最邻近距离空间分析法在食品安全风险监测中的应用  

Application of spatial analysis by nearest neighbor method in food safety risk surveillance

作　　者：梁辉[1] 王博远[2] 邓小玲[1] 周少君[1] 肖革新 何伦发[2] 刘志婷[1] 李映来[2] 陈语嫣 张瀚中[2] 

机构地区：[1]广东省疾病预防控制中心,广东广州511430 [2]中山市疾病预防控制中心 [3]国家食品安全风险评估中心

出　　处：《华南预防医学》2017年第4期317-321,共5页South China Journal of Preventive Medicine

基　　金：2015年广东省公益研究与能力建设专项资金项目(2015A020218002)

摘　　要：目的分析食品安全风险监测中样品采集地点的空间分布模式。方法收集某省22个监测城市2015年食品安全风险监测的281个蔬菜及其制品的采样地点数据,通过GIS地理编码技术,将采样地点数据信息转换成经纬度坐标导入某省电子地图数据库中制作采样地点专题图,计算采样地点平均最邻近距离与空间随机模式下的平均最邻近距离的期望值之比,即最邻近距离系数(NNI),通过NNI值判断采样地点的空间分布特征。结果 2015年某省22个监测城市的食品安全风险监测采样地点NNI值的范围在0.002~1.086之间,平均值为0.335。18%(4/22)的监测城市NNI<0.1;68%(15/22)的监测城市0.1<NNI<0.5;9%(2/22)的监测城市0.5<NNI<1;5%(1/22)的监测城市NNI>1。空间随机性Z检验P>0.05的只有1个城市(NNI=1.086,Z=0.57,P=0.28),其余21个监测城市的NNI值均<1,P<0.05。结论该省食品风险监测样本在空间上不满足随机独立的前提条件,运用经典统计方法推断的总体污染情况将存在偏倚。最邻近距离空间分析方法应用于食品安全风险监测,能够分析样本的空间分布模式,检验其是否符合随机性,为保证监测数据代表性提供了一种新的手段。Objective To analyze the spatial distribution pattern of sampling locations of food safety risk surveillance. Methods Data were collected from 281 sampling sites of vegetables and their products for food safety risk surveillance of 22 monitoring cities in a province in 2015. The sampling location data were converted into latitude and longitude coordinates by GIS Geocoding technology, and then, a sampling location thematic map was created. After calculating the average nearest neighbor distance of the sampling location, the nearest neighbor index （NNI）, namely the ratio of the average nearest neighbor distance to the expected value of the nearest neighbor distance of sampling location in each monitoring city, was calculated to analyze the pattern of the spatial distribution pattern. Results The NNIs of the sampling locations of the 22 monitoring cities ranged from 0.002 - 1. 086, with an average NNI of 0.335. The NNIs were less than 0.1 in 18% （4/22） monitoring cities, ranged from 0. 1 to 0.5 in 68% （15/22） cities, ranged from 0.5 to 1 in 9% （2/22）cities, and were bigger than 1 in 5% （ 1/22 ） cities. Spatial randomness test showed that only one city＇s P value was greater than 0.05 （ NNI = 1. 086, Z = 0. 57, P = 0. 28 ） , while the NNIs of the remaining 21 cities were less than 1, P 〈 0.05. Conclusion The samples of these cities were neither random nor independent, and the bias would be caused using the classical statistical method to infer the overall pollution situation. The nearest neighbor spatial analysis method can analyze the spatial distribution pattern of sampling locations in different cities, and find the cluster sampling locations, to avoid biased sample data and improve the scientific level of sampling scheme for food safety risk monitoring.

关 键 词：食品安全 选择偏倚 空间分布模式 最邻近距离分析法 

分 类 号：TS201.6[轻工技术与工程—食品科学] R155.5[轻工技术与工程—食品科学与工程]