Machine Learning Methods and Visual Observations to Categorize Behavior of Grazing Cattle Using Accelerometer Signals

Accelerometers worn by animals produce distinct behavioral signatures, which can be classified accurately using machine learning methods such as random forest decision trees. The objective of this study was to identify accelerometer signal separation among parsimonious behaviors. We achieved this objective by (1) describing functional differences in accelerometer signals among discrete behaviors, (2) identifying the optimal window size for signal pre-processing, and (3) demonstrating the number of observations required to achieve the desired level of model accuracy,. Crossbred steers (Bos taurus indicus; n = 10) were fitted with GPS collars containing a video camera and tri-axial accelerometers (read-rate = 40 Hz). Distinct behaviors from accelerometer signals, particularly for grazing, were apparent because of the head-down posture. Increasing the smoothing window size to 10 s improved classification accuracy (p < 0.05), but reducing the number of observations below 50% resulted in a decrease in accuracy for all behaviors (p < 0.05). In-pasture observation increased accuracy and precision (0.05 and 0.08 percent, respectively) compared with animal-borne collar video observations.

Predicting feed intake in confined beef cows

Six existing equations (three for nonlactating and three for lactating; NRC, 1987, Predicting feed intake of food-producing animals. Washington, DC: The National Academies Press, National Academy of Science; doi: 10.17226/950; NRC, 1996, Nutrient requirements of beef cattle, 7th Revised Edition: Update 1996. Washington, DC: The National Academies Press; doi: 10.17226/9791; Hibberd and Thrift, 1992. Supplementation of forage-based diets. J. Anim. Sci. 70:181. [Abstr]) were evaluated for predicting feed intake in beef cows. Each of the previously published equations are sensitive to cow-shrunk BW and feed energy concentration. Adjustments in feed intake prediction are provided for level of milk yield in NRC (1987. Predicting feed intake of food-producing animals. Washington, DC: The National Academies Press, National Academy of Science; doi: 10.17226/950) and NRC (1996 Nutrient requirements of beef cattle, 7th Revised Edition: Update 1996. Washington, DC: The National Academies Press; doi: 10.17226/9791) equations. The equation published in 1996 used data generated between 1979 and 1993. Our objectives were to validate the accuracy of the published equations using more recent data and to propose alternative prediction models. Criteria for inclusion in the evaluation dataset included projects conducted or published since 2002, direct measurement of feed intake, adequate protein supply, and pen feeding (no metabolism crate data). After removing outliers, the dataset included 53 treatment means for nonlactating cows and 32 treatment means for lactating cows. Means for the nonlactating dataset were dry matter intake (DMI) = 13.2 ± 2.9 kg/d, shrunk body weight (SBW) = 578 ± 83.9 kg, body condition score = 5.7 ± 0.73, and Mcal net energy for maintenance (NEm)/kg of feed = 1.27 ± 0.15 Mcal/kg. Means for the lactating dataset were DMI = 14.6 ± 2.24 kg/d, SBW = 503 ± 73.4 kg, body condition score = 4.7 ± 0.58, and Mcal NEm/kg feed = 1.22 ± 0.16. Simple linear regression was used to determine slope, intercept, and bias when observed DMI (y) was regressed against predicted DMI (x). The NRC (1996. Nutrient requirements of beef cattle, 7th Revised Edition: Update 1996. Washington, DC: The National Academies Press; doi: 10.17226/9791) nonlactating equation underestimated feed intake in diets moderate to high in energy density with intercept differing from 0 and slope differing from one (P ≤ 0.01). Average deviation from observed values was 2.4 kg/d. Similarly, when the NRC (1996. Nutrient requirements of beef cattle, 7th Revised Edition: Update 1996. Washington, DC: The National Academies Press; doi: 10.17226/9791) equation was used to predict DMI in lactating cows, the slope differed from one (P < 0.01) with average deviation from observed values of 3.0 kg/d. New models were developed by pooling the two datasets and including a categorical variable for stage of production (0 = nonlactating and 1 = lactating). Continuous variables included study-average SBW0.75 and diet NEm, Mcal/kg. The best-fit empirical model accounted for 68% of the variation in daily feed intake with standard error of the estimate Sy root mean squared error = 1.31. The proposed equation needs to be validated with independent data.

