Abstract A time series of 23 Landsat Thematic Mapper (TM) band 6 thermal infrared images over the period 1986–1996 is used to quantify variability of sea surface temperature (SST) along the central coast of Maine, a morphologically complex region of bays, estuaries, and islands. An iterative regression scheme using coregistered, temporally coincident, daily composites of Advanced Very High Resolution Radiometer (AVHRR) Pathfinder SST data is used to scale the TM digital numbers in each scene to SST, approximating an atmospheric correction. This approach provides temporally concurrent match-ups, even for Landsat scenes more than 10 years old and over 1000 data points to most regressions. Analysis of the TM scenes by year–day delivers temporal resolution sufficient for insight into overall seasonal pattern and allows identification of recurring seasonal features within the study area. The dominant seasonal patterns is a cross-shelf SST gradient of coldest water nearshore in winter which reverses sign in summer and disappears in spring and fall. Differences in summer SST are evident between four adjacent bays, attributable to differences in residual circulation, freshwater input, and flushing. Recurrent frontal zones evident in summer are identified and compare well to available but noncoincident in situ hydrographic data.