High-resolution air temperature mapping in a data-scarce, arid area by means of low-cost mobile measurements and machine learning

The availability of gridded, screen-level air temperature data at an effective spatial and temporal resolution is important for many fields such as climatology, ecology, urban planning and design. This study aims at providing such data in a data-scarce, arid city within the greater Cairo region (Egypt), namely the Sixth of October, where, to our knowledge, no such data are available. By using (i) air temperature data, collected from mobile measurements, (ii) multiple spectral indices, (iii) spatial analysis techniques and (iv) random forest regression modelling, we produced air temperature maps (for both daytime and nighttime) at 30-m spatial resolution for the entire city. The proposed method is systematic and relies on low-cost instrumentation and freely-available satellite data and hence it can be replicated in similar data-scarce, arid areas to allow for better spatial and temporal monitoring of air temperature.