Estimation of urban air temperature spatial patterns based on sensors network observations and satellite derived predictors

Besides new economical, managerial and social challenges associated with growing cities, the modifications caused in the energy budget of the urban surface intensifies the existing urban heat island (UHI). UHI can vary temporally and spatially according to meteorological conditions, landscape and urban typologies. Urban cover and form, as well as anthropogenic activities, pose an important effect on the city’s thermal behaviour that influence UHI and therefore the quality of life of the citizens. In this study, we focus on quantifying the air temperature spatiotemporal patterns across the urban and peri-urban area of Heraklion, Greece at a grid of 100 m x 100 m cells. We use point air temperature observations from the Wireless Sensors Network of Heraklion and interpolate spatially by means of sophisticated geostatistical modelling parameterized with satellite derived predictors. Regression kriging interpolation technique is implemented over the study area, using different predictors to minimize the uncertainty in air temperature estimation. We deal for multicollinearity between predictors and spatio-temporal correlations between measurements. A maximum magnitude of UHI ~ 4 oC has been observed between 04:00-05:00 (UTC+3). Cross-validations indicate a mean MAE ~0.86 oC in the estimated air temperature maps.