A multi-step modelling approach to evaluate the fuel consumption, emissions, and costs in forest operations

Abstract

Assessing the financial, energy and emission levels associated with alternative harvesting prescriptions, techniques, and technologies is crucial to sustainable forest management. This study aims at developing an overall predictive system able to estimate productivity, fuel consumption and emissions for a set of different forest harvesting technologies as a function of stand characteristics, site conditions and silvicultural prescriptions. The system consists of two separate group of models, one for felling and processing, and the other for extraction. In the first model, six quantitative (DBH, tree volume, harvest, engine power, purchase price of the machine) and five qualitative independent variables (forest species, management, treatment, operation, type of machine, type of fuel) were considered. In the second model, four qualitative and six quantitative independent variables were applied. Both models have been trained on a large database of real field data using linear and nonlinear approaches based on artificial intelligence. The dataset was randomly separated into two subsets, one using for training the models and the other for validating them. The models reliably predicted fuel consumption (r = 0.84–0.95), energy emission and cost (r = 0.88–0.98), per hour and per hectare, and they were successfully validated. Despite being estimated from a heterogeneous input dataset, constituted by quantitative and qualitative variables, both models proved to be efficient, robust and generally representative. The input variables are intuitive and suitable for practical use by both entrepreneurs and policy makers at the institutional level.