Linking dairy system, breed, and welfare indicators to infrared-predicted fine milk composition and cheese making traits in Parmigiano Reggiano cheese production area

Milk technological traits result from the interaction among dairy system, genetics, and cow welfare, factors that also shape public perception of product quality and sustainability. Fourier Transform Mid-Infrared (MIR) spectroscopy is a valid and rapid tool for monitoring milk composition and predicting technological traits, yet few studies have evaluated the combined effects of dairy systems and animal welfare (AW) within a single production system. This study investigated the effects of dairy system, breed, and AW scores based on the Italian official AW assessment (CReNBA) on fine milk composition and cheese-making traits predicted by MIR spectroscopy within the Parmigiano Reggiano production area. Data included 2,175,148 test-day records, 1,869,687 MIR spectra, and farm-level and CReNBA data from 983 farms. Clustering based on feeding and housing systems, geographical area, breed (Holstein Friesian – HF, dairy Crossbred – Cr, Brown Swiss – BS, Simmental – Si, Reggiana – Re), milk yield, and herd size identified 4 dairy systems: traditional Apennines (TA), traditional Po Plain (TP), modern without (MWTMR) or with (MTMR) total mixed ration. CReNBA indicators included management (MGT), structure and equipment (SAE), and animal-based measures (ABM), and were classified into low, medium, and high scores according to compliance with AW standards. The predicted traits, i.e. protein and macromineral profiles and cheese-making traits, were obtained using gradient boosting machine learning method developed on an external calibration population (R2 from 0.58 to 0.88) and analyzed using mixed models with fixed effects of dairy system, breed, AW class, lactation stage, parity, month, and year, and random effects of animal and farm. Breed, dairy system, and their interaction resulted significant for almost all traits, suggesting specific breed attributes and a production context-dependent expression of breed potential. Milk yield increased from traditional to modern systems for all breeds, except Re, which performed better in TP farms. Fat content showed the opposite trend to milk yield, while protein and casein concentrations increased toward modern systems, with breed-specific proportions. Somatic cell count (SCC) was lower in TA farms, but Si showed lower values in TP and Re in MTMR. BS had higher fat, protein, and urea content. Indeed, Re and Si showed higher true protein and casein, with Re performing better in traditional systems and Si in modern ones. Minerals declined to varying extents from traditional to modern systems. Cheese-making traits were improved in modern systems with higher cheese yields and more standardized coagulation, although Si showed better performance in TA. Regarding AW, milk yield was more affected than milk composition, increasing with higher MGT and ABM scores. Higher scores in different areas were associated with higher lactose content and casein index, lower SCC, reduced urea, lower αS2-CN and α-LA, increased mineral content, improved cheese yield (mainly due to greater water retention), and slightly faster coagulation with firmer curd. Overall, breed performance was strongly linked to production context. Albeit to a limited extent, AW influenced fine milk composition and technological traits, confirming the value of routine MIR-based assessments for monitoring dairy herd performance. This research was conducted within the INTAQT project funded by the EU Horizon 2020 programme (grant No. 101000250).