et enc c eel M i 
  
  
V(a), V(Q),2C0V(a,Q) ,2(COV(a,R), and 2COV(Q,R) are only influenced by the Table © 
regressions and by the estimated P and are not influenced by V(P) from the 
independent photographs. Thus these five terms cannot become closer to zero 
by making additional photo measurements to reduce V(P). —B 
COMPONE 
The term V(R) in equation 4 (Table 4) is, itself, not entirely made up 
of components affected by V(P). Only four terms b^V(e)V(P), V(b)e?V(P), 
b^ e? V(P) , and V(b)V(e)V(P) contributing to V(R) are influenced by V(P) and b2V(e) 
these make up less than 10$ of V(R) or about 3$ of the total V(B). Even V(b)? 
if the sample size used in estimating P was infinitely large, it would not V(b) V(e 
be possible to reduce V(B) by more than 3 units to less than 13. Thus the b? e2V(E 
minimum standard error possible with the current regressions is about/13 or bV(e)P? 
3.60. Increasing the number of paired observations in the two regressions to V(b)e?I 
reduce the variancesof the estimatesof the intercepts and slopes would permit V(b) V(e 
a reduction in the V(B) in future surveys. Specifically, reducing the term Total 
  
V(b)e?P?in. Table 5 by increasing the sample size in the destructive sampling 
could substantially reduce V(B) and increase the accuracy of the predictive 
regression. 
However, the standard error of the estimate of the biomass using equa- Aldred, 
tion 4 is not particularly high. At 10.4% of mean biomass it compares well 
with North American experiences where standard errors of 0.61-0.91m mean 
crown diameter and 15-20$ of mean volume (quoted here in the absence of Aldred, 
similar information on wood biomass) result from the use of 76-122m flying 
heights and 1:650-1:1250 photo scales(Avery, 1958; Kippen & Sayn-Wittgenstein, ; 
1964; Aldred & Kippen, 1967; Aldrich, 1979). Aldrick 
The statistical procedure demonstrated in this paper provided a means Ä 
to estimate woody biomass over large areas using aerial photography. The Avery 
method permits use of pre-existing data sets, in which woody biomass previous- : 
ly had been correlated to crown diameter, and thus avoided the expense and : 
difficulty of repeating the destructive sampling. In contrast to true double Dirsch] 
sampling, the user must recognize the assumption that the regressions remain 
valid wherever the method is used. The approach may also be applicable to 
other situations where use of existing data is desirable or else it is not Draper 
possible to repeat previously completed studies. : 
  
  
Freese, 
ACKNOWLEDGEMENTS 
Special thanks are due to the UNESCO-IPAL and KREMU field staff who Heller 
carried out the painstaking work of destructive sampling. Logistical assistance 
for part of the destructive sampling work was provided by Dr. J. Schwartz of 
the UNESCO-TLMP Project at Ngurunit. All aerial photography and subsequent 
analysis was provided by KREMU. Husch, 
Table 4, Value of each term of equation 4 when estimating the variance Kippen, 
of the biomass, V(B), for the test stand. 
Loetsch 
COMPONENT OF V(B) VALUE PERCENT CONTRIBUTION OR 
REDUCTION 
Miller, 
V(a) 7.07 45.0 
V(Q) 2. 33 14.8 
V(R) 32.83 209.1 Mood. / 
200V(a,Q) -3.21 -20.0 : 
2C0V(a,R) 713.73 -87.5 
2COV(Q, R) -9.59 - 61.1 
Total 15.70 100.0 
  
420 
55 À—— —À j A ei or : ex d 
3 LE EN - EN s - ——mA 00 0 s 
= nM MEE mm. ;,#"" 2 2 in un a