Three different ways for estimating Green Oak Leaf Roller dynamics type: OLS, MEP, and Almost-Bayesian approaches

• Main Author: L.V. Nedorezov
• Format: Article
• Language: English
• Published: International Academy of Ecology and Environmental Sciences 2019-03-01
• Series: Computational Ecology and Software
• Subjects: discrete logistic model; parameter estimation; ordinary least squares; method of extreme points; analysis of deviations; almost-Bayesian approach
• Online Access: http://www.iaees.org/publications/journals/ces/articles/2019-9(1)/estimating-Green-Oak-Leaf-Roller-dynamics-type.pdf

The generalized discrete logistic model (GDLM) of population dynamics was used for fitting of the known empirical time series on the green oak leaf roller (Tortrix viridana L.) fluctuations in European part of Russian Federation (Korzukhin and Semevsky, 1992). The model was assumed to demonstrate satisfactory data approximation if and only if the set of deviations of the model and empirical data satisfied several statistical criterions (for fixed significance levels). Distributions of deviations between theoretical (model) trajectories and empirical datasets were tested for symmetry (with respect to the ordinate line by Kolmogorov-Smirnov, Mann - Whitney U-test, Lehmann - Rosenblatt, and Wald - Wolfowitz tests) and the presence or absence of serial correlation (the Swed-Eisenhart and ‘‘jumps up-jumps down’’ tests). Stochastic search in a space of model parameters show that the feasible set (set of points where all used tests demonstrate correct/required results) is not empty and, consequently, the model is suitable for fitting of empirical data. It is also allowed concluding that observed regime of population dynamics isn’t cyclic (if length of cycle is less than 1500 years) and can be characterized by the fast decreasing autocorrelation function (with further small fluctuations near zero level). Feasible set allows constructing almost-Bayesian estimations of GDLM parameters. For the situation when model parameters are stochastic variables algorithm of calculation of model trajectories is presented.