This paper investigates the influence of atmospheric variables (net solar radiation, wind speed, precipitation and vapor mixing ratio, all of which are at or near the sea surface) on the annual and seasonal cycle of near surface air minus sea surface temperature (Tair–Tsst) over the global ocean. The importance order of these variables is discussed using several statistical methods and two global data sets. After demonstrating that neither Tair nor Tsst exhibit any skill in determining difference between the two, a regression tree model (the so-called Generalized, Unbiased, Interaction Detection and Estimation (GUIDE) algorithm) is used to investigate influences of the atmospheric variables mentioned above in regulating Tair–Tsst. Overall, net solar radiation (sum of net shortwave and longwave radiation) at the sea surface is found to be the most important variable in driving the seasonal cycle of Tair–Tsst over the global ocean when the nonlinear relationship between Tair–Tsst and atmospheric variables is taken into account. This is true for both annual and seasonal (May through August) or monthly (November and December) timescales. Similar to the GUIDE results, a simple linear regression analysis also confirms that the net solar radiation explains most of the variance in the seasonal cycle of Tair–Tsst over most (≈50%) of the global ocean. The importance of the net solar radiation in controlling Tair–Tsst is even more significant in the regions surrounding the Kuroshio and the Gulf Stream current systems. The results presented in this paper have various implications for air–sea interaction and ocean mixed layer studies.