Livestock Informatics Toolkit: A Case Study in Visually Characterizing Complex Behavioral Patterns across Multiple Sensor Platforms, Using Novel Unsupervised Machine Learning and Information Theoretic Approaches.

Large and densely sampled sensor datasets can contain a range of complex stochastic structures that are difficult to accommodate in conventional linear models. This can confound attempts to build a more complete picture of an animal's behavior by aggregating information across multiple asynchronous sensor platforms. The Livestock Informatics Toolkit (LIT) has been developed in R to better facilitate knowledge discovery of complex behavioral patterns across Precision Livestock Farming (PLF) data streams using novel unsupervised machine learning and information theoretic approaches. The utility of this analytical pipeline is demonstrated using data from a 6-month feed trial conducted on a closed herd of 185 mix-parity organic dairy cows. Insights into the tradeoffs between behaviors in time budgets acquired from ear tag accelerometer records were improved by augmenting conventional hierarchical clustering techniques with a novel simulation-based approach designed to mimic the complex error structures of sensor data. These simulations were then repurposed to compress the information in this data stream into robust empirically-determined encodings using a novel pruning algorithm. Nonparametric and semiparametric tests using mutual and pointwise information subsequently revealed complex nonlinear associations between encodings of overall time budgets and the order that cows entered the parlor to be milked.

Diurnal and nocturnal activity budgets of zoo elephants in an outdoor facility.

The present study examined the activity budgets of 15 African elephants (1 bull, 6 cows, 2 male juveniles, 2 female juveniles, and 4 male calves) living at the San Diego Zoo Safari Park during the summers of 2010 and 2011. Onsite behavioral data (n = 600 hr) were collected for approximately 12 weeks from 0400 to 0830 and 1100 to 2400 during the 2010 and 2011 summer season. Foraging was the most common behavior state during the day followed by resting, and walking. During the evening hours, the elephants spent majority of their time foraging, resting, and sleeping. The average rate of self-maintenance behavior events (dust, wallow, etc.) increased from 0600 to 0700, 1100 to 1500, and from 1700 to 1900. Positive social behavior events (touch other, play, etc.) remained high from 0500 to 2300, with peaks at 0600, 1300, 1500, and 1900. Negative social events occurred at low rates throughout the day and night, with peaks at 0600, 1900, and 2200. The majority of positive behavior events during the daylight and nighttime hours involved the mother-calf pairs. Furthermore, the calves and juveniles initiated approximately 60% of all social events during the daytime and 57% of all social interactions at night. The results of this study demonstrate the differences between diurnal and nocturnal activity budgets of a multi-age and sex elephant herd in a zoological facility, which highlights the importance of managing elephants to meet their 24 hr behavioral needs.

Judgement bias of group housed gestating sows predicted by behavioral traits, but not physical measures of welfare.

Judgement bias testing has emerged as a potential tool for assessing affective states in animals. Researchers infer an animal's affective state based on an animal's response to an ambiguous stimulus that is intermediate to both the rewarded and punished conditioned stimuli. Animals can be classified as "optimistic" or having a positive affective state if the animal displays behaviors that suggest an increased expectation of reward in the face of ambiguous stimuli. Alternatively, animals can be classified "pessimistic" or having a negative affective state if the animal displays behaviors that suggest an increased expectation of punishment in the face of ambiguous stimuli. Recent reports in multiple species question what factors influence performance in judgement bias testing, and which may allow for erroneous conclusions regarding individual affective state. In order to better understand this concern, 25 female swine were subjected to behavioral assessments at critical rearing stages to determine response variability. These same individuals were then assessed for physical measures of welfare and judgement bias using the "go/no-go" task as breeding adults. Sows which were more aggressive approached the ambiguous, but not the positive, stimulus significantly faster than others. Both optimistic and pessimistic biases were observed despite all sows living in enriched housing, and, sows with more positive physical welfare measures (fewer skin lesions and healthy body condition) did not exhibit more optimistic judgement biases. Our data demonstrate that behavior traits, such as aggressiveness, can affect a sow's performance in a judgement bias test, while measures of physical health did not. We suggest that individual differences in behavior (e.g., bold-aggressive behavioral syndrome, or, proactive coping style) generate different emotional responses and can contribute to the animal's overall affective state more so than physical ailment. Our findings highlight the complexity of how different factors impact an animal's overall affective state and support the need for complementary measures in future JBT studies, including personality assessment.

Providing height to pullets does not influence hippocampal dendritic morphology or brain-derived neurotrophic factor at the end of the rearing period.

