4 -3 -2 -1 12 3 4 



NORMALIZED DEVIATIONS FROM PREDICTIONS 



Figure 2. — Location of prediction points for three fractions of Normal Distribution. 



Each of these three records is associated wiin one of the three portions of the 

 error distribution characterizing the deviations about prediction (fig. 2). The first 

 record representing the 60 percent of the population (approximately the center of the 

 distribution) is given a prediction to which is added the average value of the 

 deviations in that portion of the normal distribution. This "biased" point is 

 indicated by A in figure 2. The second record representing the upper 25 percent of 

 the error distribution is given a prediction corresponding to point B, and likewise, 

 the record for the lower 15 percent is given a prediction corresponding to point C. 

 By this method, the weighted average prediction for the three records is still unbiased. 



Which of these two procedures is followed is controlled by whether the number of 

 tree records is greater than 1,3 50. Hence, a stand described by 150 tree records at the 

 start would go through two projection cycles using the record- tripling approach before 

 switching to the use of a single random deviate for each record. 



Mortality 



Random fluctuations in mortality are notorious problems in analyzing forest growth. 

 Unfortunately, our records that could be used to assess the distribution of the random 

 variable for mortality are barely adequate to estimate an average mortality rate. The 

 variability of m.ortality rates through time, in response to fluctuations in climatic 

 stress, extreme winds, and other destructive agents is currently unavailable. Later 

 versions of stand growth prognosis programs will, we hope, be able to assess the effects 

 of the stochastic nature of mortality. 



13 



