Land use regression modeling of microscale urban air temperatures in greater Vancouver, Canada

Abstract

BackgroundExtreme heat events have been associated with excess morbidity and mortality worldwide. Previous research mainly evaluated extreme heat exposures at the municipal and local scales, but individuals are exposed in much smaller areas. The goal of this study was to assess whether land use regression (LUR) models could be developed for air temperature using measurements collected by a pedestrian. MethodsMicroscale air temperature (<100 m) was measured during 42 sampling runs across 20 routes in greater Vancouver, Canada. Six independent variables were considered as potential predictors for LUR model construction for each run and for greater Vancouver as a whole. All models were evaluated using a spatial leave-ten-out cross-validation (LTOCV) approach. ResultsThe most predictive LUR variables were Distance to Large Water Body, Distance to Major Road, Normalized Difference Water Index (NDWI), and Sky-View Factor (SVF). On average, the best individual route models explained 39% of the variation in microscale air temperatures for the 20 routes. The overall model explained only 10% of the variation in the 20 combined routes. ConclusionMobile air temperatures were associated with geographic and built environment features at the microscale. The collected data were used to build moderately predictive LUR models for some locations, but could not be used to successfully model the entire study area.