The purpose of this study was to examine the impact of physical geographic factors and climate conditions on human longevity. The centenarian rate (CR) in 2005 was computed for Japan's 47 prefectures, whose geography and climate vary greatly. Several pathways, such as excess winter mortality, land use and agricultural production, possibly linking physical and climate factors with extreme longevity, were explored. The probability of becoming a centenarian varies significantly among the Japanese prefectures. In particular, the computation of CR(70) demonstrated that the actual probability for individuals 70 years old in 1975 of becoming centenarians in 2005 was 3 times higher, on average, in Okinawa, both for males and females, than in Japan as a whole. About three quarters of the variance in CR(70) for females and half for males is explained by the physical environment and land use, even when variations in the level of socio-economic status between prefectures are controlled. Our analysis highlighted two features which might have played an important role in the longevity observed in Okinawa. First, there is virtually no winter in Okinawa. For instance, the mean winter temperature observed in 2005 was 17.2 degrees C. Second, today, there is almost no rice production in Okinawa compared to other parts of Japan. In the past, however, production was higher in Okinawa. If we consider that long term effects of harsh winters can contribute to the mortality differential in old age and if we consider that food availability in the first part of the 20th century was mainly dependent on local production, early 20th century birth cohorts in Okinawa clearly had different experiences in terms of winter conditions and in terms of food availability compared to their counterparts in other parts of Japan. This work confirms the impact of climate conditions on human longevity, but it fails to demonstrate a strong association between longevity and mountainous regions and/or air quality.