Search for: All records

In strong winds, air flow detaches from the ocean surface in the lee of wave crests and creates a low‐pressure zone on the wave’s leeward face. The pressure difference between the wave’s rear and front face modulates the momentum input from wind to waves. Numerical wave models parameterize this effect using a so‐called sheltering coefficient. However, its value and dependence on wind speed are not well understood, particularly with background swell waves. To bridge this gap, we conducted laboratory experiments with winds up to Category 4 hurricane force blown over various mechanically generated wave conditions (pure wind sea, mixed waves with directional spreading, and monochromatic unidirectional waves) and measured the wind, waves, and stress at a sufficient frequency to resolve wind‐wave variability over the long‐wave phase. We analyze the results in the context of Jeffreys’s sheltering theory and find two regimes: (a) from low‐to‐moderate wind forcing (10 m s⁻¹ < U₁₀ < 33 m s⁻¹), the aerodynamic sheltering increases with wind speed, consistent with previous studies; (b) in hurricane conditions (U₁₀ > 33 m s⁻¹), the aerodynamic sheltering decreases with wind at a rate depending on wave state. Further, we isolate the short wind waves from the longer paddle waves and find that the aerodynamic sheltering by longer waves leads to a phase‐dependent variability of the short wind‐waves’ local steepness, which is evidenced by the sheltering coefficient’s value derived from wind and wave measurements. Our results emphasize the need for further measurements of aerodynamic sheltering and improving its representation in models.

 

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. 

During the summer of 2016, a boreal summer intraseasonal oscillation (BSISO) event was observed over Southeast Asia and the South China and Philippine seas. Precipitation anomalies associated with this event propagated northward at a speed of 0.5–1° per day from July to August. To understand the mechanisms, a regional atmosphere‐ocean coupled system with the Weather Research and Forecasting (WRF) model and the Hybrid Coordinate Ocean Model (HYCOM) is used to study this BSISO event. The 50‐day‐long coupled simulations reasonably capture large‐scale northward propagation of the event. Coupled simulations with altered air‐sea interaction and atmosphere‐only simulations with prescribed sea surface temperature illuminate the insignificant role of air‐sea interaction within the computation domain in the northward propagation the for this event. Diagnostics of the coupled simulation as well as the ECMWF‐Interim reanalyses indicate that convection and barotropic vorticity are largely in phase north of 5°N during the event. The BSISO convection is accompanied by moisture anomalies whose magnitudes increase as the BSISO propagates northward. Analysis of the moisture budget shows that positive horizontal advection leads positive moisture anomalies on the intraseasonal time scale north of 10°N. The vorticity‐convection relationship, the lead‐lag relationship between moisture and its horizontal advection, and the latitude dependence of each for this BSISO event are consistent with general features of BSISO events composited with ECWMF‐Interim reanalysis and satellite precipitation data sets.