| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* Department of Genetics and Biotechnology, Faculty of Agricultural Sciences, Aarhus University, Tjele DK 8830, Denmark
Sustainable Livestock Systems Group, Scottish Agricultural College, Dairy Research Centre, Midpark House, Bankend Road, Dumfries, DG1 4SZ, United Kingdom
1 Corresponding author: Peter.Lovendahl{at}agrsci.dk
 |
ABSTRACT |
|---|
 TOP ABSTRACT ACKNOWLEDGMENTSREFERENCES |
|---|
Key Words: activity meter • heritability • estrus • fertility
The need for genetic improvement of dairy cow fertility has been emphasized lately in several reports as reproductive performance has declined in the recent decades (Miglior et al., 2005; Weigel, 2006). This is partly a consequence of the correlated response between reproductive efficiency and selection for higher milk yield (Veerkamp et al., 2000; Royal et al., 2002). However, traditional fertility traits (e.g., nonreturn rate, calving interval) have low heritabilities, which delay genetic progress. The period from calving to commencement of luteal activity determined from milk progesterone has been proposed as a genetic indicator of fertility (Royal et al., 2002; Petersson et al., 2007). However, electronic activity tags used for estrus detection represent another source of fertility information and deserve further study as initially suggested by Løvendahl and Chagunda (2006).
Detection of estrus in dairy cows and heifers using electronic pedometers or activity tags is based on distinct behavioral changes associated with estrus. Restlessness and general physical activity increase markedly during estrus (Farris, 1954; Van Eerdenburg et al., 2002). Other changes in behavior are "standing to be mounted" and "mounting." Pedometers and activity tags are designed to identify the restlessness and elevated physical activity of cows. Estrus detection based on activity measurements involves dedicated software supplied as part of the farm-management systems (e.g., AlproWin, DeLaval, Tumba, Sweden) together with the activity tags as a complete package. Algorithms for detection of estrus are part of the software and details are usually not disclosed to end users. Theoretical work on detection algorithms (De Mol and Woldt, 2001; Firk et al., 2002; Roelofs et al., 2005a,b; Løvendahl and Chagunda, 2009) show that detection efficiency varies greatly (50 to 100%) depending on the complexity of the method and the chosen criteria of success (Firk et al., 2002; Roelofs et al., 2005a).
Although the use of electronic estrus detectors is already widespread, information on their possible use in genetic evaluations for improving female fertility is scarce (Løvendahl and Chagunda, 2006). To implement electronic estrus detection records in breeding schemes, their predictive value in terms of estimated genetic parameters needs to be assessed. The current study, which is based on experimental herd data, aims at presenting genetic parameters for days to first estrus and other traits describing intensity and duration of estrus based on activity tag records.
This experiment was conducted in the experimental herd at the Danish Cattle Research Centre (Tjele, Denmark) between July 2001 and December 2008. The herd included 3 different breeds: Jersey, Holstein, and Red Danes. All cows and heifers had their pedigrees traced back at least 3 generations in the Nordic Cattle Database (NordicEBV, Skejby, Denmark) to establish a relationship matrix for the genetic analysis. The cow herd was divided into 3 groups, each assigned to an automated milking system (VMS, DeLaval, Tumba, Sweden). At any time each of the first 2 cow groups included approximately 55 cows, half of them purebred Holstein, the other half purebred Red Dane; the last group included 42 Jersey cows. The Red Danes in the 2 mixed groups were replaced by Holsteins during the fall of 2007. Culled cows were replaced with young stock to maintain group size over the experimental period. Cows were housed all year and fed a TMR ad libitum with supplementary concentrates during milking in the automated milking system. Details of the phenotypic aspects of the study were reported by Løvendahl and Chagunda (2009) and are briefly described below.
The experiment included both maiden heifers and cows in first to fourth parity. Maiden heifers (parity 0) were kept in a separate building in pens holding up to 12 heifers at any time. Heifers were fed a TMR ad libitum. Recording of activity data for the study was initiated in July 2001 (heifers in February 2004) and ended in December 2008. The activity data were obtained from heifers between 365 and 545 d of age and from lactating cows between 5 and 155 DIM.
