[1] The retrieval of Arctic sea surface temperatures (SSTs) using satellite radiometric imagery has not been well documented owing to the paucity of match-ups with in situ data. SST algorithms developed in temperate regions lead to positive biases in high latitudes due to an overestimation of atmospheric IR absorption. The composite arctic sea surface temperature algorithm (CASSTA) presented in this paper was developed from concurrent satellite and shipborne radiometric data collected in the North Water Polynya between April and July 1998. This algorithm considers three temperature regimes: seawater above freezing, the transition zones of water and ice, and primarily ice. These regimes, which are determined by advanced very high resolution radiometer (AVHRR) calibrated brightness temperatures, require different calculations for temperature estimates. For seawater above freezing, a specific Arctic SST algorithm was produced through a linear regression of AVHRR against in situ data. Areas consisting mainly of ice use an established ice surface temperature (IST) algorithm. The transition zone uses a combination of the Arctic SST and IST algorithms. CASSTA determines the Channel 4 brightness temperature for each pixel in a calibrated AVHRR image and then applies the appropriate algorithm to create a thermal image. The mean deviation of CASSTA compared to in situ data was 0.17 K with a standard deviation of 0.21 K. This represents a significant improvement over SST values using McClain coefficients for temperate waters, which overestimate the same data set by an average of 2.40 K. Application of CASSTA to the North Water imagery gives superior results compared to existing SST or IST algorithms.Citation: Vincent, R. F., R. F. Marsden, P. J. Minnett, K. A. M. Creber, and J. R. Buckley (2008), Arctic waters and marginal ice zones: A composite Arctic sea surface temperature algorithm using satellite thermal data,