When
Where
Location Change!
Harvill 107 - 25 Mar 2024: Apologies to all for the late notice, but this week's talk has been canceled.
Seminar Format
Available in-person and via Zoom webinar. Contact the department to subscribe to the email list (zoom link provided in announcement).
Abstract
Understanding the heterogeneous characteristics of aquifers is essential for effective groundwater management and protection. The hydrogeology community has developed hydraulic tomography as a method to collect aquifer response data from various pumping tests and integrate the collected data into an inverse model for heterogeneity characterization.
In addressing karst hazards in Kentucky, we implemented a U-Net image segmentation model, a specific type of deep learning convolutional neural network (CNN), to identify sinkholes from high-resolution elevation data. CNNs are widely applied in many scientific disciplines, providing a robust solution for various problems.
I will begin with an overview of hydraulic tomography, elucidating the utilization of an SSLE inverse model to process collected data for hydraulic parameter estimation. Subsequently, I will describe the development, training, and testing of the U-Net model. Finally, I will explore the interconnections between hydraulic tomography and deep neural networks.
Bio
Dr. Junfeng Zhu is a senior hydrogeologist with Kentucky Geological Survey, University of Kentucky. He is also an adjunct faculty member in the Department of Earth and Environmental Sciences, University of Kentucky. His research focuses on aquifer characterization and modeling, karst hydrogeology, geohazard, hydrogeophysics, and machine learning in hydrology. He received his B.S. and M.S. degrees in Hydrogeology and Engineering Geology from Nanjing University, China in 1995 and 1998. In 2001, he attended the University of Arizona’s Hydrology and Water Resources program under Dr. Yeh and received his PhD degree in Hydrology in 2005.
Junfeng Zhu Email | Website | Google Scholar