Using dynamic modelling to enhance the assessment of the beef water footprint

Current water footprint assessment methods make a meaningful assessment of livestock water consumption difficult as they are mainly static, thus poorly adaptable to understanding future water consumption and requirements. They lack the integration of fundamental ruminant nutrition and growth equations within a dynamic context that accounts for short- and long-term behaviour and time delays associated with economically significant beef-producing areas. The current study utilised the System Dynamics methodology to conceptualise a water footprint for beef cattle within a dynamic and mechanistic modelling framework. The problem of assessing the water footprint of beef cattle was articulated, and a dynamic hypothesis was formed to represent the Texas livestock water use system as the initial step in developing the Dynamic Beef Water Footprint model (DWFB). The dynamic hypothesis development resulted in three causal loop diagrams (CLD): cattle population, growth and nutrition, and the livestock water footprint, that captured the daily water footprint of beef (WFB). Simulations and sensitivity analysis from the hypothesised CLD structures indicated that the framework was able to capture the dynamic behaviour of the WFB system. These behaviours included key reinforcing and balancing feedback processes that drive the WFB. It is extremely difficult to identify policy interventions (i.e., management strategies) for complex systems, like the U.S. beef cattle system, because there are many actors (i.e., cow-calf, stocker, feedlot) and interrelated variables that have delayed effects within and across the supply chain. Identification and understanding of feedback processes driving water use over time will help to overcome policy resistance for more sustainable beef production. Thus, the causal loops identified in the current study provide a system-level insight for the drivers of the WFB within and across each major segment of the beef supply chain to address freshwater concerns more adequately. Further, the nutrient scenarios and sensitivity analysis revealed that the high versus low nutrient composition of pasture, hay, and concentrates resulted in a significant difference in the WFB (2 669 L/kg boneless beef, P < 0.05). The WFB was sensitive to changes in nutrient composition and specific water demand (m3/t) for each production phase, not only phases with high levels of concentrate feed use. As models evolve, there is potential for the DWFB to integrate precision livestock data, further improving quantification of the WFB, precision water-efficient strategies, and selection of water-efficient livestock.

Grassland-based ruminant farming systems in China: Potential, challenges and a way forward

With an increasing demand for high-quality, eco-friendly food products and growing concerns over ecological conservation, the development of ecology-based alternatives for ruminant production in China is urgently needed. This review discusses the capabilities for integrating grassland grazing into existing livestock farming systems to meet the contemporary human needs for high-quality foods and ecologically stable environments. Additionally, this review provides a critical analysis of the challenges and future directions associated with grassland-based ruminant farming systems. Integrating nutritional manipulation with grazing manipulation is critical for improving the productivity of grassland-based ecosystems and natural ecological functions. Biodiversity is the primary determinant of grassland ecosystem functions, while the composition and function of rumen microbiomes determine ruminant production performance. Future studies should focus on the following aspects: 1) how livestock grazing regulates grassland biodiversity and the mechanisms of grassland biodiversity maintenance, offering an important scientific basis for guiding grazing manipulation practices, including grazing intensity, livestock types, and grazing management practices; to 2) characterize the microbial ecology within the rumen of grazing ruminants to offer clarified instruction for the nutritional manipulation of grazing ruminants. Our recommendation includes creating a transdisciplinary system that integrates ecology, animal nutrition, and animal behavior to develop grassland-based ruminant farming systems sustainably, thereby achieving high-quality animal production and environmentally sustainable goals.

Modifying the National Research Council weight gain model to estimate daily gain for stockers grazing bermudagrass in the southern United States

