BPBrian Stephen Powell

Brian Stephen Powell

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Brian Stephen Powell has a total of 50 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Physics‐Informed Neural Networks as Differentiable Surrogates for 4D‐Variational Data Assimilation

    by Kevin Egan, Brian Powell

    Biogeochemical forecasting requires state estimation methods that accommodate sparse observations, nonlinear dynamics, and discontinuous ecological processes. Four‐dimensional variational data assimilation addresses this challenge but depends on tangent‐linear and adjoint models that are costly to derive and maintain alongside evolving forward codes, and that are not defined where biogeochemical formulations are non‐differentiable. We propose a framework in which a physics‐informed neural network (PINN) replaces the traditional numerical model and its linearizations, remaining differentiable through automatic differentiation while embedding simplified governing equations as structural guardrails and computing the forward, tangent‐linear, and adjoint operators simultaneously from a single computational graph. Applied to nitrate‐phytoplankton‐zooplankton dynamics as a controlled testbed, our approach achieves an approximately six‐fold speedup over traditional numerical models in twin experiments while maintaining comparable state estimation accuracy across observation sampling scenarios. Jacobian analysis shows that the learned sensitivities are consistent with the embedded ecological relationships where these agree with the data, while departing where simplified formulations prove inadequate. This framework demonstrates that PINN‐based surrogates can maintain dynamical consistency in variational data assimilation while reducing implementation burden, offering a path toward more flexible and maintainable forecasting systems for increasingly complex biogeochemical models.

  • Open Access

    Physics-Informed Neural Networks as Differentiable Surrogates for 4D-Variational Data Assimilation

    by Kevin Egan, Brian Powell

    Biogeochemical forecasting requires state estimation methods that accommodate sparse observations, nonlinear dynamics, and discontinuous ecological processes. Four-dimensional variational data assimilation (4D-Var) addresses this challenge but depends on tangent-linear and adjoint models that are costly to derive and maintain alongside evolving forward codes, and that are not defined where biogeochemical formulations are non-differentiable. We propose a framework in which a physics-informed neural network (PINN) replaces the traditional numerical model and its linearizations, remaining differentiable through automatic differentiation while embedding simplified governing equations as structural guardrails and computing the forward, tangent-linear, and adjoint operators simultaneously from a single computational graph. Applied to nitrate-phytoplankton-zooplankton (NPZ) dynamics as a controlled testbed, our approach achieves an approximately six-fold speedup over traditional numerical models in twin experiments while maintaining comparable state estimation accuracy across observation sampling scenarios. Jacobian analysis shows that the learned sensitivities are consistent with the embedded ecological relationships where these agree with the data, while departing where simplified formulations prove inadequate. This framework demonstrates that PINN-based surrogates can maintain dynamical consistency in variational data assimilation while reducing implementation burden, offering a path toward more flexible and maintainable forecasting systems for increasingly complex biogeochemical models.

  • Open Access

    Physics-Informed Neural Networks as Differentiable Surrogates for 4D-Variational Data Assimilation

    by Kevin Egan, Brian Powell

    Biogeochemical forecasting requires state estimation methods that accommodate sparse observations, nonlinear dynamics, and discontinuous ecological processes. Four-dimensional variational data assimilation (4D-Var) addresses this challenge but depends on tangentlinear and adjoint models that are costly to derive and maintain alongside evolving forward codes, and that often fail for the non-differentiable formulations common in biogeochemical systems. We propose a framework in which a physics-informed neural network (PINN) replaces the traditional numerical model and its linearizations, remaining differentiable through automatic differentiation while embedding simplified governing equations as structural guardrails and computing all three operators simultaneously from a single computational graph. Applied to nitrate-phytoplankton-zooplankton (NPZ) dynamics as a controlled testbed, our approach achieves a 6.5-fold speedup over traditional numerical models in twin experiments while maintaining equivalent state estimation accuracy across observation sampling scenarios. Jacobian analysis reveals how the network reconciles embedded structural constraints with training data, highlighting that the PINN preserves ecologically meaningful relationships while adapting where simplified formulations prove inadequate. This framework demonstrates that PINN-based surrogates can maintain dynamical consistency in variational data assimilation while reducing implementation burden, offering a path toward more flexible and maintainable forecasting systems for increasingly complex biogeochemical models.

