Estimation of gestating sows' welfare status based on machine learning methods and behavioral data.

Estimating the welfare status at an individual level on the farm is a current issue to improve livestock animal monitoring. New technologies showed opportunities to analyze livestock behavior with machine learning and sensors. The aim of the study was to estimate some components of the welfare status of gestating sows based on machine learning methods and behavioral data. The dataset used was a combination of individual and group measures of behavior (activity, social and feeding behaviors). A clustering method was used to estimate the welfare status of 69 sows (housed in four groups) during different periods (sum of 2 days per week) of gestation (between 6 and 10 periods, depending on the group). Three clusters were identified and labelled (scapegoat, gentle and aggressive). Environmental conditions and the sows' health influenced the proportion of sows in each cluster, contrary to the characteristics of the sow (age, body weight or body condition). The results also confirmed the importance of group behavior on the welfare of each individual. A decision tree was learned and used to classify the sows into the three categories of welfare issued from the clustering step. This classification relied on data obtained from an automatic feeder and automated video analysis, achieving an accuracy rate exceeding 72%. This study showed the potential of an automatic decision support system to categorize welfare based on the behavior of each gestating sow and the group of sows.

Prediction of the daily nutrient requirements of gestating sows based on sensor data and machine-learning algorithms.

Precision feeding is a strategy for supplying an amount and composition of feed as close that are as possible to each animal's nutrient requirements, with the aim of reducing feed costs and environmental losses. Usually, the nutrient requirements of gestating sows are provided by a nutrition model that requires input data such as sow and herd characteristics, but also an estimation of future farrowing performances. New sensors and automatons, such as automatic feeders and drinkers, have been developed on pig farms over the last decade, and have produced large amounts of data. This study evaluated machine-learning methods for predicting the daily nutrient requirements of gestating sows, based only on sensor data, according to various configurations of digital farms. The data of 73 gestating sows was recorded using sensors such as electronic feeders and drinker stations, connected weight scales, accelerometers, and cameras. Nine machine-learning algorithms were trained on various dataset scenarios according to different digital farm configurations (one or two sensors), to predict the daily metabolizable energy and standardized ileal digestible lysine requirements for each sow. The prediction results were compared to those predicted by the InraPorc model, a mechanistic model for the precision feeding of gestating sows. The scenario predictions were also evaluated with or without the housing conditions and sow characteristics at artificial insemination usually integrated into the InraPorc model. Adding housing and sow characteristics to sensor data improved the mean average percentage error by 5.58% for lysine and by 2.22% for energy. The higher correlation coefficient values for lysine (0.99) and for energy (0.95) were obtained for scenarios involving an automatic feeder system (daily duration and number of visits with or without consumption) only. The scenarios including an automatic feeder combined with another sensor gave good performance results. For the scenarios using sow and housing characteristics and automatic feeder only, the root mean square error was lower with gradient tree boosting (0.91 MJ/d for energy and 0.08 g/d for lysine) compared with those obtained using linear regression (2.75 MJ/d and 1.07 g/d). The results of this study show that the daily nutrient requirements of gestating sows can be predicted accurately using data provided by sensors and machine-learning methods. It paves the way for simpler solutions for precision feeding.

New technologies, such as sensors and automatons, are being developed in agriculture to reduce workload or help farmers make management decisions. The most common approach to the analysis of the huge amount of data generated by these technologies is to use machine-learning algorithms, to detect health or welfare problems for example. The hypothesis was that these automatically collected data and algorithms could also serve to predict the nutrient requirements of gestating sows, usually calculated based on complex models that require a lot of on-farm input data. The predictions of 22 scenarios were compared based on different combinations of sensor data, with the prediction of a nutritional model for gestating sows. The results of nine algorithms applied to the different scenarios were also compared. The results suggested that feeder data, alone or in combination with another sensor, predicted nutrient requirements with high accuracy. Data from other sensors combined with additional information about the sow (i.e., age and body weight) also led to high prediction accuracy. The difference between the algorithms evaluated was relatively significant, but all showed acceptable prediction results, especially non-linear algorithms. In conclusion, this work demonstrated the possibility of accurately predicting daily nutrient requirements for each sow using sensor data and machine-learning algorithms.

