i4cast’s water level forecasts on the Rhine River

Artificial intelligence techniques applied by i4cast® generate consistently more accurate water level forecasts on the Rhine River , with the benefit of up to 21 days of more reliable forecasts!

Read on for more details about this study.

Generic forecast

Global or regional forecasts are critical for monitoring weather conditions and are successfully used in different applications.

ELWIS, for example, provides numerical forecasts for the entire river Rhine, being the main online German waterway information portal. It is a high quality service, widely accepted and used in the transport and navigation sector in the region.

However, when it comes to activities in which decision-making assertiveness is associated with climatic conditions in micro regions, the challenge becomes different. In this case, the application of artificial intelligence techniques associated with numerical predictions can generate more accurate and adequate information for each type of activity.

In the case of the Rhine River, this challenge is mainly fueled by increasingly unpredictable weather patterns that include prolonged periods of heavy rains or drought conditions along the river, both of which can cause large-scale disruptions to one of the shipping waterways. most important in Europe - decrease in transported cargo and surcharges.

It is to solve specific needs associated with the impact of environmental conditions that we developed i4cast® - a weather intelligence system that applies Machine Learning techniques to offer even more assertive forecasts.

Here you can learn more about how they are made and the difference between generic predictions and high resolution predictions.

The i4cast® on the Rhine River

Severe fluctuations in water levels on the Rhine River in recent years pose a recurring challenge for those who rely on river transport, especially in Kaub, one of the Rhine's narrowest points in Germany.

The Rhine meanders over 1,300 kilometers across much of the European continent, from Switzerland to emptying into the North Sea at Rotterdam. It is a crucial transport route for commodities, transporting cargo barges through some of Europe's most important industrial zones.

i4cast® uses multiple data sources from the entire catchment area of the River Rhine and its tributaries, provided by WSV (German Federal Administration of Waterways and Navigation) and DWD (German Weather Service), to provide 10-day forecasts at various strategic points along the river.

Based on this wide range of data, i4sea has developed a machine learning model that reproduces water flow patterns at locations upstream of the area of interest – which improves over existing public forecasts in the region, particularly with the increase effectiveness in extreme flood or prolonged drought events.

As a result, i4cast® provides 21 days of better water level based navigation decisions for Rhine River locations compared to existing predictions.

Check out the analysis for the stations in Kaub and Koln, focusing on the third day of forecast. This is the average navigation time between the two stations and, therefore, a critical period for decision-making on barge loading.

i4cast® forecast reliability in Kaub

An analysis of forecast accuracy, within a 10 cm range, showed a 5% gain in assertiveness for i4cast® over public forecasts for a 3-day forecast. Which means 18 days of better decisions a year in Kaub.

i4cast® forecast reliability in Koln

The same analysis performed for the station in Koln, showed a gain in assertiveness by 6% of i4cast® in relation to public forecasts for a 3-day forecast. Which adds up to 21 days of better decisions a year in Koln.

Therefore, the i4cast® model delivers more accurate results than existing public forecasting services much more consistently at specific locations. Like Kaub and Koln, where generic forecasts can present large errors in certain forecasts - especially in low-level river situations, which are critical for decision-making on vessel loading.

AI-based predictions

Using deep learning techniques to predict water levels, i4cast® can provide 10-day forecasts across the Rhine region and with similar accuracy for many unique river locations that are not covered by public forecasts. Like Maxau and, on the Main River, Raunheim, Frankfurt and places customized according to demand.

If your operations are impacted by environmental conditions in river environments, request a demo.