The reproduction strategy at the Danish Cattle Research Centre aimed at initiating AI at first estrus after 15 mo of age in Holstein and Red Dane heifers and at 13 mo in Jersey heifers. The voluntary waiting period for AI after parturition was 35 d. Artificial insemination was repeated until established pregnancy or culling. Artificial induction of estrus was used if heat had not been detected before 120 DIM. Animals were equipped with electronic activity tags fitted on neckbands (Alpro version 6.60, DeLaval). Data, as counts per hour, were stored and used for the calculation of estrus traits applying an algorithm based on deviations from exponentially smoothed hourly counts for each individual cow (Løvendahl and Chagunda, 2009). Briefly, activity data for every hour, every day for each cow were ln transformed before being exponentially smoothed to provide 24 hourly baseline values. Deviations from reference values were again smoothed exponentially. An episode of high activity was initiated when 3 consecutive values exceeded a set threshold (
) and ended when another 3 consecutive values fell below the threshold. To protect against too many short episodes, a new episode was only allowed to start after the smoothed deviation value had been below zero for 3 consecutive hours. The duration of the pulse was counted as hours from start to end. The intensity or strength of the episode was measured as the mean of the 3 highest deviation values during the episode. The episodes were numbered starting from 1 at each new parity. The number of days from calving to first detected high-activity episode was seen as a measure of days to first detectable estrus (DFE). Over the following episodes, regularity was measured as days from start of the previous episode to start of the present episode. Days to first episode and regularity were ln transformed before further analysis. Direct activity counts were considered unsuitable for further analysis as activity tags were not uniformly calibrated (data not shown). The effect of threshold settings on detection and error rates was evaluated using a learning data set from 461 successful AI records giving either a calf or a confirmed pregnancy. Detection rate was the proportion of high-activity periods that occurred when there was an estrus that resulted in a successful insemination. Error rate was the proportion of high-activity periods that were not associated with any estrus. On the complete data set, threshold settings also affected genetic parameters for DFE. The optimal threshold setting for genetic analysis was assumed to give the highest heritability estimate for DFE. From this step, episode characteristics were subjected to further analysis as repeated records from the same animal using up to 5 episodes per animal per parity.
A single-trait mixed model was used to assess effects of systematic factors and co-variance components for individual animal (permanent environment) and genetic effect on traits. The model [1] also included a random year-season term:
 |
Variation in the trait variables y were modeled as dependent on systematic effects a given in the design matrix X including breed (Red Dane, Holstein, Jersey), age group (heifers, cows), parity within the cow age group (1, 2, 3, 4), a breed by age group interaction, and episode number (1, ..., 5) and episode number by age group interaction. Random effects from animals across and within parities were in b1 with incidence matrix Z1 with genetic effects in b2 with relationship matrix Z2. Random effects of year-months (used as "short seasons") were in s with incidence matrix W and random residuals in e assumed to be normally distributed. Additive genetic variance 
permanent animal variance across parities 
and within parities 
were obtained together with year-season variance 
considered as noise, and residual variance 
. The model was also applied to data restricted to the first occurring activity episode in each parity, in which case the within-parity variance component was omitted, and with a further restriction to first parity alone. Estimates of variance components were obtained using average information REML in the DMU package (Madsen and Jensen, 2005).
Holstein, Red Dane, and Jersey heifers all had their first activity episode at an average age of 14 mo (Table 1). In cows, the interval from calving to first detected estrus episode was 39, 44, and 45 d for Red Dane, Holstein, and Jersey, respectively, when considered as simple group means (Table 1). When considered across breeds, activity episodes lasted 8.12 h in cows and 9.24 h in heifers with a peak activity of 1.03 ln units in both age groups (Table 1). The regularity of estrus episodes was shorter and less variable in heifers than in cows when expressed in days between successive episodes, indicating that heifers have more short intervals possibly coming from non-estrus-related episodes (Table 1).
|
|
|
|
The number of days from calving to first estrus showed a heritability of 0.12 to 0.18. The repeatability for the interval from calving to first estrus was 0.18, showing that this trait is predominantly determined by the genetic component. Previous experimental studies have based detection of commencement of luteal activity on progesterone measurements in milk and obtained moderate heritability estimates. Royal et al. (2002) obtained an estimate of h2 = 0.17 in British Holstein-Friesian cows when 3 milk samples were obtained per week, and Petersson et al. (2007) obtained a higher estimate (h2 = 0.30) by expressing the trait as percentage of cows with luteal activity, using the same data set. In comparison, estimates of heritability for days to first AI or successful AI based on field records are low, around h2 = 0.03 (Roxström et al., 2001a, b; Philipsson and Lindhe, 2003; Wall et al., 2003), although higher estimates (h2 = 0.07 – 0.10) were reported by Jamrozik et al. (2005). Characteristics of estrus behavior had low heritability for duration at around 0.02 to 0.08, and for strength between 0.04 and 0.06. Few other genetic studies have considered these traits. Roxström et al. (2001a), who obtained low heritability estimates (h2 ~0.02) of heat intensity from subjectively scored field data, confirm the findings of the present study.
