Journal of Biosystems Engineering. September 2018. 247-254



  • Introduction

  • Materials and Methods

  •   Case study

  •   Data collection

  •   Data analysis

  •   Spatial analysis

  • Results and Discussion

  •   Patterns of vehicle movement using point density

  •   Patterns of vehicle movement determined by Euclidean distance

  •   Area and extent

  •   Comparison of point-density and Euclidean-distance maps

  • Conclusions

  • Conflict of Interest


In the past few years, although livestock-farming productivity has increased, it has been confronted with economic and health issues; infectious diseases can damage livestock as well as humans living on livestock farms (Liverani et al., 2013). To minimize the damage from infectious livestock diseases at the early stages, preventive measures are needed. Airborne diseases spread from livestock farms, manure, and transportation of livestock. Certain diseases such as foot-and-mouth disease (FMD) and influenza can spread easily from these sources. FMD can be spread through direct contact with an infected animal or through indirect contact, such as through objects, vehicles, people, or airborne spreading (Weber et al., 2008).

Since 2000, there have been four outbreaks of FMD in Korea. In 2010 and 2011, because of FMD, a total of 3.35 million pigs were slaughtered to prevent the virus from spreading (Seo et al., 2015). Every year, there are economic losses and risk to public health because of the FMD virus. The various transmission routes of this disease have not yet been blocked owing to lack of monitoring and preventive measures. Multiple factors should be considered when preventative measures are taken, and forecasting is done regarding FMD. However, there are a substantial number of difficulties with field experiments, including lack of basic information on virus spread. Although a counterplan for direct transmission has been prepared in the past, it is difficult to execute a plan for blocking indirect and airborne transmission, which can arise from a number of sources. One of these sources is transmission from moving livestock vehicles. Airborne diseases not only spread from livestock farms but can also spread during the process of transporting livestock and manure. Although on-farm evaluations and control of diseases have been performed in the past, the evaluation and control of disease in the process of transporting livestock and manure still needs to be thoroughly investigated. This study used a variety of methods to address this issue.

Jeju Island (South Korea) has many livestock farms and is therefore subject to the threat of disease spread. A proper forecasting plan and preventive measures are needed to reduce damage before a disease outbreak occurs. Even though livestock on farms is regularly vaccinated on Jeju Island, livestock diseases continue to persist, especially when the livestock is being transported. During disease occurrence, paths on which vehicles move regularly can become dangerous to the local people. FMD can spread over several kilometers by air depending on the air velocity (Gloster et al., 2007). Preventive measures that have been used to reduce FMD spread include the use of preventive vaccinations within 3 km of infected farms (Martínez-López, Perez & Sánchez-Vizcaíno, 2010). The pattern of livestock-moving vehicles can become a serious threat; therefore, it is necessary to predict the most common route or pattern for future application of disinfectant measures. This method of prediction and spraying can be a preventive measure for controlling airborne diseases.

In the past 10 years, the geographic information system (GIS) has raised great expectations for addressing natural disasters, forecasting disease occurrence, and predicting many events. There are some tools that help in predicting new data from old data or in making models from old data. ArcGIS software can be used to give results in statistical probabilities by showing maps. In this study, we analyzed global positioning system (GPS) data from livestock vehicles on Jeju Island. GPS analysis of tracing data is a very active research area and a numerous algorithms have been developed for the clustering of trajectories (Necula, 2015); moreover, various definitions of distance functions between traces have been introduced by Pelekis et al. (2011). Sun and Zhou (2005) used cluster analysis to segment speed–density data and determine the regime boundaries for typical (two-regime and three-regime) speed–density models. GPS collars allowed Bailey et al. (2001) to determine alterations in cattle- grazing patterns as a result of dehydrated molasses supplement being introduced to a foothills rangeland. Although the use of GPS trackers for traffic analysis is increasing, activity-based analysis using GPS equipment for data collection has had major issues (Obuhuma and Moturi, 2012). Much research has focused on data from wearable GPS recorders because of easy logging activity and validation with users (Kochan et al., 2006). Another question arises about the accuracy of GPS devices. Agouridis et al. (2004) determined the accuracy of the horizontal-position data collected by GPS collars under both static and dynamic conditions. Both static and dynamic tests were performed to better assess the accuracies, capabilities, and limitations of using GPS collars to track animal movement in a grazed watershed. The dynamic performance of GPS receivers also needs to be considered. Cole (2004) designed and constructed an apparatus to assess the dynamic performance of GPS receivers for application to the field of precision agriculture. There are different kinds of GPS devices; some of them are of low quality and exhibit large errors. The use of GPS for tracking vehicles and determining vehicles’ dynamic properties was evaluated by Haugen et al. (2000) based on the accuracy of several GPS receivers as well as the accuracy of the calculated dynamic vehicle properties. GPS (ASCEN GPS742) also exhibited less error; therefore, we used it in this study.

