Variation of daily warm season mortality as a function of micro-urban heat islands

Abstract

Background: Little attention has been paid to how heat-related health effects vary with the micro-urban variation of outdoor temperatures. This study explored whether people located in micro-urban heat islands are at higher risk of mortality during hot summer days.

Methods: Data used included (1) daily mortality for Montreal (Canada) for June-August 1990-2003, (2) daily mean ambient outdoor temperatures at the local international airport and (3) two thermal surface images (Landsat satellites, infrared wavelengths). A city-wide temperature versus daily mortality function was established on the basis of a case-crossover design; this function was stratified according to the surface temperature at decedents' place of death.

Results: The risk of death on warm summer days in areas with higher surface temperatures was greater than in areas with lower surface temperatures.

Conclusions: This study suggests that measures aimed at reducing the temperature in micro-urban heat islands (eg, urban greening activities) may reduce the health impact of hot temperatures. Further studies are needed to document the variation of heat-related risks within cities and to evaluate the health benefits of measures aimed at reducing the temperature in micro-urban heat islands.

Figure 1

Associations between daily mean ambient temperature, in degrees Celsius, during summer months, 1990–2003, evaluated at lag 0 days and daily non-accidental mortality in Montreal, according to two categories of surface temperatures at the places of deaths (created with satellite images of summer 1990 and 2001). The strata were created with a cut-point at the 75th percentile of the distribution of the surface temperatures at the geographical centroid of the postal codes of the places of death. The black line represents the response function among decedents who were located in “hot areas” at the time of death whereas the grey line is for those who were located in “cooler areas”. Dashed lines represent the 95th confidence limits and are relative to the reference value of 20°C, as indicated by the vertical line. Vertical arrows crossing the x-axis locate the four knots used to create the spline functions.

Figure 2

Risk of dying at mean daily ambient temperature of 20°C (lag 0) compared with 20°C, in Montreal during summer 1990–2003, for categories of residential dwelling values (proxy for the socioeconomic status) and categories of surface temperatures at place of death (from a satellite image of summer 1990 or 2001). Deaths that occurred in long-term care facilities were excluded owing to missing former residential dwelling values. There were 5459 deaths with missing dwelling values.