A climatology of global sea surface temperature (SST), based on observations for the 1951-80 period, is presented in the form of maps on a 5 deg. latitude X longitude grid. A 1 deg. latitude X longitude grid is used for selected regions. The data are based on the Meteorological Office Main Marine Data Bank (MOMMDB), coupled in certain regions after 1960 with the Consolidated Data Set (CDS) of the US Fleet Numerical Oceanography Center, which together amount to about 60 million SST reports. Monthly anomalies from the 30-year means are presented for the periods 1968-77 and 1982-83. Climatological values of air minus sea temperature and air-sea energy exchange are also presented. A development of the latter was used to apply approximate provisional corrections to analysed monthly mean SST values for the 1856-1941 period, which are believed to have been based mainly on uninsulated bucket data. Further refinements suggested that these provisional corrections were slightly too positive before about 1900 and slightly too low in the early 20th Century. Decadal seasonal mean anomalies are printed for 1866-1985 based on the provisional corrections, but a set of seasonal time-series and low-pass filtered time-series of SST for 1856-1989 for selected regions, including one representative of the entire world ocean, incorporates the refined corrections, as well as being based on an updated MOMMDB. These time-series are presented in parallel with series of corrected nighttime marine air temperatures for comparison. The largest fluctuation in marine temperature is a substantial temperature drop around 1902-03 with the low values persisting for about 10 years, and this was evident in most regions. Information is included on geographical patterns of data coverage, along with statistics of the quantities of data underlying the climatology and other fields. Standard deviations and l-month lag correlations of the monthly SST fields for 1951-80 are also presented. All values in the Atlas Plates, with the complete 1 deg. resolution climatology are available on a documented CD-ROM disk or magnetic tape. The disk and tape also hold, if different, corresponding values based on the updated MOMMDB and with the refined corrections to SST incorporated where appropriate. In addition, they hold complete sequences of uncorrected and corrected monthly SST anomalies for 1859-1989 on a 5 deg. latitude X longitude grid derived from this updated data bank, with the refined corrections also in a separate file.


We are grateful for support in data-processing from S. Ary, J. Fried, F.B. Tourneur and Wu Z.-X. (MIT); A. Jackson, R.J. Lavery, H. Lawes, R. Marles, J. Minhinick and P.M. Stephenson (Meteorological Office). Dr P. Caton (Meteorological Office) designed much of the analysis and quality-control schemes. Constructive comments were provided by Dr R. Selkirk (MIT). Administrative support was provided by K. Huber and S. Black (MIT). L. Cohen of MIT Press obtained very helpful peer reviews for the Atlas.

Sea-air interaction studies at MIT have been funded since 1974 by the Climate Dynamics Program of the US National Science Foundation, most recently under Grant ATM-8517107, and the US Department of Energy. The Foundation requests that we make the following statement on behalf of the US authors: 'Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the Foundation'. The US Department of Energy under Grant DOE ACO2 76EV12195 funded the purchase by MIT of computational and graphic facilities necessary for calculation of the heat fluxes, and for production of the maps in the Atlas. MIT obtained financial assistance from the US Naval Ocean Research and Development Activity, to complete compilation of the Atlas. The UK Department of the Environment and the UK Department of Energy provided funding to the Meteorological Office for the publication of the Atlas, under the Intergovernmental Panel on Climate Change (IPCC) contract. Production of the CD-ROM disk was funded through the US National Climate Program Office under the IPCC contract.

Last, but not least, we acknowledge with gratitude the many 'unknown mariners', without whose perseverance and courage in braving the ocean's fury to make the observations, this Atlas would not have been possible.


