In 1997 the Tropical Prediction Center began operational Gulfstream-IV jet aircraft missions which deployed dropsondes for the purpose of improving numerical model guidance of hurricanes threatening the USA.
Some of the sondes deployed near to Hurricane Erika and Hurricane Linda were used by the UK Meteorological Office global model. Four 5-day forecasts were rerun without dropsonde data (two for each hurricane) and mean track forecast errors were found to be some 30% higher in the runs which used the dropsonde data. Forecast strength of the hurricanes was also lower in the runs which included dropsonde data.
Most of the worsening of performance can be attributed to one of the four forecasts which was rerun (for Hurricane Erika). In this case, sondes were not dropped equally in all quadrants of the hurricane, but only to the north of the circulation. This appears to have resulted in a forecast overstrengthening of the ridge to the north of the hurricane which erroneously prevented recurvature of the hurricane by the model forecast.