Dolines are the most numerous karst features in Slovenia, characterized by circular depressions ranging from a few meters to over a kilometer in diameter. Found in limestone, dolomite, carbonate breccia, and conglomerate, they form through processes like dissolution, collapse, suffosion, and cave transformation. Using lidar data from Slovenia’s nationwide laser scanning project, researchers manually labeled parts of a digital elevation model (DEM) to identify dolines. A modified U-net machine learning algorithm was trained on this data and used to infer doline locations across the DEM. The resulting binary mask was converted into an ESRI Shapefile and manually verified to address errors, especially in less common features like collapse dolines.

The study cataloged 471,192 dolines, primarily solution dolines (470,325), with an average depth of 9 m, a diameter of 42 m, and a volume of 14,098 m³. Collapse dolines (314) are deeper (mean depth 49 m) and larger (mean volume 1.2 million m³). Suffosion dolines (553) form in blind valleys and poljes. Dolines are most abundant on level surfaces, with densities up to 500/km², and absent from polje floors and steep slopes. This dataset provides a foundation for further studies on karst geology and topography. The catalog is available at https://dolines.org.

Aknowledgements:

Mihevc, A., Mihevc, R., 2021. Distribution and morphological characteristics of dolines in Slovenia defined by lidar data sets and machine learning. Acta carsologica

https://dolines.org/

Zenodo: DOI: https://zenodo.org/record/5856456