Application of Sentinel-2 and EnMAP new satellite data to the mapping of alpine vegetation of the Karkonosze Mountains

Abstract

Abstract Effective assessment of environmental changes requires an update of vegetation maps as it is an indicator of both local and global development. It is therefore important to formulate methods which would ensure constant monitoring. It can be achieved with the use of satellite data which makes the analysis of hard-to-reach areas such as alpine ecosystems easier. Every year, more new satellite data is available. Its spatial, spectral, time, and radiometric resolution is improving as well. Despite significant achievements in terms of the methodology of image classification, there is still the need to improve it. It results from the changing needs of spatial data users, availability of new kinds of satellite sensors, and development of classification algorithms. The article focuses on the application of Sentinel-2 and hyperspectral EnMAP images to the classification of alpine plants of the Karkonosze (Giant) Mountains according to the: Support Vector Machine (SVM), Random Forest (RF), and Maximum Likelihood (ML) algorithms. The effects of their work is a set of maps of alpine and subalpine vegetation as well as classification error matrices. The achieved results are satisfactory as the overall accuracy of classification with the SVM method has reached 82% for Sentinel-2 data and 83% for EnMAP data, which confirms the applicability of image data to the monitoring of alpine plants.