Validating a next generation GreenFeed unit to measure methane emissions in sheep : a thesis presented in partial fulfilment of the requirements for the degree of Master of Science in Animal Science, School of Agriculture and Environment, Massey University, Manawatū, Palmerston North, New Zealand

Abstract

Enteric methane (CH₄) from sheep is a major contributor to agricultural greenhouse gas (GHG) emissions in pasture-based systems such as New Zealand, and reliable, scalable measurement is essential for mitigation research, inventory development, and low-emission breeding. Respiration chambers (RC) provide high-precision continuous measurement but cannot be used under grazing conditions, creating demand for field-compatible alternatives such as the GreenFeed (GF) system. Although GF has been widely validated in cattle, sheep-specific validation, particularly for cattle designed units equipped with new-generation tunable diode laser (TDL) sensors remains limited. This study therefore compared CH₄ emission estimates from a cattle-designed TDL-GF unit against RC measurements in sheep under controlled indoor conditions. Twelve Romney ewe lambs (<1 year) were allocated to either a standard ryegrass–white clover pasture or a diverse pasture (ryegrass–clover–plantain–chicory) to generate variation in emissions. Methane production was measured in open-circuit RC over two consecutive 24-h periods (with one additional day due to technical issues) and subsequently estimated using GF over three days using eight scheduled 5-min spot samples distributed across a 24-h cycle. Agreement and bias were assessed using regression-based methods following the St-Pierre framework, with residuals (RC − GF) regressed against mean-centred GF values to test mean (intercept) and proportional (slope) bias. Carbon dioxide (CO₂) estimates were analysed as a diagnostic comparison. Eleven animals generated paired RC–GF data (one lamb, #7306 did not engage with GF bait and produced no valid visits). GreenFeed CH₄ production showed a significant positive relationship with RC measurements (R² = 0.65; P < 0.01), indicating moderate agreement and substantial capture of between-animal variation. Mean bias for CH₄ production was not significant (P = 0.27), but proportional bias was detected (negative slope; P < 0.01), demonstrating that disagreement between methods increased with emission level. For CH₄ yield, agreement was weaker (R² = 0.38; P = 0.04) with no significant mean bias (P = 0.28) but significant proportional bias (P = 0.003). In contrast, GF performed poorly for CO₂, showing weak association with RC (R² = 0.28; P = 0.09) and substantial mean and proportional bias (both P < 0.001), consistent with greater sensitivity of CO₂ to short-term variation and intermittent sampling in sheep. Diet influenced CH₄ residuals (P = 0.02), with improved agreement under the lower-fibre diverse pasture and greater residual variability under the higher-fibre standard pasture. These results show that a cattle-designed GF unit equipped with TDL sensors can capture meaningful between-animal differences in CH₄ emissions in sheep without systematic mean bias, supporting its use for controlled research phenotyping and comparative evaluation of mitigation strategies. However, proportional bias and increased variability relative to RC indicate that intermittent sampling limits precision for absolute daily emission quantification, particularly when sampling density is low and under higher-fibre feeding conditions. Increasing the number and temporal spread of spot samples and validating protocols under grazing are likely to improve representativeness and robustness, while RC remain essential for high-precision quantification and method validation.