Project period: 9 25 23-9 14 26PROJECT title: harnessing physics-informed machine learning to improve image-based streamflow measurementsmeasurement of flow in rivers and streams across the nation…

The U.S. Geological Survey (USGS) awarded a $300,024 Cooperative Agreement grant under the USGS Research and Data Collection program (CFDA 15.808) to The Pennsylvania State University (Penn State) for the project "Harnessing Physics-Informed Machine Learning to Improve Image-Based Streamflow Measurements". The 3-year project aims to leverage machine learning and artificial intelligence to enhance two image-based flow measurement techniques - space-time image velocimetry (STIV) and large scale particle image velocimetry (LSPIV). The project will develop open-source Python software packages to implement these improved flow measurement algorithms, leveraging data collected from field measurements, flume experiments, and computer simulations. The deliverables include the software tools, technical reports, and peer-reviewed publications that will be shared with the research community through conferences and workshops for end-user training.

Project period: 9 25 23-9 14 26PROJECT title: harnessing physics-informed machine learning to improve image-based streamflow measurementsmeasurement of flow in rivers and streams across the nation is of great importance for the proper management of water resources and flood damage mitigation. New measurement technologies, such as those based on images and videos, have advanced to a level that they are ready to be adopted. However, there are still significant uncertainties and errors in the produced measurement data. This project seeks to use the latest advancement in machine learning (ML) and artificial intelligence (AI) to improve these image-based flow measurement techniques. In particular, the physics-informed ML will be utilized where the measurement technologies are guided by the fundamental physics of fluid mechanics. There are two main goals in this project. The first is to leverage ML to automate and enhance the space-time image velocimetry (stiv). The second is to utilize physics-informed ML to augment and improve large scale particle image velocimetry (lspiv). Open-source python packages will be developed to implement the processing algorithms for these flow measurement techniques. To enable the ML training and validation, data will be collected from field measurement, flume experiments, and computer model simulations. The deliverables include the software packages, dataset generated, technical reports, and peer-reviewed journal publications. The results and products from this project will be shared with the research community through conference presentations. In addition, at the conclusion of this project, a training workshop for the public will be offered at a national conference to teach the end users on how to use the produced tools. The project team will also help USGS staff to teach their own short courses on non-intrusive measurement techniques. This project also provides excellent opportunities for student training. One graduate student and one undergraduate student will be supported by this project to work on various aspects of the tasks. They will not only gain direct knowledge and experiences on the subject, but also can be competent future workforce for USGS and other agencies.