Comparison of the effects of selected variables on urban surface temperature

The intra-urban variation in surface temperature and its related natural and social variables region by region within a city was investigated in the study. The study area is Washington DC, USA. Data sources include one EOS Terra ASTER scene, census data and high spatial resolution (1m) color infrared DOQQ. The census tracts were used to partition the city into different regions. Variables extracted and considered in the study are: (1) urban surface temperature, (2) population density, (3) NDVI, and (4) land use/cover, particularly the percentage of urban surfaces. Mean values of each variable were calculated based on the census tracts, and their interrelationships were examined using a correlation matrix. Results indicate that urban surface temperature is mostly correlated with the percentage of urban surfaces (r = 0.857). Whereas, NDVI shows strong negative correlation with surface temperature (r= - 0.817), and its coefficient is further subject to the influence of available water bodies (i.e., water bodies have low NDVI but can lower surface temperature). Positive relation was found between population density and temperature with a reduced degree of correlation (r = 0.417), but still significant (p-value <= 0.0001 at 95% confidence level)