312 REINERS 



varies between types. A better understanding of which examples were used can 

 be found in the footnotes of Table 5. The total turnover estimate of 46 X 10 9 

 tons C/year derived from primary-production considerations was considered 

 appropriate for maximum input conditions, but the total estimate by this second 

 method is even higher (64 X 10 tons). The source of difference between the 

 two results lies in nonforest types and is due, in part, to the inclusion, in the 

 Rodin and Bazilevich data, of grass and shrub shoots in litter of grassland, 

 savanna, and shrubland. This is not an error from their point of view, but, to the 

 extent that fire and grazing remove surface mulch, the litter does not enter the 

 detritus pool in the sense used in this paper. 



The Third Approximation 



Because of the similarity in data base, it is not surprising that the total 

 estimate for the third approximation (62 X 10 9 tons) is quite similar to that of 

 the second (64 X 10 9 tons). In this case, total-litter measurements in single 

 stands or a few stands, which were presented as typical values by Rodin and 

 Bazilevich, were multiplied by Whittaker and Likens' estimate of areal cover 

 (Table 4). This method is graced by the judgment of Rodin and Bazilevich with 

 regard to representativeness of the data. They may have arranged their stands in 

 different categories, however. Again, there is no allowance for intense grazing or 

 fire in nonforest types. 



The Fourth Approximation 



This approximation is developed on a larger literature base than the Russian 

 work and rests on a number of assumptions regarding root contributions to litter 

 (Table 5). Where available, ecosystems of all stages of development are included, 

 and, for forests in particular, this estimate probably best represents current 

 conditions. The estimate for total carbon turnover, 37.5 X 10 9 tons, is the 

 lowest of all four approximations. This method suffers, as do all the others, from 

 lack of proper weighting for litter values to area coverage within the type. Much 

 information is lost by the use of simple arithmetic averages that might be 

 exploited by a finer grained division of ecosystem types within the broad 

 categories. For example, the / 14 weights applied to the extremely high values for 

 Arundo donax and grass tugai are probably out of proportion to their actual 

 areal extent. 



Possible errors in estimates for tropical forests are especially serious because 

 of their high areas and high input rates. The average given may be too low for 

 unexploited forests because timberfall, which may equal litterfall, 50 was not 

 estimated in many of the examples. On the other hand, it may approach a 

 proper compensation for current logging and slash-and-burn agricultural effects. 



Very productive grasslands are probably overweighted in comparison with 

 extensive but less productive dry steppes more common today. The level of error 