Effects of thermal conditions on gestating sows' behaviors and energy requirements.

Room temperature and individual behavior may influence the energy requirements of gestating sows. These factors are not yet integrated on a daily and individual basis in the calculation of these requirements. The objective of this study was to quantify the effect of temperatures on the sows' behaviors, especially on the level of physical activity, and on the energy requirements of gestating sows. Over four consecutive weeks, the temperature of two gestation rooms was maintained at medium temperatures (16.7 °C and 18.5 °C, respectively, for room 1 and room 2) for the first and third week, at low temperatures (14.4 °C and 15.3 °C) for the second week, and at high temperatures (31.6 °C and 31.9 °C) for the fourth week. Individual behavior was manually recorded based on videos and the data used to estimate the physical activity and social interactions of 37 gestating sows separated into two groups. The videos were analyzed over two periods of 5 h ("Feeding period" from 2300 to 0400 hours, "Resting period" from 1330 to 1830 hours). The energy requirements were calculated by the InraPorc model, modified for gestating sows, on the basis of a thermo-neutral situation and an average activity of 4 h standing per day for all the sows. The sows of one group were less active in high than low temperatures (83 vs. 103 min standing or walking over 5 h, P < 0.001). Isolation for high temperatures or huddling for low temperatures could be observed when sows were lying down. The sows spent more time lying laterally with high temperatures than low temperatures (66% vs. 52% of time spent lying, respectively, P < 0.001). Both groups reacted differently to high temperatures, in one the sows changed their activity (lying more) whereas in the other they drank more water compared to medium temperatures (11 vs. 8.5 L/d, P = 0.01). In one group, with high temperatures the sows were fed above their requirements (they should have received 110 g of feed per day per sow less, P < 0.001) and with low temperatures the same group should have received 50 g/d per sow more to fulfill their requirements. For the second group of sows, the temperatures did not significantly affect the feed requirements. In conclusion, daily ambient temperature and individual physical activity seem to be relevant information to add in nutritional models to improve precision feeding.

Ambient temperature may influence the energy requirement of gestating sows, but this factor is not yet integrated daily in the calculation of this requirement. The objective of this study was to quantify the effect of temperatures on sow's behavior, physical activity, and energy requirements on gestating sows. The 37 gestating sows were housed in two groups for which the temperature of each room was maintained at different temperatures during four consecutive weeks: the first and third weeks at 18 °C on average (medium temperature), the second week at 15.5 °C (low temperature), and the last one at 32 °C (high temperature). The sows modified their behavior regarding the room temperature even though these changes differed regarding the group of sows. Compared to medium temperature, high temperatures may induce an increase of water consumption or of the time spent lying, and of the rectal temperature of some sows. Low temperatures may induce huddling and/or an increase in aggressiveness. Low and high temperatures seem to impact energy costs even though it depends on the group of sows. Therefore, ambient temperature and individual activity are relevant information to add into nutritional models to improve their accuracy of energy requirement prediction.

ASAS-NANP Symposium: Mathematical Modeling in Animal Nutrition: Opportunities and challenges of confined and extensive precision livestock production.

Modern animal scientists, industry, and managers have never faced a more complex world. Precision livestock technologies have altered management in confined operations to meet production, environmental, and consumer goals. Applications of precision technologies have been limited in extensive systems such as rangelands due to lack of infrastructure, electrical power, communication, and durability. However, advancements in technology have helped to overcome many of these challenges. Investment in precision technologies is growing within the livestock sector, requiring the need to assess opportunities and challenges associated with implementation to enhance livestock production systems. In this review, precision livestock farming and digital livestock farming are explained in the context of a logical and iterative five-step process to successfully integrate precision livestock measurement and management tools, emphasizing the need for precision system models (PSMs). This five-step process acts as a guide to realize anticipated benefits from precision technologies and avoid unintended consequences. Consequently, the synthesis of precision livestock and modeling examples and key case studies help highlight past challenges and current opportunities within confined and extensive systems. Successfully developing PSM requires appropriate model(s) selection that aligns with desired management goals and precision technology capabilities. Therefore, it is imperative to consider the entire system to ensure that precision technology integration achieves desired goals while remaining economically and managerially sustainable. Achieving long-term success using precision technology requires the next generation of animal scientists to obtain additional skills to keep up with the rapid pace of technology innovation. Building workforce capacity and synergistic relationships between research, industry, and managers will be critical. As the process of precision technology adoption continues in more challenging and harsh, extensive systems, it is likely that confined operations will benefit from required advances in precision technology and PSMs, ultimately strengthening the benefits from precision technology to achieve short- and long-term goals.

