Contributed by Dennis Meissner and Bastian Klein
Traffic-related water-level forecasts are a fundamental part of traffic control and information systems for navigation on waterways. These forecasts allow navigational users to optimize the load capacity of their vessels as well as to take into account in time that waterways might be blocked due to floods. On behalf of the German Ministry of Transport, the Federal Institute of Hydrology (BfG), offers operational forecasting services for navigation on the Federal waterways, like the River Rhine being on of the most frequented waterways in Central Europe.
BfG is planning to move from deterministic to probabilistic forecasts within the coming years. In order to demonstrate the usability and the added value of probabilistic forecasts, BfG conducted, together with the Development Centre for Ship Technology and Transport Systems (DST), a study coupling probabilistic water-level forecasts with a cost-structure model. This economic model takes into account a variety of relevant expense factors, like fuel costs, fixed work costs, lighterage etc. The aim was to quantify the transportation costs close to reality for different forecast variants, varying risk behaviour as well as for different types of vessels and routes. The results of this study serve as an element to convince navigational forecast users of the usability and the added value of probabilistic forecasts.
As an example, the figure below compares the best deterministic (blue columns) and probabilistic forecast variant with the current (deterministic) forecast for three routes along the Rhine in terms of transport costs.
It’s clearly noticeable that, depending on the size of the vessel (JW – a small vessel up to 1250t, SV a huge vessel with a high load capacity up to 11000t) and the way of transport, different forecast variants perform best. Especially probabilistic forecasts have the potential to outperform the best deterministic forecast with respect to the economic criteria. As the water-level dependent cost-functions are very heterogeneous amongst navigational users, it is obvious that it is impossible to generate “the best” deterministic forecast for all users as their optima are very heterogeneous as well.
Figure 2 quantifies the economic differences (expressed as Euro per tons) when using a probabilistic forecast (in this example, the median) instead of the deterministic information in order to decide how much to load on a river barge (type JOWI, maximum load capacity of about 6000t) going from the port of Rotterdam to the industrial region of the Rhine-Main-area (black dots).
It’s visible that when the water-levels falls below a certain level (near mean flow conditions) ships whose load capacity was determined based on the probabilistic forecasts are to travel more cost-efficient than those using the deterministic forecast. The differences reach quite significant margins from several euros per ton. The red dots indicate the number of daily requests of the operational traffic-related forecast. As figure 2 shows, the forecast causes economic sensitivities in a water-level range in which the users are highly interested.
In the context of a current research project, the Federal Institute of Hydrology systematically analyses, adapts and tries to optimize methods of uncertainty quantification focussing on navigational users on the inland waterways. At the same time, BfG tries to convince potential users of the added (economic) benefit of probabilistic forecasts in general. Additionally, users should be motivated to include also pre-operational probabilistic forecast in their day-to-day business procedures in order to provide feedback helping to establish an iterative development process.