The energy requirements, feed intake, and performance of grazing animals vary daily due to changes in weather conditions, forage nutritive values, and plant and animal maturity throughout the grazing season. Hence, realistic simulations of daily animal performance can be made only by the models that can address these changes. Given the dearth of simple, user-friendly models of this kind, especially for pastures, we developed a daily gain model for large-frame stockers grazing bermudagrass sCynodon dactylon (L.) Pers.], a widely used warm-season perennial grass in the southern United States. For model development, we first assembled some of the classic works in forage-beef modeling in the last 50 yr into the National Research Council (NRC) weight gain model. Then, we tested it using the average daily gain (ADG) data obtained from several locations in the southern United States. The evaluation results showed that the performance of the NRC model was poor as it consistently underpredicted ADG throughout the grazing season. To improve the predictive accuracy of the NRC model to make it perform under bermudagrass grazing conditions, we made an adjustment to the model by adding the daily departures of the modeled values from the data trendline. Subsequently, we tested the revised model against an independent set of ADG data obtained from eight research locations in the region involving about 4,800 animals, using 30 yr (1991-2020) of daily weather data. The values of the various measures of fit used, namely the Willmott index of 0.92, the modeling efficiency of 0.75, the R2 of 0.76, the root mean square error of 0.13 kg d-1, and the prediction error relative to the mean observed data of 24%, demonstrated that the revised model mimicked the pattern of observed ADG data satisfactorily. Unlike the original model, the revised model predicted more closely the ADG value throughout the grazing season. The revised model may be useful to accurately reflect the impacts of daily weather conditions, forage nutritive values, seasonality, and plant and animal maturity on animal performance.

Impact of forage diversity on forage productivity, nutritive value, beef cattle performance, and enteric methane emissions

Greenhouse gas emissions from the beef industry are largely attributed to the grazing sector, specifically from beef cattle enteric methane emissions. Therefore, the study objective was to examine how forage diversity impacts forage productivity, nutritive value, animal performance, and enteric methane emissions. This study occurred over three consecutive grazing seasons (2018 to 2020) and compared two common Midwest grazing mixtures: 1) a simple, 50:50 alfalfa:orchardgrass mixture (SIMP) and 2) a botanically diverse, cool-season species mixture (COMP). Fifty-six steers and heifers were adapted to an Automated Head Chamber System (AHCS) each year (C-Lock Inc., Rapid City, SD) and stratified into treatment groups based on acclimation visitation. Each treatment consisted of four pastures, three 3.2-ha and one 1.6-ha, with eight and four animals each, respectively. Forage production was measured biweekly in pre- and postgrazed paddocks, and forage nutritive value was analyzed using near-infrared reflectance spectroscopy. Shrunk body weights were taken monthly to determine animal performance. Forage availability did not differ between treatments (P = 0.69) but tended lower in 2018 (P = 0.06; 2.40 t dry matter ha⁻¹) than 2019 (2.92 t dry matter ha⁻¹) and 2020 (P = 0.10; 2.81 t dry matter ha⁻¹). Crude protein was significantly lower for COMP in 2018 compared with SIMP. Forage acid detergent fiber content was significantly lower for the COMP mixture (P = 0.02). The COMP treatment resulted higher dry matter digestibility (IVDMD48) in 2018 and 2019 compared with the SIMP treatment (P < 0.01). Animal performance did not differ between treatments (P > 0.50). There was a tendency for the COMP treatment to have lower enteric CH₄ production on a g d⁻¹ basis (P = 0.06), but no difference was observed on an emission intensity basis (g CH₄ kg⁻¹ gain; P = 0.56). These results would indicate that adoption of the complex forage mixture would not result in improved forage productivity, animal performance, or reduced emission intensity compared with the simple forage mixture.

Effects of Replacing Cottonseed Meal with Corn Dried Distillers' Grain on Ruminal Parameters, Performance, and Enteric Methane Emissions in Young Nellore Bulls Reared in Tropical Pastures

Two experiments were conducted to evaluate the effect of replacing cottonseed meal with DDG on ruminal parameters, methane (CH4) emissions (Experiment 1), and animal performance (Experiment 2) of young Nellore bulls grazing Marandu grass during the rainy season. Four supplementation strategies were used in both experiments: (1) Mineral supplementation (MS); (2) conventional multiple supplement (energy/protein) with cottonseed meal and citrus pulp (CMS); (3) CMS with 50% cottonseed meal replaced by DDG (50DDG); and (4) CMS with 100% cottonseed meal replaced by DDG (100DDG). The 50DDG condition resulted in greater intake of dry matter (p = 0.033), organic matter (OM) (p = 0.050), forage (p = 0.035), and digestible OM (p = 0.031) than 100DDG. The supplemented animals presented greater final body weight (BW) and average daily gain than the animals consuming MS (p = 0.011), and lower pH, acetate, and acetate:propionate (p < 0.05). However, the treatments had no influence on stocking rate, gain per area, and enteric CH4 emissions (p > 0.05). Replacing cottonseed meal with DDG does not result in great variations in ruminal parameters, animal performance, and enteric CH4 emissions of grazing Nellore cattle during the rearing phase in the wet season. Both protein sources in 0.3% BW supplementation can be used to intensify beef cattle production in pastures.

