Machine Learning-Based Assessment and Forecasting of Temperature–Humidity Index for Dairy Heat Stress Using Multivariate and Panel Data Approaches in Bihar, India

In the humid, sun-drenched fields of India's dairy country, the daily rhythm of the farm is dictated by the weather. For the cows that provide the milk, the air is not just a backdrop; it is a physical force that can either support their health or slowly drain their energy. When the air grows hot and heavy, cows struggle to cool themselves, a condition known as heat stress. This is not merely a matter of comfort; it is a direct threat to their ability to produce milk, stay healthy, and reproduce. To measure this invisible burden, scientists use a single number called the Temperature–Humidity Index. Think of this index as a combined reading of the thermometer and the moisture in the air, giving a clear picture of how oppressive the environment feels to an animal. In regions like Bihar, where the summers are long and the humidity is high, understanding this number is essential for farmers trying to keep their herds productive.

A team of researchers set out to map this invisible burden across five major dairy districts in Bihar over a twenty-year period, from 2006 to 2025. They gathered a massive collection of weather data, tracking the highest and lowest daily temperatures, the amount of moisture in the air, and the speed of the wind. Their goal was to see how these conditions changed over time and to build a system that could predict when the heat would become dangerous for the cattle. Instead of relying on simple averages, they turned to advanced computer programs capable of finding complex patterns in the data. These programs, known as machine learning models, were trained to recognize the specific combinations of weather that signal danger, much like a seasoned farmer might learn to read the sky, but with the precision of a computer analyzing thousands of data points at once.

The analysis revealed that the cows in these districts are frequently exposed to conditions that push them into a state of mild to moderate heat stress. On average, the heat index hovered around a level that indicates the animals are working hard just to stay cool. The researchers found that the most critical factor was not the scorching heat of the afternoon, but the warmth that lingers through the night. When the temperature does not drop significantly after sunset, the cows cannot recover from the day's heat, leading to a cumulative burden that builds up over time. This finding challenges the idea that only the peak daytime temperatures matter; it suggests that the lack of nighttime cooling is the true driver of stress in this region.

To make sense of the diverse weather patterns across the five districts, the researchers grouped the data into three distinct types of climate. One group was characterized by warm, sticky air where humidity was high. Another was hot and dry, with intense heat but less moisture. The third was a cooler period where the animals could rest more easily. This distinction is vital because it shows that a single rule for the entire state would not work; what helps a cow in a hot, dry district might not be the right solution for one in a humid, warm area. The study confirmed that the heat stress experienced by dairy cattle is not a uniform event but a complex mix of local conditions that vary from place to place and season to season.

Using powerful computer algorithms, the team tested different methods to predict these stress levels. One method, known as Random Forest, proved to be the most accurate, correctly identifying the heat stress category in more than 99 percent of the cases. This level of precision means that the model can reliably tell farmers when conditions are shifting from safe to dangerous. The computer also identified that the minimum temperature—the lowest point reached during the night—was the single most important clue for predicting stress. While wind speed and humidity played a role, the failure of the night to cool down was the dominant signal.

Looking ahead, the researchers used these patterns to forecast what the coming years might look like. Their predictions show a clear, repeating cycle: the heat will rise steadily through the spring, reaching its peak in the summer months of May, June, and July, before cooling down again in the winter. The forecast suggests that the highest levels of stress will consistently occur in June. While the data did not show a dramatic, long-term increase in heat over the last two decades, the seasonal pattern remains relentless. The cows will face the same intense heat every year, with the most critical period arriving in the early summer.

The value of this work lies in its ability to turn vague worries about the weather into concrete, actionable information. By knowing exactly when and where the heat will be most severe, farmers and policymakers can prepare in advance. They can plan for better ventilation, provide shade, or adjust feeding times to help the animals cope. The study demonstrates that by combining old-fashioned weather records with modern computer intelligence, it is possible to build a reliable early warning system. This approach offers a way to protect the health of dairy herds and secure the livelihoods of those who depend on them, ensuring that even as the climate changes, the cows can continue to thrive in the face of the heat.