Why Gliders Appreciate Good Company: Glider Assimilation in the Oregon-Washington Coastal Ocean 4DVAR System With and Without Surface Observations

Gliders are low-power autonomous underwater vehicles used to obtain oceanic measurements in vertical sections. Assimilation of glider temperature and salinity into coastal ocean circulation models holds the potential to improve the ocean subsurface structure estimate. In this study, the impact of assimilation of glider observations is studied using a four-dimensional variational (4DVAR) data assimilation and forecast system set offshore of Oregon and Washington on the U.S. West Coast. Four test cases are compared: (1) no assimilation, (2) assimilation of glider temperature and salinity data alone, (3) assimilation of the glider data in combination with the surface observations including satellite sea surface temperature, sea surface height, and high-frequency radar surface velocities, and (4) assimilation of the surface data alone. It is found that the assimilation of glider observations alone creates unphysical eddies in the vicinity of the glider transect. As a consequence, the forecast errors in the surface velocity and temperature increase compared to the case without data assimilation. Assimilation of surface and subsurface observations in combination prevents these features from forming and reduces the errors in the forecasts for the subsurface fields compared to the other three experiments. These improvements persisted in 21-day forecasts run after the last data assimilation cycle. Plain Language Summary Ocean forecast systems, like our system for the ocean offshore Oregon- Washington, USA, use numerical models to predict future temperature, currents, sea surface height, and salt concentration. Eventually, these predictions are corrected using observations in a process called data assimilation ( DA) in order to better approximate the true state of the ocean. Assimilation of subsurface observations made by autonomous underwater vehicles called gliders can potentially improve the predictions of the ocean state below the surface. In this study, we have assimilated glider observations together with, and in absence of, surface observations. We found that assimilation of glider observations alone creates predictions that are less accurate than those obtained from a model without DA as the DA creates unphysical features. Assimilating the glider observations in tandem with surface observations prevents these features from forming and realizes local improvements to the subsurface ocean predictions that last for 21 days after the last DA correction. These results show that one has to be careful with using glider observations in DA and should always assimilate glider observations in combination with observations that cover large swaths of the ocean surface.