Advertisement

Validation of a portable, self-contained individual feeding unit for monitoring supplement intake of grazing cattle

      ABSTRACT

      Objective

      This experiment aimed to validate feed intake and bunk attendance data collected from a portable, self-contained feeding unit (SmartFeed; C-Lock Inc.) by comparing estimates collected electronically (electronic observations, EO) with direct observations and video recordings (DO).

      Materials and Methods

      To validate individual animal intake, 30 cows were allowed access to 3 portable units (2 feeding bins per unit) for 6 h. Feed in each bin was removed and weighed in an external scale at the end of each cow visit. To validate total feed disappearance, 75 cows were assigned to 1 of 3 pastures equipped with a portable unit. Feed disappearance over 24 h was assessed manually over 21 consecutive days using an external scale. Ten cows per pasture were randomly selected and identified to facilitate collection of video observational data. Two trained observers recorded animal identification and start and end times for a visit (n = 480 min/d; 3 d).

      Results and Discussion

      When regressing the EO and DO results within individual cow visit, a linear fit was detected (P < 0.01; r2 = 0.95). Methods were very strongly correlated (r = 0.91) when all visits were analyzed but decreased to strong (r = 0.78) when DO intake was ≤0.1 kg. When regressing the EO and DO results for total feed disappearance, a linear fit was detected (P < 0.01; r2 = 0.97): as DO total feed disappearance increased, the EO output linearly increased. For observational data, Cohen’s kappa coefficient for agreement between DO and EO was considered strong (0.96). The sensitivity of EO in yielding true positive results was 97%, whereas specificity in yielding true negative results was 99%.

      Implications and Applications

      These results indicate the portable, self-contained feeding units used herein are sufficiently effective for detecting cow presence at a bunk and measurement of feed consumption.

      Key words

      INTRODUCTION

      Adequate nutritional management of beef herds during each stage of production is critical to support growth, pregnancy, and lactation. Extensive beef cattle production systems depend on forage to provide a vast majority of nutrients, which can be variable in quality and may not be nutritionally complete (
      • Greene L.W.
      Designing mineral supplementation of forage programs for beef cattle..
      ). To maintain maximum long-term productivity, cattle producers must design supplementation programs according to animal stage and level of production as well as forage quality and quantity. Providing supplements to pasture-based cattle does not allow for measurement of individual animal intake, and as such, assessment of an individual’s specific response to supplementation is infeasible.
      Adoption of technology into supplementation programs represents an opportunity to facilitate unprecedented data collection of grazing cattle. Several advances in this field have improved efficiency of production, including the use of electronic ID systems, traditionally used to manage animals in feedlot settings. Recently, development of a self-contained portable unit (SmartFeed; C-Lock Inc.) has allowed researchers to characterize individual supplement intake of grazing cattle in large groups representative of commercial operations (
      • Wyffels S.A.
      • Dafoe J.M.
      • Parsons C.T.
      • Boss D.L.
      • DelCurto T.
      • Bowman J.G.P.
      The influence of age and environmental conditions on supplement intake by beef cattle winter grazing northern mixed-grass rangelands..
      ;
      • McCarthy K.L.
      • Undi M.
      • Becker S.
      • Dahlen C.R.
      Utilizing an electronic feeder to measure individual mineral intake, feeding behavior, and growth performance of cow–calf pairs grazing native range..
      ). This system is solar powered and includes a cloud-based interface, enabling researchers to detail the number of times an animal visits the feeder, length of time an animal spends at the feeder, and amount of feed consumed, providing valuable information about supplement intake in an extensive setting. Currently, several SmartFeed systems are being used across the United States (
      • Williams G.D.
      • Beck M.R.
      • Thompson L.R.
      • Horn G.W.
      • Reuter R.R.
      Variability in supplement intake affects performance of beef steers grazing dormant tallgrass prairie..
      ;
      • McCarthy K.L.
      • Underdahl S.R.
      • Undi M.
      • Becker S.
      • Dahlen C.R.
      Utilizing an electronic feeder to measure mineral and energy supplement intake in beef heifers grazing native range..
      ;
      • Dafoe J.M.
      • Wyffels S.A.
      • Parsons C.T.
      • Carter B.H.
      • DelCurto T.
      • Boss D.L.
      Techniques to estimate colostrum quality and the effects of cow age and prepartum supplement intake levels on colostrum quality and serum IgG levels..
      ;
      • Wyffels S.A.
      • Dafoe J.M.
      • Parsons C.T.
      • Boss D.L.
      • DelCurto T.
      • Bowman J.G.P.
      The influence of age and environmental conditions on supplement intake by beef cattle winter grazing northern mixed-grass rangelands..
      ;
      • McCarthy K.L.
      • Undi M.
      • Becker S.
      • Dahlen C.R.
      Utilizing an electronic feeder to measure individual mineral intake, feeding behavior, and growth performance of cow–calf pairs grazing native range..
      ); however, to date no data have been published that validate either the behavioral or intake measurements from this system. Thus, the objective of this study was to validate these measurements by comparing estimates collected electronically (EO) with those from direct observation and time-lapse video recordings (DO).

