Computational assessment of Amazon forest plots regrowth capacity under strong spatial variability for simulating logging scenarios

In this paper, we assess the regrowth capacity of tropical forest plots by developing an original computational procedure based on statistical model checking methods. We calibrate a new mathematical model of forest dynamics with respect to post-logging data, produced in the Amazon basin. Our mathematical model is determined by a mechanistic system of ordinary differential equations and integrates a new nonlinear aging term, which is necessary to reproduce the complex post-logging dynamics of the tropical forest plots. Our method is based on an efficient algorithmic procedure that explores the parameter space of the model and computes a score for each parameter value, depending on the distance between the trajectory of the model and the data. We distinguish a group of four reference plots, for which the logging was the most intense, from another group of five non-reference plots, which are fitted a posteriori. Our results provide a set trajectories of the new model, which successfully fit the non-monotone post-logging data on each forest plot, in spite of the high level of biological variability identified in the study site. Rather than a unique set of parameters, we return a small cell of parameters, extracted from the parameter space, which contains in its close vicinity several relevant sets of parameters that are equally able to reproduce the regrowth dynamics of distinct tropical forest plots. The size of the cells that should be extracted from the parameter space increases with the level of biological variability. Finally, we show how to use the calibrated mathematical model for simulating logging scenarios, so as to better understand the temporal dynamics of forest regrowth.