Estimation of Near-Surface Air Temperature From SEVIRI/Meteosat Remotely Sensed Infrared Radiation Data Using Artificial Neural Network Modeling

  • Fernanda Schuch

Student thesis: Master's Thesis


Growing urbanization over the past decades has increased energy consumption and vehicle usage across the world, which in turn has contributed to the phenomenon called urban heat island (UHI). The most important variable to characterize UHI is the urbanrural air temperature differential. This study aims at deriving a correlation between air temperatures measured at ground weather stations and brightness temperatures obtained from remotely sensed thermal infrared radiation data of the SEVIRI/Meteosat instrument. The correlation is obtained through an Artificial Neural Network model, which is shown to be superior to other approaches. While weather stations can be costly to install and maintain, satellite images have become more accessible with technological advances and offer high frequency readings with greater land coverage. It is therefore relevant to find the most accurate correlation in order to enable future studies to estimate near-surface air temperature values without the need for ground stations. The data used is from Abu Dhabi, UAE, where it is especially relevant to understand the intensity of the UHI, the comparatively high levels of urban density and anthropogenic activity (motorized vehicles and air conditioning). Two models were developed, one for a rural location and another for an urban location. Results show that the error (RMSE) between rural measured and predicted data is 1.54 °C while the urban model presented an error of 1.36 °C. Bias or mean error (ME) values are negligible for both locations.
Date of AwardDec 2017
Original languageAmerican English


  • air temperature
  • urban heat island (UHI)
  • Artificial Neural Network.

Cite this