Automation of Surface Karst Assessment Using Sentinel‑2 Satellite Imagery

Abstract

The article demonstrates the advantages of a detailed analysis of remote sensing data for karstological purposes using the Google Earth Engine cloud platform and geographic information systems. The karst area within the Kishert gypsum and carbonate gypsum karst development area in Perm krai was chosen as the study area. The article demonstrates the application of space imagery classification with learning. The purpose of imagery classification is automatic zoning of the territory by type of land cover: meadows and croplands, forests, urbanized areas. In meadows and croplands, calculation of vegetation indices has been carried out in order to delineate potentially karst hazardous areas. The idea of using vegetation indices in assessing surface karst is based on the geobotanical properties of sinkholes in the study area. The relatively high values of vegetation indices within sinkholes reflect the fact that the sides, slopes, and bottoms of sinkholes are covered with shrubby, moisture-loving vegetation. This vegetation is interpreted successfully by calculation of vegetation indices under these conditions. Based on the spatial analysis of the distribution of potentially hazardous areas, a predictive model zoning the study area according to the degree of karst hazard was constructed. As a result of the quantitative assessment of the applicability of the methodology, we can conclude that the areas of coincidence of all four vegetation indices very accurately characterize the distribution of karst forms, and so the comprehensive research of the vegetation indices is very informative in assessing the surface karst distribution.