Multi-temporal analysis of land cover changes using Landsat data through Google Earth Engine platform

Timely and accurate maps of land cover changes are crucial for understanding the evolution of Earth's features and, consequently, the relationships between individual and collective needs. Therefore, this information is extremely important to develop future planning strategies and tackle environmental issues. This paper aims to exploit the use of Google Earth Engine (GEE) platform to examine land cover changes over a period of about fiftheen years in the pilot site of Siponto, an historical municipality in Puglia, Southern Italy. Six atmospherically corrected Landsat data, two for each selected mission (5, 7 and 8), were collected: the former was acquired in fall and the latter in spring. Land cover information was automatically extracted from each image through the implementation of an innovative Landsat Images Classifications algorithm (LICA) based on spectral indices analysis. Six classes (water, built-up, mining areas, bare soil, dense and sparse vegetation) were detected from each image, with an average overall accuracy higher than 85%. Land cover changes were assessed comparing classification maps of the same season, showing bare soil areas as the most altered ones, having been converted into arable lands in consideration of the adavantageous geomorphological features of the investigated site. This is also confirmed by the historical events experienced by the area.