Characterization of the urban heat Island effect from remotely sensed data based on a hierarchical model

Abstract

This study attempts to statistically characterize the Urban Heat Island Intensity (UHII) ($\Delta T$) for 55 cities under three climate regimes – arid, snow and temperate – across the US. The study uses remotely sensed data products, daily temperature from MODIS and daily evapotranspiration from SSEBop model, to calculate the urban–rural difference in daily-mean temperature and daily-mean evapotranspiration ($\Delta T$ and $\Delta ET$ respectively) for the selected cities. By developing a hierarchical model that explains UHII using temporally-varying $\Delta ET$ and spatially-varying urban morphometric characteristics (total urban area and percentage impervious area) available for each city, we find that 89% of the spatio-temporal variability in annual $\Delta T$ can be explained. The relationship between $\Delta T$ and $\Delta ET$ is found to be negative indicating increased difference in daily means of ET ($\Delta ET$) result in increased difference in daily means of temperature ($\Delta T$) between urban and rural paracels The variation of $\Delta T$ per unit $\Delta ET$ is found to be highest in arid and snowy environments and smallest in temperate environments in the south-southeast US. The relation between $\Delta T$ and $\Delta ET$ is negative for most cities, except Madison (WI) and Sacramento (CA), across the US. Both the selected urban morphometric properties are found to be statistically significant in explaining the spatial variability in UHII, but the difference in urban–rural difference in evapotranspiration is the primary driver for UHII.