Jie Bao

and 8 more

Continuous measurement and monitoring of river or creek surface water coverage is crucial for studying the exchange fluxes between the surface and subsurface water. These fluxes directly impact carbon and nitrogen exchange and cycles, which are related to organic matter transport and reactions. While satellite and related techniques have been widely used for large-scale monitoring, they may not be accurate, sensitive, or cost-efficient for monitoring and tracking of surface water at fine-scale spatial (i.e., sub-meter) and temporal (i.e., daily) variations. This is especially true for small creeks with large plant canopy coverage. On-site in-situ sensors monitoring methods primarily yield point data, often insufficient in capturing the entire spatial distribution. Wildlife cameras have proven a cost-efficient way to continuously monitor surface water coverage of rivers and creeks. To efficiently analyze the images and/or videos from the wildlife cameras, in this study, two machine learning approaches, YOLOv8 and Mask2Former, have been applied. Both models were trained by images obtained from the public dataset ADE20k along with a small dataset from wildlife camera photos collected at the current study area. Once surface water coverage is segmented, the width of the surface water in real world can be approximated according to the wildlife camera, lens, and positioning parameters. In this study, surface water was detected and monitored by applying the proposed approaches for the six wildlife cameras in the Yakima River Basin in 2023 to 2024 in United States of America. Though Mask2Former model provides slightly better transferability, both models can accurately capture the surface water from the wildlife cameras, which are installed in significantly different environments, such as the different brightness, contrast, and varying front scene object blockages. The proposed approach enables long-term continuous monitoring and quantification of river intermittency and water availability with high accuracy and low-cost, which will benefit river ecosystem research and management.

Yunxiang Chen

and 16 more

Streambed grain sizes and hydro-biogeochemistry (HBGC) control river functions. However, measuring their quantities, distributions, and uncertainties is challenging due to the diversity and heterogeneity of natural streams. This work presents a photo-driven, artificial intelligence (AI)-enabled, and theory-based workflow for extracting the quantities, distributions, and uncertainties of streambed grain sizes and HBGC parameters from photos. Specifically, we first trained You Only Look Once (YOLO), an object detection AI, using 11,977 grain labels from 36 photos collected from 9 different stream environments. We demonstrated its accuracy with a coefficient of determination of 0.98, a Nash–Sutcliffe efficiency of 0.98, and a mean absolute relative error of 6.65% in predicting the median grain size of 20 testing photos. The AI is then used to extract the grain size distributions and determine their characteristic grain sizes, including the 5th, 50th, and 84th percentiles, for 1,999 photos taken at 66 sites. With these percentiles, the quantities, distributions, and uncertainties of HBGC parameters are further derived using existing empirical formulas and our new uncertainty equations. From the data, the median grain size and HBGC parameters, including Manning’s coefficient, Darcy-Weisbach friction factor, interstitial velocity magnitude, and nitrate uptake velocity, are found to follow log-normal, normal, positively skewed, near log-normal, and negatively skewed distributions, respectively. Their most likely values are 6.63 cm, 0.0339 s·m-1/3, 0.18, 0.07 m/day, and 1.2 m/day, respectively. While their average uncertainty is 7.33%, 1.85%, 15.65%, 24.06%, and 13.88%, respectively. Major uncertainty sources in grain sizes and their subsequent impact on HBGC are further studied.

Yunxiang Chen

and 10 more

Quantifying the multiscale feedback between hydrodynamics and biogeochemistry is key to reliable modeling of river corridor systems. However, accurate and efficient hydrodynamics models over large spatiotemporal scales have not yet been established due to limited surveys of riverbed roughness and high computational costs. This work presents a semi-automated workflow that combines topographic and water stage surveys, computational fluid dynamics modeling, distributed wall resistance modeling, and high-performance computing to simulate flow in a 30-kilometer-long reach at the Columbia River during 2011-2019. The results show that this workflow enables a high accuracy in modeling water stage at all seven survey locations during calibration (1 month) and validation (65 months) periods. It also enables a high computational efficiency to model the streamflow during a 58-month solution-time within less than a 6-day wall-clock-time with mesh number, time step, and CPU hours of about 1.2 million, 3 seconds, and 1.1 million hours, respectively. Using the well-validated results, we show that riverbed dynamic pressure is randomly distributed over all spatiotemporal scales with its cross-sectional average values approximately quantified by a normal distribution with a mean and standard deviation of -0.353 m and 0.0352 m; bed shear stress is affected by flowrate and large- and small-scale topographic features with cross-sectional maximum values following a smooth but asymmetric distribution with 90% of its value falling between 5 Pa and 35 Pa; and hydrostatic pressure is influenced by flowrate and large-scale topographic features with cross-sectional maximum values quantified by a discontinuous and skewed distribution determined by streamwise topographic variations.