It is not new to farmers that certain animal diseases can be triggered by too cold or too warm temperature in the barn or that poor water availability and humidity can predispose animals to illness. But so far there was a lack of tangible data on these relationships between farm environment, health and productivity. PROHEALTH aims at closing this gap by compiling integrated data-sets of measured farm environment parameters as well as documented health and productivity parameters for several different production diseases. Here we present first results from such a PROHEALTH study that determines the extent to which farm environmental factors contribute to the increase in respiratory disease in pigs at an intensive farm level.
PROHEALTH deployed a novel real time measurement system in the farms (Figure 1, see also the description on the PROHEALTH website).
The first main component is a network of sensors that record key environmental indicators (temperature, humidity, CO2 and water intake) and transmit the data to a cloud-based server via wireless connection. The second and complementary component is a system to collect data about clinical diseases, use of medication and animal performance. This was implemented in form of a digital pen, paired with a smartphone and digitised forms. Both parts of the system have been carefully designed to be operated by farm staff in a robust and user-friendly way within a farm environment. With the help of this integrated system, data were collected within PROHEALTH from a variety of systems and enterprises (Table 1 and 2). While health and performance data were recorded on a daily basis, environmental parameters were measured every 10 seconds and stored as hourly averages.
In order to represent environmental measurements alongside daily health data, hourly sensor data were aggregated in 5 different ways that could summarise different aspects of each day (Table 3). is generates a large number of features. To determine which features to use in the modelling stage, a genetic algorithm was used to select the feature subset which produced the best statistical model.
Here we concentrate on analyses carried out concerning specifically grow-finish pigs, which explored the relationship between environmental conditions and respiratory disease prevalence and mortality. Respiratory disease prevalence was found to be higher in spring than any other season (Figure 2), although it was consistently less than 1%.
There were also two clear peaks when respiratory disease prevalence was grouped by age, which occurred at 90-100 days of age and 160-170 days of age (Figure 3).
Analysis of statistical models indicated that CO2 concentration does not play as large of a role as humidity and temperature in affecting respiratory disease prevalence. A large number of the features determined to be linked to respiratory disease prevalence by the genetic algorithm were related to humidity; the most significant being the number of hours spent above 80% humidity 6 days prior to seeing a change in respiratory disease prevalence. is means that as the number of hours spent above 80% humidity increases, the chance of seeing an increase in respiratory disease prevalence in 6 days also increases. In terms of temperature, the most significant effect was the number of hours spent above 22°C. An increase in this would cause an increase in the chances of seeing a rise in respiratory disease prevalence within one day.
In the analysis focusing on mortality rate, the number of hours spent above 2800 ppm CO2 was an important measurement. An increase in this would increase the chances of a rise in mortality rate within one day. Also, exploring the effects of humidity on mortality rate, the number of hours spent below 33% was seen to increase the chances of seeing an increase in mortality in two days time. is model was lacking many effects relating to temperature, suggesting that it may not play as large of a role in pig mortality as humidity and CO2. However, this does not imply that temperature is not an important factor.
The results from this study provide an initial insight into the risk factors regarding respiratory disease prevalence and mortality in grow-finish pigs. At least half of the environmental features selected by the genetic algorithm were related to humidity, suggesting that it plays a large role in respiratory disease prevalence and mortality in grown-finish pigs in comparison to the other environmental measures. While CO2 concentration was included in all of the models, it does not appear to have a large impact on disease levels relative to the other measures.