Unstructured grid ocean models are advantageous for simulating the coastal ocean and river–estuary–plume systems. However, unstructured grid models tend to be diffusive and/or computationally expensive, which limits their applicability to real-life problems. In this paper, we describe a novel discontinuous Galerkin (DG) finite element discretization for the hydrostatic equations. The formulation is fully conservative and second-order accurate in space and time. Monotonicity of the advection scheme is ensured by using a strong stability-preserving time integration method and slope limiters. Compared to previous DG models, advantages include a more accurate mode splitting method, revised viscosity formulation, and new second-order time integration scheme. We demonstrate that the model is capable of simulating baroclinic flows in the eddying regime with a suite of test cases. Numerical dissipation is well-controlled, being comparable or lower than in existing state-of-the-art structured grid models.

Deep learning is a promising tool to determine the physical model that describes our universe. To handle the considerable computational cost of this problem, we present CosmoFlow: a highly scalable deep learning application built on top of the TensorFlow framework. CosmoFlow uses efficient implementations of 3D convolution and pooling primitives, together with improvements in threading for many element-wise operations, to improve training performance on Intel ^® Xeon Phi™ processors. We also utilize the Cray PE Machine Learning Plugin for efficient scaling to multiple nodes. We demonstrate fully synchronous data-parallel training on 8192 nodes of Cori with 77% parallel efficiency, achieving 3.5 Pflop/s sustained performance. To our knowledge, this is the first large-scale science application of the TensorFlow framework at supercomputer scale with fully-synchronous training. These enhancements enable us to process large 3D dark matter distribution and predict the cosmological parameters Ω _M , σ ₈ and n _s with unprecedented accuracy.

This paper describes Nemo-Nordic 2.0, an operational marine model for the Baltic Sea. The model is used for both near-real-time forecasts and hindcast purposes. It provides estimates of sea surface height, water temperature, salinity, and velocity, as well as sea ice concentration and thickness. The model is based on the NEMO (Nucleus for European Modelling of the Ocean) circulation model and the previous Nemo-Nordic 1.0 configuration by Hordoir et al. (2019). The most notable updates include the switch from NEMO version 3.6 to 4.0, updated model bathymetry, and revised bottom friction formulation. The model domain covers the Baltic Sea and the North Sea with approximately 1 nmi resolution. Vertical grid resolution has been increased from 3 to 1 m in the surface layer. In addition, the numerical solver configuration has been revised to reduce artificial mixing to improve the representation of inflow events. Sea ice is modeled with the SI3 model instead of LIM3. The model is validated against sea level, water temperature, and salinity observations, as well as Baltic Sea ice chart data for a 2-year hindcast simulation (October 2014 to September 2016). Sea level root mean square deviation (RMSD) is typically within 10 cm throughout the Baltic basin. Seasonal sea surface temperature variation is well captured, although the model exhibits a negative bias of approximately −0.5 ^∘C. Salinity RMSD is typically below 1.5 g kg⁻¹. The model captures the 2014 major Baltic inflow event and its propagation to the Gotland Deep. The model assessment demonstrates that Nemo-Nordic 2.0 can reproduce the hydrographic features of the Baltic Sea.

Numerical modeling of three-dimensional estuarine circulation is often challenging due to complex flow features and strong density gradients. In this paper the skill of a specific model is assessed against a high-resolution data set, obtained in a river-dominated mesotidal estuary with autonomous underwater vehicles and a shipborne winched profiler. The measurements provide a detailed view of the salt wedge dynamics of the Columbia River estuary. Model skill is examined under contrasting forcing conditions, covering spring freshet and autumn low flow conditions, as well as spring and neap tides. The data set provides a rigorous benchmark for numerical circulation models. This benchmark is used herein to evaluate an unstructured grid circulation model, based on linear finite element and finite volume formulations. Advection of momentum is treated with an Eulerian–Lagrangian scheme. After the model’s sensitivity to grid resolution and time step is examined, a detailed skill assessment is provided for the best model configuration. The simulations reproduce the timing and tidal asymmetry of salinity intrusion. Sharp density gradients, however, tend to be smoothed out affecting vertical mixing and gravitational circulation. We show that gravitational salt transport is underestimated in the model, but is partially compensated through tidal effects. The discrepancy becomes most pronounced when the stratification is strongest, i.e., under high river discharge and neap tide conditions