A shelf-scale assessment of GLORYS and BRAN ocean reanalysis model outputs for the southern Benguela Upwelling System

Abstract

The Benguela Upwelling System, with its high variability and productivity, is challenging to monitor owing to limited in situ ocean observations. Researchers therefore often use ocean reanalysis products to simulate oceanographic conditions. We evaluated the performance of the GLORYS and BRAN reanalyses against in situ hydrographic data obtained from four research cruises in 2019 and from multi-year moored observations. The models reproduced cross-shelf temperature and salinity distributions reasonably well (r = 0.66–0.98, p < 0.001) for most stations. At the time of the cruises, GLORYS was 3–6 °C higher and had salinities 0.2–0.4 higher than in situ data in summer, but temperatures of 1–2 °C and salinities of 0.1–0.3 lower in winter. BRAN consistently underestimated temperature by 2–5 °C and salinity by 0.2–0.7. Both models performed better at depth than in the upper layers of the water column, reproducing temperature better than salinity. At longer time-scales, both models simulated current magnitude better than temperature (r = 0.77–0.79, p < 0.001). Model performance varied with latitude, with stronger correlations in the south than in the north, particularly for salinity. Our results suggest that these reanalyses need improved representation of regional ocean–atmosphere interactions and bathymetry. Thus, caution is required when using the model outputs for monitoring and operational studies.