Results from 2016 demonstrate global improvement of tropical cyclone tracks as a result of assimilating GH dropsondes. The GH sampled hurricanes Gaston, Hermine, and Matthew as well as tropical storm Karl. Not only did the average GFS track error decrease for all of those storms when GH dropsondes were assimilated, but errors also decreased substantially for TD8, which was not sampled. Average track error for 150 cases over the North Atlantic in 2016 decreased at every hour after 36h with a peak reduction of about 15% at 72 h. Furthermore, substantial decreases in track errors were also observed in both the eastern and western North Pacific basins, which were not sampled. The reasons for the decrease in remote errors is being explored. Meanwhile, initial results from downstream HWRF retrospectives indicate that the addition of GH data can improve both track and intensity forecasts.
Retrospectives from earlier years, however, yield neutral results, and the reasons for this discrepancy are being explored. One problem is that an upper-level dry bias plagued all Vaisala dropsondes between 2010 and 2015. Though this impacted all operational dropsondes used by NOAA as well as the GH, the impacts on the GH were much more severe due to its higher cruise altitude. Another possibility is that the flight strategies were fundamentally different in SHOUT from the earlier HS3 experiment. Additional experiments are planned to investigate the extent to which these factors contributed to the disparity in results.