A model for longitudinal data sets relating wind-damage probability to biotic and abiotic factors: a Bayesian approach

To develop a statistical model framework to analyze longitudinal wind-damage records while accounting for autocorrelation, and to demonstrate the usefulness of the model in understanding the regeneration process of a natural forest. University of Tokyo Chiba Forest (UTCBF), southern Boso peninsula, Japan. We used the proposed model framework with wind-damage records from UTCBF and wind metrics (speed, direction, season, and mean stand volume) from 1905–1985 to develop a model predicting wind-damage probability for the study area. Using the resultant model, we calculated past wind-damage probabilities for UTCBF. We then compared these past probabilities with the regeneration history of major species, estimated from ring records, in an old-growth fir–hemlock forest at UTCBF. Wind-damage probability was influenced by wind speed, direction, and mean stand volume. The temporal pattern in the expected number of wind-damage events was similar to that of evergreen broad-leaf regeneration in the old-growth fir–hemlock forest, indicating that these species regenerated after major wind disturbances. The model framework presented in this study can accommodate data with temporal interdependencies, and the resultant model can predict past and future patterns in wind disturbances. Thus, we have provided a basic model framework that allows for better understanding of past forest dynamics and appropriate future management planning.