This study develops a model to estimate GHG and NH 3 emissions at both the housing and storage levels, using experimental data validated with existing data from commercial farms, covering three laying hen housing systems: conventional cages, enriched cages, and cage-free systems. Data preprocessing included dendrogram-based correlation analysis, Pareto scaling, and principal component analysis (PCA) to identify key explanatory variables (pH, temperature, ventilation rate, body mass, egg production, and relative humidity). Housing emissions were modelled using mixed linear regression and system-specific approaches, while storage emissions were estimated using literature-based coefficients. Regression models demonstrate a moderate-to-high predictive capacity for CO 2 (R 2 = 0.68) and CH 4 (R 2 = 0.63) in enriched cages and NH 3 (R 2 = 0.87) in cage-free, while a constant emission factor best represented N 2 O. However, validation with data from 30 commercial farms showed poor predictive performance due to high variability in management, scale, and environmental conditions. This study develops system-specific emission models based on operational and environmental variables measured on farms, without needing direct gas measurements. Results show the approach's feasibility and accuracy depend on the system and gas, offering guidance on reliable application. • Emission modelling using production and environmental parameters. • Models developed for conventional, enriched, and cage-free systems. • Integration of statistical preprocessing enhanced predictive capacity. • Validation with farm data revealed challenges for large-scale use.
Carranza-Díaz et al. (Sat,) studied this question.