| Background: The livestock movement is an important reason for the long-distance transmission of animal diseases,so it is of great significance to strengthen the supervision of livestock movement.The amount of pig slaughter ranks first among all kinds of livestock in China,and China is the worlds largest producer and consumer of pork.Due to the large number and high frequency of pig movement,pigs can be used as a "model organism" for the study of livestock movement in China.The amount of pig slaughter in Hunan Province accounts for about 9% of the total amount of pig slaughter in China,ranking second in all provinces,therefore,it is appropriate to take Hunans pig movement as an example to carry out the study on livestock movement.Objectives: To explore the structure of livestock movement network based on the animal health supervision database;To compare the temporal and spatial changes of pig movement before and after the outbreak of African Swine Fever(ASF),identify the crucial areas and key time of pig movement,and propose the supervision strategies based on the key risk points;To build and compare the spatio-temporal models under the framework of frequentist statistics and Bayesian statistics,discuss the optimal spatio-temporal model of pig movement,and evaluate the influencing factors of pig movement.Methods: The social network analysis(SNA)technology was applied to construct pig movement networks based on the data of the electronic issuing system of animal quarantine certificates of Hunan Province;The generalized additive model(GAM)and generalized additive mixed model(GAMM)were applied to evaluate the time trend and seasonality of pig movement;Morans I index was used to carry out spatial autocorrelation analysis of pig movement;GA(M)M and Bayesian Spatio-Temporal Model(BSTM)were applied to evaluate the spatio-temporal interaction and influencing factors of pig movement.(1)Pig movement network: The data on pig movement from 2017 to 2021 were extracted from the electronic issuing system of animal quarantine certificates of Hunan Province.The SNA technology was used to build the intra-provincial and inter-provincial pig movement networks at city-level and county-level,respectively.Then the number of pigs moved from other districts,the number of pigs moved to other districts,the number of pigs moved within the district,the shipment of pigs moved from other districts,the shipment of pigs moved to other districts,and the shipment of pigs moved within the district,were calculated according to the relationship formulas of livestock movement network indicators.The clustering pattern was evaluated using a Blondel-Guillaume-Lambiotte-Lefebvre algorithm.(2)Time trend and seasonality: GAMM and GAM whose basis functions were cyclic cubic regression splines were used to fit the time series of the shipment and number of pig movement,and the seasonality and time trend of pig movement from 2017 to 2021 were evaluated.(3)Spatial autocorrelation analysis: Morans I index was used to conduct global and local spatial autocorrelation analysis on pig movement network indicators.(4)Spatio-temporal interaction effect analysis: The spatio-temporal GAM with tensor product smooth was constructed under the framework of frequentist statistics,and the BSTM was constructed based on the integrated nested laplace approximation algorithm under the framework of Bayesian statistics.The spatio-temporal interaction models of the shipment of pig movement and the number of pig movement were evaluated.(5)Analysis of influencing factors: The data on influencing factors of pig movement from 2017 to 2020 were collected from the Statistical Yearbook of Hunan Province and the Statistical Yearbook of Hunan Rural Areas.A total of 5 influencing factors were included in this study,including the population of county-level administrative regions,the number of fertile sows at the end of the last year,rice yield,corn yield and vegetable planting area.The independent variables related to the dependent variables(including the number of pigs moved from other counties,the number of pigs moved to other counties,and the number of pigs moved within the county)were included in the spatio-temporal GAMM and BSTM.The independent variables were screened by forward stepwise regression,and the results of spatio-temporal GAMM and BSTM were compared.Results: The key import regions and key export regions in the Hunan pig movement network were different,and the network showed temporal and spatial characteristics such as seasonality,spatial autocorrelation,and spatio-temporal interaction effects.Pig movement decreased significantly during the ASF outbreak and in 2019,and resumed the upward trend in 2020 and 2021.The results of influencing factors evaluated by GAMM and BSTM are basically consistent(1)Key nodes and indicators of the pig movement in Hunan Province were evaluatedThe city-level network of intra-provincial pig movement showed that Changsha city