Jet Stream Analysis and Forecast Errors Using GADS Aircraft Observations in the DAO, ECMWF, and NCEP Models

Abstract

Peak analyzed jet stream wind speeds are compared with independent aircraft observations over Canada and the continental United States. The results permit a study of the accuracy of analyzed jet streak strength for the data-sparse 85% of the earth's surface versus the data-dense 15%. The observations come from the Global Aircraft Data Set (GADS) experiment, which since 1996 has collected flight data recorder information from every flight of 56 British Airways 747-400 aircraft. The study is timely because automated aircraft observations are reaching their near-asymptotic limits (there are not many uncovered commercial aircraft routes left), and we are about to enter a new, third-generation, satellite-sounding-instrument era. Future reanalyses will mix time periods from both eras. This study gives an estimate of the analysis accuracy of data assimilation using second-generation satellite systems. The results show that major current generation assimilation models have peak wind speed errors of −5% to −9% over data-sparse Canada compared with −2% to −5% over the data-dense continental United States. When these additional aircraft observations are assimilated as a part of the normal observational input data stream, a small but statistically significant improvement is shown for 1-month forecast experiments over two consecutive winters.