Terrain music spectrum: An effective model for landform interpretation and recognition

Terrain music, a unique way to convey topographic information, has made progress in the field of geomorphology in recent years. However, research has been limited to a single watershed scale. To overcome this limitation and capture terrain characteristics across multiple scales, this study proposed a novel geomorphological analysis method—Terrain Music Spectrum (TMS). By introducing statistical quantitative indicators to describe the characteristics of TMS, this research conducted systematic investigations into geomorphological quantification and recognition applications. Taking the Loess Plateau as the study area and utilizing Digital Elevation Model (DEM) data, TMS models were constructed for several representative sample regions, enabling quantitative characterization of watershed geomorphological features. The proposed method was further applied to landform recognition tasks. The results showed that the recognition outcomes based on the TMS method were consistent with those based on traditional terrain indicators, achieving an overall accuracy of 86.11% and a Kappa coefficient of 0.83. This indicated that the TMS method was effective in distinguishing different landform types. Further analysis revealed that the two methods exhibit complementary strengths in identifying various landform categories. Through their integration, classification performance was enhanced, with overall accuracy improving to 88.89% and the Kappa coefficient rising to 0.87. Feature importance analysis demonstrated that TMS indicators played a significant role in the integrated model. These findings suggest that the introduction of the TMS method offers a new perspective for geomorphological research and facilitates interdisciplinary innovation in terrain analysis.