ATMOSPHERIC CARBON DIOXIDE AND RADIOCARBON: I 53 



Oceanography. Those data readily identified as contaminated by local biological, 

 volcanic, or combustive activity were edited from the record. The remaining data 

 were averaged first over each day and then by month. The course of the monthly 

 mean concentration, expressed as a mixing ratio in parts per million bv volume 

 (ppm) (Fig. 1), reveals a seasonal variation of about 6 ppM, peak to peak, 

 superimposed upon a secular increase of about 0.8 ppM per year. The former 

 feature we ascribe to the seasonal growth and decay of the land biota and the 

 latter feature to retention of industrial C0 2 in the atmosphere. 



To establish the secular trend apart from the seasonal variations, we first 

 fitted several trial functions to the monthly means by the method of least 

 squares. Statistical tests of significance indicate that the best fit is achieved using 

 an oscillating power function consisting of a third-order polynomial in time to 

 represent the trend, plus sinusoidal terms representing 6- and 12-month 

 oscillations. The polynomial function is plotted in Fig. 2 along with the 

 seasonally adjusted monthly average concentrations. It suggests that variations 

 have occurred in the secular trend during the last decade. A 12-month moving 

 average of the monthly means (Fig. 3) indicates essentially the same trend. The 

 trend, derived by either of the foregoing methods, is probably close to the best 

 available estimate of recent changes in atmospheric COt in the Northern 

 Hemisphere. 



From measurements of C0 2 in air collected in 5-liter glass flasks twice per 

 month at Amundsen— Scott station at the South Pole, the secular trend of 

 atmospheric C0 2 for the polar Southern Hemisphere can also be estimated. The 

 results constitute the longest modern record of atmospheric C0 2 variations at 

 any single location. In addition to 145 3 analyses of 749 flasks exposed on 291 

 sampling dates between 1957 and 1971, a gas analyzer at the station furnished 

 continuous data from 1961 to 1963. Contaminated data were removed from the 

 flask analysis record by rejecting analyses that were more than 1.25 times the 

 standard deviation above an estimate of the mean provided by a least-squares fit 

 of the same form of oscillating power function as that used for the Mauna Loa 

 record. This procedure was reiterated until the originally skewed distribution 

 resembled a normal distribution with a standard deviation approximately equal 

 to the inherent precision of the sampling technique (0.30 ppM). 5 Several ' 

 iterated rejections beyond the predetermined cutoff yielded no significant 

 changes in either the dispersion of the retained data or the estimate of the 

 secular trend. 



The retained flask data (73% of the original data) were combined into 

 averages for each sample date and these, together with the bimonthly continuous 

 analyzer averages for 1961 and 1963, were fitted again to the oscillating power 

 function used before (Fig. 4). The secular trend was abstracted as a third-order 

 polynomial (Fig. 5). The South Pole and Mauna Loa trends when compared 

 (Fig. 6) are very similar. The C0 2 concentration at the South Pole is 

 systematically lower because the principal sources of industrial C0 2 occur in the 

 Northern Hemisphere. 



x 



{ 