Pullets reared with diverse behavioral experiences are faster to learn spatial cognition tasks and acclimate more successfully to laying environments with elevated structures. However, the neural underpinnings of the improved spatial abilities are unclear. The objective of this study was to determine whether providing structural height in the rearing environment affected the development of the hippocampus and whether hippocampal neural metrics correlated with individual behavior on spatial cognition tasks. Female Dekalb White pullets were reared in a floor pen (FL), single-tiered aviary (ST), or two-tiered aviary (TT; 5 pens/treatment). Pullets completed floor-based Y-maze and elevated visual cliff tasks to evaluate depth perception at 15 and 16 wk, respectively. At 16 wk, brains were removed for Golgi-Cox staining (n = 12 for FL, 13 for ST, 13 total pullets for TT; 2 to 3 pullets/pen) and qPCR to measure gene expression of brain-derived neurotrophic factor (BDNF; n = 10 for FL, 11 for ST, and 9 pullets for TT). Rearing environment did not affect various morphometric outcomes of dendritic arborization, including Sholl profiles; mean dendritic length; sum dendritic length; number of dendrites, terminal tips, or nodes; soma size; or BDNF mRNA expression (P > 0.05). Hippocampal subregion did affect dendritic morphology, with multipolar neurons from the ventral subregion differing in several characteristics from multipolar neurons in the dorsomedial or dorsolateral subregions (P < 0.05). Neural metrics did not correlate with individual differences in behavior during the spatial cognition tasks. Overall, providing height during rearing did not affect dendritic morphology or BDNF at 16 wk of age, but other metrics in the hippocampus or other brain regions warrant further investigation. Additionally, other structural or social components or the role of animal personality are areas of future interest for how rearing environments influence pullet behavior.

Researching Human-Cattle Interaction on Rangelands: Challenges and Potential Solutions.

Human-animal interaction (HAI) research spans across many scientific fields and animal taxa. For livestock species, HAI research tends to focus on animals that are managed in close proximity with humans such as poultry, dairy cattle, and swine. Given the nature of rangeland cattle production, HAI research with beef cattle often occurs in and around the processing environment. This high arousal context may skew behavioral and physiological responses by the animals due to the potentially negative interaction. The aim of this review is to describe cattle production on rangelands, examine the considerations and limitations of current HAI research used to evaluate interaction quality or traits of rangeland cattle, identify contexts in which rangeland cattle interact with humans, and provide recommendations for improving future HAI research with rangeland cattle. Current research delineating individual differences in response to humans by beef cattle occur during routine husbandry and management on rangelands (pragmatic) and in a research context (experimental). Human-cattle interactions can be distinguished based on the quality and goal of the interaction into four broad categories: human presence, human approach, human contact, and restraint. Limitations of HAI research with rangeland cattle are identified and reconciled by recommendations for HAI research that can take place outside of the processing environment (i.e., while cattle are ruminating, resting or grazing on rangelands).

Mind the Queue: A Case Study in Visualizing Heterogeneous Behavioral Patterns in Livestock Sensor Data Using Unsupervised Machine Learning Techniques.

Sensor technologies allow ethologists to continuously monitor the behaviors of large numbers of animals over extended periods of time. This creates new opportunities to study livestock behavior in commercial settings, but also new methodological challenges. Densely sampled behavioral data from large heterogeneous groups can contain a range of complex patterns and stochastic structures that may be difficult to visualize using conventional exploratory data analysis techniques. The goal of this research was to assess the efficacy of unsupervised machine learning tools in recovering complex behavioral patterns from such datasets to better inform subsequent statistical modeling. This methodological case study was carried out using records on milking order, or the sequence in which cows arrange themselves as they enter the milking parlor. Data was collected over a 6-month period from a closed group of 200 mixed-parity Holstein cattle on an organic dairy. Cows at the front and rear of the queue proved more consistent in their entry position than animals at the center of the queue, a systematic pattern of heterogeneity more clearly visualized using entropy estimates, a scale and distribution-free alternative to variance robust to outliers. Dimension reduction techniques were then used to visualize relationships between cows. No evidence of social cohesion was recovered, but Diffusion Map embeddings proved more adept than PCA at revealing the underlying linear geometry of this data. Median parlor entry positions from the pre- and post-pasture subperiods were highly correlated (R = 0.91), suggesting a surprising degree of temporal stationarity. Data Mechanics visualizations, however, revealed heterogeneous non-stationary among subgroups of animals in the center of the group and herd-level temporal outliers. A repeated measures model recovered inconsistent evidence of a relationships between entry position and cow attributes. Mutual conditional entropy tests, a permutation-based approach to assessing bivariate correlations robust to non-independence, confirmed a significant but non-linear association with peak milk yield, but revealed the age effect to be potentially confounded by health status. Finally, queueing records were related back to behaviors recorded via ear tag accelerometers using linear models and mutual conditional entropy tests. Both approaches recovered consistent evidence of differences in home pen behaviors across subsections of the queue.