Global sensitivity analysis of a pig fattening unit model simulating technico-economic performance and environmental impacts. (September 2018)

• Record Type: Journal Article
• Authors: Cadero, A.; Aubry, A.; Brun, F.; Dourmad, J.Y.; Salaün, Y.; Garcia-Launay, F.
• Is Part Of: Agricultural systems. Volume 165(2018)
• Page Start: 221
• Page End: 229
• Publication Date: 2018-09
• Subjects: Individual-based model -- Global sensitivity analysis -- Scaled elementary effect -- Livestock farming system -- Multiple outputs -- Life cycle assessment; Agricultural systems -- Periodicals; Agriculture -- Environmental aspects -- Periodicals; 338.16
• DOI: 10.1016/j.agsy.2018.06.016
• Languages: English
• ISSNs: 0308-521X

Abstract

Abstract: Livestock farming system (LFS) models are used to produce key technical or economic outputs. Current models simulate multicriteria performance of LFS, i.e. technical, economic and environmental outputs. Therefore, conducting sensitivity analysis (SA) of these models is increasingly challenging. We developed a pig fattening unit model which is a stochastic, discrete-event mechanistic model with a one-day time step. An individual-based model is used to represent the pigs. Our objective was to perform a global SA of this model while accounting for effects of parameters on all outputs. Due to the model's long computational time, we first performed screening SA using the Morris method to identify and exclude non-influential parameters, and then performed variance-based SA of the influential parameters using metamodels. The most influential parameters were mainly pig characteristics and the disinfection period. This study provides a generic SA sequence adapted for models with a high computational cost and multiple outputs. Highlights: A method for global sensitivity analysis of a multiple-output model was developed. Screening and metamodels reduced the model's long simulation time. Aggregated indices were used to calculate effects of parameters on all outputs. Pig characteristics contributed 90% of aggregated variance of model predictions.