Herbage mass and allowance and animal genotype affect daily herbage intake, productivity, and efficiency of beef cows grazing native subtropical grassland

The beef sector in Campos grasslands must increase animal productivity without external inputs, while reducing environmental impact. The objective of this study was to estimate herbage intake (g/metabolic body weight [MBW]/d) of straightbred (Hereford/Angus) and crossbred (F1 of Hereford × Angus) beef cows grazing subtropical native grassland with High and Low herbage allowance (HA, 5 vs. 3 kg DM/kg bodyweight [BW]) during gestation and lactation and its relationship with biological efficiency of cow-calf productivity. Herbage intake (estimated via n-alkanes C32:C33 ratio) was measured during early (Ge1, -163 d prior calving) and mid to late [Gm1 (-83) and Gm2 (-90 d prior calving)] gestation and lactation (L0, L1, and L2, 60, 47, and 31d following calving) periods in 24 to 36 cows, selected to create 8 groups (4 per block) of HA × cow genotype treatment. Cows grazed native grassland year-round, under High and Low HA (except in winter). We analyzed the effect of cow genotype (straightbred vs. crossbred cows) and HA (High vs. Low) on herbage mass and height, daily herbage intake rate (DMI), cow body condition score (BCS), calf average daily gain (ADG) and BW at weaning (BWW) and g of calf weaned/kg DMI. High allowance improved DMI during lactation periods (High 115.6 vs. Low 94.1 ± 5.3; P < 0.05 g/MBW/d). Crossbred cows decreased DMI during gestation (Crossbred 81 vs. Straightbred 94 ± 4.3; P = 0.05 g/MBW/d) compared with straightbred cows. Crossbred and High HA improved biological efficiency, 40.0 vs. 26.2 and 36.0 vs. 29.7 g of calf/kg DMI, respectively. High allowance increased herbage mass and sites with greater canopy height that allow greater DMI, positively associated with cow BCS at weaning, calf ADG, BWW, and g of calf/kg DMI. Crossbred cows reduced DMI during gestation showing no greater annual DMI. Animal productivity and biological efficiency can be improved using High HA and crossbred cows, which should decrease the environmental impact of cow-calf systems.

Understanding intake on pastures: how, why, and a way forward

An assessment of dietary intake is a critical component of animal nutrition. Consumption of feed resources is the basis upon which feeding strategies and grazing management are based. Yet, as far back as 1948, researchers have lauded the trials and tribulations of estimation of the phenomenon, especially when focused on grazing animals and pasture resources. The grazing environment presents a unique situation in which the feed resource is not provided to the animal but, rather, the animal operates as the mechanism of harvest. Therefore, tools for estimation must be developed, validated, and applied to the scenario. There are a plethora of methods currently in use for the estimation of intake, ranging from manual measurement of herbage disappearance to digital technologies and sensors, each of which come with its share of advantages and disadvantages. In order to more firmly grasp these concepts and provide a discussion on the future of this estimation, the Forages and Pastures Symposium at the 2020 ASAS-CSAS-WSASAS Annual Meeting was dedicated to this topic. This review summarizes the presentations in that symposium and offers further insight into where we have come from and where we are going in the estimation of intake for grazing livestock.

Application of the California Net Energy System to grazed forage: feed values and requirements

The California Net Energy System (CNES) has been successfully used for many years to generate estimates of grazing animal energy requirements, supplemental needs, and energy value of grazed forage diets. Compared to pen feeding situations, validation of feed nutritive value estimates or animal performance projections are extremely difficult in grazing animals because many of the system inputs are constantly changing. A major difficulty in applying this or any energy accounting system in the field is acquiring accurate estimates of forage intake. We discuss the various equations available to estimate forage intake for grazing animals with emphasis on beef cows. Progress has been made in recent years although there remains substantial discrepancy among various equations, particularly in the upper range of forage digestibility. Validation work and further development is needed in this area. For lactating cows, our conclusion is that the adjustment of intake for milk production (0.2 kg increase in forage intake per kg of milk produced) needs to be increased to a minimum of 0.35. A particular challenge with the CNES for grazing beef cows is the dramatic interaction that can occur between genetic potential for production traits and nutrient availability. Examples from literature are provided and a case study is presented demonstrating that energy requirements are dynamic and depend on nutrients available in grazing systems. The CNES is a useful tool in grazing beef cattle management although there remains substantial opportunity and need to improve inputs and validate the system in grazing situations.

