Abstract. We present a new ultra-high resolution 1/30° (∼3 km) regional ocean model of the eastern Australian region and evaluate the performance of this model against two 1/10° (∼10 km) models as well as a suite of satellite and in situ observations. We evaluate model biases in the context of (i) submesoscale-permitting (∼3 km) vs. mesoscale-permitting (∼10 km) horizontal resolution and (ii) differences between version 5 and version 6 of the Modular Ocean Model (MOM5 vs. MOM6) to assess the added value in each case and determine the suitability of our higher resolution model for scientific research. There are some consistent biases shared by the two regional MOM6 configurations, and also in the higher resolution configuration that are not seen in the lower resolution models. These biases are further investigated with two sets of sensitivity experiments to understand the effect of submesoscale eddy parameterization and imposed dynamic viscosity at a submesoscale-permitting resolution. The high-resolution simulation has much higher variance compared to the lower resolution simulations across all evaluation metrics, indicating that the greater spectrum of length scales also manifests in more variability in the temporal domain. The two MOM6 regional configurations of differing resolution appear to be more aligned than the regional (MOM6) and global (MOM5) 1/10° configurations in most results, reflecting the substantial changes made to the MOM between version 5 and version 6. Importantly, we also show that higher resolution is not a panacea: in regions where key dynamics are quasi-linear and well-captured captured by coarser grids (e.g., the EAC jet), further refinement may offer limited benefit – and actually degrade the performance if parameterizations are not appropriately tuned.
Reilly et al. (Tue,) studied this question.