Interest and investment in precision technologies are growing within the livestock sector. Though these technologies offer many promises of increased efficiency and reduced inputs, there is a need to assess the opportunities and challenges associated with precision technology implementation in livestock production systems. In this review, precision livestock measurement and management tools are explained in the context of a logical and iterative five-step process that highlights the need for systems computer modeling to realize anticipated benefits from these technologies and avoid unintended consequences. This review includes key case studies to highlight past challenges and current opportunities within operations that house animals in a central area or building with sufficient infrastructure (confined livestock production systems) and other operation settings that utilize large grasslands that contain far less infrastructure (extensive livestock production systems). The key to precision livestock management success is training the next generation of animal scientists in computer modeling, precision technologies, computer programming, and data science while still being grounded in traditional animal science principles.

ASAS-NANP symposium: mathematical modeling in animal nutrition: limitations and potential next steps for modeling and modelers in the animal sciences.

The field of animal science, and especially animal nutrition, relies heavily on modeling to accomplish its day-to-day objectives. New data streams ("big data") and the exponential increase in computing power have allowed the appearance of "new" modeling methodologies, under the umbrella of artificial intelligence (AI). However, many of these modeling methodologies have been around for decades. According to Gartner, technological innovation follows five distinct phases: technology trigger, peak of inflated expectations, trough of disillusionment, slope of enlightenment, and plateau of productivity. The appearance of AI certainly elicited much hype within agriculture leading to overpromised plug-and-play solutions in a field heavily dependent on custom solutions. The threat of failure can become real when advertising a disruptive innovation as sustainable. This does not mean that we need to abandon AI models. What is most necessary is to demystify the field and place a lesser emphasis on the technology and more on business application. As AI becomes increasingly more powerful and applications start to diverge, new research fields are introduced, and opportunities arise to combine "old" and "new" modeling technologies into hybrids. However, sustainable application is still many years away, and companies and universities alike do well to remain at the forefront. This requires investment in hardware, software, and analytical talent. It also requires a strong connection to the outside world to test, that which does, and does not work in practice and a close view of when the field of agriculture is ready to take its next big steps. Other research fields, such as engineering and automotive, have shown that the application power of AI can be far reaching but only if a realistic view of models as whole is maintained. In this review, we share our view on the current and future limitations of modeling and potential next steps for modelers in the animal sciences. First, we discuss the inherent dependencies and limitations of modeling as a human process. Then, we highlight how models, fueled by AI, can play an enhanced sustainable role in the animal sciences ecosystem. Lastly, we provide recommendations for future animal scientists on how to support themselves, the farmers, and their field, considering the opportunities and challenges the technological innovation brings.

Modeling in the animal sciences has received a boost by large-scale adoption of sensor technology, increased computing power, and the further development of artificial intelligence (AI) in the form of machine learning (ML) and deep learning (DL) models. Together with open-source programming languages, extensive modeling libraries, and heavy marketing, modeling reached a larger audience via AI. However, like most technological innovations, AI overpromised. By adopting an almost singular model-centric view to solving business needs, models failed to integrate with existing business processes. Models, especially AI, need data and both need humans. Together, they need room to learn and fail and by offering them as the end-solution to a problem, they are unable to act as sparring partners for all relevant stakeholders. In this article, we highlight fundamental model limitations exemplified via AI, and we offer solutions toward a more sustainable adoption of AI as a catalyst for modeling. This means sharing data and code and placing a more realistic view on models. Universities and industry both play a fundamental role in offering technological prowess and business experience to the future modeler. People, not technology, are the key to a more successful adoption of models.

