Abstract
The temperature-humidity index (THI) remains one of the most widely used tools for assessing heat stress in dairy farming; however, its application is often limited by methodological inconsistencies and insufficient integration with welfare indicators. This study proposes a unified analytical framework for evaluating thermal load at the herd level by combining daily THI values with productivity, feed intake, and clinical indicators such as mastitis and lameness. The analysis was based on two years of herd-level data from a commercial dairy farm with naturally ventilated barns. General linear models (GLM) were applied to assess both direct and delayed effects of heat stress and to compare model reproducibility across years. The results confirmed that maximum daily THI had the strongest association with milk composition and dry matter intake, while cumulative heat load and elevated night-time THI contributed to increased mastitis and lameness incidence. The inclusion of welfare indicators substantially improved the explanatory power of THI-based models, providing a more biologically relevant assessment of heat stress. The proposed framework enhances the accuracy of herd-level monitoring and supports the development of predictive models for welfare-oriented management in dairy systems.