and Zhuzhou city had the largest number of pigs moved from other cities,Shaoyang city,Yueyang city and Hengyang city had the largest number of pigs moved to other cities,Shaoyang city and Yongzhou city had the largest number of pigs moved within cities.The city-level network of inter-provincial pig movement showed that Yongzhou city had the largest number of pigs moved to other provinces,and South China was the main destination of inter-provincial pig movement in Hunan Province.Network clustering showed that the division of node clusters was related to geographical location,and cities in the same cluster were geographically adjacent.The county-level network of intra-provincial pig movement showed that there was a positive correlation between the number of pigs moved from other counties and in-degree centrality(r=0.76,P<0.05),and there was a positive correlation between the number of pigs moved to other counties and out-degree centrality(r=0.77,P<0.05).The county-level network of inter-provincial pig movement showed that there was a positive correlation between the number of pigs moved to other provinces and out-degree centrality(r=0.83,P<0.05).(2)The time trend of province-level indicators of pig movement in Hunan Province from 2017 to 2021 was exploredThe shipment of pigs moved within Hunan Province showed the same upward trend in2017,2018,2020 and 2021,and a downward trend in 2019(P<0.05).The number of pigs moved within the province showed upward trends in 2017,2018,2020 and 2021(P<0.05),but the growth rates in 2020 and 2021 were higher than those in 2017 and 2018.The time trends of city-level network indicators of pig movement in Hunan Province from 2017 to 2021 were exploredExcept for betweenness centrality and eigenvector centrality,the city-level network indicators of intra-provincial pig movement decreased significantly during the ASF outbreak in Hunan Province(i.e.,from October 2018 to February 2019).From March 2019 to December 2021,the number of pigs moved from other cities and the number of pigs moved to other cities both showed a trend of decreasing first and then increasing.After the ASF outbreak,the proportion of the number of pigs moved within the city to the total number of pigs moved within Hunan Province increased.Except for February and March 2020,this proportion is higher than 60% from October 2018 to December 2021,and reached 99.9% and97.1% in November and December 2018.The number of pigs moved to other provinces in November and December 2018 was 0.After the ASF epidemic in Hunan Province,the number of pigs moved to other provinces showed an overall upward trend.The seasonal trend of county-level network indicators of pig movement in Hunan Province from 2017 to 2021 were exploredThe county-level network of intra-provincial pig movement showed that the number of pigs moved from other counties,the number of pigs moved to other counties,and the number of pigs moved within the county all showed a seasonal trend of increasing from February to November and decreasing from December to February of the following year.In addition,The county-level network of inter-provincial pig movement showed that the number of pigs moved to other provinces showed a seasonal trend of increasing from February to July and decreasing from August to February of the following year.(3)The spatial autocorrelation of pig movement in Hunan Province was analyzedThe indicators of intra-provincial and inter-provincial pig movement networks all had positive spatial autocorrelation(P<0.05).In 2017-2021,the high-high clusters of the number of pigs moved from other counties and the number of pigs moved to other counties gradually migrated from central Hunan to to southern Hunan;the high-high clusters of the number of pigs moved within the county were always located in central Hunan Province;and the highhigh clusters of the number of pigs moved to other provinces were mainly located in the southern Hunan and the scope was expanding year by year.The low-low clusters of these four indicators were all located in the northwest of Hunan Province.(4)The spatiotemporal interaction effect of pig movement in Hunan Province from2017 to 2021 was exploredThe time effect,space effect and spatio-temporal interaction effect on intra-provincial and inter-provincial pig movement networks were all statistically significant(P<0.001).BSTMs showed that the number of pigs moved from other counties,the number of pigs moved to other counties,the number of pigs moved within the county,and the number of pigs moved to other provinces all had the interaction effects of structured space and structured time.