ABSTRACT. We conducted a spatial analysis of low pathogenic H5N2 avian influenza (AI) outbreaks, that affected 41 chicken farms in Japan in 2005. A statistically significant (p=0.001) cluster of AI-positive farms was identified in the central part of Ibaraki Prefecture. Inside the AI cluster, the density was high for both chicken farms and chicken population, the proportion of layer finisher type farms was high and the farm size was large. We considered it important to take precautions for AI outbreaks in densely chicken-populated areas and to implement appropriate movement control around the affected farms to prevent transmission among farms located within small distances in the case of AI outbreaks. Spatial scan statistics are applicable in veterinary epidemiology to detection of high risk areas for animal diseases.KEY WORDS: avian influenza, cluster, spatial scan statistic.J. Vet. Med. Sci. 71 (7): [979][980][981][982] 2009 On the 26th of June 2005, a H5N2 subtype of avian influenza (AI) virus was recovered from a layer chicken flock in Ibaraki Prefecture, Japan, that showed a slight decrease in egg production. Nine more virus-positive farms and 32 seropositive farms (by haemagglutination inhibition tests) were detected over the course of 26 weeks. Since the virus strain was low in pathogenicity, with an intravenous pathogenicity index [14] of 0.0, and did not induce any particular clinical signs in the chickens of most flocks [15,16], the disease spread silently and the timing of virus introduction was unclear. This strain was considered to be newly introduced into Japan, taking into account its genetic characteristics and the results of national AI monitoring implemented periodically prior to the outbreaks [9]. Of 41 affected farms, 40 were located in Ibaraki Prefecture, approximately 60 km north of Tokyo. In this prefecture, almost all the poultry raised were chickens; 70% of these were laying hens, and 30% were broilers.When animal disease outbreaks occur, effective steps to explore the risk factors include investigation of the clustering of the affected farms and identification of their location and size. Nevertheless, it is generally difficult to recognize clusters visually when the locations of the cases and noncases (not affected by AI) are irregularly scattered. On the other hand, the Geographic Information System offers an opportunity to explore the associations between potential risk factors and disease incidence in combination with spatial analyses, and complements traditional visual approaches. Among the spatial analyses, spatial scan statistics enable researchers to locate statistically significant geographical clusters (the locations of the clusters and their size) [10]. In public health epidemiology, spatial scan statistics have been applied for detecting disease clusters to enable investigation of risk factors or planning of intervention trials [2,7,9]. On the other hand, few spatial analyses are reported in animal diseases in Japan [13], and there are no reports concerning the clustering ...