A number of endeavours in meteorology and oceanography require information about global sea surface temperature patterns and their variation with time: the planning of oceanographic expeditions designed to study particular phenomena such as upwelling and oceanic productivity; the reduction of remote sensing data from satellites which attempt to measure sea surface temperatures; the construction of numerical models of the ocean and atmosphere, both separately and jointly; the development of schemes to relate sea surface temperature to succeeding and preceding atmospheric patterns, which are part of the larger field of long-range forecasting; investigation of the El Niño-Southern Oscillation phenomenon which in many regions gives rise to the largest time variability in the tropical atmosphere and ocean; the computation of the exchange of energy between the atmosphere and ocean which is an important factor in understanding climatic fluctuations; the establishment of any systematic variations with time in the observations such as those which may be associated with the injection of trace constituents into the atmosphere, e.g. aerosol from volcanoes and carbon dioxide from the burning of fossil fuel or biomass; studies of the linkage between sea surface temperature patterns and certain large-scale processes underlying e.g. drought in the Sahel or the Indian Monsoon; the reduction of palaeo-oceanographic observations, such as those made on deep sea cores or corals, to environmental parameters; and the exploration of the general field of ocean fluctuations, or which the surface temperature patterns are only one manifestation.

All of these efforts require a global sea surface temperature field whose values may be compared from one place to another and from one time to another. The global fleet of merchant and naval vessels make and record observations of sea and air temperature every 6 hours. A major problem is to take these observations of sea water temperature, made from a variety of different types of ships using methods of measuring which have varied with time, and produce as far as possible a homogeneous data set. While the largest signal in the time variation of sea surface temperature, the El Niño, may give rise to departures from long-term averages amounting to 6 deg. C and persisting for several months in certain areas of the tropical Pacific, the temperature variations in zonal seasonal averages in middle and high latitudes on a 100-year time-scale are less than one degree Celsius; thus the utmost care has to be taken in the construction of the averages and the deviations from the averages. Much of the text which accompanies the Atlas is devoted to a description of the procedures which have been devised to minimize this problem.

The work leading to this Atlas began at Massachusetts Institute of Technology (MIT) as a study of air-sea interaction over the Pacific, encouraged by the papers of Namias (1975), but the investigations were soon extended to the globe using sea surface temperature data provided by the Meteorological Office at Bracknell. The paper of Sawyer (1965) on physical causes of long- term atmospheric circulation anomalies, in which the influence of sea surface temperature anomalies was shown to be substantial, was a major stimulus to the Meteorological Office work of compiling reliable data sets of sea surface temperature from their Main Marine Data Bank, as were the subsequent papers of Murray and Ratcliffe (1969) and Ratcliffe and Murray (1970) on the use of sea surface temperature for long-range forecasting. In 1980 MIT acquired a separate global marine data set covering the period 1949-79 (the so-called Consolidated Data Set or CDS) from the US Navy Fleet Numerical Oceanography Center at Monterey, California. The CDS was used to construct a new set of sea surface temperature averages as well as energy flux values at the ocean-air interface. The latter was possible because cloudiness, relative humidity and wind data were included in the CDS. These parameters are also available in the Meteorological Office Main Marine Data Bank (MOMMDB), but have not yet been analysed, and in the Comprehensive Ocean-Atmosphere Data Set (COADS) which was released by Slutz et al. in 1985. During 1982-83 the Meteorological Office marine temperature data sets were improved and updated. This entailed re-extraction of the data from the MOMMDB, which had become fully operational, followed by application of revised quality- control criteria. Gaps in the MOMMDB data were filled using MIT's CDS values. By this time it was clear that some of the corrections needed for reduction of the older sea surface temperature data obtained from uninsulated buckets could be established to a useful extent by the use of heat-loss rates calculated using data derived from the CDS. Application of a first attempt at these corrections was completed in 1987, along with the creation of a globally complete blended climatology based largely on the quality-controlled MOMMDB data.

Completion of the final text represents a milestone in the collation and application of sea surface temperature data, and in the beneficial collaboration between the Meteorological Office and MIT.


The original version of this Atlas was sent out for peer review by L. Cohen of MIT Press. The reviews were supplied to us in May 1987 and a revised text, taking account of the reviewers' comments, was dated February 1988 and was passed to MIT Press in anticipation of US funding. This funding had not materialized before November 1989 when an arrangement was made to print the Atlas at HMSO using funds allocated to the Intergovernmental Panel on Climate Change (IPCC) by the UK Department of the Environment and the UK Department of Energy.

Further refinements were made to the corrections in 1989, and the final time-series presented in Plates 292-301 have incorporated these new corrections. However the decadal seasonal means in Plates 244-291 make use of the previous set of corrections.