110 BACASTOW AND KEELING 



with the rising demand for fossil fuels. The prime causes for rising fuel demand 

 are a rise in human living standards and increasing human population, and either 

 of these will also raise the demand for food and timber during the next 100 

 years. The area of forests and perennial grass land will probably diminish at such 

 a rate that the world biomass will be reduced at least as fast as C0 2 fertilization 



40 



promotes an increase. 



Thus, the two preferred models almost surely overestimate the possible 

 increase in biomass. The upper limit on biomass increase is probably so small 

 that any increase is negligible compared to the input of industrial C0 2 . If the 

 biomass increase in the model is stopped in 1970, the predicted atmospheric 

 C0 2 concentration in 2070 is about 30% higher than that quoted above; for the 

 first model the predicted concentration is 8.3 and for the second model 8.0 

 times the preindustrial level. As shown by Table 4, these predictions are not 

 greatly changed if the oceanic parameters are varied between limits suggested by 

 the uncertainty in observational data. Varying the ocean parameters has still less 

 influence on prediction if the biota is assumed to increase in mass after 1970. 



As a final test of the relative roles of the oceans and land biota in long-range 

 prediction, we considered the case where past variations in atmospheric C0 2 are 

 explained solely by ocean uptake, i.e., the case |3 = 0, N m0 /N a0 = 15.0, 

 r am = 7.9 years, and T^ m = 1500 years. According to this extreme model, the 

 atmospheric C0 2 concentration in 2070 will be 6.5 times the preindustrial level. 

 Comparing this result with the previous two models, we see that the predicted 

 atmospheric C0 2 concentration in 2070 is almost sure to fall between 6 and 8 

 times the preindustrial value, regardless of whether the biota increases or not. 



Returning to the preferred models, the instantaneous and total partitioning 

 of industrial C0 2 between major reservoirs, also shown in Tables 3a and 3b, 

 furnishes a measure of the nonlinearity of the model. A linear model (see 

 Eqs. 21 and 23 of the preceding paper, Ekdahl and Keeling 7 ) would predict 

 almost time invariant partitioning. As can be seen, the partitioning departs from 

 near constancy only after the end of this century. Thus, to make short-term 

 predictions of the atmospheric C0 2 increase, as Machta has done, it is not 

 essential to know how the industrial C0 2 is shared between the oceans and land 

 biota: regardless of which reservoir is taking up most of the C0 2 , the airborne 

 fraction will probably not change significantly. This near constancy in fraction 

 depends, of course, on the plausible assumptions that fossil-fuel combustion 

 continues to increase at nearly the same rate and that the land biota continues to 

 respond in the same manner for the next few years as during the recent past. 



The instantaneous rate of change of C0 2 partial pressure with increasing 

 total carbon in surface water relative to initial conditions 



s- _ dr m /P m o (15) 



~ dSC/SCo 



we refer to in the tables as the surface ocean buffer factor. It is more indicative 

 of the ocean surface water affinity for industrial C0 2 than is the evasion factor, 



