Modelling the temporal variation in the seed production of North American trees

Many aspects of temporal variation in tree seed production (e.g., the proability distribution, periodicity, unimodality) are poorly understood. In this paper, we used 32 annual seed production records from 22 species to show that there are no discernible endogenous cycles, and there is a modest (but seldom significant) tendency for a high seed production year to be followed by an unusually low production year. Finally, we found that all of the records conformed to a single lognormal probability distribution, although our ability to discriminate among species, given short and extremely variable records, is admittedly very limited. We used the lognormal to develop the distribution of the sums of local seed production events (summed across 4 years) as an aid in predicting postharvest or postfire tree regeneration success. Our conclusion is that reliable (defined as 90% of the time) adequate stocking at the edge of an area source requires that the species of interest must comprise a very large fraction of the total basal area per area. Indeed, if the species constitutes less than about 50% of the source, neither burns nor even very narrow strip cuts will be reliably stocked.