Elsevier

Livestock Science

Volume 183, January 2016, Pages 1-3
Livestock Science

Short communication
Temperature as a predictor of fouling and diarrhea in slaughter pigs

https://doi.org/10.1016/j.livsci.2015.11.007Get rights and content

Highlights

  • We included data from 24 separate batches of slaughter pigs.

  • We modeled probability of pen fouling and diarrhea given pen temperatures.

  • We showed good prediction performances up to three days before the events.

  • We recommend including other types of observations in more complicated models.

Abstract

The PigIT Project aims at improving welfare and productivity of slaughter pigs by integration of various sensor systems for alarm purposes. Here we present an exploratory analysis to assess the predictive value of temperature sensor data with respect to pen fouling and diarrhea. We recorded the temperature at two locations in 8 pens between November 2013 and December 2014. A single logistic regression model was made to express the probability of either diarrhea or fouling per pen per day, and was reduced via backwards elimination. The predictive performances were evaluated by the area under the receiver operating characteristics curve (AUC). Indiscriminant prediction of either event reached an AUC of 0.80. Similar performances were seen when predicting each of the events on their own using the same model, with AUC values at 0.78 and 0.81 for diarrhea and fouling, respectively. Thus, temperature information seems to provide predictive value in relation to fouling and diarrhea. It would be meaningful to combine this information with other available data by using more advanced models to achieve an optimal predictive power.

Introduction

General and political interest in production and animal welfare is currently at an all time high in Denmark and other western countries. Denmark alone produces nearly 30 million pigs annually, distributed between just over 3000 farms (Landbrug og Fødevarer, 2014). Large herds are at increased risk of infectious diseases (Claes et al., 2002), and infectious disease will be more likely to persist for larger herds (Evans et al., 2010). It is further known that pig health and productivity is affected by a range of stress factors. One such factor is the temperature, where especially diurnal changes can cause a stress response, resulting in slower growth and higher feed intake (Lopez et al., 1991). Temperature is further known as a key factor for the onset of pen fouling (Aarnink et al., 2006), where the pigs will rest in the excretion area and in return excrete in the resting area.

Here, we wish to evaluate the potential of pen level temperature measurements for predicting pen level outbreaks of two undesired events in pig production, namely diarrhea and pen fouling. An effective prediction of such undesired events would allow the farmer to react proactively to a problem, thus improving the overall health and welfare of the herd, and in return secure a higher production for the farmer.

Section snippets

Data source

The data used for this study were collected for the PigIT Project1 in the finisher unit of a commercial Danish pig farm. Temperature data was collected continuously in 8 pens in two separate sections. Each pen contained 18 pigs at insertion, sorted by sex and size. Two neighboring pens (a double-pen) always shared feed and water supply. Data from two such double-pens were included from each of the two sections. Data were collected between November 20th 2013 and December 12th

Results and discussion

Table 1 summarizes the logistic regression model for predicting any of the two events, when reduced to including only significant or borderline significant variables. It is seen that high rates of both temperature increase and decrease, measured near the corridor, is associated with a higher risk of undesired events. This could indicate that the pigs are generally sensitive to sudden changes in temperature, which concievably could cause them to become stressed (Lopez et al., 1991).

Conclusion

It is shown that temperature data recorded at the pen level contains information, which is applicable to prediction of pen fouling and diarrhea up to 3 days before these events occur. The area under the receiver operating characteristics curves for indiscriminant, diarrhea, and fouling predictions are 0.80, 0.78 and 0.81, respectively. However, the logistic regression method used in this study is not likely to be the best method for practical purposes, as the achieved trade-off between

Future scope

In the PigIT project, we are currently collecting data on water and feed consumption, live weight and section humidity as well as pen level temperature, and there are plans for including automatic monitoring of pig activity. The information contained in this data could conceivably be combined with the temperature data presented in this paper, using a number of methods, such as multivariate dynamic linear modeling (West and Harrison, 1997), (naïve) Bayesian networks, artificial neural networks

Acknowledgments

This research was carried out with support from the Danish Council for Strategic Research (The PigIT Project, Grant number 11-116191). We further wish to thank the anonymous farmer, the Danish Pig Research Center and the technical staff at Aarhus University for installing and supervising the sensors and for taking care of the daily observations in the herd.

References (10)

There are more references available in the full text version of this article.

Cited by (15)

  • Automatic counting and positioning of slaughter pigs within the pen using a convolutional neural network and video images

    2021, Computers and Electronics in Agriculture
    Citation Excerpt :

    However, fouling is a multifaceted problem with multiple risk factors (Larsen et al., 2018b), which adds to the difficulties when trying to prevent it. Previous studies have tried to detect and/or predict undesired events, including pen fouling, based on sensor data on pen-level temperature and/or drinking behaviour, which were collected in the pens of grower/finisher pigs (Jensen and Kristensen, 2016; Jensen et al., 2017; Dominiak et al., 2019). The models developed in these studies were, however, not designed to distinguish between different types of events, and consequently only unspecified alarms could be raised.

  • Predicting pen fouling in fattening pigs from pig position

    2020, Livestock Science
    Citation Excerpt :

    This is because pen fouling is often a response to heat stress (Larsen et al., 2018). Heat stress will cause the pigs to seek towards the excretion area, which is typically a few degrees colder than the resting area (e.g. Randall et al. (1983) and Jensen and Kristensen (2016)). This effect, however, hardly explains the fact that in our study, the fourth batch of pigs, which grew from September to November 2016, experienced as many fouling events as the three previous batches combined.

  • Learning patterns from time-series data to discriminate predictions of tail-biting, fouling and diarrhoea in pigs

    2019, Computers and Electronics in Agriculture
    Citation Excerpt :

    One key observation is that the sensitivities were consistently lower than the specificities for all the behavioural predictions task in this study (see Table 5). This is consistent with the literature (Jensen et al., 2017; Jensen and Kristensen, 2016): for a sensitivity of 80%, specificities of approximately 60% were reported for fouling and diarrhoea predictions. Low specificities are indicative of alarms raised when no event occurred (False Positive), this is not necessarily indicative of poor model performance.

  • How do stocking density and straw provision affect fouling in conventionally housed slaughter pigs?

    2017, Livestock Science
    Citation Excerpt :

    Further, the results by Savary et al. (2009) state the importance of keeping the correct climate when trying to prevent fouling also when using other measures. Also, temperature measurements of the straw-bedded area might prove effective in predicting events of fouling, which has already been shown for temperature measurements at other locations of the pen (Jensen and Kristensen, 2016). Straw has been shown to have other positive effects for slaughter pigs such as reducing the risk of tail biting (Van de Weerd et al., 2005; Scollo et al., 2013; Larsen et al., 2017a) and gastric ulceration (Di Martino et al., 2013; Herskin et al., 2016; Jensen et al., 2017), improving growth rate (Pedersen et al., 2015) and increasing the expression of species-specific behaviour such as exploration (Jensen et al., 2015).

  • A multivariate dynamic linear model for early warnings of diarrhea and pen fouling in slaughter pigs

    2017, Computers and Electronics in Agriculture
    Citation Excerpt :

    This could be done, for example, via cameras above the pen, combined with software designed for activity analysis. It is worth pointing out that in our previous study (Jensen and Kristensen 2016), an AUC of 0.80 was achieved for indiscriminate warnings of diarrhea and pen fouling when only using temperature summaries and a −3/+1 prediction window. In this study, using the same prediction window, we achieved an AUC 0.75 when only temperature data were included.

View all citing articles on Scopus
View full text