Modeling shade use of grazing dairy cows using sensor-derived data and machine learning.

Area of Science:

• Animal Science
• Environmental Physiology
• Agricultural Engineering

Background:

• Heat stress significantly impacts dairy cow health, welfare, and productivity, especially for pasture-grazing animals.
• Solar radiation is a primary contributor to heat load in grazing dairy cows.
• Shade provision is a key strategy for mitigating heat stress, but large-scale data on its use is lacking.

Purpose of the Study:

• To quantify shade availability and utilization at a herd level in grazing dairy cows.
• To analyze shade use patterns using sensor-derived data and machine learning approaches.
• To identify environmental and resource-based factors influencing dairy cow shade selection.

Main Methods:

• Utilized global positioning system (GPS) location data from lactating dairy cows across four New Zealand farms over one summer month.
• Integrated 15-minute weather data and aerial light detection and ranging (LiDAR) imagery to predict shaded areas.
• Employed nine distinct machine learning models to quantify and predict cow shade use patterns.

Main Results:

• Cow shade use was predictable and influenced by multiple interacting environmental variables.
• Shade utilization increased significantly when solar radiation surpassed 2 MJ/m² per hour.
• Optimal shade use occurred with at least 1.3 m² of shade per cow, particularly when temperatures exceeded 24°C and wind speed was below 4.2 m/s.

Conclusions:

• Machine learning and GPS data provide a scalable method for analyzing shade use in commercial dairy herds.
• Specific environmental thresholds (solar radiation, temperature, wind speed) and shade availability per cow dictate shade utilization.
• Findings support data-driven decisions for optimizing shade infrastructure investment and management to improve dairy cow welfare and production.