Estrus was assumed when the deviation between actual and baseline activity exceeded a given threshold for 3 consecutive hours. The baseline activity profile was calculated by exponential smoothing of activity recorded in each of 24 hourly periods. The deviation from the baseline was thereby adjusted for circadian variation and so expressing relative changes. Similar approaches have been used previously (At-Taras and Spahr, 2001; Roelofs et al., 2005a). Further exponential smoothing was effective in limiting the random noise on single hourly deviations. More protection against false positives was enforced by requiring the smoothed deviations to exceed the threshold for at least 3 consecutive hours. This restriction may in turn also have removed atypical and short estrus episodes. Such atypical episodes were found more frequently at spontaneous multiple ovulations (more than one dominant follicle producing oocytes) and in cows with low BCS or very high yield (Lopez et al., 2005). Thus, optimal detection algorithm settings will depend on whether the aim is detection of "normal" estrus episodes or both the normal and the atypical episodes.
The present study recorded activity in 1-h periods that allowed estimation of estrus duration, which lasted, on average, 8.1 h in cows and 9.2 h in heifers, respectively. In a comparable housing system, the duration of estrus was somewhat longer, 11.8 h when recorded by visual observation at 3 h intervals, and almost similar (10.0 h) when recorded by pedometers in 2-h periods (Roelofs et al., 2005b). The intensity or strength of estrous behavior was determined as a relative deviation from a moving average for the individual cow in ln units, which in linear scale expresses folds of increase (Table 3). The measured units are not fully comparable across studies because units are defined by the manufacturers and because tags are attached either to a neckband or to the leg (Maatje et al., 1997). A further complication to units is that activity tags, even from the same manufacturer, may not be calibrated. However, log-transformation is effective in stabilizing the variation and the deviations from baseline express the relative increase in percentage or fold. Indeed, for the chosen threshold of 0.75 the mean strength was equivalent to a 3-fold increase, which is similar to findings in other studies (Firk et al., 2002).
Red Dane cows had their first high-activity (estrus) episode at 39 DIM, 5 d before Holstein and 6 d before Jersey cows. In a comparative study, Brown Swiss had a calving interval 22 d shorter than that of Holsteins, and similarly Jerseys had on average a calving interval 14 d shorter than Holsteins (Garcia-Peniche et al., 2005). In the current study, breeds differed both in BCS and milk yield (Friggens et al., 2007), and differences between breeds in days to first estrus could partly be attributed to differences in BCS (Løvendahl and Chagunda, 2009). This is in agreement with previously reported negative relationships between fertility, BCS, and yield both at the genetic and phenotypic levels (Veerkamp et al., 2000; Wassmuth et al., 2000; Pryce et al., 2002; Royal et al., 2002).
The results of the present study have shown that automated heat detection using electronic activity tags may not only be of value in estrus detection, but could also be helpful in recording of fertility traits for genetic evaluations. Higher heritability was obtained compared with field data from AI services. A key point in a genetic selection strategy is to obtain large volumes of inexpensive and unbiased data allowing estimation of reliable breeding values. By using electronic activity tags higher heritability for days to first estrus could be obtained as shown in the present study. Other equipment giving unbiased information about fertility traits could have been used in similar ways, such as on-farm progesterone measurements (Friggens and Chagunda, 2005; Løvendahl et al., 2009). In conclusion, the results suggest that activity tag-based fertility traits may provide valuable information to dairy herd improvement programs.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTACKNOWLEDGMENTSREFERENCES |
|---|
Received for publication September 22, 2008. Accepted for publication May 26, 2009.
|
REFERENCES |
|---|
|
TOPABSTRACTACKNOWLEDGMENTSREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
