A novel statistical methodology to overcome sampling irregularities in the forest inventory data and to model forest changes under dynamic disturbance regimesThis article is part of a larger document. View the larger document here.
Forest inventory datasets offer unprecedented opportunities to model forest dynamics under evolving environmental conditions but they are analytically challenging due to irregular sampling time intervals of the same plot, across the years. We propose here a novel method to model dynamic changes in forest biomass and basal area using forest inventory data. Our methodology involves the following steps: 1) parameterize transition matrices for biomass using Gibbs sampling, 2) incorporate dynamic disturbance and forest growth scenarios and 3) simulate transient dynamics and stationary states using Markov chain model. We extend this method to further include changes in natural disturbance regimes and land-use practices, to predict the impact of changing climate and forest management practices. We apply this methodology on North American forests. We first assess the predictive power of the methodology, without including changing disturbance regimes, in two independent ways: (a) the first years of the dataset are used to predict the later years, and (b) the long-term predictions of two random partitions are compared. The model predicts consistent short-term increases in biomass. We then investigate the consequences of global warming scenarios including changes in forest fire rate in hardwood forests as well as possible growth enhancements due to increasing CO2 and temperature. We conclude that ongoing increasing biomass trends are relatively unaffected in the short term by changing disturbances regimes. Overall, our original data-intensive methodology provides both descriptions of the short-term dynamics as well as predictions of forest development on a longer timescale.