Determining the location of a deceased mother tree and estimating forest regeneration variables by use of microsatellites and spatial genetic models

Abstract

In this paper we first mathematically formulate spatial genetic models that rely on dispersal kernels, using the genetic inhomogeneous Poisson process. On the basis of mapped and genotyped data pertaining to adult and juvenile trees we are able to estimate three fundamental variables of population dynamics: individual female reproductive success, seed dispersal, and pollen flow. The model was applied to a secondary Fagus crenata stand in northern Japan, regenerated after shelterwood harvesting. Highly polymorphic microsatellite data revealed that most of the juveniles around one adult tree were not that tree’s progeny and that some minor alleles were clustered there. These data suggested that another mother tree had formerly been present in the vicinity, produced offspring there and died. Inferring its genotype and location, we applied the genetic inhomogeneous Poisson process. Results confirmed that we would have wrongly assessed the regeneration if we had been unaware of the existence of the dead mother. The average distances for seed dispersal and pollen flow were 18 and 193 m, respectively. The contribution of outside mothers, simultaneously assessed using the dispersal variables in the models, ranged from 10 to 50% depending on their positions relative to preserved adults. Individual female reproductive success varied as much as fiftyfold among the eight preserved adults. Our comprehensive approach, utilizing currently available genetic information, mathematical models, and previous forestry records, helped elucidation of the past forest-regeneration processes.

Appendix

Materials, methods, and results of isozyme analysis

In June 2004, we collected fresh leaf tissues from:

• 1 sixty-one preserved F. crenata adults in part of the compartment, including all the 49 mapped adults; and

• 2 a total of 275 F. crenata juveniles (all 203 stems with DBH ≥ 3 cm and randomly sampled 72 smaller juveniles taller than 2 m) distributed throughout the plot.

Samples were collected separately from multiple stems growing from the same base to check whether they originated from stump sprouts or from different seeds. The samples were kept at approximately 4°C before being transferred to the laboratory, then stored in a deep freezer at −80°C until the enzymes were extracted. Isozymes of ten enzymes were analyzed by vertical slab polyacrylamide gel electrophoresis, and we detected 13 polymorphic genetic loci (6Pg, Aap1, Aap2, Amy, Dia, Fum, Gdh, Got1, Got3, Mdh, Pgi, Pgm1, Pgm2). Details of the procedures used in the isozyme analysis are given elsewhere (Kitamura et al. 1997 and references therein).

Among twenty-four multiple-stem clusters, nineteen consisted of an identical genotype, two had identical genotypes with the adjacent adult, and three were mixtures of different genotypes. In this study, the first type was regarded as single individuals. The second type were very likely to be sprouting stems from the adjacent adults, thus were excluded from the offspring population. The third type probably originated from clustered seeds cached by rodents. Consequently, we had a total of 231 mapped and genotyped individual F. crenata juveniles.

For microsatellite analysis we again collected fresh leaf tissue. This time we excluded the twenty-three adults located relatively far from the plot and added three that were closer. For juveniles we collected one sample from each “single individual” identified as above, except for six that had already died.

In 2004, we applied a version of a spatial genetic model that was under development and obtained unreasonable estimates. Unreasonable estimates were again obtained when the current genetic inhomogeneous Poisson process was applied. We obtained one-component τ = 404 m pollen flow, long-distance seed dispersal of σ ₁ = 5 (15%) and σ ₂ = 26 m, and positive female reproductive success was observed for A4, A5, A6, and A7 only. The maximum log-likelihood was −1,244, which was only a slight improvement on random mating of −1,288.