Planejamento otimizado da colheita florestal: modelagem para projeção de sortimentos da indústria e agendamento sequencial da colheita

Abstract

The planning of forest harvesting aligned with industrial production is essential to optimize the use of timber resources, reduce economic losses, and increase the profitability of forestry companies. In this context, this study aimed to integrate two main approaches to harvest management and planning: (i) modeling the diameter distribution for volumetric projection of assortments and (ii) optimization of sequential harvest scheduling through mathematical programming in operations research. In the first stage, two fitting methods for the threeparameter (3P) Weibull probability density function were compared: maximum likelihood estimation (MLE) and method of moments (MM), combined with two projection methods: Parameter Prediction Method (PPM) and Parameter Recovery Method (PRM), thereby generating four diameter projection flows (F). In the second stage, a Integer Linear Programming (ILP) model was developed, integrating Production Regulation (R) constraints with Harvest Sequencing (S) constraints to maximize Net Present Value (NPV). In the modeling section, the analyses of the Weibull function included the quotient Kolmogorov–Smirnov test (KSq) and prediction error statistics for each plot, as well as the projections of each flow (F) applied to the final production of industrial assortments. In the harvest optimization phase, analyses were performed through simulations of three scenarios defined by the constraints: C1 (R), C2 (R+S), and C3 (control: company standard). For both parts of the study, information from commercial plantations supplying a sawmilling industry in the state of Alagoas was used. The data were processed in R, and optimization was solved using the Gurobi solver. The results indicated that the moments-based method combined with the parameter recovery projection (F4 = MM + PRM) showed the highest statistical accuracy in predicting industrial assortments. At harvest age, F4 presented the best results for both the highest-value assortment (0.04%) and the sum of products from sawmills I + II (–7.45%) and total volume (–9.56%). In terms of harvest sequencing, scenario C2 showed the best performance, with a 5.6% increase in multiproduct volume and gains of up to R$ 750.00 per hectare compared to the company’s current practice (C3). Therefore, it is concluded that combining MM fitting with PRM projection improves volumetric prognosis, and the use of this information in the proposed ILP optimization model constitutes a complete and accurate approach to optimizing harvest sequencing by synchronizing forest production with industrial demand.