Forage crops rank third in Canada in terms of cultivated area, feeding over one million dairy cattle, four million beef cattle [1], and 500,000 sheep [2]. However, they are also among the main sources of greenhouse gas (GHG) emissions in the agricultural sector. For instance, corn silage contributes 1.44% (1.02 MT CO2 eq/year) [3] of total agricultural emissions, while cultivated hay accounts for 6.3% (4.47 MT CO2 eq/year) [3]. Animals consuming these forages produce enteric methane (CH4) emissions, representing 38% (22.8 MT CO2 eq/year) of total agricultural emissions [4].
Forage systems consist of several interconnected subsystems: soil, plants, harvesting, storage, and feeding. Suboptimal management of these subsystems can not only increase GHG emissions but also harm farm profitability. Integrated management of these systems is crucial to ensure improvements in one subsystem do not negatively affect another. While some practices exist to reduce GHG emissions, such as improving fiber digestibility (reducing emissions by 13%) or incorporating tannin-rich plants (reducing emissions by 18%) [5], producers and agronomists face challenges in implementing them due to a lack of suitable tools.
Artificial intelligence (AI) has the potential to assist producers and agronomists in this task, but its application in forage systems is limited by the high cost of collecting the necessary data. To address this issue, this project proposes developing an expert system leveraging Large Language Models (LLMs) to efficiently and comprehensively generate a knowledge base and serve as an inference engine. Existing deterministic agronomic models, such as Holos and STICS, will also be used to generate additional data. These models will be employed in a probabilistic framework to estimate GHG emissions and plant growth.
Furthermore, a multi-objective optimization module will be developed to simultaneously maximize farm profitability and GHG emission reductions. This will be integrated into a prototype within MSF's decision-support tool (DST). The DST will identify best practices (e.g., selecting forage species, improving silage conservation quality, optimal cutting times) and propose tailored strategies to reduce CH4 emissions while maintaining economic viability.
Each recommendation will be customized to the specific conditions of the farm, facilitating the adoption of GHG reduction strategies while considering the economic and environmental constraints of producers.
