Demonstrating the complementarity of observations in an operational ocean forecasting system

We have performed a series of near-real-time observing system experiments with FOAM, the Met Office operational ocean forecasting system. These were conducted in parallel to the operational suite and identical to it except that certain observation types were excluded. At the start of each month the parallel system was reset to the operational restart and a run started with a different observation type excluded: in February -XBT; March -TAO/TRITON; April -Jason-2 altimeter; May -all altimeter; June -AVHRR sea-surface temperature (SST) data; and July -Argo data. We show that the existing ocean observing systems offer a good deal of complementary information. All components of the observing system offer unique, independent information that help initialise, and improve, operational ocean forecasts. Withholding XBTs causes little impact on globally averaged metrics, for example RMS innovations. However, locally we see long-lasting temperature impacts (similar to 1 degrees C) near XBT transects. Withholding TAO/TRITON data has a regional impact increasing the tropical Pacific RMS temperature and salinity innovations by 39 and 60%, respectively. Withholding Jason-2 data results in a global 4% increase in the RMS sea-surface height (SSH) innovations, and small-scale changes in temperature and salinity of about 2 degrees C at 100m depth, and similar to 0.2 psu near the surface. Withholding all altimeter data leads to a 16% increase of the RMS SSH innovations. We also see changes in other model variables similar in magnitude to those from withholding Jason-2, but more widespread. Withholding AVHRR SST data produces significant changes around 1 degrees C in model temperature at the surface and within the surface mixed layer, but there is little or no effect below this. Withholding Argo data for 1 month leads to a 5% increase in the temperature and salinity innovations, and changes in SSH of up to 5 cm.