Evaluation of a decision support system for precision feeding of gestating sows.

Precision feeding (PF) with the daily mixing of 2 diets with different lysine content (high (H) or low (L)) was previously reported for growing pigs to reduce protein intake and N excretion compared with a conventional feeding (CF) based on a single diet (C). Using a simulation approach based on farm data, the objective of the present paper was to describe and evaluate a decision support system for the PF of gestating sows allowing the daily distribution of a tailored ration to each sow. Two datasets, 1 of 2,511 gestations (farm A) and 1 of 2,528 gestations (farm B), reporting sows' characteristics at insemination and objectives at farrowing were used as inputs for a Python model. This model, mainly based on InraPorc, calculates the nutrient requirements of each sow over gestation and simulates the impact of PF in comparison to CF. Simulated diets L, H, and C contained 3.0, 6.5, and 4.8 g/kg of standardized ileal digestible lysine (SID Lys) and 2.0, 3.3, and 2.5 g/kg of standardized total tract digestible phosphorus (STTD-P), respectively. The influence of farm, parity, gestation week, and their interactions, on calculated SID Lys and STTD-P requirements was analyzed applying a mixed model. The calculated SID Lys and STTD-P requirements increased markedly in the last third of gestation (P < 0.01) and were higher for primiparous than for multiparous sows, unless after week 14 for STTD-P requirement. The calculated SID AA and mineral requirements were lower for farm B than farm A (respectively, 2.94 vs. 3.08 g/kg for SID Lys and 1.30 vs. 1.35 g/kg for STTD-P, P < 0.01). On average, feed L represented 86% and 92% of the feed projected to be delivered by the PF strategy in farms A and B, respectively. Compared to CF, average calculated dietary SID Lys content was lowered by 27% and 32% with PF, for farms A and B, respectively, while average calculated dietary phosphorus content was lowered by 13% and 16%. The simulated proportions of sows in excess and deficient in SID Lys were reduced with PF. Compared to CF, the PF strategy allowed for a 3.6% reduction in simulated feed cost per sow during gestation, and reduced nitrogen and phosphorus intake (by 11.0% and 13.8%, respectively) and excretion (by 16.7% and 15.4%, respectively). To conclude, these simulations indicate that PF of gestating sow appears to be relevant to meet the amino acid requirement while reducing feed cost, and supplies and excretion of nitrogen and phosphorus.

Predicting physiological imbalance in Holstein dairy cows by three different sets of milk biomarkers.

Blood biomarkers may be used to detect physiological imbalance and potential disease. However, blood sampling is difficult and expensive, and not applicable in commercial settings. Instead, individual milk samples are readily available at low cost, can be sampled easily and analysed instantly. The present observational study sampled blood and milk from 234 Holstein dairy cows from experimental herds in six European countries. The objective was to compare the use of three different sets of milk biomarkers for identification of cows in physiological imbalance and thus at risk of developing metabolic or infectious diseases. Random forests was used to predict body energy balance (EBAL), index for physiological imbalance (PI-index) and three clusters differentiating the metabolic status of cows created on basis of concentrations of plasma glucose, ß-hydroxybutyrate (BHB), non-esterified fatty acids (NEFA) and serum IGF-1. These three metabolic clusters were interpreted as cows in balance, physiological imbalance and "intermediate cows" with physiological status in between. The three sets of milk biomarkers used for prediction were: milk Fourier transform mid-IR (FT-MIR) spectra, 19 immunoglobulin G (IgG) N-glycans and 8 milk metabolites and enzymes (MME). Blood biomarkers were sampled twice; around 14 days after calving (days in milk (DIM)) and around 35 DIM. MME and FT-MIR were sampled twice weekly 1-50 DIM whereas IgG N-glycan were measured only four times. Performances of EBAL and PI-index predictions were measured by coefficient of determination (R2cv) and root mean squared error (RMSEcv) from leave-one-cow-out cross-validation (cv). For metabolic clusters, performance was measured by sensitivity, specificity and global accuracy from this cross-validation. Best prediction of PI-index was obtained by MME (R2cv = 0.40 (95 % CI: 0.29-0.50) at 14 DIM and 0.35 (0.23-0.44) at 35 DIM) while FT-MIR showed a better performance than MME for prediction of EBAL (R2cv = 0.28 (0.24-0.33) vs 0.21 (0.18-0.25)). Global accuracies of predicting metabolic clusters from MME and FT-MIR were at the same level ranging from 0.54 (95 % CI: 0.39-0.68) to 0.65 (0.55-0.75) for MME and 0.51 (0.37-0.65) to 0.68 (0.53-0.81) for FT-MIR. R2cv and accuracies were lower for IgG N-glycans. In conclusion, neither EBAL nor PI-index were sufficiently well predicted to be used as a management tool for identification of risk cows. MME and FT-MIR may be used to predict the physiological status of the cows, while the use of IgG N-glycans for prediction still needs development. Nevertheless, accuracies need to be improved and a larger training data set is warranted.

