Detecting and mapping karst landforms using object-based image analysis: Case study: Takht-Soleiman and Parava Mountains, Iran

Abstract

This study presents a novel, semi-automated approach for integrating decision rules and object-based image analysis (OBIA) methods for identifying and mapping karst zones and landforms. We developed a multi-resolution segmentation approach using an Approximate Gaussian function to compute the degree of fuzzy memberships of object-based features and applied it to Sentinel-2 satellite images and a digital elevation model. The object based features and decision rules were applied to identify and detect karst landforms in the semi-automated approach. The efficiency of each technique was examined based on two case studies in Takht-Soleiman and Parava-Biston in Iran using a fuzzy synthetic evaluation (FSE) approach and ground control points. The validation of the karst landform detection and delineation yielded high accuracies for the six prominent landforms, namely Dolin (96.8%), Ouvala (99.2%), Lapiez (95.1%), Canyon (98.3%), Polje (96.1%) and Karren (97.4%), respectively. Based on the research outcome, we conclude that the combined use of spatial (e.g. shape index, compactness, asymmetry), spectral (e.g. brightness, mean and standard deviation) and textural (grey-level co-occurrence matrix, GLCM) features allows us to detect and map karst landforms efficiently. This fuzzy rule object-based approach can enhance the accuracy of geomorphological and geological maps and allows for a regular update of the usually labor-intensive geological mapping campaigns.