An Evaluation Of The Accuracy Of Some Objective Techniques For Predicting Tropical Cyclone Movement In The Western North Pacific

Leoncio A. Amadore

Some important facts about tropical cyclones in general, and typhoons in the Philippine Area of Responsibility (PAR), in particular, are mentioned. A brief summary of different objective techniques of forecasting tropical cyclone movement is presented. It has been found out by meteorologists that these forecast techniques gave a variety of forecast positions even when applied to the same cyclone. It is therefore, one of the purposes of this study to evaluate the accuracy of seven different forecasting techniques in the PAR in order to determine which of these forecast techniques could help best the typhoon forecaster in the area. Using the best-track cyclone positions and post-analysed weather charts, 24-hr forecast positions of 28 tropical cyclones of 1968 to 1970, in the area bounded by 5°N - 25°N and 115°E-145°E, were made. The techniques tested are: Persistence, Climatology, (P+C)/2, Arakawa, Miller-Moore, Wang and Tse. The arithmetic mean and standard deviation of latitude and longitude forecast errors and the mean resultant errors of each technique were computed. The results showed that: 1) there is positive bias in Arakawa's latitude and longitude forecasts. 2) there is positive bias in the latitude forecast and negative bias in the longitude forecast of the Miller-Moore technique. 3) Persistence technique got the lowest mean resultant error followed by (P + C)/2, Miller-Moore, and Arakawa technique. 4) Climatology technique has the largest mean resultant error. 5) the results of this investigation are generally in good agreement with the results of other researchers. Some differences are noted in the arithmetic mean and standard deviation of errors of the 1969 sample and the 1968 and 1970 samples of Climatology and Persistence techniques. This is due to the fact that most of the cyclones of 1969 recurved east of the Philippines while the 1968 and 1970 cyclones moved westward or northwestward. A certain degree of subjectivity entered into the calculation of Wang's and Tse's techniques, hence the large forecast errors obtained. If the Arakawa and Miller-Moore techniques are corrected for their bias, the mean resultant errors are expected to reduce. Giving equal weightage to Persistence and Climatology techniques in combination is only based on general experience that the two methods are complementary. A more rational way of combining the two forecasts is presented. Using the standard deviation of forecast errors obtained for Persistence and Climatology forecasts and knowing the correlation between the two, a new forecast combination is derived for the area under study: For latitude: F3=0.6 P + 0.4 C For Longitude: F3=0.8 P + 0.2 C. The standard deviation of errors of the new forecast combination is less than the standard deviation of errors of either Persistence or Climatology or the (P+C)/2 forecast. This new combination obtained the least mean resultant forecast error during the 1971 tropical cyclone season as against Persistence, Climatology, (P+C)/2 techniques and the Weather Bureau official forecasts.