BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Memento EPFL// BEGIN:VEVENT SUMMARY:EESS talk on "How to beat your teachers in hydrologic machine lear ning" DTSTART:20211026T121500 DTEND:20211026T131500 DTSTAMP:20260428T033701Z UID:b763f5fec5a95059d42e98355a00672adb0f47707566de886f49db54 CATEGORIES:Conferences - Seminars DESCRIPTION:Dr Chaopeng Shen\, Associate Professor\, visiting professor at EAWAG\, Civil and Environmental Engineering\, College of Engineering\, Pe nnsylvania State University (US)\nAbstract:\nDeep learning (DL) models tra ined on hydrologic observations are recently shown to be highly performant . Used directly\, however\, they inherit certain flaws of their supervisin g data. In other words\, these models are students that cannot exceed thei r teachers (supervising data). For example\, satellite data has global cov erage but low resolution/accuracy\, while in-situ networks are spatially i mbalanced. For another example\, we cannot predict a variable at large sca les if we do not have extensive observations for it and observational nois e propagates into the trained models. While some have shown that adding ph ysical constraints could be beneficial\, the benefit has so far been limit ed to minor-to-modest gains in performance. Here we explore several pathwa ys for machine learning models to exceed their teachers. First we explore learning from multiple data sources (soil moisture) at different scales\, creating a multiscale forecast model that breaks the confines of the indiv idual supervising dataset. Second we demonstrate how we connect machine le arning with physics-based models to predict unobserved variables that help determine future trends of the water cycle. Third we show how network mod els can be leveraged to learn physics rather than purely making prediction s. Overall\, there are substantial new paths to take for hydrology to bene fit from big data machine learning apart from elevating the prediction acc uracy.\n\nShort biography:\nChaopeng Shen is Associate Professor in Civil Engineering at The Pennsylvania State University and is currently visiting Swiss Federal Institute of Aquatic Science and Technology (EAWAG) as an a cademic guest. He received the Ph.D. degree in environmental engineering f rom Michigan State University\, East Lansing\, MI\, USA\, in 2009. His PhD research focused on computational hydrology and he developed the hydrolog ic model Process-based Adaptive Watershed Simulator(PAWS)\, which was late r coupled to the community land model to study the interactions between hy drology and ecosystem. His recent efforts focused on harnessing the big da ta and machine learning opportunities in advancing hydrologic predictions and connecting physics with machine learning. He has written technical\, e ditorial\, review and collective opinion papers on hydrologic deep learnin g to call to attention the emerging opportunities for scientific advances. In addition\, his research interests also include floodplain systems\, sc aling issues\, process-based hydrologic modeling\, and hydrologic data min ing. He is currently an Associate Editor of the Water Resources Research a nd also an Associate Editor in Frontiers in AI. LOCATION:https://epfl.zoom.us/j/63900222242?pwd=OXluejhzTklCbkdWakkvaUFCSG Vndz09 STATUS:CONFIRMED END:VEVENT END:VCALENDAR