Numerical studies of ice flow have consistently identified the grounding zone of outlet glaciers and ice streams (the region where ice starts to float) as crucial for predicting the rate of grounded ice loss to the ocean. Owing to the extreme environments and difficulty of access to ocean cavities beneath ice shelves, field observations are rare. Estimates of melt rates derived from satellites are also difficult to make near grounding zones with confidence. Therefore, numerical ocean models are important tools to investigate these critical and remote regions. The relative inflexibility of structured grid models means, however, that they can struggle to resolve these processes in irregular cavity geometries near grounding zones. To help solve this issue, we present a new nonhydrostatic unstructured mesh model for flow under ice shelves built using the Firedrake finite element framework. We demonstrate our ability to simulate full ice shelf cavity domains using the community standard ISOMIP+ Ocean0 test case and compare our results against those obtained with the popular MITgcm model. Good agreement is found between the two models, despite their use of different discretisation schemes and the sensitivity of the melt rate parameterisation to grid resolution. Verification tests based on the Method of Manufactured Solutions (MMS) show that the new model discretisation is sound and second-order accurate. A main driver behind using Firedrake is the availability of an automatically generated adjoint model. Our first adjoint calculations, of sensitivities of melt rate with respect to different inputs in an idealised grounding zone domain, are promising and point to the ability to address a number of important questions on ocean influence on ice shelf vulnerability in the future.

We present a deep learning method to downscale low-resolution geophysical fields by merging them with high-resolution data. The downscaling was performed using an ensemble of convolutional neural networks (CNNs), whose prediction values are the average values of the outputs of 20 CNNs. Academic experiments were conducted on simulated ocean data in the Gulf Stream region, given by the outputs of the NATL60 model. The CNNs forced with low-resolution (120 × 120 km) sea surface high (SSH) data and mesoscale resolution (12 × 12 km) sea surface temperature (SST) data allowed us to obtain mesoscale resolution sea surface currents with good accuracy. Sensitivity experiments have shown that taking SST into account significantly increases the accuracy of the high-resolution velocity retrieval, even when noise is added to the SSH data. The velocity information embedded in the transport equation modeling the SST advection is taken into account by the CNN, which greatly increases the resolution of ocean currents provided by SSH. In the present work, we only consider spatial downscaling by assuming that SSH and SST are daily observations. The method we developed is generic and can be used to improve the resolution of a wide variety of large-scale fields by merging them with high-resolution fields.

Small-scale eddies play an important role in preconditioning and restratifying the water column before and after mixing events, thereby affecting deep water formation variability. Results from a realistic eddy-resolving (∼5 km local horizontal resolution) ocean model suggest that small-scale temperature fluxes due to turbulent potential to kinetic energy conversion are the main driver of mixed layer restratification during deep convection in the Labrador Sea interior and the West Greenland Current. This resupply of heat due to turbulent upward buoyancy fluxes exhibits a large interannual variability imposed by the atmospheric forcing. Eddy fluxes only become active in periods of strong buoyancy loss, while being quiescent otherwise. In a low-resolution (∼20 km) control simulation the modeled turbulence is strongly reduced and the associated modeled and parameterized heat fluxes are too weak to increase stratification.