Slope-driven edge analysis of high-resolution LiDAR data for automated detection of cultural terraces in Slovenia

Cultural terraces were often constructed to improve agriculture. Some terraces are still in use, while others have been abandoned. Knowledge of their locations is important for their preservation or potential reuse. There have been several attempts worldwide to create a register of terraces. In Slovenia, a suitable register has not yet been created due to heavy overgrowth and significant differences in cultural terrace types across different regions of the country. This research proposes detecting terraces using a LiDAR digital elevation model, geoinformation tools, and additional spatial data. The method detects sharp changes in slope data and creates polygons where such changes are detected in close proximity. The main advantage of the method is that it does not require any training samples yet still provides accurate results despite the diversity of terraced areas. We applied the method in Slovenia and achieved an accuracy of 91 percent, a precision of 76 percent, and a recognition value of 66 percent in one test area, and 92, 47, and 65 percent in another designated test area. To achieve higher accuracy, the input settings can be adapted to regional characteristics, which confirms earlier findings that terraces in Slovenia exhibit high diversity.