Hinterland transport encompasses the flow of goods between the deep sea ports and the inland. It enables maritime shipping and entails a significant portion of end-to-end transportation costs. Multimodal transport is the utilization of multiple modes of transport (e.g. barges, trains) that can potentially decrease unit transportation costs and the environmental impact of hinterland transport through economies of scale. In this dissertation, we focus on containerized hinterland transport from a multimodal perspective. First, we provide an explanatory study and a framework that portrays the current practices extensively, which is based on 12 in-depth interviews made with various actors of hinterland transport varying in 1-4 hours in duration. Second, we focus on the existing coordination challenges between the actors in practice that decrease the efficiency and the competitive power of multimodal transport. We propose a method to model these challenges in a holistic manner and propose a qualitative method to analyse such models in depth. Third, we focus on the impact of uncertainty in container arrivals and delays, which have not yet been studied quantitatively in literature. We model the barge planning problem as a 2-stage stochastic mixed integer program (MIP) and analyse the impact of uncertainty on total costs and modal split. We carry out numerical experiments using the real data of an inland terminal and show that the delays reduce the flexibility in planning and have a significant impact on costs. Finally, we propose a method which combines data science and optimization to improve barge planning under uncertainty. More specifically, a decision tree is trained on historical data, which is used to predict the delays in container arrivals by using the container specific information such as terminal name, whether the container is refrigerated or not and deep sea vessel call size. These predictions are fed into the scenario generation process of the 2-stage stochastic MIP, which enables more informed decisions while planning barge calls. We conduct a numerical experiment that uses the real data of an inland terminal and the Port of Rotterdam, and show that significant improvements are possible using the existing information better.