  • Open Access

    Data Driven Reconstruction of Upper Ocean Profiles for Improved State Estimation in the Philippine Sea

    by Guangpeng Liu, Brian Powell

    Despite significant advances in observational systems such as the global Argo array of autonomous profiling floats, the spatiotemporal coverage of subsurface ocean observations remains limited compared to the dense data provided by satellite platforms. This study develops a data‐driven framework to reconstruct synthetic profiles of upper ocean temperature and salinity by training a self‐attention‐based neural network with satellite‐derived sea surface height (SSH) and sea surface temperature anomalies, using 17 years of collocated Argo float measurements. Daily synthetic profiles for the upper 650 m of the Philippine Sea were generated for the entirety of 2010 and assimilated into a regional ocean model via 4D‐Var data assimilation. Results show overall improved effectiveness of state estimation when synthetic profiles are utilized. Diagnostic variables like temperature, salinity, SSH, horizontal velocity all show improvement. Synthetic profiles of subsurface temperature had an overall positive impact on SSH analysis and forecast, especially in regions east of the Luzon Strait and the southern domain influenced by the North Equatorial Current. In the vertical range of 150–600 m, the impact of synthetic profiles on various observations was promising, leading to substantial reductions in the analysis and forecast error.

  • Patterns of Ocean Acidification Emergence in the Hawaiian Islands Using Dynamically Downscaled Projections

    by L. Hošeková, T. Friedrich, B. S. Powell, C. Sabine

    This study presents the first dynamically downscaled projections of ocean acidification (OA) for the Main Hawaiian Islands using coupled Regional Ocean Modeling System and Carbon, Ocean Biogeochemistry, and Lower Trophics models integrated with Coupled Model Intercomparison Project Phase 6 (CMIP6) outputs from the Community Earth System Model 2. We analyze three Shared Socioeconomic Pathways (SSP1‐2.6, SSP2‐4.5, and SSP3‐7.0) and introduce a climate novelty metric to assess the extent to which future OA conditions exceed historical variability by comparing the magnitude of projected changes to past variability. Our results indicate unprecedented levels of OA within the next three decades across all scenarios, with aragonite saturation state (), pH, and substrate‐to‐inhibitor ratio (bicarbonate to free hydrogen ions [HCO 3 − ]/[H + ]) projected to decline significantly. By 2100, under SSP3‐7.0, novelty could exceed reference variability by a factor of 12. Spatial analysis reveals heterogeneous OA impacts, with windward coastlines consistently exhibiting higher novelty levels. Importantly, we find contrasting spatial patterns of OA indices due to varying sensitivities to temperature and dissolved inorganic carbon, resulting in higher novelty in northern areas and higher pH and substrate‐to‐inhibitor ratio novelty in southern regions.


  • Simulating Vibrio vulnificus in the Ala Wai Canal using a coupled microbial-hydrodynamic numerical model

    by Emma S. Nuss, Brian Powell, Conor Jerolmon, Olivia D. Nigro, Andrian P. Gajigan, Shaun Giancaterino, Grieg F. Steward


  • Open Access

    Patterns of ocean acidification emergence in the Hawaiian Islands using dynamically downscaled projections

    by Lucia Hosekova, Tobias Friedrich, Brian Stephen Powell, Christopher L. Sabine

    This study presents the first dynamically downscaled projections of ocean acidification (OA) for the main Hawaiian Islands using coupled Regional Ocean Modeling System (ROMS) and Carbon, Ocean Biogeochemistry and Lower Trophics (COBALT) models integrated with Coupled Model Intercomparison Project Phase 6 (CMIP6) scenario outputs from the Community Earth System Model 2 (CESM2). We analyze three Shared Socioeconomic Pathways (SSP1-2.6, SSP2-4.5, and SSP3-7.0) and introduce a novelty metric to quantify departure from historical variability. Our results indicate unprecedented levels of OA within the next three decades across all scenarios, with aragonite saturation state (ΩA), pH, and substrate-to-inhibitor ratio projected to decline significantly. By 2100, under SSP3-7.0, ΩA novelty could exceed reference variability by a factor of 12. Spatial analysis reveals heterogeneous OA impacts, with windward coastlines consistently exhibiting higher novelty levels. Importantly, we find contrasting spatial patterns of OA indices due to varying sensitivities to temperature and dissolved inorganic carbon, resulting in higher ΩA novelty in northern areas and higher pH and substrate-to-inhibitor ratio novelty in southern regions.

