Weighted Consensus Approach to Tropical Cyclone 96-H and 120-H Track Forecasting

摘要

A long-range (96 h-120 h) weighted position consensus for tropical cyclone tracks is evaluated for 24 western North Pacific storms in 2006. The first weighted position technique simply weights the 96-h, 108-h, and 120-h dynamical model positions inversely to their distances from the 60-h, 66-h, and 72-h consensus positions. The second weighted consensus technique uses the same weighting factors but is applied to the forecast motion vectors to assess 96 h- 120 h track errors. The weighted position consensus yields modest reductions in error relative to an unweighted position consensus at 96 h 120 h and produces smoother track forecasts. Weighted position consensus errors are reduced when the COAMPS model and the Air Force Weather Agency MM5 model are removed from the unweighted consensus used to form the weighting factors. Including the Japan and ECMWF model tracks also improves the weighted position consensus performance. The weighted motion vector consensus achieves dramatic improvements over an unweighted position consensus (9.9% at 96 h and 5.6% at 120 h). Most of the improvement over an unweighted position consensus is from using a motion vector consensus rather than a position consensus since large improvements are also achieved with an unweighted motion vector consensus.