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
Companies are making significant moves towards the clean energy transition in various ways. In order to reduce carbon emissions, certain companies are investing in several rescue methods in 2024. Renewables production is growing over time as the…
Amazon is one of the leading global e-commerce platforms for consumers. The brand boosts its online shopping experience for users in multiple ways. Surveys show that Amazon regulates several business solutions to increase its products and services…
Globally, countries are facing challenges in the transformation of green energy. The economies worldwide target to achieve net zero emissions. Hence, countries are introducing several ways to reduce environmental carbon emissions. Developing…
China is the leading nation in the energy market. Since the clean energy transition focuses on renewables, China surpasses other nations in renewables production. In a mission to strengthen its energy security, China moves forward with oil and gas…
There will be a surge in copper demand in the energy market in 2024. The growing demand for copper contributes to different usages in the energy market. China stands out in the copper exports of 2024. The Chinese economy is booming, with an…
Team Lead - Research