Our approach differs from those used in previous works. We analyzed previous years’ GPS data for moving vehicles and used them for predictive purposes. The objective of this study was to predict common moving patterns of livestock vehicles using GPS and GIS.

Materials and Methods

Case study

Jeju Island, which is located in the south of South Korea, was selected for this study because of its many livestock farms, making the airborne spread of disease very possible. Also, it is easier to control disease on Jeju Island than in other parts of South Korea owing to its relative isolation.

Vehicles on Jeju Island transfer livestock and manure using different paths. To predict the common path or pattern of moving livestock vehicles, this study was conducted between March and August 2012 and between March and August 2013.

Data collection

Data-measuring GPS devices (ASCEN GPS742, Ascen Korea Technology, Seoul Korea) were placed in 30 livestock vehicles. The size of the portable GPS device was 70 × 44 × 20 mm3, and it also had a data-logger function with a capability of 250,000 points. A 30-s time interval was set on all GPS devices. ON/OFF settings were selected according to the respective state of the vehicle’s engine. Vehicle drivers were taught to carry the GPS devices and also to recharge their batteries. After the completion of this task, the data were transferred to a personal computer.

Data analysis

PS data include vehicle coordinates, time, and speed, thereby allowing vehicle movement patterns to be recreated. Also, the coordinates of livestock farms and manure-keeping sites were taken for analysis. In 2012, there were 320 livestock farms and 148 manure-keeping sites. In 2013, a slight difference was observed as livestock farms decreased to 306 and manure-keeping sites increased to 166.

Owing to the large amounts of GPS data, an analysis was performed by selecting the data from three vehicles for 2012 and 2013. Data were transferred from Microsoft Excel to ArcGIS 10.1 to predict the movement pattern of livestock vehicles. The road data for Jeju Island were used as a base map, and a projected coordinate system (Korea_ 2000_Korea_Central_Belt) was used in all analyses. The roads on Jeju Island were divided into three types: primary, secondary, and tertiary. This made it easier to analyze the types of roads used by different vehicles at different times.

Spatial analysis

To conduct spatial analyses, the point-density and Euclidean-distance ( techniques were used in ArcGIS (10.1, ESRI, New York, USA). Spatial analysis was performed with maps prepared using these techniques. Figure 1 shows the process followed in GIS analysis for calculating predicted and actual maps. The GPS coordinates of vehicle moving patterns were divided into time fields, and analysis of point density and Euclidean distance was performed on monthly data. Five point-density maps and five Euclidean distances (with each map covering five months) were prepared. Maps were categorized as high-, middle-, and low-density maps, with the high-density maps showing a common moving path or route that is used most of the time by vehicles. Five monthly maps were added by a raster calculator using the map-algebra tool, and the final map was taken for further analysis. A raster calculator was designed to execute a single-line algebraic expression using multiple tools and operators. The common patterns for 2012 were considered as predicted common patterns for 2013 and then these predictions were compared to the actual common patterns for 2013.
Figure 1.

Flowchart showing the process followed in GIS analysis for calculating predicted and actual maps.

Statistical analysis was performed using Graph Pad Prism 5 software, and the mean was calculated from the data.

Results and Discussion

Patterns of vehicle movement using point density

Figure 2 shows the analysis of each vehicle for the 2012 and 2013 data. In 2012, the high-density lengths for vehicles 9121, 9772, and 9115 were 22, 31, and 52 km, respectively. In 2013, the high-density lengths for these vehicles were 18, 41, and 46 km, respectively, as shown in Figure 3. High density shows us how frequently the vehicles move on the same roads regularly. There is a clearly elevated risk of spreading disease around these roads.
Figure 2.

Point-density maps of three vehicles for predicted and actual common patterns. The predicted pattern for 2013 is calculated from 2012 data.
Figure 3.

Length of predicted and actual common patterns of point-density maps. PR: primary roads.

The 2012 maps are described as “predicted common patterns,” and the 2013 maps are described as “actual common patterns.” Point density shows us which roads were used most of the time throughout the year. Common patterns can be predicted for the future by selecting high-density patterns. Some slight differences can be observed in the maps mentioned above; one map occupied lesser road area than the other owing to changes in the patterns of vehicles. To compare both patterns, the high-density lengths were calculated.

The primary roads on Jeju Island are the main roads, which are subject to considerable amounts of construction and human activity; this means that the first priority should be assigned to these areas when evaluating and executing a plan for the control of airborne disease. Vehicles also use these primary roads to transport manure and livestock.

