Date of Award

1-2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

College/School

College of Science and Mathematics

Department/Program

Earth and Environmental Studies

Thesis Sponsor/Dissertation Chair/Project Chair

Mark Chopping

Committee Member

Stefan Robila

Committee Member

Aparna Varde

Committee Member

Bruce Cook

Abstract

Arctic shrub expansion threatens to accelerate permafrost thaw through complex feedbacks, yet whether shrubs primarily indicate or drive degradation remains unresolved. This dissertation integrates deep learning analysis of two decades of satellite imagery with LiDAR canopy structure and radar soil moisture data to reveal that shrubs play a dual role: young, expanding shrubs signal active permafrost thaw, while mature, tall shrubs stabilize underlying permafrost through insulation. By demonstrating that vertical canopy structure predicts thaw depth better than cover extent alone, this work establishes a scalable framework for monitoring permafrost vulnerability across the rapidly changing Arctic. Arctic shrub expansion is accelerating alongside permafrost degradation; nevertheless, the mechanisms driving this transformation and its feedbacks with permafrost dynamics remain poorly understood at fine spatial scales. This dissertation develops and applies deep learning approaches to map decadal changes in Arctic vegetation and surface hydrology using very high-resolution QuickBird and WorldView satellite imagery (2002–2022) across Alaska's North Slope, integrating these observations with airborne LiDAR canopy structure data and radar-derived soil moisture to examine vegetation-permafrost interactions. Five deep learning architectures (CNN, ResNet, VGG, U-Net, Vision Transformer) were trained on pansharpened multispectral images and validated against field reference data. ResNet and VGG achieved the highest performance for shrub mapping (81–86% accuracy, 51–59% F1 scores, R² = .78–.81 with field data), while U-Net provided the most balanced performance across wet tundra and surface water classes. Model performance declined systematically with increasing shrub density (r = −.85 for accuracy), reflecting greater difficulty distinguishing dense canopies from surrounding vegetation. Continuous probability distributions revealed significant shrub cover increases 0.04–0.89% yr-1(p < .001), wet tundra declines 0.10–0.50% yr-1, and surface water decreases 0.01–0.09% yr-1. Shrub expansion occurred primarily on dry, bare tundra with optimal soil moisture, while moisture-driven retreat was observed along streams and riverbanks. The inverse relationship between shrub expansion and wet tundra decline (R² = .36, r = −.60, p < .001) indicates vegetation redistribution rather than uniform greening. Small and medium lakes (< 10 ha) declined moderately (0.1–0.4% yr−1), while large lakes showed slight expansion, demonstrating size-dependent hydrological responses. Integration with UAVSAR soil moisture data revealed scale-dependent relationships between shrubs and permafrost. Across all sites, shrub cover correlated positively with active-layer thickness (ALT) (R² = .59, r = .77, p < .01), reflecting regional climate gradients. However, within cold North Slope sites, this relationship reversed (R² = .54, r = −.73, p < .01), indicating that established shrubs associate with shallower thaw depths. Weighted multiple regression controlling for soil moisture showed shrub cover was negatively associated with ALT (β = −0.14 m per unit cover, p < .001), demonstrating that mature shrub communities stabilize permafrost through insulation when hydrological conditions are constant. NASA G-LiHT LiDAR-derived canopy height explained substantially more variance in permafrost indicators than horizontal cover alone (R² = .76 vs. .59 for ALT). Taller shrubs (> 1.5 m) are associated with shallower active layers. In comparison, shorter shrubs are associated with deeper thaw, suggesting that vertical structure captures mechanisms of vegetation-permafrost interaction that are not evident from cover extent alone. These findings demonstrate that Arctic shrubs function simultaneously as indicators of past permafrost degradation (expanding into recently thawed terrain) and drivers of future stability (insulating permafrost through mature canopies), with the dominant role determined by community age and regional climate context. Integrating vegetation structure metrics with cover estimates offers a scalable framework for monitoring permafrost vulnerability and projecting carbon-climate feedbacks across the circumpolar Arctic. However, sensor calibration challenges and limited temporal depth of very high-resolution imagery remain essential constraints.

File Format

PDF

Download

Share

COinS