Using in situ spectra to explore Landsat classification of glacier surfaces

摘要: Abstract Glaciers are key to understanding the world's hydrological cycle as well regional and global climate change. Glacier surfaces metamorphose into a range of zones which have implications for surface energy balance can be used proxy glacier's mass balance. Multispectral images (in particular Landsat data) been extensively classify study glacier surfaces. This uses full-spectrum in situ reflectance data from Midtre Lovenbreen (Svalbard) Langjokull (Iceland) combined with ETM+ imagery inform explore classification. Qualitative comparison unsupervised classifications shows that while visible near-infrared (VNIR, ∼350–1350 nm) compared shortwave-infrared (SWIR, ∼1500–2500 nm) is important identifying vs. non-glacier, VNIR wavelengths crucial distinguishing classes. Principal component analysis (PCA) spectra indicates only two PCs required segment “clean” third PC necessary on ash-covered glaciers. Simulation led new linear band combinations (LCs) transferable both temporally spatially, comparisons drawn existing techniques. For clean glaciers, LC1 contains primary information, it through combination LC2 delineations begin successfully drawn. Landsat's has sufficient spectral resolution accomplish this task, but not case ash or other significant debris cover. Further studies should investigate sensor-specific indices, suitability various quantitative clustering techniques, impact pixel/footprint size may these conclusions.