A statistical forecasting model applied to container throughput in a multi-port gateway system: the Barcelona-Tarragona-Valencia case

Abstract

This contribution investigates the container throughput flow in a multi-port gateway system: Barcelona-Tarragona-Valencia (BTV). First, the paper examines the recent dynamics of the total and transhipment flow showing a relevant shifting of traffic share from Barcelona to Valencia. A novel model based on a two-state Markov model in conjunction with a Monte Carlo experiments is implemented to estimate the predictions of annual growth in container throughput. Verification tests show how the predictions are reasonably good with error metrics similar to other methods based on timeseries analysis (trend projections and ARIMA). The strength of the method relies in the statistical nature of the predictions provided (i.e., mean and data dispersion). The method is considered suitable for short-term forecasting. The practical application of the method considers separately the import/export and transhipment container throughput, revealing a different dynamics in both container flows.