      MATERIALS AND METHODS

      All protocols were approved by the Mississippi State University Institutional Animal Care and Use Committee (#21-023). Each cow was equipped with an electronic ID tag containing a unique passive transponder (High-Performance ISO Half Duplex Electronic ID Tag, Allflex USA) attached to the left ear for individual identification to measure supplement intake and visits to the electronic feeder.

      Device

      The SmartFeed system was used to deliver supplement and measure intake (Figure 1). The portable self-contained unit features a stainless-steel feed bin (79 × 71 × 86 cm; 2 per unit) suspended on 2 load cells, radio frequency identification (RFID) reader and antenna, adjustable metal framework to limit access to one animal at a time, data acquisition system that records RFID tags, and feed bin weights at 1 Hz. For each visit to the bin, the system recorded the cow RFID, bin number, initial and final times, and beginning and ending mass in the bin. Three units (6 feeding bins) were tested during this study.
      Figure 1
      Figure 1A self-contained portable unit (SmartFeed; C-Lock Inc.) that was used to characterize individual supplement intake of grazing cattle in large groups representative of commercial operations. The exterior (A) and interior (B) of the unit are shown.

      Data Collection

      To validate feed intake on an individual animal basis, 30 nonlactating, pregnant crossbred cows were confined in a 35 × 10 m drylot pen with access to the 6 feeding bins. Supplement (50% soybean hulls, 50% corn gluten feed) was offered ad libitum in feed bins, whereas trained observers monitored one bin per person beginning immediately after the delivery of fresh feed and continuing for 6 h. The feed in each bin was removed and weighed in an external scale at the end of each visit, resulting in 230 cow visits across 6 feeding bins. The estimated feed consumption by the monitoring system (EO) was then compared with the consumption estimated manually (DO).
      To validate total feed disappearance, 75 nonlactating, pregnant crossbred cows were ranked by age and BCS (age = 6.1 ± 0.2 yr; BCS = 5.4 ± 0.05) and assigned to one of three 10-ha pastures (25 cows per pasture) equipped with an electronic feeder for a 21-d observation period. Feeders were placed near established supplement delivery locations, whereas cows were adapted to the system for 7 d before the initiation of data collection and provided with supplement at 0830 h to meet ad libitum intake. Supplement consisted of 50% dried distillers grain and 50% fine mixing salt to limit intake to approximately 0.5 kg/cow daily based on previous research with group-housed, self-fed grazing animals and this electronic feeder (
      • Reuter R.R.
      • Moffet C.A.
      • Horn G.W.
      • Zimmerman S.
      • Billars M.
      Technical Note: Daily variation in intake of a salt-limited supplement by grazing steers..
      ). Feed disappearance over 24 h was assessed manually over 21 consecutive days by removing residual feed from each bunk and using an external scale (DO) and compared with the sum of feed disappearances for each bunk recorded by the monitoring system at each cow visit (EO). Samples from weekly offered and nonconsumed feed were used for moisture analysis as in the study by
      • Parsons I.L.
      • Johnson J.R.
      • Kayser W.C.
      • Tedeschi L.O.
      • Carstens G.E.
      Characterization of feeding behavior traits in steers with divergent residual feed intake consuming a high-concentrate diet..
      .
      For video data collection, 10 cows per pasture were randomly selected and individually identified using an adhesive marker (Estrotec) before the initiation of video recordings. To facilitate collection of observational data, cows were evaluated on 3 consecutive days for 8 h on d 10 through 12. Video surveillance cameras (JVC Everio S GZ-MS120; Americas Corp.) were positioned approximately 0.5 m above feed bins in each unit and had the capability to store output internally for up to 15 h. Cameras were retrieved in the evening and replaced each morning during observation periods. Internal clocks on the video cameras and the data acquisition system were synchronized at the onset of the experiment. Two trained observers independently recorded animal identification number, bin number, and the start and end times for each bunk visit. The start and end times for a bunk-visit event were recorded once an animal had completely passed its RFID through the metal framework of the portable unit and was above the feeding surface of the bin. There was no attempt to quantify the orientation of animal’s head during a bunk-visit event (Figure 2). Raw data were summarized for each minute of the day (n = 480/d), recording cow presence (1) or absence (0) at the feed bin as determined by the SmartFeed system (EO) and video (DO).
      Figure 2
      Figure 2View of 2 of the bunks within a single SmartFeed (C-Lock Inc.) unit that were used to capture electronic and time-lapse-video observations.

