Accurate classification of cirques is essential for studying paleoglacier activities. Longitudinal-profile-based classification has advantages over other methods such as expert classification and the methods that utilize morphometric parameters. Longitudinal profiles method first deploys exponential function to fit the longitudinal profiles of individual cirque samples and then fits a linear classifier based on the exponential coefficient and the cirque height of the cirque sample set to classify cirque candidates as cirques or non-cirques, However, the existing studies have applied and evaluated longitudinal profile based classification using only small number (i.e., several dozens) of cirque sample-set collected within small study areas. In this study we evaluated the applicability of the longitudinal profile method to a larger number of glacial cirques from a larger area. The cirque sample set (256 cirques and 101 non-cirques) of this study was extracted from the southeastern Tibetan Plateau. The original linear classifier fitted in previous studies, and the linear as well as non-linear classifiers fitted from the new sample set were evaluated. The classification accuracy results reveal that the longitudinal profile based classification method was applicable, and that with the non-linear classifiers shows the improved performance than the refitted linear classifier, when both were better than that with original linear classifier.
|Published - 17 Apr 2023