Norovirus (NoV) is a major cause of gastroenteritis [1] during the autumn and winter seasons in Japan as well as in other temperate climate regions [2]. NoV patients span all age groups, and the mode of virus transmission is thought to be person-to-person, food-borne, or water-borne [3], [4]. Most outbreaks are thought to occur by secondary attacks through person-to-person infection by the fecal-oral route, contact with contaminated fomites or surfaces, or by inhalation of virus particles from vomitus [5], [6]. The most severe cases are found in young children or patients with chronic diseases and may result in death [7].

NoV is highly communicable with a small infectious dose [8]. Prolonged viral shedding even in asymptomatic people has been reported [7], [9]. NoV shows frequent mutations, and repeated infections could occur because of its molecular changes [10]–[12]. NoV is classified into five genogroups (GI to GV) and multiple genotypes on the basis of the phylogeny of the capsid sequence. In GII, 19 genotypes are recognized, with GII.4 as the leading cause of NoV-associated acute gastroenteritis in humans [13].

In Japan, NoV epidemics are monitored by not only outbreaks but also individual cases through sentinel surveillance, namely, the National Epidemiological Surveillance of Infectious Diseases (NESID) of National Institute of Infectious Diseases (NIID) [14]. The largest epidemic of NoV in the history of national surveillance of gastroenteritis since its launch in 1981 was found in the 2006–07 season and it was caused by the emergence of a GII.4 subtype termed 2006b [15]. Recurring epidemics have led to interest in how the initiation of epidemics occurs and how they are disseminated. Clarifying the patterns of epidemic diffusion is important for considering effective monitoring and surveillance as well as possible prevention. In the present study, we investigated a nation-wide NoV epidemic trend by focusing on the timing of increase of NoV cases for the 2006–07 to 2008–09 seasons using sentinel gastroenteritis surveillance data and the kriging method in GIS (geographic information system).

In each of the seasons examined, 2006–07 to 2008–09, an increase in NoV cases first occurred in the south and gradually moved north. The average period of this progress was 12.7 weeks, or about 3 months. GII.4 was always the most frequent genotype in each of the three years, but ranged from 60.4 to 88.9%. The reason for this south to north pattern is not certain, although it could suggest climate related effects. The southern area of Japan, such as Fukuoka (except Okinawa), and the middle area, such as Tokyo, lie in the temperate zone and have a yearly average temperature of 17°C. In contrast, the northern area, Hokkaido, lies in the subarctic zone and has a yearly average temperature of 9°C (Fig. S1a shows the monthly averages of daily maximum temperature (2006–2009)). A previous study has concluded that temperature and humidity had highly consistent effects to the NoV epidemic in England and Wales, although a complex interplay may exist between host, viral, and climatic factors [21]. The study has showed that the increases in NoV cases were associated with cold weather, dry temperature, low population immunity, and the emergence of novel genogroup 2 type 4 antigenic variants.

In Japan, the southern, middle, and northern areas have distinguishing climatic factors that could explain the period of the NoV epidemic initiation, which begins in September in the south and reaches the north in December (Fig. S1).

Particularly, when the maximum temperatures were similar in the southern and middle areas between 27.8 to 29.0 degrees and relatively lower in the northern area at under 23.4 degrees (from September onward), differences were found in the relative humidity; it rapidly decreases (−11.3%) in the southern area (Fukuoka) after slightly increasing towards September, and in the middle area (Tokyo) it modestly decreases (−6.7%), and in the northern area (Hokkaido) it slightly decreases (−2.0%) but continues to be high with small increases after October (Fig. S1a, and b). Because NoV cases first increased in the southern area, a warm temperature lowered in association with a relatively rapid change in humidity could be the cause for the increase. The NoV epidemic is also thought to be associated with cold weather [2]. However, most of the epidemic peaks appeared by the end of December, but not in January when the lowest temperature of the year occurs, indicating that there are other climate factors involved [21].

The emergence of a GII.4 subtype, 2006b, brought the largest NoV epidemic. The number of sentinel cases at the peak week was 22.81 (per sentinel site), and it has modestly decreased but remained relatively high in the following two seasons, as 19.33 and 15.88, respectively. 2006b was dominant in the 2006–07 season and remained so in the 2007–08 and the 2008–09 seasons in Japan [15], [22], [23]. The modest decrease of the number of sentinel cases at the peak week may suggest the development of herd immunity after a period of high prevalence, although such immunity maybe limited for short-term [12]. Additionally, 2006b was also detected in the 2005–06 season [22], which would suggest that 2006b remained persistent over the seasons. However, persistency or endemicity of NoV still remains unclear, because NoV cultures are not possible. Meanwhile the south to north pattern of the increase in cases could also suggest a possible migration of the epidemic as well as a dynamic virus transmission beginning from the south area. Although other factors such as difference in population density or standards of medical services may affect the observed geographic pattern of infections, the data obtained in this study do not support this suggestion. Nevertheless timely monitoring should enhance the surveillance of and preparedness against epidemics.

This study has some limitation regarding the data. The sentinel gastroenteritis data are mainly collected from pediatric clinics therefore adult data would have been underreported. Nevertheless, this is the first study to our knowledge that considered the nationwide patterns of NoV epidemics from the national sentinel data.