      Data Analysis

      The GLM procedure of SAS (SAS Institute Inc.) was used to regress EO and DO results within visit or day for individual intake and total feed disappearance, respectively, to determine the best regression fit: linear, quadratically, or cubically. Pearson (r), Kendall (τ), and Spearman’s (ρ) rank correlation coefficients were also determined between EO and DO results, using the CORR procedure of SAS (SAS Institute Inc.). Estimated concordance correlation coefficients (CCC) with 95% confidence limits were used to determine the association between EO and DO results, representing concurrent measures of correlation, accuracy, and precision (
      • Lin L.I.
      A concordance correlation coefficient to evaluate reproducibility..
      ). Correlation coefficients (r, ρ, or CCC) were considered very strong if ≥0.80, strong if <0.80 and ≥0.60, and moderate if <0.60 and ≥0.40 (
      • Akoglu H.
      User’s guide to correlation coefficients..
      ). Kendall rank correlation (τ) yields smaller coefficient values (by approximately 70%) compared with Spearman’s rank (ρ) correlation (
      • Capéraà P.
      • Genest C.
      Spearman’s ρ is larger than Kendall’s τ for positively dependent random variables..
      ). Additionally, a Bland-Altman assessment for agreement was used to compare the 2 methods. The bias between the 2 methods was estimated by the mean difference (D) and the variation around bias estimated as the SD. A 95% limit of agreement (LOA) was defined as D ± 1.96SD (
      • Bland J.M.
      • Altman D.G.
      Measuring agreement in method comparison studies..
      ;
      • Yellareddygari S.K.R.
      • Gudmestad N.C.
      Bland-Altman comparison of two methods for assessing severity of Verticillium wilt of potato..
      ).
      For observational data, Cohen’s kappa coefficient of agreement of detection of cow at a bin between EO and DO was calculated using the FREQ procedure of SAS (SAS Institute Inc.). Kappa was considered almost perfect if ≥90, strong if <0.90 and ≥0.80, moderate if <0.80 and ≥0.60, and weak if <0.60 and ≥0.40. Sensitivity (likelihood that a cow present at the bin is detected by the monitoring system) and specificity (likelihood that a cow that is absent from the bin is detected as absent by the monitoring system) were calculated as in the study by
      • Trevethan R.
      Sensitivity, specificity, and predictive values: Foundations, pliabilities, and pitfalls in research and practice..
      and considered satisfactory if ≥80%. Significance was set at P ≤ 0.05 and tendencies declared if P > 0.05 and ≤0.10.

