6 
Introduction 
Strictly speaking, such surge graphs apply only under stationary conditions when circulation dynamics 
in the German Bight, after one or two days, have adapted to the wind field (KOLTERMANN and LANGE, 
1975). 
Tide tables for marine harbours based on astronomical predictions have been in use for a long time 
(ANON., 1878). Also the influence of weather on water levels had been recognised very early (ORTT, 
1897). But it took a refined synoptic method to obtain fairly reliable wind forecasts for the open sea 
(SCHERHAG, 1948), which later, since the middle of the 20th century when high-performance remote 
data transmission became available, have led to relatively good water level forecasts for the German 
coastal waters and marine harbours (MÜLLER-NAVARRA, 2009a). Today, water levels in the German 
Bight are predicted using both empirical methods and numerical models (MÜLLER-NAVARRA et al., 
2003), which are linked by a human interface (man-machine mix, MMM). 
Analysing the prediction error, expressed as root mean square error (RMSE), for different prediction 
periods in the time from 2003 to mid-2009, it has been found that the quality of model predictions for a 
period of up to 15 hours can be improved by applying a man-machine mix (MMM). Beyond that, it has 
not been possible to further improve the direct model output (DMO) of the 2-dimensional model 
(DMO-2D) by human intervention (Fig. 2). The 2D model has been optimised for water level prediction 
applications; the RMSE of the 3-dimensional model (DMO-3D) generally is more than 2 cm greater in 
most cases. Besides, because of its longer computation time, the 3D model does not use weather 
predictions prior to the twelfth prediction hour. 
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 
forecast time [h] 
Fig. 2: RMSE (cm) of DMO (-2D and -3D, version 3:1/2003 -12/2007, version 4:1/2008 - 8/2009) and MMM. 
Cuxhaven harbour. Number of cases for each forecast time: MMM: about 1500, DMO-2D: 700, DMO-3D: 350. 
It will be demonstrated in this paper, on the example of the river Elbe with its heavy traffic of seagoing 
ships, that a marked improvement of water level predictions for the harbour of Cuxhaven can be 
achieved by applying model output statistics (MOS), i.e. by subsequently applying statistics-based 
corrections. The model output statistics (MOS) method has been used successfully in meteorological 
operational forecasting since the 1970s (GLAHN et al., 1972). This novel method of predicting wind 
surge will be called BSH-MOS in the following.