(5)The results of GAMM and BSTM on the factors affecting the movement of pigswere basically consistentBoth GAMM and BSTM showed that population was a positive factor affecting the number of pigs moved from other counties.For every 10 million increase in population,GAMM showed the number of pigs moved from other counties increased by 10.8%(P<0.001),and BSTM showed that it increased by 7.5% [95% confidence interval(Credible Interval,Cr I): 6.1%-8.9%].Both GAMM and BSTM showed that the number of fertile sows at the end of the last year and the rice yield were positive factors affecting the number of pigs moved to other counties.The number of pigs moved to other counties increased by 31.3%(P<0.001)in GAMM and 25.2%(95%Cr I: 16.5%-33.9%)in BSTM for every 1 million heads increase in the number of fertile sows at the end of the last year,respectively.In addition,the number of pigs moved to other counties increased by 2.1%(P<0.001)in GAMM and 4.2%(95%Cr I:2.6%-5.7%)in BSTM for every 1 million tons of increase in rice yield,respectively.Both GAMM and BSTM showed the number of fertile sows at the end of the last year and the rice yield were positive factors affecting the number of pigs moved within the county.The number of pigs moved within the county increased by 17.3%(P<0.001)in GAMM and14.0%(95% Cr I: 4.2%-23.9%)in BSTM for every 1 million heads increase in the number of fertile sows at the end of the last year,respectively.In addition,the number of pigs moved within the county increased by 1.9%(P<0.001)in GAMM and 3.4%(95% Cr I: 1.6%-5.2%)in BSTM for every 1 million tons of increase in rice yield,respectively.Both GAMM and BSTM showed that population was a negative factor affecting the number of pigs moved to other provinces,and the number of fertile sows at the end of the last year and the rice yield were its positive influencing factors.The number of pigs moved to other provinces decreased by 23.8%(P<0.001))in GAMM and 32.8%(95% Cr I: 18.3%-47.1%)in BSTM for every 10 million increase in population,respectively.The number of pigs moved to other provinces increased by 31.7%(P<0.001)in GAMM and 40.3%(95%Cr I: 21.9%-58.4%)in BSTM for every 1 million heads increase in the number of fertile sows at the end of the last year,respectively.In addition,the number of pigs moved to other provinces increased by 4.5%(P<0.001)in GAMM and 6.9%(95% Cr I: 4.0%-6.9%)in BSTM for every 1 million tons of increase in rice yield,respectively.Conclusions: The application of SNA and spatio-temporal statistical analysis to animal quarantine certificate data can reveal the spatio-temporal characteristics and influencing factors of the livestock movement network.(1)Animal quarantine certificate data and SNA technology can be used to construct a livestock movement network.Calculating the node indicators of the livestock movement network and identifying the crucial areas and key time of livestock movement are conducive to the accurate implementation of supervision and management strategies.(2)GAM/GAMM can be used to obtain seasonal and temporal trends of livestock movements.There are seasonal trends of pig movement in Hunan Province,and it is necessary to strengthen the supervision and management of intra-provincial pig movement before New Years Day and inter-provincial pig movement in summer.In 2021,the pig movement in Hunan Province has recovered to the pre-ASF epidemic levels,and the number of pigs movement has reached a 5-year high.The intra-regional pig movement increases,while the inter-regional pig movement decreases,and the main inter-provincial destination for pig movement is South China,indicating that the regional prevention and control of major animal diseases in the central and southern regions achieves remarkable results.(3)The pig movement in Hunan Province presents spatial clusters and spatiotemporal interaction effects.The hot spots for pig movement have shifted from the central Hunan to the southern Hunan after the ASF epidemic,and the key supervision areas need to be changed.The main pig-producing areas and main pig-marketing areas in Hunan Province are separated,and the trend of this separation is more obvious after the ASF epidemic.Differentiated and precise supervision strategies need to be adopted for the main pigproducing areas and main pig-marketing areas.(4)The influencing factors of pigs moving in and out are different.It is necessary to improve the risk analysis of populous cities and counties to enhance the early warning ability.At the same time,it is necessary to focus on the slaughter and quarantine of pigs and shift from “transporting pigs” to “transporting meat” in major counties of pig breeding and rice planting(5)Both spatio-temporal GAM/GAMM and Bayesian spatio-temporal model can identify the time effect,spatial effect,spatio-temporal interaction effect and influencing factors of livestock movement network.The two results are basically consistent,but the modeling process of BSTM is cumbersome,which is not convenient for beginners to use.Spatio-temporal GAM/GAMM cannot be modeled based on administrative units and hence is not suitable for practical applications.The appropriate spatio-temporal model should be selected according to the data characteristics and study purposes in the supervisory work of livestock movement. |