  • Remote Impacts of Cyclonic Eddies on Productivity Around the Main Hawaiian Islands

    by K. Feloy, B. S. Powell, T. Friedrich

    The Hawaiian Island chain lies on the southern limb of the oligotrophic North Pacific Subtropical Gyre, an area characterized by low productivity. In this region, productivity is controlled by several factors, including the island mass effect and mesoscale eddies generated by wind stress curl in the lee of the islands. This study identifies high‐chlorophyll events that occur periodically off the northern coasts of the Main Hawaiian Islands. These events are present in the satellite record and can be reproduced using a regional dynamical model. Model results indicate events are characterized by increases in total phytoplankton and total zooplankton, and a shift in phytoplankton size structure. We show these events are uniquely driven by the presence of cyclonic mesoscale eddies located downstream, on the opposite side of the island chain. While these eddies are known to impact productivity locally, we reveal that nutrients upwelled by these eddies can also be transported around the islands, counter to the mean background flow. These results demonstrate how leeward cyclonic eddies can have far‐reaching, remote impacts on productivity around the Hawaiian Islands.


  • Open Access

    Modeling the dispersal of the cryptogenic alga Chondria tumulosa (Rhodophyta, Ceramiales) in the Papahānaumokuākea Marine National Monument

    by James T. Fumo, Brian S. Powell, Randall K. Kosaki, Alison R. Sherwood

    The cryptogenic nuisance alga Chondria tumulosa was first observed in 2016 at Manawai (Pearl and Hermes Atoll) in the Papahānaumokuākea Marine National Monument. It has since spread across the atoll, growing in thick mats and smothering benthic habitat. In September 2021 the species was observed at Kuaihelani (Midway Atoll), ~130 km to the northwest. Due to its growth habit and recent spread, considerable concern has been raised and management of the species may be warranted. We used publicly available oceanographic data and the Connectivity Modeling System software to assess how the potential for successful dispersal of C. tumulosa is affected by particle properties and oceanographic conditions. We found the likelihood of successful transit to be linked to particle density, oceanographic conditions at the time of release, and release location. Further modeling explicitly targeted the capacity of both reproductive tetraspores as well as drifting fragments to disperse. Model results indicated tetraspores of C. tumulosa are unlikely to survive the transit from Manawai to Kuaihelani, as none arrived at Kuaihelani above the depth limit of the species and those arriving below successfully settled at a rate of only 0.02%. In contrast, fragments modeled as rafting on marine flotsam such as macroalgae and marine debris settled at a rate of 3.85%. Rafting fragments also settled ~600 km further to the southeast (towards the Main Hawaiian Islands) than tetraspores. This study identified oceanographic conditions and particle properties likely to aid dispersal of C. tumulosa to Kuaihelani and suggests that fragments rafting on marine flotsam may accelerate its spread.


  • Open Access

    Contrasting Central Equatorial Pacific Oxygen Isotopic Signatures of the 2014/2015 and 2015/2016 El Niño Events

    by S. Stevenson, K. M. Cobb, M. Merrifield, B. Powell, S. Sanchez, J. Nusbaumer, G. O’COnnor, A. Atwood

    Paleoclimate reconstructions of El Niño/Southern Oscillation (ENSO) behavior often rely on oxygen isotopic records from tropical corals ( δ 18 O). However, few reef‐based observations of physical conditions during El Niño events exist, limiting our ability to interpret coral δ 18 O. Here we present physical and geochemical measurements from Palmyra Atoll (5.9°N, 162.1°W) from 2014–2017, along with a data assimilation product using the isotope‐enabled Regional Ocean Modeling System (isoROMS). Coral δ 18 O signals are comparably strong in 2014–2015 and 2015–2016; notably, over 50% of the signal is driven by seawater δ 18 O, not temperature. If a constant seawater δ 18 O:salinity relationship were present, this would imply a comparable salinity anomaly during both events. However, salinity changes are much larger during 2014–2015, indicating a highly nonstationary relationship. isoROMS then shows that advection strongly influences δ 18 O during both the 2014–2015 and 2015–2016 El Niño, driving differences in the salinity/seawater δ 18 O relationship. This demonstrates the need for considering ocean dynamics when interpreting coral δ 18 O.