Modeling climatic effects on milk production

• Published in: Computers and electronics in agriculture 2024-10, Vol.225, p.109218, Article 109218
• Main Authors: Katsini, Lydia, Muñoz López, Carlos André, Bhonsale, Satyajeet, Roufou, Styliani, Griffin, Sholeem, Valdramidis, Vasilis, Akkermans, Simen, Polanska, Monika, Van Impe, Jan
• Format: Article
• Language: English
• Subjects: Climate change; Dairy industry; Data-driven modeling; Forecasting; Impact modeling; Partial least squares; Principal component analysis; t-SNE
• ISSN: 0168-1699
• DOI: 10.1016/j.compag.2024.109218

With the increasing food demand, it is essential to plan for a future where climate change affects the food sector. The policy making necessary to ensure the food demand must be driven by a thorough assessment of the impact climate change has on the food sector. Such an assessment can be facilitated by building impact models focusing on different aspects of the food industry. In this paper, we build an impact model to demonstrate the effect the climate has on the dairy (raw milk) industry. This impact model is built using raw milk data collected from more than 200 dairy farms in two climate zones over a 6-year timespan. The analysis using dairy variables like milk volume, fat, lactose, protein, somatic cell counts, and total bacterial counts (TBC) revealed correlations with climate variables such as temperature, humidity, wind, radiation, and precipitation. To enable prediction of these dairy variables over different climate scenarios, regression and forecasting models are built individually. Rather than temperature, the most significant predictors are the radiation, humidity, wind, and precipitation. These models are tested for their generalizability over multiple climate zones. With prediction errors always below 20%, these models can be exploited to quantitatively assess the effect of climate change on milk production. Overall, this research contributes to the design of strategies towards a resilient, climate change-proof food system. [Display omitted] •Milk volume, fat, protein, SCC and TBC from hundreds of farms are analyzed.•Correlations with temperature, humidity, wind and rain are identified.•Principal component analysis reveals correlations between milk variables.•Milk traits predictive models are built and tested for generalization.•The most impactful predictors are radiation, humidity, wind and rain.