Land-Use Regression Models of Parcel Level Intra-Urban Surface Temperature in Three Cities in Massachusetts.

摘要

Surface temperatures have been shown to vary substantially at the intra-urban scale. Models are needed that allow fine-scale spatial predictions of exposure to surface temperatures to conduct intraurban exposure assessment and health analyses. The aims of this work were to (1) develop land use regression models of surface temperature at the parcel level in three cities in Massachusetts (Boston, Springfield and Worcester); (2) explore the relative importance of spatial and temporal factors on intraurban surface temperature; and (3) compare models between cities. Thermal imaging data at a 30 m × 30 m resolution were obtained during flyovers from winter 2014 - summer 2016. In each city, image data were available for >30 days over the study period, and were used to calculate mean parcel surface temperature. Parcel scale regression models were developed for each city and for warm and cool seasons. Independent covariates in land use regression models included land use classification, year built, property value, impervious surface area ratio, normalized difference vegetation index (NDVI), wind direction, wind speed, temperature, relative humidity, population density and distance from major water bodies. Preliminary results indicate that regression models explained between 32 to 48% of the variability in mean parcel surface temperature. Proximity to major water bodies explained the greatest amount of the surface temperature variation in each city (adjusted-R~2=0.11-0.28), with differential effects by season. Land use and NDVI explained an additional ～10% of the variation, but had considerable differences in magnitude and strength of association between cities. Significant housing predictors included year built and building value. Our models can be used to predict fine scale surface temperature based on land use, housing, and meteorology, expanding our understanding of the drivers of urban-heat islands and associated health effects.