multi-year ice extent at that time of year, is not as prone to algorithm errors (e.g., Comiso, 2002).