Reliability of sealed surfaces detection using Copernicus data

Abstract

Over the last 50 years there have been significant changes in land cover and land use, mainly those classified as artificial. This process, and its generalisation on a global scale, affect directly the basic functions of the soil, increasing other problems such as the loss of biodiversity, pollution, soil degradation, flooding and the effects of climate change. In the study area (Mazarrón, Region of Murcia) this problem is exemplary: the binomial urban development associated with the increase of beach tourism and intensive agriculture (under greenhouses) alter drastically the nature of the soil. The aim of this paper is twofold: to establish a supervised classification model that distinguishes, with an assumable error, the different classes established, highlighting those considered as sealed surfaces and, in addition, to make a comparison with the latest Corine Land Cover available information (2018). For this purpose, Sentinel 2A satellite images were selected and a maximum likelihood classification was performed. To validate the results, a confusion matrix was developed and an overall accuracy of 89% was obtained. Finally, a significantly underestimation by the Corine Land Cover of 75% of the sealed surfaces was observed, mainly due to its resolution.