CO 2 concentration data in the atmosphere are widely known to possess a seasonal cycle, largely due to plant photosynthesis and respiration, superimposed upon an upward trend that is largely due to increasing fossil fueI use. In this paper we assess the information contained in the seasonal component of atmospheric CO 2 data by applying modern techniques of time series decomposition to monthly average CO 2 observations at three locations. At Mauna Loa and South Pole, which have the longest time series, the amplitudes of the seasonal components are found to be increasing with time, from • 5.6 ppm in 1958 to •6.2 ppm in 1978 at Mauna Loa and from • 1.0 ppm in 1965 to ~ 1.3 ppm in 1978 at the South Pole. We consider four possible causes of the CO 2 seasonal behavior--changes in the seasonal pattern of fossil fuel use, increasing vegetation, increasing global photosynthetic activity, and changes in ocean temperature--and conclude that it is most likely that the CO 2 seasonal behavior reflects an increase in global photosynthetic activity. 1. INTRODUCTION Many geophysical time series contain seasonal variation. One example is CO2 concentration data in the atmosphere, which are widely known to possess a seasonal cycle, largely due to plant photosynthesis and respiration, superimposed upon an upward trend that is largely due to increasing fossil fuel use [Keeling et al., 1976a, b]. The upward trend has received great publicity because of predictions that further increases in CO2 may have the potential to produce changes in global climate [Hansen et al., 1981; Kukla and Gavin, 1981]. Although the seasonal behavior of the data has been studied as well [Machta, 1972; Hall et al., 1975; Bacastow et al., 1981a, b; Pearman and Hyson, 1980, 1981], it has received somewhat less attention than the upward trend since the seasonal component has been regarded as a relatively stable periodic fluctuation of the concentrations. However, because the seasonal oscillations reflect, at least in part, the effect of vegetation on atmospheric CO2, they provide important information about the properties of plant processes on the earth.During the past several decades, substantial effort has been devoted to developing statistical procedures for describing seasonal variation in time series. Thus far the major area of application has been economic time series [Zellner, 1978], but the methodology is not tied to economic applications and has potential widespread applicability in many other disciplines, including geophysics.In this paper we will describe SABL [Cleveland and Terpenning, 1981; Cleveland et al., 1982], a recently developed set of procedures for decomposing time series into three components: trend, seasonal, and the remaining variation, which is called the irregular. We will use the SABL methodology to decompose atmospheric CO2 in order to assess the seasonal variation.