Prediction of the indoor temperatures of an urban area with an in-time regression mapping approach.

Excess mortality has been noted during high ambient temperature episodes. During such episodes, individuals are not likely to be uniformly exposed to temperatures within cities. Exposure of individuals to high temperatures is likely to fluctuate with the micro-urban variation of outdoor temperatures (heat island effect) and with factors linked to building properties. In this paper, a GIS-based regression mapping approach is proposed to model urban spatial patterns of indoor temperatures in time, for all residential buildings of an urban area. In July 2005, the hourly indoor temperature was measured with data loggers for 31 consecutive days, concurrently in 75 dwellings in Montreal. The general estimating equation model (GEE) developed to predict indoor temperatures integrates temporal variability of outdoor temperatures (and their 24 h moving average), with geo-referenced determinants available for the entire city, such as surface temperatures at each site (from a satellite image) and building characteristics (from the Montreal Property Assessment database). The proportion of the variability of the indoor temperatures explained increases from 20%, using only outdoor temperatures, to 54% with the full model. Using this model, high-resolution maps of indoor temperatures can be provided across an entire urban area. The model developed adds a temporal dimension to similar regression mapping approaches used to estimate exposure for population health studies, based on spatial predictors, and can thus be used to predict exposure to indoor temperatures under various outdoor temperature scenarios. It is thus concluded that such a model might be used as a means of mapping indoor temperatures either to inform urban planning and housing strategies to mitigate the effects of climate change, to orient public health interventions, or as a basis for assessing exposure as part of epidemiological studies.