Surface winds from atmospheric reanalysis lead to contrasting oceanic forcing and coastal upwelling patterns

Ocean surface winds determine energy, material and momentum fluxes through the air-sea interface. Accounting for wind variability in time and space is thus essential to reliably analyze and simulate ocean circulation and the dynamics of marine ecosystems. Here, we present an assessment of surface winds from three widely used atmospheric reanalysis products (NCEP/NCAR, ERA-Interim and JRA-55) and their corresponding ocean forcing data sets (CORE v2.1, DFS v5.2 and JRA55-do), which include corrections for use in ocean simulations. We compared wind patterns most relevant to ocean circulation (surface wind stress, its curl and estimates of induced vertical upwelling velocity) across global and regional scales, with added emphasis on the main Eastern Boundary Upwelling Ecosystems (EBUEs). All products provided consistent large-scale patterns in surface winds and wind stress, although agreement was reduced for indices involving the calculation of spatial derivatives, like wind stress curl and Ekman pumping. Fidelity with respect to a reference reanalysis based on blended satellite and buoy observations (CCMP v2.0) improved in more recent, higher resolution products like JRA-55 and ERA-Interim. Adjustments applied when deriving ocean forcing data sets from atmospheric reanalysis robustly improved wind speed and wind stress vectors, but degraded wind stress curl (and implied Ekman upwelling) in two of the three ocean forcing products considered (DFS v5.2 and CORE v2.1). At regional scales, we found significant inconsistencies in equatorial and polar regions, as well as in coastal areas. In EBUEs, upwelling favorable winds were weaker in atmospheric reanalysis products and ocean forcing data sets than estimates based on CCMP v2.0 and QuikSCAT. All reanalysis products featured lower amplitude seasonal cycles and contrasting patterns of low frequency variability within each EBUE, including the presence of sudden changes in mean upwelling only for some products. Taken together, our results highlight the importance of incorporating uncertainties in wind forcing into ocean simulation experiments and retrospective analysis, and of correcting reanalysis products for ocean forcing data sets. Despite the continued improvement in the quality of wind data sets, prevailing limitations in reanalysis models demonstrate the need to confirm global products against regional measurements whenever possible and improve correction strategies across multiple ocean-relevant wind properties.