Identifying Moraine-Dammed Glacial Lakes Using Moraine Accumulation Characteristics and Vision Transformer

Moraine-dammed glacial lakes are naturally formed by the accumulation of moraine debris in high mountain glacier environments. Due to their remote locations and the challenges in identification, these lakes often elude systematic and comprehensive surveys. However, under the influences of glacier melting and climate change, they can potentially cause catastrophic outburst floods, threatening the safety of downstream communities and the stability of ecosystems. Therefore, precise identification and monitoring of these lakes are crucial for disaster early warning and risk management.

The aim of this study is to develop a novel method based on multi-source remote sensing data and Vision Transformer technology for effectively identifying moraine-dammed glacial lakes. Traditional remote sensing methods face numerous challenges in these high mountain environments, such as confusion with similar water bodies and the impact of complex terrain. Our proposed method focuses on utilizing moraine accumulation characteristics, a key factor in the formation of moraine-dammed lakes. By analyzing the relationship between glacier movement and lake formation, we aim to more accurately identify potential dammed lakes, thereby reducing misidentifications.

We are using high-resolution satellite imagery and terrain data, combined with the Vision Transformer model for feature extraction. This model is capable of efficiently processing a large amount of complex spatial data and identifying specific geographical and geomorphological features. We are focusing on changes at the glacier front and terrain changes related to lake formation. Through this approach, we aim to extract key features directly related to the formation of moraine-dammed glacial lakes, thus improving the accuracy of identification.

Additionally, we are establishing a database containing samples of known moraine-dammed glacial lakes to train and validate our model. By comparing it with existing databases of moraine-dammed glacial lakes, we aim to further test the effectiveness and reliability of our method. We are anticipating that this research will provide a new technological approach for monitoring moraine-dammed glacial lakes, with significant scientific importance and practical value in understanding the mechanisms of lake formation, assessing potential risks, and developing effective disaster prevention measures.