Accurate maps of regional surface water features are integral for advancing ecologic, atmospheric and land development studies. The only comprehensive surface water feature map of Alaska is the National Hydrography Dataset (NHD). NHD features are often digitized representations of historic topographic map blue lines and may be outdated. Here we test deep learning methods to automatically extract surface water features from airborne interferometric synthetic aperture radar (IfSAR) data to update and validate Alaska hydrographic databases. U-net artificial neural networks (ANN) and high-performance computing (HPC) are used for supervised hydrographic feature extraction within a study area comprised of 50 contiguous watersheds in Alaska. Surface water features derived from elevation through automated flow-routing and manual editing are used as training data. Model extensibility is tested with a series of 16 U-net models trained with increasing percentages of the study area, from about 3 to 35 percent. Hydrography is predicted by each of the models for all watersheds not used in training. Input raster layers are derived from digital terrain models, digital surface models, and intensity images from the IfSAR data. Results indicate about 15 percent of the study area is required to optimally train the ANN to extract hydrography when F1-scores for tested watersheds average between 66 and 68. Little benefit is gained by training beyond 15 percent of the study area. Fully connected hydrographic networks are generated for the U-net predictions using a novel approach that constrains a D-8 flow-routing approach to follow U-net predictions. This work demonstrates the ability of deep learning to derive surface water feature maps from complex terrain over a broad area.

Accurate maps of built structures around stream channels, such as dams, culverts, and bridges, are vital in monitoring infrastructure, risk management, and hydrologic modeling. Hydrologic modeling is essential for research and decisionmaking related to infrastructure and development planning, emergency management, ecology, and developing hydrographic data. Technological advances in remote sensing afford increasingly fine-scale elevation data, such as the U.S. Geological Survey 1-meter digital elevation models (DEMs), that can accurately model the Earth's surface characteristics and related hydrologic dynamics. A long-standing challenge in flow modeling is the presence of built structures in an elevation model that resist flow in a way that does not reflect actual dynamics, such as culverts, bridges, and dams (Figure 1a). This challenge is exacerbated in fine-scale elevation data as more built structures are resolved. Here we present a test of the extensibility of a culvert and dam detection workflow, culvert-net (CN). CN was developed using a large dataset of field-validated culverts, bridges, and dam locations for Alexander County, North Carolina, USA, supplemented by manual review and identification of additional features. In this workflow, the CN model is tested on a new study area in western Michigan, USA, where culverts and associated hydrography have recently been manually compiled. A. B.

This study demonstrates that flow resistance in rivers, quantified via calibration of a modified Manning's equation, can inform its estimation from channel hydrometric information available from regional or global datasets for a limited set ( N = 14) of rigorously calibrated river reaches in Alaska and with less certainty in a large set ( N = 754) of river reaches across the United States developed from U.S. Geological Survey (USGS) Acoustic Doppler Current Profiler (ADCP) discharge measurements. The channel morphometry datasets tested include the National Aeronautics and Space Administration (NASA) Surface Water and Ocean Topography (SWOT) River Database (SWORD) and U.S. Geological Survey (USGS) Scale Specific Sinuosity (S3) river channel morphometry metrics. Multiple regression prediction models of the baseline (near bankfull) Manning's n flow resistance coefficient were developed from the SWORD and S3 reach data. Significant predictive variables included but were not limited to water‐surface slope, water‐surface variation (width and elevation), the number of active channels, maximum width, sinuosity and along channel meander length. The study points to the possibility of global a priori estimation of flow resistance in rivers for application in remote sensing of river discharge from river morphology, water‐surface width, slope and height measurements made by SWOT and other satellites.