The ocean can act as a rain gauge for tropical precipitation

We use ocean salinity observations to track the surface flux of water and thus check/improve atmospheric reanalysis estimates of net freshwater flux. Oceanographers have long recognized that the nearly 80% of tropical precipitation that falls on the ocean leaves its signature in the dilution of the upper layers. However, limitations in salinity data coverage and lack of information about salt transport have hindered application of this approach. Salinity data coverage has dramatically improved in the last 15 years with the expansion of Argo, deployment of tropical moorings and development of satellite sensing. The lack of information about salt transport divergence becomes increasingly important at seasonal and longer timescales. It would still interfere with the ability of ocean general circulation model simulations to act as 'virtual rain gauges' on seasonal and longer time-scales. Sequential data assimilation, in contrast, proceeds through a series of brief forecasts, for each of which advection is relatively unimportant. Thus, errors in flux are revealed by errors in local salt storage. In this study we use salt budget information produced by the sequential data assimilation within the Simple Ocean Data Assimilation v3 (SODA3) ocean reanalysis system to show that there are freshwater flux errors as large a 1-2 mm/dy for the period 2004-2016 in several popular atmospheric reanalyses. Additional experiments provide a check on our results by demonstrating that when we use our results to correct net surface flux the atmosphere/ocean imbalance in water exchange is greatly reduced.