60 EKDAHL AND KEELING 



Since both trends were established on the basis of the same calibrating gases, 

 the close agreement of results for each hemisphere does not prove the accuracy 

 of the long-term increase in CO2 . Analysis of our principal surveillance gas 

 (Fig. 6) indicates that the calibration has not significantly drifted, but analyses 

 of additional surveillance gases suggest an upward drift, perhaps as much as 0.8 

 ppM. Since the secular increase is about 9ppM, the atmospheric CO2 variation 

 may thus be underevaluated by as much as 10%. Manometric calibrations of 

 reference gases, now in progress, should eventually establish a correction for any 

 drift in the calibrating gases and reduce the relative uncertainty to about 5%. 

 These calibrations will also better establish the absolute values of the reported 

 mixing ratios. 



From the average of the Hawaiian and South Pole trends, we estimate that 

 the global average secular increase in the atmospheric concentration of CO2 

 from 1959 to 1969 was 2.34% ± 0.20 of the presumed preindustrial value of 290 

 ppM. This is 49% of the industrial input as estimated by Keeling. The relative 

 uncertainty in this last figure is about 25% if we include the uncertainty in 

 industrial input data. 



STRATEGY FOR INTERPRETING 

 THE ATMOSPHERIC C0 2 RECORD 



Owing to the complexity of the carbon cycle, real difficulties arise in 

 devising a geochemical model to approximate its behavior. The model that we 

 have formulated to interpret the new observational data on atmospheric C0 2 is 

 more detailed than most of those used previously, and its properties are 

 sufficiently complicated to deserve careful examination. As a preliminary test of 

 these properties, we will now consider the two major classes of perturbations to 

 which the carbon system can be subjected by a steadily working external carbon 

 source. The first phenomenon, illustrated by the attenuation of cyclic 14 C 

 variations, is essentially oscillatory in character, whereas the second, illustrated 

 by the partitioning of industrial C0 2 , is essentially exponential. 



The mathematical transfer function which we derive to study the attenua- 

 tion problem has the useful feature that it can afterward be modified to predict 

 industrial CO2 partitioning. An additional useful feature is that this function is 

 analytical and relatively simple to evaluate, whereas the calculation of 

 year-to-year predictions, considered later, must be carried out by tedious 

 numerical approximations of the governing equations. The two methods, one 

 analytic and the other numerical, are essentially independent procedures, and, 

 by requiring that they give concordant predictions, we reduce the risk of 

 retaining computer programming blunders in our calculations. 



