Supervised machine learning classification in Google Earth Engine: Time series analysis in Akkuyu, Turkey

Abstract

This paper tests the potential of the Google Earth Engine platform and Sentinel-2 data to classify and monitor land use and land cover nearby a nuclear power plant construction which is being eroded in Akkuyu district of Mersin Province in Turkey. After classification study, the performance of different supervised machine learning classification algorithms are compared. According to the accuracy assessment results, the random forest classifier has performed slightly better classification than other classifiers. According to the numerical results, Akkuyu NPP project has doubled the footprint of built-up areas. The project has brought a significant deforestation and filling areas on the seashore.