Ocean model utility dependence on horizontal resolution

Abstract

This thesis examines the change in ocean model utility with changing horizontal resolution. Oceans are a crucial part of the climate system, with numerical models offering important insights into our mechanistic understanding. We use a 30 year integration (1978 to 2007) of the NEMO model at 1º, 1/4º and 1/12º to investigate the impact of modelling choices associated with horizontal resolution changes. Changes in degrees of freedom associated with the increasing resolution allow alternative energy dissipation pathways, with potential impact on model accuracy. We develop a measure of utility based on an estimate of the accuracy, as well as a penalisation which scales with resolution. Overall, accuracy is thought to increase with resolution, and we examine the associated change in utility on a range of model fields.

The exploration of the NEMO model assesses the surface mixed layer, deep (>2000m) to surface (<2000m) communication through the ocean interior and the changes in the meridional overturning with topographic interactions. Assessing these areas, we illustrate potential changes in the energy pathways in the system. We investigate the surface in terms of the mixed layer depth globally, but also investigating a case study in the Southern Ocean. We find that the mixed layer does not change significantly with resolution, and that NEMO compares well with observations. Minor changes with resolution are attributed to increased numbers of fronts with increasing resolution. When the mixed layer is assessed, we see no significant change with resolution, and so find that 1º has the highest utility. For our case study, we investigate the zonally asymmetric deepening of the mixed layer in the Southern Ocean. We find that the stratification set by the advection is key, and confirm this using the 1D Price-Weller-Pinkel model.

The communication between the surface and the deep ocean is assessed by looking at the steric height variability, and specifically its covariance between the surface and the deep. We find that there are large changes with resolution, and attribute these to the higher resolutions' ability to include eddy effects. This suggests that the Gent-McWilliams scheme that is active at low resolution fails to capture this. We look at the low and high frequency parts of the variance, finding that strongly eddying regions dominate the high frequency steric height covariance, confirming the importance of eddies. The ratio between the surface and the deep steric height shows poor utility in both ORCA1 and ORCA025, while we find seasonal leakage obscuring our accuracy measure for the steric height.

The overturning is assessed in density space, and we notice a strengthening of the anti-clockwise component in the Southern Ocean. Decomposing the transport into its baroclinic and barotropic components, we find that changes in the baroclinic overturning can account for this. The lack of western boundaries in the Southern Ocean suggests that eddies, as well as interaction with topography, are especially important here, and we investigate the change in the balance of forcing in terms of the associated vortex stretching. We assess this in terms of the bottom pressure torque, but find the major changes in the baroclinic component of the bottom pressure torque. We find that increasing the resolution still leads to increased utility, particularly in the barotropic and baroclinic density space overturning case.

The major implications of our results are that low resolution is appropriate for fields such as the mixed layer depth, but increasing the resolution is seen to improve the mean overturning through allowing eddy activity.

More information

Identifiers

Catalogue record

Export record