734 JOURNAL OF FORESTRY 



recognized that these local aggregate differences could be reduced to 

 zero. It is felt that the differences from 1 to 3 per cent which ap- 

 pear must be accepted as inevitable in any system of tables which is 

 not prohibitively cumbersome. 



The average deviations also will be seen to be materially less than 

 those found in the case of the Forest Service table (Table 1). 



It appears, therefore, that the three new tables represent a material 

 gain in accuracy over the old. This gain has been made in three ways: 



(A) By working to a fixed top cutting limit. 



(B) By employing improved modern technique, the frustum form 

 factor method. 



(C) By segregating into three sites. 



The desirability of A and B has already been discussed ; C, how- 

 ever, is a new expedient, and it is therefore well to try to evaluate 

 the gain resulting therefrom, independent of A and B. It is fair to 

 ask, that is, whether a single table calculated from taper curves, by 

 the frustum form factor method would not have been sufficiently ac- 

 curate. 



Such a table was therefore prepared. It is self-evident that the aver- 

 age site of such a table would be between I and II, since there are 

 relatively few of the data in the lower site classes. It is, therefore, ob- 

 vious that the table should apply fairly well to these two higher sites, 

 and that the maximum error should occur in using it for Site III ; 

 both higher and lower sites should show an error but that in the for- 

 mer should be less marked than the latter. 



Two groups only were therefore checked against the table, those 

 forming the highest and lowest site qualities, as these should show 

 the maximum plus and minus error resulting from the single table. 

 The results follow, arranged for ready comparison with those obtained 

 from the same groups and the segregated tables. 



The range of error would therefore be increased by the substitution 

 of a single table from 5 per cent to 22 per cent. It would perhaps 

 be possible to modify this single table in such a way as to make it 

 more accurately represent a site intermediate between the highest and 

 lowest, and if so, the maximum error might be reduced from the 18 

 per cent above indicated to about one-half of 22 per cent or 11 per 

 cent, but this latter figure still appears far too high to be acceptable, 

 and the need for the three tables is therefore established. 



