Accurate, up-to-date delineation of hydrographic features is essential for hydrologic modeling, water-resource management, and climate resilience. However, existing workflows for the National Hydrography Dataset (NHD) rely heavily on Digital Elevation Models (DEMs) and manual editing, limiting scalability, consistency, and the timeliness of updates across diverse landscapes. This dissertation addresses the limitations of data-driven hydrography mapping—specifically in transferability, adaptability, and physical consistency—by developing a geospatial deep learning framework for scalable and automated hydrographic delineation.
The first study addresses the limitation of transferability by leveraging transfer learning with convolutional neural networks pre-trained on ImageNet to improve streamline segmentation accuracy and generalization across heterogeneous landscapes. The second study tackles the challenge of adaptability through Model-Agnostic Meta-Learning (MAML), enabling models to rapidly adapt to new regions with minimal labeled data using Alaska’s IfSAR dataset. Building upon these advances, the final study confronts the issue of physical consistency by developing a multimodal, multitask deep learning framework that fuses DEM, SAR, optical, thermal, and learned embeddings to jointly predict streamlines and flow directions. This framework integrates complementary information sources to enforce topological coherence while enhancing scalability and robustness for continental-scale mapping.
Collectively, this research establishes an integrative, CyberGIS-enabled deep learning approach that advances the automation, transferability, and scalability of national hydrography production—supporting next-generation water-data initiatives and sustainable environmental management.