Search for: All records

Abstract The wind shear stress at the ocean surface drives momentum exchange across the air-sea interface regulating atmospheric and oceanic phenomena. Theoretically, the mean wind stress acts in a reference frame moving with the ocean surface; however, the relative motion between the air and ocean surface layers is conventionally neglected in bulk transfer formulae. Recent developments improving air-sea momentum flux quantification advocate for explicitly defining the air-sea relative wind, especially in the regime of low wind forcing, where surface currents may approach a significant fraction of the total wind speed. Yet, in practice, this new approach is typically applied using opportunistic definitions of the near-surface current. Here, we build on this recent work and propose a general framework for the bulk air-sea momentum flux that directly accounts for vertical current shear and surface waves in quantifying the stress at the interface. Our approach partitions the stress at the interface into viscous skin and (wave) form drag components, each applied to their relevant surface advections, which are quantified using the inertial motions within the sub-surface log layer and the modulation of waves by currents predicted by linear theory, respectively. The efficacy of this approach is demonstrated using an extensive oceanic dataset from the Coastal Endurance Array (Ocean Observatories Initiative) offshore of Newport, Oregon (2017–2023) that includes co-located measurements of direct covariance wind stress, directional wave spectra, and current profiles. As expected, our framework does not alter the overall dependence of momentum flux on mean wind forcing, and we found the largest impacts at relatively low wind speeds. Below 3 m s$$^{-1}$$, accounting for sub-surface shear reduced form drag variation by 40–50% as compared to a current-agnostic approach; as compared to a shear-free current, i.e., slab ocean, a 35% reduction in form drag variation was found. At this wind forcing, neglecting the currents led to systematically overestimating the form stress by 20 to 50%—an effect that could not be captured by using the slab ocean approach. This framework builds on the existing understanding of wind-wave-current interaction, yielding a novel formulation that explicitly accounts for the role of current shear and surface waves in air-sea momentum flux. This work holds significant implications for air-sea coupled modeling in general conditions. 

Full Changelog: https://github.com/Cloud-Drift/scipy-2024-poster/compare/1.3...1.4 

What's Changed 🔎 type comparison with isinstance by @philippemiron in https://github.com/Cloud-Drift/clouddrift/pull/491 🧹 Update type comparison in ragged.py by @selipot in https://github.com/Cloud-Drift/clouddrift/pull/497 🧹 Update random number generator in gdp1h and gdp6h adapters by @selipot in https://github.com/Cloud-Drift/clouddrift/pull/496 🐛 fix locationtype bug in dataset by @kevinsantana11 in https://github.com/Cloud-Drift/clouddrift/pull/494 Update .zenodo.json by @selipot in https://github.com/Cloud-Drift/clouddrift/pull/498 ++ increment version (v0.40.0) by @kevinsantana11 in https://github.com/Cloud-Drift/clouddrift/pull/499 Full Changelog: https://github.com/Cloud-Drift/clouddrift/compare/v0.39.0...v0.40.0 

Implementing artificial neural networks is commonly achieved via high-level programming languages such as Python and easy-to-use deep learning libraries such as Keras. These software libraries come preloaded with a variety of network architectures, provide autodifferentiation, and support GPUs for fast and efficient computation. As a result, a deep learning practitioner will favor training a neural network model in Python, where these tools are readily available. However, many large-scale scientific computation projects are written in Fortran, making it difficult to integrate with modern deep learning methods. To alleviate this problem, we introduce a software library, the Fortran-Keras Bridge (FKB). This two-way bridge connects environments where deep learning resources are plentiful with those where they are scarce. The paper describes several unique features offered by FKB, such as customizable layers, loss functions, and network ensembles. The paper concludes with a case study that applies FKB to address open questions about the robustness of an experimental approach to global climate simulation, in which subgrid physics are outsourced to deep neural network emulators. In this context, FKB enables a hyperparameter search of one hundred plus candidate models of subgrid cloud and radiation physics, initially implemented in Keras, to be transferred and used in Fortran. Such a process allows the model’s emergent behavior to be assessed, i.e., when fit imperfections are coupled to explicit planetary-scale fluid dynamics. The results reveal a previously unrecognized strong relationship between offline validation error and online performance, in which the choice of the optimizer proves unexpectedly critical. This in turn reveals many new neural network architectures that produce considerable improvements in climate model stability including some with reduced error, for an especially challenging training dataset.