      RESULTS AND DISCUSSION

      When regressing the EO and DO results for kilograms of intake per individual cow visit, a linear fit was detected (P < 0.01; r2 = 0.95; Figure 3A): as DO intake increased, the EO output linearly increased. This outcome was noted when all visits were analyzed or only those with DO intake ≤0.1 kg (Figure 3A and 3B). Cattle intake varies according to supplement type, with free-choice mineral intake highly variable due to small amounts consumed (
      • Tait R.M.
      • Fisher L.J.
      Variability in individual animal’s intake of minerals offered free-choice to grazing ruminants..
      ;
      • McCarthy K.L.
      • Undi M.
      • Becker S.
      • Dahlen C.R.
      Utilizing an electronic feeder to measure individual mineral intake, feeding behavior, and growth performance of cow–calf pairs grazing native range..
      ). Hence, this analysis aimed to determine whether the electronic monitoring system was adequate for all types of supplement, including those consumed in small quantities. Accordingly, Pearson’s, Spearman’s rank, and Kendall correlations were detected (P < 0.01) between EO and DO observations within all individual cow visits (Figure 3A). The strength of the correlation based on Spearman’s rank and CCC varied based on intake and method of assessment. The DO and EO methods were very strongly correlated (r = 0.91, CCC = 0.95) when all visits were analyzed but decreased to strong (r = 0.78, CCC = 0.81) when DO intake was ≤0.1 kg. This difference may be due to rapid exchanges of animals at the bunks and human presence within the feeder leading to disruptions in normal feeding patterns. The CCC is used as a reproducibility index to account for accuracy and precision between methods (
      • Lin L.I.
      A concordance correlation coefficient to evaluate reproducibility..
      ), whereas based on the strength of CCC, the system described herein is expected to be effective in monitoring individual feed consumption. The Bland-Altman analysis for kilograms of intake per individual cow visit indicated D, SD, and LOA were estimated as 0.01, 0.03, and −0.05 to 0.06 (lower to upper LOA), respectively (Figure 4A). When intake was ≤0.1 kg, the analysis revealed D, SD, and LOA were estimated as 0.003, 0.02, and −0.04 to 0.05 (lower to upper LOA), respectively (Figure 4B). The estimated positive bias for both analyses indicates the DO method gives greater results compared with EO.
      Figure 3
      Figure 3Relationship between direct (DO) and electronic (EO) observations obtained per cow visit. Spearman’s rank (ρ) and Kendall (τ) rank correlations between methods are reported in panel A (all visits) and panel B (observations with DO intake ≤0.1 kg).
      Figure 4
      Figure 4Bland-Altman plots to analyze the agreement between direct (DO; M1) and electronic (EO; M2) observations obtained per cow visit. Bias between methods was estimated by the mean difference (D) and the variation around bias estimated as SD. A 95% limit of agreement (LOA) was defined as D ± 1.96SD (
      • Bland J.M.
      • Altman D.G.
      Measuring agreement in method comparison studies..
      ;
      • Yellareddygari S.K.R.
      • Gudmestad N.C.
      Bland-Altman comparison of two methods for assessing severity of Verticillium wilt of potato..
      ). For all visits (A), D, SD, and LOA were estimated as 0.01, 0.03, and −0.05 to 0.06 (lower to upper LOA), respectively. When DO intake was ≤0.1 kg (B), the D, SD, and LOA were estimated as 0.003, 0.02, and −0.04 to 0.05 (lower to upper LOA), respectively.