The differences between the predicted and actual lengths were calculated by using the following formula (

percent difference=measured1-measured2measured1+measured22×100%    (1)

Where “measured1” is the predicted pattern length for 2013 (taken from 2012 data), and “measured2” is the actual pattern length for 2013 data.

The percentage differences (Eq. 1) in the lengths of the predicted and actual common patterns of point density for vehicles 9121, 9772, and 9115 were 20%, −27%, and −12%, respectively. Positive values indicate an increase in the length of the predicted common pattern, and negative values indicate a decrease. This increase or decrease may be due to changes in the numbers of livestock farms and manure-keeping sites. The number of manure-keeping sites increased in 2013, so this may have increased the vehicle paths. In the results we can see differences in paths of vehicles between 2012 and 2013.

In our study, the lengths of the primary roads used in the common patterns for 2012 by vehicles 9121, 9772, and 9115 were 11, 15, and 33 km, respectively. In 2013, these values were 9, 22, and 28 km, respectively. The area around these primary roads should be considered sensitive areas whenever disease occurs and should be given priority for disease prevention measures.

The percent differences between the lengths of the predicted and actual primary roads used in common patterns, as obtained using point density, for vehicles 9121, 9772, and 9115 were 20%, −38%, and 16%, respectively.

Patterns of vehicle movement determined by Euclidean distance

The percent different in the lengths between the predicted and actual common patterns of Euclidean distance for vehicles 9121, 9772, and 9115 were 36%, −29%, and 22%, respectively. In 2012, the lengths of the primary roads used by vehicles 9121, 9772, and 9115 were 17, 25, and 20 km, respectively. In 2013, these values were 9, 15, and 22 km, respectively.

Using ArcGIS, we also obtained five different Euclidean- distance monthly maps; all maps were then added by using the map-algebra tool. The final map was considered as a predicted common pattern for the year 2013 and then compared with the actual common patterns for that year (Figure 4). In this case, the Euclidean distance of the GPS coordinates yielded a map with differently colored zones, thereby indicating that, if a disease occurs, the blue zones are more threatened because they represent the most common patterns.
Figure 4.

Euclidean distance maps of three vehicles for predicted and actual common patterns. The predicted pattern for 2013 is calculated from 2012 data.

The blue zones in the predicted and actual pattern maps were measured and compared again, and the primary roads occupied by the blue zone were measured, as shown in Figure 5. The percent differences between the predicted and actual lengths of the primary roads used for vehicles 9121, 9772, and 9115, as measured by Euclidean distance, were 61.5%, 50%, and −9.5%, respectively. A positive sign in percentage difference indicates that the length of the predicted pattern is greater than that of the actual pattern and a negative sign indicates that the length of actual pattern is greater than that of the predicted pattern.
Figure 5.

Length of predicted and actual common patterns of Euclidean distance maps. PR: primary roads.

Area and extent

The extents of the point-density and Euclidean- distance maps were similar; however, the areas of the common patterns differed between the predicted and actual maps. As given in Table 1, the areas occupied by the predicted common patterns were 41, 81, and 89 km2 for vehicles 9121, 9772, and 9115, respectively; the areas occupied by actual common patterns were 19, 195, and 90 km2 for vehicles 9121, 9772, and 9115, respectively; and the areas occupied by both patterns for vehicles 9121, 9772, and 9115 were 19, 79, and 88 km2, respectively. Therefore, these areas should receive first preference for any defensive strategy against airborne disease.

Table 1. Area and extent of the predicted and actual common-pattern maps

Predicted pattern extent* Area (km2)Actual pattern extent*Area (km2)
Vehicle 9121126.392533 W126.489196 E41126.408543 W126.464789 E19
33.306275 N33.264188 S33.306282 N33.273571 S
Vehicle 9772126.620219 W126.709233 E81126.577098 W126.772853 E195
33.534259 N33.442959 S33.531804 N33.433198 S
Vehicle 9115126.43405 W126.585722 E89126.434046 W126.589575 E90
33.30986 N33.252282 S 33.309795 N33.253647 S
*Extent of area of point-density maps and Euclidean-distance maps.

Comparison of point-density and Euclidean-distance maps

By using the point-density tool, the average similarity between the predicted and actual common patterns for the three vehicles was determined to be 80%, which means that the difference was 20%. However, by using the Euclidean-distance tool, the average similarity between the predicted and actual patterns for the three vehicles was determined to be 72%, which means the difference was 28%. Consequently, the point-density tool may be more useful for predicting common patterns in the future owing to the smaller difference found between the predicted and actual common patterns.


