高速鐵路運量預測之巨觀方法：臺灣高速鐵路案例回顧與模式重建

An Aggregate Approach for High-Speed Rail Ridership Forecasting: Model Development Based on Case Revisit of Taiwan High Speed Rail

高速鐵路運量預測一直是備受爭議的課題，特別是運量高估仍顯見於全球眾多高速鐵路計畫。由於高速鐵路計畫涉大規模的資本投資，運量高估也導致後續營運上的財務困難。本研究重行檢視臺灣高鐵規劃階段所採用之運量預測方法，探討實際運量和預測運量之間落差的可能原因，進而發展運量預測方法，含直接旅運需求模式和運具分配模式，並利用臺灣高鐵通車前可取得之巨觀資料進行模式估計。模式預測結果與實際運量之間約有10%的誤差，與過去的預測方法相較，已是顯著的提升；然而仍顯示社經背景變動條件下長期需求預測的困難度。檢視過去的方法並可發現，其對於誘發性旅運需求的預測約占臺灣高鐵總預測運量的三分之一，可能是導致運量高估的主因之一。 Ridership forecasting for High-Speed Rail (HSR) systems is a contentious issue as significant ridership overestimation has been observed in several HSR systems worldwide. This situation has resulted in severe financial difficulties for their long-term operations against the involved large-scale capital investment. This study uses the Taiwan HSR (THSR) data collected since the start of its operation in 2007 to revisit THSR planning and its associated approaches used for ridership forecasting, seeking to investigate the likely factors that lead to the gap between the forecasted and actual ridership. A forecast framework is then proposed that consists of a direct travel demand prediction model and a mode split model that are constructed with regard to several methodological aspects identified from the revisit. The models are estimated using historical data before THSR operation, which are generally obtainable and observable at an aggregate level, thereby affording the feasibility and easiness in practice. The forecasting result is approximately 10% from the actual ridership and can be viewed as an evident improvement from previous forecasts. The result highlights the difficulty of long-term travel demand prediction if significant changes of socioeconomic backgrounds over the prediction horizon are involved. It also indicates that the prediction of the induced travel demand, which accounts for approximately one-third of the forecasted ridership of the THSR in the previous approaches, can be one of the major sources of overestimation.

Ridership forecasting for High-Speed Rail (HSR) systems is a contentious issue as significant ridership overestimation has been observed in several HSR systems worldwide. This situation has resulted in severe financial difficulties for their long-term operations against the involved large-scale capital investment. This study uses the Taiwan HSR (THSR) data collected since the start of its operation in 2007 to revisit THSR planning and its associated approaches used for ridership forecasting, seeking to investigate the likely factors that lead to the gap between the forecasted and actual ridership. A forecast framework is then proposed that consists of a direct travel demand prediction model and a mode split model that are constructed with regard to several methodological aspects identified from the revisit. The models are estimated using historical data before THSR operation, which are generally obtainable and observable at an aggregate level, thereby affording the feasibility and easiness in practice. The forecasting result is approximately 10% from the actual ridership and can be viewed as an evident improvement from previous forecasts. The result highlights the difficulty of long-term travel demand prediction if significant changes of socioeconomic backgrounds over the prediction horizon are involved. It also indicates that the prediction of the induced travel demand, which accounts for approximately one-third of the forecasted ridership of the THSR in the previous approaches, can be one of the major sources of overestimation.