Effect of temporal autocorrelation on apparent growth rate variation in forest tree census data and an alternative distribution function of tree growth rate

Distributions of forest tree growth rates often display wide variation and positive skewness even within a narrow range of tree sizes. Temporal autocorrelation of growth rates is also commonly observed in forest trees. We suggest that temporal autocorrelation causes the positive skew in growth rate distributions. We propose the ‘passage’ distribution of growth rates, defined by the density of trees passing a particular size over a given inter‐census interval, as a superior description of growth performance of trees with temporally autocorrelated growth. We relate the passage distribution to the conventional ‘resident’ distribution of growth rates, defined by the growth rates of trees within a particular size range at a given census. The passage distribution is approximated by the resident distribution weighted by the absolute growth rate. For common rain forest tree populations, the average of the passage distribution is nearly twice that of the resident distribution. Simulation models which use growth rates estimated from resident distributions and assume a Markov process are valid only to reconstruct the dynamics of the overall size structure of a stand. When examining the influence of life history strategy on forest dynamics, models must incorporate the effect of temporal autocorrelation on growth rates.