Herbage dry matter intake estimation of grazing dairy cows based on animal, behavioral, environmental, and feed variables

Information about the individual herbage DMI (HDMI) of grazing dairy cows is important for an efficient use of pasture herbage as an animal feed with a range of benefits. Estimating HDMI, with its multifaceted influencing variables, is difficult but may be attempted using animal, performance, behavior, and feed variables. In our study, 2 types of approaches were explored: 1 for HDMI estimation under a global approach (GA), where all variables measured in the 4 underlying experiments were used for model development, and 1 for HDMI estimation in an approach without information about the amount of supplements fed in the barn (WSB). The accuracy of these models was assessed. The underlying data set was developed from 4 experiments with 52 GA and 50 WSB variables and one hundred thirty 7-d measurements. The experiments differed in pasture size, herbage allowance, pregrazing herbage mass, supplements fed in the barn, and sward composition. In all the experiments, cow behavioral characteristics were recorded using the RumiWatch system (Itin and Hoch GmbH, Liestal, Switzerland). Herbage intake was estimated by applying the n-alkane method. Finally, HDMI estimation models with a minimal relative prediction error of 11.1% for use under GA and 13.2% for use under WSB were developed. The variables retained for the GA model with the highest accuracy, determined through various selection steps, were herbage crude protein, chopped whole-plant corn silage intake in the barn, protein supplement or concentrate intake in the barn, body weight, milk yield, milk protein, milk lactose, lactation number, postgrazing herbage mass, and bite rate performed at pasture. Instead of the omitted amounts of feed intake in the barn and, due to the statistical procedure for model reduction, the unconsidered variables postgrazing herbage mass and bite rate performed at pasture, the WSB model with the highest accuracy retained additional variables. The additional variables were total eating chews performed at pasture and in the barn, total eating time performed at pasture, number of total prehension bites, number of prehension bites performed at pasture, and herbage ash concentration. Even though behavioral characteristics alone did not allow a sufficiently accurate individual HDMI estimation, their inclusion under WSB improved estimation accuracy and represented the most valid variables for the HDMI estimation under WSB. Under GA, the inclusion of behavioral characteristics in the HDMI estimation models did not reduce the root mean squared prediction error. Finally, further adaptation, as well as validation on a more comprehensive data set and the inclusion of variables excluded in this study such as body condition score or gestation, should be considered in the development of HDMI estimation models.

An agent-based model of cattle grazing toxic Geyer's larkspur

By killing cattle and otherwise complicating management, the many species of larkspur (Delphinium spp.) present a serious, intractable, and complex challenge to livestock grazing management in the western United States. Among the many obstacles to improving our understanding of cattle-larkspur dynamics has been the difficulty of testing different grazing management strategies in the field, as the risk of dead animals is too great. Agent-based models (ABMs) provide an effective method of testing alternate management strategies without risk to livestock. ABMs are especially useful for modeling complex systems such as livestock grazing management, and allow for realistic bottom-up encoding of cattle behavior. Here, we introduce a spatially-explicit, behavior-based ABM of cattle grazing in a pasture with a dangerous amount of Geyer's larkspur (D. geyeri). This model tests the role of herd cohesion and stocking density in larkspur intake, finds that both are key drivers of larkspur-induced toxicosis, and indicates that alteration of these factors within realistic bounds can mitigate risk. Crucially, the model points to herd cohesion, which has received little attention in the discipline, as playing an important role in lethal acute toxicosis. As the first ABM to model grazing behavior at realistic scales, this study also demonstrates the tremendous potential of ABMs to illuminate grazing management dynamics, including fundamental aspects of livestock behavior amidst ecological heterogeneity.