Information Technology for Land Degradation Assessment Based on Remote Sensing

Abstract

Since the launch of ESA Copernicus program, satellite data of high resolution became publicly available and methods and tools for their automated processing to solve a wide range of applications have developed rapidly. An important scientific task is to assess land degradation and achieve zero levels of degradation. There are many methods for determining land degradation. Known approaches to the tasks of environmental land monitoring usually use the same methodology for all types of land cover. The paper represents the approach to the calculation of land degradation based on remote sensing data and modelling results taking into account the specifics of land degradation for different land cover and land use types. Our method is based on the classification of different land cover and land use types from satellite imagery and application of different schemes of land degradation assessment for each of them. We consider forest cuts as land degradation for forests and assess them using deep learning models. Land degradation for croplands is estimated by comparison of real leaf area index (LAI) and ideal LAI, calculated with the bio-physical crop development model. And land degradation for grassland is determined with a traditional approach based on vegetation index NDVI extracted from satellite imagery. The proposed approach was implemented for the territory of Ukraine.

Keywords

• geospatial data analysis
• machine learning
• land degradation
• remote sensing
• land cover