Exploration of individual variability to better predict the nutrient requirements of gestating sows1.

Sows often receive the same feed during gestation even though their nutrient requirements vary during gestation and among sows. The objective of this study was to report the variability in nutrient requirement among sows and during gestation, in order to develop a precision feeding approach. A data set of 2,511 gestations reporting sow characteristics at insemination and their farrowing performance was used as an input for a Python model, adapted from InraPorc, predicting nutrient requirement during gestation. Total metabolizable energy (ME) requirement increased with increasing litter size, gestation weeks, and parity (30.6, 33.6, and 35.5 MJ/d for parity 1, 2, and 3 and beyond, respectively, P < 0.01). Standardized ileal digestible lysine (SID Lys) requirement per kg of diet increased from weeks 1 to 6 of gestation, remained stable from weeks 7 to 10, and increased again from week 11 until the end of gestation (P < 0.01). Average Lys requirement increased with increasing litter size (SID Lys: 3.00, 3.27, 3.50 g/kg for small, medium and large litters, P < 0.01) and decreased when parity increased (SID Lys: 3.61, 3.17, 2.84 g/kg for parity 1, 2, and 3++, P < 0.01). Standardized total tract digestible phosphorus (STTD-P) and total calcium (Total-Ca) requirements markedly increased after week 9, with litter size, and decreased when parity increased (STTD-P: 1.36 vs. 1.31 g/kg for parity 1 and parity 3 and beyond; Total-Ca: 4.28 vs. 4.10 g/kg for parity 1 and parity 3 and beyond, P < 0.01). Based on empirical cumulative distribution functions, a 4-diets strategy, varying in SID Lys and STTD-P content according to parity and gestation period (P1 from weeks 0 to 11, P2 from weeks 12 to 17), may be put forward to meet the requirements of 90% of the sows (2 diets for multiparous sows: P1: 2.8 g SID Lys/kg and 1.1 g STTD-P/kg; P2: 4.5 g SID Lys/kg and 2.3 g STTD-P/kg; and 2 diets for primiparous sows: P1: 3.4 g SID Lys/kg and 1.1g STTD-P/kg; P2: 5.0 g SID Lys/kg, 2.2 g STTD-P/kg). Better considering the high variability of sow requirement should thus make it possible to optimize their performance whilst reducing feeding cost and excretion. Feeding sows closer to their requirement may initially be achieved by grouping and feeding sows according to gestation week and parity, and ultimately by feeding sows individually using a smart feeder allowing the mixing of different feeds differing in their nutrient content.

Dynamic modeling of nutrient use and individual requirements of lactating sows1.

