Evaluation of Bluelink hindcast BRAN 3.5 at surface and 1000 m

Abstract

We present an evaluation of a 19-year numerical hindcast, Bluelink ReANalysis 3.5 (BRAN 3.5), derived from the Australian Bluelink programme. The evaluation is made for the surface and at 1000m depth using Global Drifter Program (GDP) drifters and Argo floats, respectively. It is aimed at comparing the model’s ability to correctly simulate the Lagrangian mixing processes. These mixing and dispersal processes are often parameterised in terms of an eddy diffusivity, and we use diffusivity as a metric of comparison. Because the data are spatially sparse and non-uniform in time, the hindcast was sampled at the data locations to provide a model velocity corresponding to each observed velocity. Pseudo-Eulerian mean velocity and eddy kinetic energy fields were computed from bin-averaging of the Lagrangian velocities, and eddy diffusivities were computed from single-particle statistics. The mean field at 2.5m in BRAN is largely dominated by Ekman transport. The mean velocities at 12.5m compare well with the GDP drifter mean velocities, except that the Antarctic Circumpolar Current (ACC) is stronger in GDP than in BRAN. At the surface, EKE in BRAN tends to be marginally higher than in GDP. Modelled eddy diffusivity compares well with observed diffusivity, except in the boundary currents and near the equator. At 1000m, BRAN tends to overestimate the mean field, EKE and eddy diffusivity at low latitudes and underestimate them at higher latitudes.