Machine Learning Analysis of Hydrologic Exchange Flows and Transit Time Distributions in a Large Regulated River

Abstract

Hydrologic exchange between river channels and adjacent subsurface environments is a key process that influences water quality and ecosystem function in river corridors. High-resolution numerical models were often used to resolve the spatial and temporal variations of exchange flows, which are computationally expensive. In this study, we adopt Random Forest (RF) and Extreme Gradient Boosting (XGB) approaches for deriving reduced order models of hydrologic exchange flows and associated transit time distributions, with integrated field observations (e.g., bathymetry) and hydrodynamic simulation data (e.g., river velocity, depth). The setup allows an improved understanding of the influences of various physical, spatial, and temporal factors on the hydrologic exchange flows and transit times. The predictors also contain those derived using hybrid clustering, leveraging our previous work on river corridor system hydromorphic classification. The machine learning-based predictive models are developed and validated along the Columbia River Corridor, and the results show that the top parameters are the thickness of the top geological formation layer, the flow regime, river velocity, and river depth; the RF and XGB models can achieve 70% to 80% accuracy and therefore are effective alternatives to the computational demanding numerical models of exchange flows and transit time distributions. Each machine learning model with its favorable configuration and setup have been evaluated. The transferability of the models to other river reaches and larger scales, which mostly depends on data availability, is also discussed.

Combined Effects of Stream Hydrology and Land Use on Basin‐Scale Hyporheic Zone Denitrification in the Columbia River Basin

“…In some cases, both approaches can be combined to gain further insight and predictability. For example, the model can be used to reveal the dominant process or features through variable importance analysis (Ren et al, 2020(Ren et al, , 2021Ward et al, 2022).…”

“…3 of 22 While physically based numerical models can represent explicit mechanisms and simulate HZ denitrification at varying spatial and temporal scales, these models are computationally expensive (Ren et al, 2021) and require various data sources for model calibration (Chen et al, 2021). As an alternative, machine learning approaches show high performance with limited data and capture complex relationships between inputs and outputs (Mori et al, 2019).…”

From legacy contamination to watershed systems science: a review of scientific insights and technologies developed through DOE-supported research in water and energy security

“…Zonation is also powerful for understanding the co-variability of above-and belowground processes, as well as identifying relevant field experimental sites throughout the domain. For example, research at the Hanford Site is using river channel morphology zonation (hydromorphic units, Ren et al (2021)) as a framework for a transferable understanding of river-groundwater exchange fluxes and transit times.…”

Big Data Analytics for Long-Term Meteorological Observations at Hanford Site

“…It builds one tree at a time, with each tree learning from and improving upon the previous one by minimizing the error. Ren et al [31] compared the RF and GBM model performance, and found that GBM has a better performance than RF. The GBM model had a better predictive capability than RF models in genomic selection.…”