This paper uses Naïve, moving average, exponential smoothing,seasonal ARIMA (SARIMA) as well as GM (1,1) models to model outbound air travel demand of Taiwan to three major destinations - Hong Kong, Japan and the USA and compare their forecasting performances.Monthly time series data from Jan 1996 to Dec 2006 are used in this research. According to the indicators of forecasting performance, MAPE (mean absolute percentage error) and MAE (mean absolute error), the SARIMA method appears to be the superior model for forecasting seasonal air travel demand both in short- and long-term forecasting. The results also reveal that the forecasting capability of short-term forecasting appears to be more reliable than that of long-term forecasting.