      When regressing the EO and DO results for total feed disappearance, a linear fit was detected (P < 0.01; r2 = 0.97; Figure 5): as DO total feed disappearance increased, the EO output linearly increased. Accordingly, Pearson’s, Spearman’s rank, and Kendall correlations were detected (P < 0.01) between EO and DO observations within all days, indicating that the monitoring system was sufficiently precise in the measure of feed consumption (r = 0.98, CCC = 0.99). The Bland-Altman analysis for total feed disappearance indicated D, SD, and LOA were estimated as 0.01, 0.60, and −1.17 to 1.19 (lower to upper LOA), respectively (Figure 6). Similar to the analysis of individual intake, the estimated positive bias for total feed disappearance indicates the DO method gives greater results compared with EO. Percent DM did not differ (P = 0.23) between offered and nonconsumed feed throughout the 21-d observation period (90.0 vs. 88.3%, respectively; SEM = 0.9%); however, numerical differences in total feed disappearance between EO and DO observations may be partially attributed to slight loss or gain of moisture within the 24-h observations periods.
      Figure 5
      Figure 5Relationship between direct (DO) and electronic (EO) observations for total feed disappearance over 21 d. Spearman’s rank (ρ) and Kendall (τ) rank correlations between methods are reported.
      Figure 6
      Figure 6Bland-Altman plot to analyze the agreement between direct (DO; M1) and electronic (EO; M2) observations for total feed disappearance over 21 d. Bias between methods was estimated by the mean difference (D) and the variation around bias estimated as SD. A 95% limit of agreement (LOA) was defined as D ± 1.96SD (
      • Bland J.M.
      • Altman D.G.
      Measuring agreement in method comparison studies..
      ;
      • Yellareddygari S.K.R.
      • Gudmestad N.C.
      Bland-Altman comparison of two methods for assessing severity of Verticillium wilt of potato..
      ). For total feed disappearance, D, SD, and LOA were estimated as 0.01, 0.60, and −1.17 to 1.19 (lower to upper LOA), respectively.
      Cohen’s kappa coefficient for agreement between DO and EO results for presence of cow at a bunk was considered strong (0.96). The sensitivity of EO in yielding true positive results was 97%, whereas specificity in yielding true negative results was 99%. Differences in EO versus DO may be due to rapid animal exchanges at the feeder (<1 s), whereas the EO system may overestimate animal permanence time at the bunk. Meal-based estimates of feeding behavior were not evaluated herein including meal duration, frequency, or size. Visits to the feed bunk are separated by nonfeeding intervals of variable lengths, which can be combined into meals once a relevant meal criterion is applied (
      • Mendes E.D.M.
      • Carstens G.E.
      • Tedeschi L.O.
      • Pinchak W.E.
      • Friend T.H.
      Validation of a system for monitoring feeding behavior in beef cattle..
      ). Further research is warranted to validate meal criterion used by the system, as it is beyond the scope of this study. Competition at the feeder may influence supplement intake patterns (
      • Reuter R.R.
      • Moffet C.A.
      • Horn G.W.
      • Zimmerman S.
      • Billars M.
      Technical Note: Daily variation in intake of a salt-limited supplement by grazing steers..
      ), and factors including size and type of cattle, desired supplement intake, and location of the feeder in the pasture will determine optimal stocking rate. Additional research is warranted to describe the influence of animal competition at the feeder and appropriate meal duration setting within the system.

