Generation and emission of particulate matter (PM) from dairy farming have a potential effect on the health and welfare of the animals, farm workers and even the neighbors. Monitoring accuracy of the PM concentration depends much on the number and location of sampling points as well as the sampling interval (SI). Most PM studies used intermittent sampling methods, such as sampling the concentration for a couple of days in a season or in several seasons, which were unable to accurately reflect the actual PM concentration level and variation inside the intensive dairy building. To determine the reasonable SI of PM sensors, this study developed an Internet of Things (IoT)-based monitoring system for PM concentration in an intensive naturally ventilated dairy barn, in which a 17-point continuous concentration monitoring of PM less than 2.5 μm in aerodynamic diameter (PM_2.5) and the total suspended particulate (TSP) was carried out in autumn and winter, and its 5-minute mean values were regarded as relatively true values (RTV). Using error analysis, the daily averaged PM concentration with 30 min and 1, 2, 3, 6, 12 h SI in autumn and winter and the hourly mean PM concentration with 10, 15, 20, 30 min and 1 h SI during the day (05:00-23:00), night (23:00-05:00) and daily management periods (05:00-07:00, 13:00-15:00, 21:00-23:00) were first computed, respectively; then their relative errors ( E_r ) with RTV were counted within 5% and 10% range; and finally, the maximum accepted SI for daily and hourly mean PM concentration measurements were determined based on acceptance criteria in bioanalytical method (66.7%).The results showed that within 5%, when the SI for TSP concentration were set within 2 h (in autumn) and 1h (in winter), and they were within 3 h (in autumn) and 1 h (in winter) for PM_2.5 measuring, respectively. It can accurately obtain the daily average PM concentration of the naturally ventilated dairy barn in autumn and winter. When the SI were at 20 min (in autumn) and 15 min (in winter) in daytime, and 30 min (in autumn) and 15 min (in winter) in nighttime for the TSP measurements, and 30 min for PM_2.5 daytime and nighttime (in autumn and winter), an accurate monitoring could be obtained on hourly mean PM concentration and its fluctuations. When the sampling interval for TSP was 10 min, and the interval for PM_2.5 was on 20 min in autumn and winter, respectively, the measurement data can reflect the impact of daily management on the PM concentration inside the barn. The findings of this study can be applied as a standardized procedure to continuously track the PM concentration in an intensive naturally ventilated dairy barn.

Particulate matter (PM) concentration can be real-time monitored to assess the environmental risks and make emission reduction measures in dairy barns. However, a great challenge is remained on arranging as few sampling points as possible to accurately monitor the PM concentration in an intensive barn, particularly with the rapid development of large dairy barns in China. This study aims to design an appropriate monitoring layout with the optimal PM sampling number and location. An online monitoring system was built to continuously detect the PM concentration inside a naturally ventilated dairy barn using the Internet of Things (IoT) and sensing technologies. A total of 17 sampling points were set inside three relatively independent sections of the investigated barn. The total suspended particle (TSP) and PM_2.5 concentrations were monitored in real time for the six months during the field test. The uniformity of PM concentration was evaluated on the spatial distribution of PM concentration among the three sections and the difference in sampling heights. The systematical clustering and error analysis were also performed on the sampling of PM concentration. The optimal sampling was determined to compare the measurement with the six regular ones under three environmental controls (namely, EC1: Fans and spraying, EC2: Fans, EC3: No fans and no spraying). The average PM concentration from the 17 sampling points was treated as the true value during data analysis. Results showed that no significant difference was found for the TSP and PM_2.5 concentration among the three measuring sections of the barn (P>0.05). TSP concentration sampled at the height of 9.0 m was significantly lower than that at the 1.5 and 2.5 m heights (P<0.05). There was no statistical difference in the PM_2.5 concentration among different sampling heights (P>0.05). The concentrations of TSP and PM_2.5 sampled at the height of 2.5 m were uniformly distributed among three sections of the barn. The sampling point setting down the ridge opening (approximately 9.0 m above the floor surface) was necessary for the TSP concentration monitoring. In TSP and PM_2.5 concentrations, the sum of absolute errors between the true values and the optimized sampling under three ECs were 6.4%-22.6% and 4.7%-14.2%, respectively, indicating all smaller than those of six regular monitoring (P<0.05). Generally, the number of PM sampling points was appropriately reduced to consider the monitoring costs and practical operability. The final PM monitoring was determined with the optimized sampling number and location in a naturally ventilated dairy barn. Six PM sampling points were set inside a dairy barn: one sampling point 1.0−2.0 m down the ridge openings in the central of the barn, two sampling points at a 2.5 m height above the cubicles, and three sampling points distributed at milking alley, feed delivery alley and manure alley at a 2.5 m height, respectively. Among them, the three sampling points down the ridge opening and above the cubicles should be diagonally arranged along the barn. The final PM sampling can be expected to achieve both the accuracy and economy of PM real-time monitoring for a naturally ventilated dairy barn.