RESULTS AND DISCUSSION 



Crown Weights 



Dominants Greater Than 1 Inch D.B.H. 



Equations for live crown weight are presented in table 1 and for dead crown weight 

 in table 2. In the tables, EXP is the base of natural logarithms. For all species, an 

 equation containing d.b.h. as the only independent variable is presented because often 

 d.b.h. may be the only information available for estimating weight. For some species, 

 d.b.h. alone provided the best-fitting equation; however, where addition of tree height, 

 crown length, or crown ratio to the equation improved fit, these equations are also 

 presented. Although the equations in tables 1 and 2 were judged as "best fitting," 

 other equation forms and combinations of independent variables gave good fits of the 

 data. 1 found, as Crow (1971) reported, that the best-fitting equation varies by data 

 sets. Several curve forms can fit about equally well. 



Intercept regression parameters that were statistically nonsignificant were some- 

 times retained in the equations when the fit for trees of small d.b.h. was improved. 

 Although the regression constants in tables 1 and 2 are significant at a confidence 

 level of at least 0.95, unreasonable predictions beyond the range of sample data are 

 possible. Extrapolation of equations beyond about 40 inches d.b.h. risks substantial 

 error. 



The data for Douglas-fir were particularly difficult to fit; hence, two equations 

 covering different ranges^in d.b.h. are presented. Data gathered by Fahnestock (1960) 

 and myself are probably from different populations. Live crown weight populations 

 within the Pacific Northwest are known to be different (Woodard 1974) . 



This study showed that to achieve a good fit for prediction, actual deviation 

 between observed and predicted observations should be examined. It can be difficult 

 to find a function that fits the data well throughout a large range of d.b.h. High 

 R'^ values, as an indicator of good fit, can be deceiving. An examination of the litera- 

 ture on crown weight prediction suggests that some studies would profit from a reevalua- 

 tion of prediction functions to improve accuracy, especially at the outer ends of their 

 data range. 



In equations for 8 of the 11 species, crown ratio or crown length accounted for 

 a significant reduction (at the 0.95 percent probability level) in residuals beyond 

 that accounted for by d.b.h. Crown ratio was more effective at reducing residual 

 variation than either height or crown length alone or in combination, as similarly 

 found by Loomis and others (1966). Site index reduced residual variation beyond d.b.h., 

 height, crown length, and crown ratio for only western larch. Although site is known 

 to influence live crown weight per tree (Tadaki 1966; Brown 1965), this study and work 

 by Storey and others (1955) indicates that d.b.h. together with crown ratio or crown 

 length can largely account for site effects. Trees per acre reduced residual variation 

 for four species and basal area for two species. These measures of stand density were 

 ineffective variables for the other five species. Even though stand density was 

 significant for about one-half of the species, the actual reduction in residual varia- 

 tion beyond that accounted for by the tree dimension variables was very small. 



9 



