Abstract:
Not only does the temporal and spatial evolution of glaciers have a great impact on river runoff, living environment and surface morphology, but glaciers themselves have a strong response to climate change as well. Under the background of global warming, it is of great significance to monitor the long-term evolution of glaciers. In this paper, TM image and OLI image are used to extract glacier information in Meili Snow Mountain Area through unsupervised classification, supervised classification, ratio threshold method, snow cover index (NDSI), object-oriented method based on multi-scale segmentation and glacier recognition method based on neural network. The results show that: it is of relatively good effect to extract information of bare ice area and moraine covered area glacier, by means of the glacier recognition method based on neural network; extraction accuracy is the highest. Subsequently, based on ENVI deep learning module, the glacier information of Meili Snow Mountain Area in 1989, 1998, 2009 and 2019 is interpreted by using neural network classification method, and then the interpretation is visually corrected combined with Google Earth and DEM data. Finally, the glacier boundary change map of Meili Snow Mountain Area from 1989 to 2019 is obtained. The results show that the glacier in Meili Snow Mountain Area retreated by 23.77 km
2 from 1989 to 2019, with an average annual retreat rate of 0.79 km
2; and the area relative retreat rate is 17.03%, with an average annual relative retreat rate of 0.57%.