      APPLICATIONS

      Collectively, results indicate the described system is a useful tool to provide reasonable estimates of supplement intake of pasture-based, group-housed cattle. Capturing individual animal intake data with this technology provides the ability to further examine the influence of supplemental feeding on a per-animal basis while the cattle remain in a realistic production setting. Facilities using the system should be aware of factors that may influence eating patterns from the unit including topography, size and type of cattle, and stocking rate.

      ACKNOWLEDGMENTS

      Financial support for this research was provided by Zinpro Corporation (Eden Prairie, MN). Appreciation is expressed to Robert Dobbs, Douglas Shaw, and Hunter McCool (Prairie Research Unit, Mississippi State University) for their assistance during this study.

      LITERATURE CITED

        • Akoglu H.
        User’s guide to correlation coefficients..
        https://doi.org/10.1016/j.tjem.2018.08.001
        30191186
        Turk. J. Emerg. Med. 2018; 18: 91-93
        • Bland J.M.
        • Altman D.G.
        Measuring agreement in method comparison studies..
        https://doi.org/10.1177/096228029900800204
        10501650
        Stat. Methods Med. Res. 1999; 8: 135-160
        • Capéraà P.
        • Genest C.
        Spearman’s ρ is larger than Kendall’s τ for positively dependent random variables..
        https://doi.org/10.1080/10485259308832551
        J. Nonparametr. Stat. 1993; 2: 183-194
        • Dafoe J.M.
        • Wyffels S.A.
        • Parsons C.T.
        • Carter B.H.
        • DelCurto T.
        • Boss D.L.
        Techniques to estimate colostrum quality and the effects of cow age and prepartum supplement intake levels on colostrum quality and serum IgG levels..
        https://doi.org/10.1093/tas/txaa121
        33381737
        Transl. Anim. Sci. 2020; 4: S137-S140
        • Greene L.W.
        Designing mineral supplementation of forage programs for beef cattle..
        https://doi.org/10.2527/jas2000.00218812007700ES0013x
        J. Anim. Sci. 2000; 77: 1
        • Lin L.I.
        A concordance correlation coefficient to evaluate reproducibility..
        https://doi.org/10.2307/2532051
        2720055
        Biometrics. 1989; 45: 255-268
        • McCarthy K.L.
        • Underdahl S.R.
        • Undi M.
        • Becker S.
        • Dahlen C.R.
        Utilizing an electronic feeder to measure mineral and energy supplement intake in beef heifers grazing native range..
        https://doi.org/10.1093/tas/txz065
        32704942
        Transl. Anim. Sci. 2019; 3: 1719-1723
        • McCarthy K.L.
        • Undi M.
        • Becker S.
        • Dahlen C.R.
        Utilizing an electronic feeder to measure individual mineral intake, feeding behavior, and growth performance of cow–calf pairs grazing native range..
        https://doi.org/10.1093/tas/txab007
        33659862
        Transl. Anim. Sci. 2021; 5
        • Mendes E.D.M.
        • Carstens G.E.
        • Tedeschi L.O.
        • Pinchak W.E.
        • Friend T.H.
        Validation of a system for monitoring feeding behavior in beef cattle..
        https://doi.org/10.2527/jas.2010-3489
        21512115
        J. Anim. Sci. 2011; 89: 2904-2910
        • Parsons I.L.
        • Johnson J.R.
        • Kayser W.C.
        • Tedeschi L.O.
        • Carstens G.E.
        Characterization of feeding behavior traits in steers with divergent residual feed intake consuming a high-concentrate diet..
        https://doi.org/10.1093/jas/skaa189
        J. Anim. Sci. 2020; 98
        • Reuter R.R.
        • Moffet C.A.
        • Horn G.W.
        • Zimmerman S.
        • Billars M.
        Technical Note: Daily variation in intake of a salt-limited supplement by grazing steers..
        https://doi.org/10.15232/pas.2016-01577
        Prof. Anim. Sci. 2017; 33: 372-377
        • Tait R.M.
        • Fisher L.J.
        Variability in individual animal’s intake of minerals offered free-choice to grazing ruminants..
        https://doi.org/10.1016/S0377-8401(96)01007-3
        Anim. Feed Sci. Technol. 1996; 62: 69-76
        • Trevethan R.
        Sensitivity, specificity, and predictive values: Foundations, pliabilities, and pitfalls in research and practice..
        https://doi.org/10.3389/fpubh.2017.00307
        29209603
        Front. Public Health. 2017; 5: 307
        • Williams G.D.
        • Beck M.R.
        • Thompson L.R.
        • Horn G.W.
        • Reuter R.R.
        Variability in supplement intake affects performance of beef steers grazing dormant tallgrass prairie..
        https://doi.org/10.15232/pas.2017-01720
        Prof. Anim. Sci. 2018; 34: 364-371
        • Wyffels S.A.
        • Dafoe J.M.
        • Parsons C.T.
        • Boss D.L.
        • DelCurto T.
        • Bowman J.G.P.
        The influence of age and environmental conditions on supplement intake by beef cattle winter grazing northern mixed-grass rangelands..
        10.1093/jas/skaa217
        32658282
        J. Anim. Sci. 2020; 98
        • Yellareddygari S.K.R.
        • Gudmestad N.C.
        Bland-Altman comparison of two methods for assessing severity of Verticillium wilt of potato..
        https://doi.org/10.1016/j.cropro.2017.07.019
        Crop Prot. 2017; 101: 68-75