Statistical Modeling of Urban Heat Island Intensity in Warsaw, Poland Using Simultaneous Air and Surface Temperature Observations

Urban heat island (UHI) is one of the most distinctive characteristics of urban climate. The objective of this study is to apply a statistical modeling of the nocturnal atmospheric UHI based on the relationship between observed air temperature from ground stations and remotely sensed temperature of the urban surface. The goal of the approach is to limit input data for the developed modeling method in order to assure transferability of the methodology in different cities. Time series of surface temperature and normalized difference vegetation index are obtained from the MODIS instrument for a 10-year period (2008-2017). The air temperature is collected from the in-situ observational network of 21 stations. The studies are conducted for different locations with gradual changes in urbanization in order to assess the impact of urbanization on the relationship between simultaneous air and surface UHI. The urbanization is described by commonly available land cover metrics. Results showed that the proposed approach provides satisfactory AUHI modeling results for the locations with the least degree of urbanization. The best results are obtained with a simple linear regression model with the iterative procedure to minimize the mean absolute gross error (MAGE). The lowest MAGE for modeled UHI is 1.18 °C with 69% of the variance explained. The strongest linear relationship between simultaneous SUHI and AUHI is noted for those station pairs whose surroundings have the highest differences in urbanization, and the highest UHI intensities are observed. The strength of the SUHI/AUHI linear relationship decreases gradually with the increasing urbanization of the stations' surroundings.