Rock Glacier Kinematics: A Proxy for Assessing Periglacial Dynamics and Ground Ice Content on the Tibetan Plateau

Rock glaciers are ice-debris landforms widely distributed in the mountainous periglacial realm worldwide. They serve as important indicators for permafrost which is defined by its underground temperature and invisible in most other cases, especially for regions such as the Tibetan Plateau where in-situ observations are limited in spatial coverage due to the harsh and remote environment. Surface kinematics of rock glaciers, which manifests the characteristic permafrost creep process occurring at depth, has become an accessible and quantifiable feature with the application of remote sensing methods.金In a case study focusing on a puzzling periglacial landform near the Kunlun Pass, we addressed a long-standing issue concerning its geomorphological classification from a kinematic perspective. We employed the Interferometric Synthetic Aperture Radar (InSAR) technique to quantify the temporal and spatial variations of the downslope creeping velocities. By critically analyzing the influences that the mechanical processes imposed on the landform and piecing our observations together, we identified the landform as a debris-mantled-slope-connected rock glacier, with gelifluction processes occurring on the surface as small-scale and discrete events.To fill in the knowledge gap about rock glaciers on the plateau, we compiled a rock glacier inventory over the West Kunlun Mountains of China by manually interpreting the Google Earth imagery and the displacement maps generated by InSAR. This kinematic approach can only produce conservative mapping results due to the limitations of InSAR measurements. Therefore, we adopted semantic image segmentation powered by a deep learning model, i.e., DeepLabv3+, to infer rock glaciers that had been missed in the manual delineation. Finally, within the 124,000 km2 study area, we identified 413 rock glaciers: 290 of them were outlined by manual detection and 123 of them were mapped by deep learning. Their mean and maximum downslope velocities are 24 cm yr-1 and 127 cm yr-1, respectively.We investigated the potential water storage of the rock glaciers situated in Khumbu Valley, Nepal, by developing a velocity-constrained model to infer their ice contents. We adopted a rheological model based on adaptations of Glen's flow law and assumed a homogeneous two-layer structure for rock glaciers that consists of an ice-free active layer and an ice-rich permafrost core. The velocity constraints applied to the model were derived from InSAR measurements. The inferred ice fraction of the rock glaciers in Khumbu Valley ranges from 71.0% to 75.3%. Extrapolating from our findings in Khumbu Valley, the total amount of water stored in rock glaciers could be ~10 billion m3 over the Nepalese Himalayas.In summary, this thesis advanced the understanding of surface dynamics and classification of periglacial landforms from the perspectives of geomorphology and kinematics. We produced a rock glacier inventory using remote sensing and deep learning in a mountain range where previous investigations are scarce. We presented results of estimating rock glacier ice contents from a rheological model and surface velocity measurements. Products from this thesis and extended work will improve the knowledge of rock glaciers and the permafrost environment in the Tibetan Plateau where climatic conditions are various and fast changing.石冰川是廣泛分佈於全球山地冰緣範圍內的冰-岩混合地貌。它們是多年凍土的重要指標：多年凍土由地下溫度定義，在大多數情況下是不可見的；特別是對於青藏高原等地區，由於環境惡劣和偏遠，現場觀測的空間覆蓋範圍有限。隨著遙感方法的應用，石冰川的表面運動學特徵表現出發生在深處的多年凍土蠕變過程，並已成為可獲取和可量化的研究變量。在一個關注崑崙山埡口附近的冰緣地貌的案例研究中，我們從運動學的角度解決了一個長期存在的關於其地貌分類的問題。我們採用干涉合成孔徑雷達 (InSAR) 技術來量化沿坡向運動速度的時間和空間變化。通過綜合分析形變過程對地貌的影響並與我們的觀察結果相結合，我們將該地貌判定為岩屑-披覆-斜坡連接的石冰川，地表上發生的融凍泥流過程是小規模的離散事件。為了填補青藏高原地區石冰川的知識空缺，本文通過人工解譯谷歌地球圖像和InSAR生成的位移圖，繪製了中國西崑崙山石冰川編目。由於 InSAR 測量的局限性，這種運動學方法只能產生保守的製圖結果。因此，我們採用了由深度學習模型（即 DeepLabv3+）驅動的語義圖像分割來推斷人工描繪中遺漏的石冰川。最終，在124,000平方公里的研究區域內，我們識別了413條石冰川：其中290條通過人工檢測勾勒出輪廓，123條通過深度學習自動繪製。它們的平均和最大沿坡向運動速度分別為24 cm yr-1 和127 cm yr-1。我們開發了通過速度約束的流變模型來推斷石冰川含冰量，並進一步研究了位於尼泊爾昆布山谷的石冰川潛在蓄水量。我們採用了基於格倫流動定律改編的流變模型，並假設岩石冰川具有均勻的雙層結構，由無冰活動層和富含冰的多年凍土核心組成。我們採用InSAR測量的石冰川表面速度作為模型約束。模型推斷昆布山谷石冰川的含冰量範圍為71.0%至75.3%。根據我們在昆布山谷的發現進行推斷，尼泊爾喜馬拉雅山脈石冰川中儲存的總水量可能約為100億立方米。綜上所述，本文從地貌學和地表動力學的角度促進了對冰緣地貌的地表過程和分類的理解。 我們使用遙感和深度學習在先前對石冰川的調查很少的山脈中製作了石冰川編目。我們展示了從流變模型和表面速度測量值估算石冰川含冰量的結果。青藏高原氣候條件多樣且變化迅速，本論文和擴展工作的成果將增進對青藏高原的石冰川和多年凍土環境的理解。HU, Yan.Ph.D. Chinese University of Hong Kong 2021.Includes bibliographical references (leaves )Abstracts in English and Chinese.Title from PDF title page (viewed on ...