Abstract
The current method of applying the gentan probability methodology assumes that the mean and variance of the gentan probability distribution are equal to the mean and variance of the observed felling age. In this paper it is shown that this is only true if the forest is “normal” in the long run equilibrium sense that SUZUKI used the term. At all other times this assumption leads to gross inaccuracies in estimating the parameters of the distribution and in the felling predictions. To ascertain the mean and variance of the gentan probability distribution maximum likelihood techniques that take account of the censoring of the data are required. In this paper such a method is developed and applied to a number of data sets. It is found that the size of the forecasting error is substantially reduced compared to that generated if the gentan probability methodology is applied in the presently accepted way. This suggests that attempts to explain the overestimates in the forecasts by arguing that the depressed timber price causes some forest owners to cease forest management are in need of reassessment. Some evidence is presented which suggests that the size of the error of the method when correctly applied is not increasing through time.
