Using precision livestock farming (PLF) technologies to assess the impact of environmental stressors on animal welfare and production efficiency on modern dairy farms

Using precision livestock farming (PLF) technologies to assess the impact of environmental stressors on animal welfare and production efficiency on modern dairy farms

In modern dairy farming systems, heat stress is still a significant challenge. Dairy cows will encounter sub-optimal welfare which can result in production decline, diseases and even mortality, especially for high-producing cows with lower heat tolerance. The frequency and magnitude of heat stress events or heat waves are predicted to keep increasing in coming decades associated with global warming. Therefore, greater attention is being paid to alleviating the effects of heat stress on dairy cows and livestock generally. Modelling and on-farm experiments have been undertaken in many countries to assess the influence of heat stress on livestock using modern computer technologies and other hi-tech tools. At the same time, mitigation approaches such as optimal shed structure, new cooling facilities, targeted feeding regimes, improved farm management and genetic selection have all been studied extensively. However, due to differences between farm conditions and varying heat tolerance of different breeds and coping ability, the results from different heat stress models provided a variety of thresholds for on-farm decision support. Therefore, determination of accurate heat stress thresholds to facilitate practical mitigation options are still difficult.

This study was initiated by summarizing the progresses achieved by previous studies on intensively kept dairy cows in relation to measuring, assessing and mitigating their heat stress. By taking comparative analysis of the published studies about thermal indices, animal responses and mitigation solutions, a range of recommendations were given for developing more accurate assessment and designing of more effective mitigation options. The review suggested that for achieving accurate and applicable thresholds of heat stress, it is necessary to establish monitoring systems embedded into routine farm management systems, which can be an add-on unit of current robotic milking system (RMS). The robust monitoring system would measure real-time data from the ambient environment, animal responses, as well as the operation pattern of mitigations. Furthermore, by facilitating big-data analysis techniques to be used on individual farms, (or for individual animal) it might be possible to implement self-calibration procedure for the assessment, thresholds and control algorithms responding to varied cow’s production status, farm management factors and local climate changes.

The follow-up research presented in this thesis demonstrated the possibility of establishing more accurate heat stress threshold by taking advantage of the routinely collected datasets on robotic dairy farms and local weather stations. The dairy farm observed in this study situated in a subtropical climate region, held around 150 lactating cows and applied RMS with semi-free traffic. The farm management system recorded specific production, health and behaviour information of each individual animal over 5-year period (2013-2017), which was utilized for the analysis in this study. The historical climate conditions were measured by local weather station with dataset accessible on a government website, which provided the data of daily thermal parameters for this research. Furthermore, data-loggers were also positioned on farm from April 2016 to November 2017 to measure thermal parameters hourly.

By using the collected information, this study compared the performance of published thermal comfort indices (TCIs) as the indicators of cows’ responses to heat stress. These TCIs included temperature humidity index (THI), black globe humidity index (BGHI), environmental stress index (ESI), equivalent temperature index (ETI), heat load index (HLI), respiration rate index (RR) and comprehensive climate index (CCI). The comparison also included the basic thermal parameters: dry bulb temperature (Tdb), relative humidity (RH), wet bulb temperature (Twb) and dew point temperature (Tdp). The strength of their correlation with daily milk yield (DMY) and milk temperature (MT) was tested statistically. The regression analysis using climate dataset from local weather station and on-farm data-loggers were also compared to validate the accuracy of online data source. The statistical analysis found similar performance between TCIs and Tdb. It was also found that the inaccuracy of online data source, due to spatial variability between on-farm measurement and local weather station, could be neglected when modelling the association between TCIs and MT. A general threshold with significant decline of DMY was identified as THI>64 for cows with DMY around 31 kg/cow/day.

As Tdb can provide sufficient accuracy in the prediction of heat stress, the dynamic thresholds of daily minimum and mean temperature (Tmin and Tmean) were then established using individual information of 126 cows. The dataset was grouped according to the age, body weight (BW) and days in milk (DIM) of cows. Specific thresholds for different groups were identified using single broke-line regression between temperature and DMY or MT. Machine learning model was applied to transform these thresholds of different group into a decision tree of dynamic thresholds, which achieved overall 94% accuracy with the thresholds of Tmin, and 79% accuracy with the thresholds of Tmean. Moreover, for the whole herd, multiple broken-line regression was applied, which established four stages of heat stress including as thermal comfort stage (Tmin < 5 oC, Tmean < 9 oC), mild heat stress (Tmin: 5-6 oC, Tmean: 9-11 oC), effective heat stress (Tmin: 6-14 oC, Tmean: 11-16 oC) and critical heat stress (Tmin > 14 oC, Tmean > 16 oC) based on the change of DMY and MT.

To gain more understanding of the heat stress influence on animal behaviours in RMS, extra dependent variables were imported into new models involving rumination time (RT), time of milking (TM), miking frequency (MF), milking duration (MD), milking speed (MS), and milk yield per milking (MY). A new index – rumination efficiency index (REI) was created to evaluate the efficiency of rumination. According to the multiple broken-line regression, 5 minutes reduction of RT, 0.08 kg/cow/hour reduction of REI and 1% increase of low efficiency miking (LEM) were found to be associated with raising 1 oC of Tmean. It was also demonstrated that cows could not adjust their pattern of milking behaviour (e.g. visiting time pattern) coping with heat stress. Statistically, 86% of their milking event happened between 07:00 AM and 09:00 AM. However, REI and RMS performance can be improved by adjusting the pattern of milking behaviour such as milking interval (MI). The financial comparison between current pattern and adjusted pattern estimated that nearly $400 daily benefit could be gained.

In addition, this study also analysed the cumulative and lag effect of heat stress which were time-related. For the short-term effect, an intensity duration index (IDI) was defined by multiplying the mean temperature of the heat stress period with the duration of the period. Multiple levels of heat stress were then identified by IDI with different decline rate of DMY from -0.01 to -0.13 kg/cow/IDI. For long-term heat stress, the lag and cumulative effect was demonstrated by the negative correlation between the duration of heat stress during dry-off period and the production performance of the subsequent lactation period. The lag effect was found to be 3-4 days, while the cumulative effect could last for about 2 months. The regression between DMY and the average temperature of the period with heat stress during the 2 months before test day (HSmean) was found to perform stronger correlation (R2 equals 0.73-0.77) than the regression between DMY and same day’s temperature (R2 equals 0.65-0.68).