Research On Distribution Characteristics And Forecast For Tropical Cyclone Gale In Guangxi

China is one of the countries frequently affected by Tropical Cyclones (TCs) in theworld.Gale weather of TCs is one of the immediate disaster inducements. Guangxi is on the southof Nanling Mountains and the north of Beibu Gulf Ocean. TCs gale causes great loss of lives andproperties every year to the people in coastal region of Guangxi. As the development of oceanexploitation, the distribution of TCs gale and the prediction for wind speed around TCs centerplay an important role on disaster prevention and forecast for Guangxi.Based on theories of natural disaster, atmosphere science and artifial neural network, thecharacteristics and disasters of TCs for Guangxi were analyzed firstly. Secondly, using the dataof Tropical Cyclones Year Books from1960to2012and wind speed of weather stations, thenumber of affected stations, TCs types from Information Center of Guangxi Weather Bureau,spatial and temporal distribution characteristics and factors of TCs gale in Guangxi were studied.Based on33-year TCs information in1980-2012and NCEP/NCAR reanalysis data, takingclimatology and persistence and earlier physical quantities predictors selected by correlationanalysis，Stepwise Regression (SR) and Multidimensional Scaling (MDS) methods as modelinputs, SR forecast model and neural network ensemble forecast model were built respectively forwind speed around TCs center (the forecast period is24h in the future), which could help to findforecast model with good reference for strong wind speed in coastal weather stations ofGuangxi.The research results showed that:(1) The number of53-years TCs that caused gale weather in Guangxi from1960to2012was233; and TCs of typhoon or above intensity accounted for49.36%.On temporal distribution, TCsgale experienced prominent annual inter-decadal variations in Guangxi. The frequency of galeoccurrence in1960s and1970s was relatively high, while in1980s and20s century was low; thecycle was4-6year. The occurrences of TCs gale change evidently with season, and the highestfrequency was in July and August. On spatial distribution, TCs gale mainly occurred in the southof23.5°N. The occurrence of gale in the south of Guangxi was much more than the northwest’s.Extreme wind speed was large along the coast while was small in inland. The places whose windspeed was greater than20m/s were Weizhou Island, Beihai and Fang Chenggang. The mainreasons for distributional difference of Guangxi gale were intensity and track of TCs, geographicposition and topographic condition of Guangxi. (2) For primary selection of predictors of TCs gale, physical quantities predictors were addedinto the models except traditional climatology and persistence predictors, and prediction errorwith two models reduced by about30%. For selection process of predictors, based on correlationanalysis and RS, MDS method was also used for selecting residual predictors, which ensured theintegrity of prediction information.(3) Taking TSc data of South China Sea from July to September in1980-2007as modelingsamples, TSc in2008-2012as independent samples for forecast verification, SR forecast modeland MDS neural network ensemble forecast model have been established for TSc gale at36gridpoints around TCs respectively. The results by two models showed strong prediction ability forindependent samples. Mean absolute error of24h forecast with SR forecast model for windvelocities centers was1.72m/s, and mean absolute error with MDS neural network ensembleforecast model was1.6m/s. Taking July as example, prediction abilities for29MDS neuralnetwork ensemble forecast models (81%) among36grid points around TCs centers were superiorto that of SWR models. Therefore, from the comparisons of general mean absolute errors, forecastcapacity for wind speed at36grids and prediction error at different actual wind speed, predictionability for gale prediction using MDS neural network ensemble forecast models was superior tothat of SWR models, which can meet business requirements and show better reference for galeforecast in coastal region in Guangxi，Guangdong and Fujian meteorological stations.