100 BACASTOW AND KEELING 



this analysis, we abandon the observed Suess effect as a control on the choice of 

 parameters and use instead estimates of the effective size of the ocean surface 

 layer to limit the predicted Suess effect. Our final search for a good predictive 

 model thus consists of choosing mutually acceptable values of the two 

 parameters N m0 /N a0 and (S, which most directly involve the ability of the 

 oceans and biota to take up industrial C0 2 . In these calculations we employ a 

 six-reservoir model with two atmospheric reservoirs (Fig. 1) and with a 

 heliomagnetic variation in 14 C source as discussed in a later section. To model 

 the biota, we use the perturbation coefficients defined in Eq. C.35 of 

 Appendix C. 



To establish the best value for the Suess effect, we use recent data listed in 

 Table 1 for the industrial period. The average relative to standard, weighted by 

 the number of samples, we denote by the symbol Su 4 8 because the average date 

 is 1948. The value of Su 48 is -2.11%. (If the older data of Fergusson 19 had 

 been included, the value would have been somewhat less negative.) The 

 corresponding Suess effect for 1954, as deduced by solving a model that exactly 

 predicts the observed Su 48 , is -2.51% relative to standard or -2.81% relative to 

 preindustrial conditions. If one uses the same model but includes the effect of a 

 heliomagnetic variation as discussed below, the predicted value of Su 48 is 

 changed slightly to -2.14%, and the 14 C/C ratio in 1954, relative to 

 preindustrial conditions, is changed to —2.90%. 



To model industrial C0 2 , it would seem reasonable to define the ocean 

 surface layer to include all the water that turns over at least once each year, i.e., 

 approximately the water above the lowest depth attained annually by the 

 thermoeline. In temperate zones the thermocline 3 ' extends to about 100 m. It 

 is shallower in the tropics but is deeper in polar regions where it may extend to 

 the ocean bottom. Since 60 m of ocean water contains about the same total 

 amount of carbon as the atmosphere, we evidently should assign to the model an 

 ocean surface layer to atmosphere carbon ratio, N m0 /N a0 , of about 2. If we do 

 so, however, we fail to take into account a layer of water between the 

 thermocline and the 1000-m depth which is known to circulate more rapidly 

 than the deep ocean below. The two-layer model can approximately reflect the 

 exchange capability of this intermediate layer if the model's deep-ocean to 

 surface-layer transfer time, Tj m , is decreased from 1500 years or N m0 /N a0 is 

 increased from 2. 



The two curves in Fig. 5 show that the predicted correspondence of the 

 atmospheric CO; increase and the Suess effect (expressed as Su 48 ) for variable 

 T dm and N m o/N a o are strikingly close together. Apparently, for the purpose of 

 calculating short-term atmospheric C0 2 variations, it does not matter precisely 

 how the waters below the ocean surface layer are modeled. If we match the 

 predicted atmospheric C0 2 increase to the observed value, we find rj m equal to 

 about 100 years or N m0 /N a0 equal to about 15. These values almost surely 

 exaggerate the role of intermediate water. Also, the corresponding model value 

 of Su 48 is only -1.3% (Suess effect of -1.7% in 1954 relative to preindustrial 