Nutrient requirements of sows during lactation are related mainly to their milk yield and feed intake, and vary greatly among individuals. In practice, nutrient requirements are generally determined at the population level based on average performance. The objective of the present modeling approach was to explore the variability in nutrient requirements among sows by combining current knowledge about nutrient use with on-farm data available on sows at farrowing [parity, BW, backfat thickness (BT)] and their individual performance (litter size, litter average daily gain, daily sow feed intake) to estimate nutrient requirements. The approach was tested on a database of 1,450 lactations from 2 farms. The effects of farm (A, B), week of lactation (W1: week 1, W2: week 2, W3+: week 3 and beyond), and parity (P1: 1, P2: 2, P3+: 3 and beyond) on sow performance and their nutrient requirements were evaluated. The mean daily ME requirement was strongly correlated with litter growth (R2 = 0.95; P < 0.001) and varied slightly according to sow BW, which influenced the maintenance cost. The mean daily standardized ileal digestible (SID) lysine requirement was influenced by farm, week of lactation, and parity. Variability in SID lysine requirement per kg feed was related mainly to feed intake (R2 = 0.51; P < 0.001) and, to a smaller extent, litter growth (R2 = 0.27; P < 0.001). It was lowest in W1 (7.0 g/kg), greatest in W2 (7.9 g/kg), and intermediate in W3+ (7.5 g/kg; P < 0.001) because milk production increased faster than feed intake capacity did. It was lower for P3+ (6.7 g/kg) and P2 sows (7.3 g/kg) than P1 sows (8.3 g/kg) due to the greater feed intake of multiparous sows. The SID lysine requirement per kg of feed was met for 80% of sows when supplies were 112 and 120% of the mean population requirement on farm A and B, respectively, indicating higher variability in requirements on farm B. Other amino acid and mineral requirements were influenced in the same way as SID lysine. The present modeling approach allows to capture individual variability in the performance of sows and litters according to farm, stage of lactation, and parity. It is an initial step in the development of new types of models able to process historical farm data (e.g., for ex post assessment of nutrient requirements) and real-time data (e.g., to control precision feeding).

Housing of Cull Sows in the Hours before Transport to the Abattoir-An Initial Description of Sow Behaviour While Waiting in a Transfer Vehicle.

In modern pig production, sows are transported by road to abattoirs. For reasons of biosecurity, commercial trucks may have limited access to farms. According to Danish regulations, sows can be kept in stationary transfer vehicles away from the farm for up to two hours before being loaded onto the commercial truck. We aimed to describe the behaviour of sows in transfer vehicles. This preliminary, exploratory study included data from 11 loads from a total of six Danish sow herds. Selection of animals to be slaughtered was done by the farmers. Clinical registrations were made before collection of the sows, after which they (in groups of 7-13) were mixed and moved to the transfer vehicle (median stocking density: 1.2 sow/m²), and driven a short distance to a public road. The duration of the stays in the transfer vehicles before being loaded onto the commercial trucks ranged from 6-59 min. During this period, the median frequency of aggressive interactions per load was 18 (range: 4-65), whereas the median frequency of lying per load was 1 (range: 0-23). The duration of the stay correlated positively with the frequency of aggressive interactions (rs = 0.89; n = 11; p < 0.001) and with the frequency of lying (rs = 0.62; n = 11; p < 0.05). Frequency of aggressive interactions correlated positively with the temperature inside the transfer vehicle (rs = 0.89; n = 7; p < 0.001). These preliminary results are the first to describe the behaviour of cull sows during waiting in transfer vehicles, and may suggest that this period can be challenging for sow welfare, especially for longer stays and during hot days.

Social housing improves dairy calves' performance in two cognitive tests.

Early social housing is known to benefit cognitive development in laboratory animals. Pre-weaned dairy calves are typically separated from their dam immediately after birth and housed alone, but no work to date has addressed the effect of individual housing on cognitive performance of these animals. The aim of this study was to determine the effects of individual versus social housing on two measures of cognitive performance: reversal learning and novel object recognition. Holstein calves were either housed individually in a standard calf pen (n = 8) or kept in pairs using a double pen (n = 10). Calves were tested twice daily in a Y-maze starting at 3 weeks of age. Calves were initially trained to discriminate two colours (black and white) until they reached a learning criterion of 80% correct over three consecutive sessions. Training stimuli were then reversed (i.e. the previously rewarded colour was now unrewarded, and vice-versa). Calves from the two treatments showed similar rates of learning in the initial discrimination task, but the individually housed calves showed poorer performance in the reversal task. At 7 weeks of age, calves were tested for their response to a novel object in eight tests over a two-day period. Pair-housed calves showed declining exploration with repeated testing but individually reared calves did not. The results of these experiments provide the first direct evidence that individual housing impairs cognitive performance in dairy calves.