Abstract:Tropical cyclones(TCs) cause substantial adverse impacts on economies and human life.Forecasting the landfall of TCs is an important part in the prediction of TC tracks.Previous studies have focused mainly on the moving paths of TCs,especially the moving direction before landfall.However,not only the location but the timing of landfall is crucial.A successful prediction emerges only if these two aspects are combined.After landing,slow-moving TCs induce large amounts of precipitation locally,causing serious problems to the people living in the area.This study focuses on the changes in TC moving speed in the TC tracks,and analyzes the moving speed of TCs along with their temporal and spatial distribution changes in the hope to help improve TC track predictions.The study is based on the TC best track dataset for the western North Pacific over the period 1970-2009 provided by the National Meteorological Bureau and Shanghai Typhoon Institute(CMA-STI),along with TC track prediction error data for 2005-2010,and applies a percentile method to determine the anomaly index of the TC moving speed.The characteristics of the temporal and spatial distributions of anomalous TC moving speed and its change during the last 40 years over the western North Pacific are analyzed.The effect of TC moving speed prediction error on track prediction error and the relationship between large-scale steering flow and the error for TC moving speed prediction are studied.The results show that:(1) The cumulative probability at the 95%(5%) quantile thresholds for TC moving speed and its variation of anomalies in the western North Pacific are 10.8 m·s-1(1.43 m·s-1) and 2.42 m·s-1(-1.72 m·s-1),respectively;(2) Fast-moving or accelerated-moving TCs mostly appear in the area of the Japan Sea,while slow-moving or decelerated-moving TCs occur mainly in the region of the South China Sea;(3) The seasonal variation of anomalous-moving TC represents that the frequency of fast-moving(slow-moving) TCs is highest in May(October) and that of accelerated-moving TCs tends to be highest in June;(4) In the last 6 years (2006-2011),the prediction error of TC moving speed contributes on average about 41.6% of TC track prediction error;(5)In the case study of a TC with large track prediction error that is deduced more by moving speed prediction error,the results show that the weak large-scale environmental steering flow makes the TC move slowly.If the forecasted large-scale environmental steering flow is strong,the predicted TC moving speed will be faster,which may lead to large track prediction error.