PS6-7. Modeling Feed Intake in Lactation Sows Using Temperature and Humidity Covariates

Abstract Intensive genetic selection has increased reproductive output in modern sows; however, voluntary feed intake (FI) during lactation often limits milk production, particularly under heat stress. This study evaluated lactation FI trajectories and environmental covariates using 17,171 daily FI records from 898 sows (P1=503, P2=138, P3 + =257) farrowing between May and July 2024 on a commercial farm in Washington, Iowa. Records with negative intake, incomplete lactations, or missing farrowing dates were excluded, and data were truncated at 22 days in lactation. Nine alternative models, including generalized Michaelis–Menten (GMM) and cubic polynomial mixed-effects functions with varying random-effect structures, were compared using conditional R², residual standard deviation (RSD), and cross-validated RMSE. The cubic polynomial with three sow-specific random effects provided the best overall performance (R² = 0.8473; RSD = 0.9099 kg/d, RMSE=1.01), outperforming the best GMM specification (R²=0.8218, RSD=0.9832 kg/d, RMSE=2.33). The polynomial model accommodated a modest post-peak decline in FI during late lactation, which was not fully captured by the asymptotic GMM form. Across parities, FI increased to a concave peak near day 18 (P3 + P2 P1) and then slightly declined. Environmental covariates included daily maximum air temperature (MT), degree-hours above 24 °C (DGH24), and daily maximum dew point (MD). Models allowing environmental covariates to interact with lactation day consistently outperformed constant additive effects and proportional modifiers, indicating lactation-day–dependent environmental sensitivity. Among single covariates, MD provided the greatest improvement in model fit. When both MT and MD were included with day interactions, model fit improved further (ΔAIC=48; likelihood-ratio P 0.001), indicating that temperature contributed additional information beyond absolute humidity. Marginal effects showed that a 1 SD increase in MD reduced FI by 0.129 kg/d, whereas a 1 SD increase in MT increased FI by 0.075 kg/d when MD was held constant. When both increased simultaneously, predicted FI declined by 0.054 kg/d. These findings suggest that moisture-related heat load accounts for most intake suppression, while temperature exerts a smaller conditional modifying influence. Overall, parity and environmental heat load meaningfully shaped both the level and form of lactation FI trajectories. Incorporating parity together with dew point and temperature into precision feeding and thermal-management strategies may improve sow performance under warm conditions. Results are based on a single commercial herd during the warm season, and environmental variables were derived from external ambient data rather than in-barn measurements; broader validation under diverse environmental and management conditions is warranted.