Social isolation can be thought of as a withdrawal from or inability to participate in society (Okamoto, 2016).  Reasons for a person’s limited participation in society can be plentiful, yet scholars have made connections between social isolation and psychological and physical health, loneliness, and early mortality (Holt-Lundstad et al).  Furthermore, it is argued that “social isolation heightens sensitivity to risk” in the human and non-human world (Cacioppo et al).  Within Japan’s dense urban populations, social isolation is a concern for single occupant households and the elderly.

In the prefectures that make up the Tokyo Metropolitan Region, 37% of households are comprised of a single occupant.  Additionally, 11% of the total households are single occupants over 65 years of age.  Although living alone does not ultimately determine social isolation, for the purposes of our study we have considered it to indicate increased risk.  Noting the literature’s connections between social isolation, health and sensitivity to hazards, the Tokyo metro area raises some concerns, both because of high proportions of single-livers and because of the ever-present risk of natural disaster.  Moreover, with large populations of elderly populations, many of whom live alone, vulnerability to social isolation is a serious concern.

Our analysis employed a social isolation vulnerability index of our own design to identify the spatial distribution of populations that we determined to be most at risk of social isolation.  Clusters of high risk populations were most concentrated in peripheral areas, likely as a result of our inclusion of proximity to metro rail lines as a determinant of decreased isolation risk.  Peripheral high risk populations also faced increased hazard exposure from hill slopes along the western reaches of Saitama, Tokyo, and Kanagawa prefectures.  Inundation hazard due to flood, tsunami, or sea level rise affected high risk populations at coastal areas in Chiba and Kanagawa; however, inundation hazard was also a major risk for Tokyo’s densely populated central core, where populations were determined to be low risk to social isolation.

Our study has several limitations that must be addressed.  First, none of our social vulnerability index indicators necessarily determine social isolation.  Our study does not account for the variability in social capital across the region which would likely be a better indicator of vulnerability.  Finding data pertaining to one’s social capital at the scale used in our analysis would likely be challenging and is certainly beyond the scope of our research.  Moreover, our indicators for social isolation do not include persons with disabilities or reduced mobility.  This factor was not addressed because of data availability issues at our scale of study, a more focused approach might benefit from including it as a factor.  Additionally, income, shelter cost, and weekly work hours could be beneficial indicators of vulnerability, but were not included in our study as they were unavailable.  Second, our index considered proximity within 1 km of a metro rail line as an indicator of reduced vulnerability.  There is inherent uncertainty in this measure as close proximity to a rail line does not necessarily indicate easy access to a metro rail station.  Third, medical facilities were selected from a building database found through Open Street Map, an open source and editable database of geospatial data.  As such, some of this data may not be accurate or credible and therefore needs to be addressed as a source of uncertainty.  Lastly, and arguably most significantly, our index was created according to our own judgement and is ultimately somewhat arbitrary.  A deeper and more comprehensive exploration of the literature could help to reduce uncertainty in developing a useful metric, however with time constraints we relied on a limited amount of literature and our best judgment.

With consideration to the limitations mentioned above, we believe this analysis to be a useful starting point to exploring social isolation in the greater Tokyo area.  Further study could benefit from a more focused, local approach that explores variation at a larger scale.  Japan’s census data is available aggregated to 1 km, 500 m, and 250 m grid cells and could be useful in conducting a more thorough analysis that incorporates and addresses the modifiable areal unit problem that follows census data produced in various scales.

Our study relied heavily on census data as a measure of vulnerability, yet scholars have explored connections between natural disasters and social isolation.  Inoue and others note a relationship between the Great East Japan Earthquake and tsunami in March 2011 with increased social isolation.  Notably, with displacement post-disaster, individuals with limited social capital are likely to experience increased social isolation (Inoue et al, 2014) affecting physical and psychological well being.  Although our study only briefly explored the spatial relationships of communities at heightened risk of exposure to both social isolation and natural hazards, we did observe a connection between areas at high risk to social isolation and natural hazard risk.  This suggests that a deeper exploration of these spatial relationships would be valid.