
Deep Learning Estimates Watershed Subsurface Permeability using Stream Discharge
By Aarti Dhapte , 21 November, 2022
Scientists from Pacific Northwest National Laboratory, Oak Ridge National Laboratory, and Los Alamos National Laboratory trained deep neural networks (DNNs) to estimate subsurface permeability from stream discharge hydrographs in a study published in Frontiers in Earth Science.
In watershed models, subsurface permeability is a critical parameter that determines the contribution of subsurface flow to stream flows. It is difficult and expensive to directly measure permeability at the spatial extent and resolution required by the watershed models. Researchers often use inverse modeling to evaluate permeability. Compared to groundwater monitoring data, the wide availability of stream surface flow data provides a novel data source for integrated surface and subsurface hydrologic models to infer soil and geologic features.
First, they trained the DNNs to map the correlations between soil and geologic layer permeabilities and simulated stream discharge from the examined watershed's integrated surface-subsurface hydrologic model. The DNNs outperformed the classic inverse modeling method in terms of permeability estimates. Using observed stream discharge from the research site, the DNNs assessed the permeability of a natural watershed (Rock Creek Catchment in the headwaters of the Colorado River). The watershed model accurately predicted the stream flows, with permeability calculated by DNNs. This study highlights the importance of upcoming deep learning methods for assisting with integrated watershed modeling by enhancing parameter estimation, which will eventually reduce uncertainty in predictive watershed models.
The permeability of subsurface rocks and soils measures how well liquids flow through them. It is essential in determining subsurface flow and transport mechanisms in watersheds. However, measuring permeability directly at the size and precision required by watershed models is challenging and expensive. Stream flow monitoring data, on the other hand, is publicly available. The connections between permeability and stream flow open a new avenue for evaluating subsurface permeability. Deep learning, a sort of artificial intelligence, was used in this study. Deep understanding assesses a watershed's subsurface permeability from stream discharge data more precisely than standard methods. This enhancement will aid in calibrating watershed models and minimize the uncertainty in stream discharge prediction.
Traditional Machine Learning and Deep Learning Performance Comparison

Latest News

Leading technology giants Amazon, Google, and Meta have teamed up with other big energy users in a historic action at the CERAWeek 2025 energy conference in Houston to promote an ambitious initiative: doubling world nuclear power capacity by 2025…

China's refining sector showed resilience and adaptability in the first months of 2025, raising crude oil throughput by 2.1 percent over the year before. According to the National Bureau of Statistics, this increase resulted in a processing volume of…

In March 2025, Elsevier launched ScienceDirect AI, a world-renowned leader in science and health information science. This state-of-the-art generative AI tool can potentially transform the industry with its revolutionary advances in research. This…

The global electric vehicle (EV) industry has reached a pivotal milestone. The average price of lithium-ion battery packs has dropped below the USD 100 per kilowatt-hour (kWh) threshold, a benchmark long considered critical for achieving price parity…

In 2025, China will focus on self-sufficiency in high-tech industries by further integrating domestically produced semiconductor chips into the country\'s electric vehicle (EV) industry. By doing so, China aims to reduce reliance on foreign…

Team Lead - Research
Latest News