From this analysis, we can determine the common patterns of livestock vehicles using previous year’s data. In particular, the point-density tool exhibits a smaller difference between the predicted and actual common patterns compared with the Euclidean-distance tool. The difference in the patterns of the vehicles is likely due to changes in the numbers of livestock farms and manure- keeping sites. When the manure-keeping sites and livestock farms were the same, the common patterns of 2012 and 2013 were similar; however, differences arose when the numbers of these sites were changed. In this study, a model for prediction was not established owing to the lack of other variables. In the future, to obtain more accurate results and to accurately predict the patterns of vehicle movement, more dependent and independent variables will be required to make a suitable model for prediction.

Conflict of Interest

The authors have no conflicting financial or other interests.


This research was supported by the Korea Institute of Planning and Evaluation for Technology in Food, Agriculture, Forestry and Fisheries (IPET) through the Research Centre Support Program (Project No. 717001-7) and by the Advanced Production Technology Development Program (315006-02-2-WT011), Ministry of Agriculture, Food and Rural affairs (MAFRA).



Agouridis, C. T., T. S. Stombaugh, S. R. Workman, B. K. Koostra, D. R. Edwards and E. S. Vanzant. 2004. Suitability of a GPS collar for grazing studies. Transactions of the ASAE 47(4): 1321.


Arc GIS. 2016. ArcGIS pro user guide: Euclidean- distance. New York, USA. ESRI Inc. Availabel at :


Bailey, D. W., D. D. Kress, D. C. Anderson, D. L. Boss and E. T. Miller. 2001. Relationship between terrain use and performance of beef cows grazing foothill rangeland. Journal of Animal Science 79(7): 1883-1891.


Cole, J. T., T. S. Stombaugh, and S. A. Shearer. 2004. Development of a test track for the evaluation of GPS receiver dynamic performance. In: 2004 ASAE/CSAE Annual International Meeting, Paper No. 041060. Ottawa, Ontario, Canada: August 2004.


Gloster, J., P. Williams, C. Doel, I. Esteves, H. Coe and J. F. Valarcher. 2007. Foot-and-mouth disease–Quantification and size distribution of airborne particles emitted by healthy and infected pigs. The Veterinary Journal 174(1): 42-53.


Haugen, L., P. D. Ayers, M. Vance and A. Anderson. 2000. Using GPS for vehicle tracking and dynamic property monitoring. In: 2000 ASAE Annual International Meeting, Paper No. 20003018608. Milwaukee, Wisconsin, USA: July 2000.


Kochan, B., T, Bellemans, D. Janssens and G. Wets. 2006. Dynamic activity-travel data collection using a GPS- enabled personal digital assistant, In : Proceedings of the Ninth International Conference on Applications of Advanced Technology in Transportation (AATT), pp. 319-324, Chicago, Illinois, USA: August 2006.


Liverani, M., J. Waage, T. Barnett, D. U. Pfeiffer, J. Rushton, J. W. Rudge, M. E. Loevinsohn, I. Scoones, R. D. Smith, B. S. Cooper, L. J. White, S. Goh, P. Horby, B. Wren, O. Gundogdu, A. Woods and R. J. Coker. 2013. Understanding and managing zoonotic risk in the new livestock industries. Environmental Health Perspectives 121(8): 873-877.


Martínez-López, B., A. M. Perez and J. M. Sánchez- Vizcaíno. 2010. A simulation model for the potential spread of foot-and-mouth disease in the Castile and Leon region of Spain. Preventive Veterinary Medicine 96(1-2): 19-29.


Necula, E. 2015. Analyzing traffic patterns on street segments based on GPS data using R. Transportation Research Procedia 10: 276-285.


Obuhuma, J. I. and C. A. Moturi. 2012. Use of GPS with road mapping for traffic analysis. Internation Journal of Scientific and Technology Research 10(1): 120-128.


Pelekis, N., I. Kopanakis, E. E. Kotsifakos, E. Frentzos and Y. Theodoridis. 2011. Clustering uncertain trajectories. Knowledge and Information Systems 28(1): 117-147.


Seo, I. H., I. B. Lee, S. W. Hong, H. S. Noh and J. H. Park. 2015. Web-based forecasting system for the airborne spread of livestock infectious disease using computational fluid dynamics. Biosystems Engineering 129: 169-184.


Sun, L. and J. Zhou, 2005. Development of multiregime speed-density relationships by cluster analysis. Transportation Research Record: Journal of the Transportation Research Board, 1934: 64-71.


Weber, T. P. and Stilianakis, N. I. 2008. Inactivation of influenza A viruses in the environment and modes of transmission: a critical review. Journal of Infection 57(5): 361-373.

페이지 상단으로 이동하기