<![CDATA[EAS Seminar Series - Dr. Jemma Fadum]]>36678 Aquatic ecosystems are constantly responding to the interactive effects of both local and global stressors. There are of course numerous and diverse ways of framing and studying these ecosystem responses, all with their own unique merits and challenges. In this seminar I’ll discuss one potential framework: viewing ecosystem scale function through the lens of biogeochemistry. Centered around the impacts of sustainable aquaculture, I’ll present case studies from Lake Yojoa (Honduras) and Whycocomagh Bay (Nova Scotia) and share how layering different analytical and modeling techniques (from metagenomics to remote sensing to theoretical models) has shed light on the cryptic biogeochemical impacts of industrial aquaculture; findings which can be applied to a broader suite of anthropogenic disturbances as well. I will conclude with demonstrating how this body of research has been particularly impactful because of concerted efforts to combine both fundamental and applied research questions in order to pursue rigorous basic science while also meaningfully working with local communities.
*Refreshments: 12-12:30 PM, ES&T Atrium
]]> tbuchanan9117912014322026-10-05 11:57:1217912021802026-10-05 12:09:4000eventAquatic ecosystems are constantly responding to the interactive effects of both local and global stressors. There are of course numerous and diverse ways of framing and studying these ecosystem responses, all with their own unique merits and challenges. In this seminar I’ll discuss one potential framework: viewing ecosystem scale function through the lens of biogeochemistry. Centered around the impacts of sustainable aquaculture, I’ll present case studies from Lake Yojoa (Honduras) and Whycocomagh Bay (Nova Scotia) and share how layering different analytical and modeling techniques (from metagenomics to remote sensing to theoretical models) has shed light on the cryptic biogeochemical impacts of industrial aquaculture; findings which can be applied to a broader suite of anthropogenic disturbances as well. I will conclude with demonstrating how this body of research has been particularly impactful because of concerted efforts to combine both fundamental and applied research questions in order to pursue rigorous basic science while also meaningfully working with local communities.
*Refreshments: 12-12:30 PM, ES&T Atrium
]]>2026-11-19T11:00:00-05:002026-11-19T12:00:00-05:002026-11-19T12:00:00-05:002026-11-19 16:00:002026-11-19 17:00:002026-11-19 17:00:002026-11-19T11:00:00-05:002026-11-19T12:00:00-05:00America/New_YorkAmerica/New_Yorkdatetime2026-11-19 11:00:002026-11-19 12:00:00America/New_YorkAmerica/New_Yorkdatetime<![CDATA[]]>681348681348image<![CDATA[Fadum Headshot]]>image/png17912021122026-10-05 12:08:3217912021122026-10-05 12:08:32<![CDATA[]]><![CDATA[EAS Seminar Series - Dr. Douglas Brinkerhoff]]>36678 Machine-learning surrogates for ice dynamics are often motivated by thebhigh computational cost of traditional flow models. Here, I describe a somewhat unexpected outcome of pursuing that approach: attempting to buildneural-network emulators ultimately led back to classical numericalmethods—albeit with several ideas borrowed from modern machine learning.Initial work focused on graph-based neural operators and diffusion-style architectures designed to emulate shallow-shelf ice dynamics. These models emphasized properties that are also desirable in physical solvers: rotational and scale invariance, discretization independence, and information transport mechanisms that do not rely on fixed receptive fields. While such architectures produced promising emulators, the design process revealed strong parallels with well-established numerical techniques.These insights ultimately motivated the development of Glide, a GPU-accelerated ice-dynamics solver built around nonlinear geometric multigrid and discretization-invariant transport operators. Many structural features of modern neural architectures have direct numerical analogues:multigrid V-cycles resemble U-Net hierarchies, restriction and prolongationcorrespond to pooling and upsampling, and stencil-based flux operators actas physics-consistent convolutional layers. Unlike neural emulators, however, this approach preserves exact physical constraints, delivers predictable convergence properties, and is amenable to implicit time-stepping schemes. As with PINNs and surrogates, the model remains differentiable for inverse problems through adjoint methods, while being less expensive to differentiate through time because of the lack of hidden state.The resulting solver achieves orders-of-magnitude performance gains through GPU-native implementation while retaining the robustness of classical PDE methods. More broadly, this work suggests that the interaction between machine learning and scientific computing need not always produce neural surrogates—sometimes the most productive outcome is a re-examination of classical algorithms through the lens of modern ML architectures.
*Refreshments: 12-12:30 PM, ES&T Atrium
]]> tbuchanan9117859283282026-08-05 11:12:0817864575752026-08-11 14:12:5500eventMachine-learning surrogates for ice dynamics are often motivated by thebhigh computational cost of traditional flow models. Here, I describe a somewhat unexpected outcome of pursuing that approach: attempting to buildneural-network emulators ultimately led back to classical numericalmethods—albeit with several ideas borrowed from modern machine learning.Initial work focused on graph-based neural operators and diffusion-style architectures designed to emulate shallow-shelf ice dynamics. These models emphasized properties that are also desirable in physical solvers: rotational and scale invariance, discretization independence, and information transport mechanisms that do not rely on fixed receptive fields. While such architectures produced promising emulators, the design process revealed strong parallels with well-established numerical techniques.These insights ultimately motivated the development of Glide, a GPU-accelerated ice-dynamics solver built around nonlinear geometric multigrid and discretization-invariant transport operators. Many structural features of modern neural architectures have direct numerical analogues:multigrid V-cycles resemble U-Net hierarchies, restriction and prolongationcorrespond to pooling and upsampling, and stencil-based flux operators actas physics-consistent convolutional layers. Unlike neural emulators, however, this approach preserves exact physical constraints, delivers predictable convergence properties, and is amenable to implicit time-stepping schemes. As with PINNs and surrogates, the model remains differentiable for inverse problems through adjoint methods, while being less expensive to differentiate through time because of the lack of hidden state.The resulting solver achieves orders-of-magnitude performance gains through GPU-native implementation while retaining the robustness of classical PDE methods. More broadly, this work suggests that the interaction between machine learning and scientific computing need not always produce neural surrogates—sometimes the most productive outcome is a re-examination of classical algorithms through the lens of modern ML architectures.