Project Description


The project's overarching objective is to develop a hydrodynamic model properly embedded in the observational network that allows accurate, seamless forecasting of total water depths in the Dutch North Sea. The model resolves all relevant physical processes in the North Sea, including the coastal waters; it will exploit tide gauge and radar altimeter water levels, in-situ temperature/ salinity profiles, and remotely sensed sea surface temperature data in the framework of data assimilation; and its output is complemented by a realistic accuracy description. Here, we identify the deficiencies of the current operational systems, define the main objectives of the three separate subprojects, and illustrate the subprojects’ mutual interaction.

The Challenge

For the largest ships, the access to the main Dutch harbors and their approach routes is already restricted to certain stages of the tide, referred to as tidal windows. Yet, ships still become bigger and ship drafts deeper! Consequently, more and more ships will have restricted access with smaller tidal windows. This occurs in a challenging environment that is not only one of the busiest sailing areas in the world, but that is also very dynamic; with propagating rhythmic sea floor patterns, occurring at different spatial-temporal scales, and with a sea surface exhibiting large tidal motions, waves, and suffering several severe storms a year. However, ship owners are looking for profit optimization within the constraints imposed by the safety limits for ship draft, even when carrying harmful substances such as crude oil. Therefore, the challenge is to optimize the use of the available waterway capacity without compromising safety, and at the same time avoiding prohibitively expensive dredging campaigns.

What is needed to face the challenge

The key element in the required infrastructure to face this challenge is a system that predicts the evolution of the sea floor topography and water levels, respectively. The former is subject of the on-going NWO-TTW project 13275 “SMARTSEA”. The latter is addressed in this project. The heart of a water level prediction system is a hydrodynamic model linked to a strong and extended observational network. Given the large environmental, economic, and/or social consequences in case of accidents, it is crucial that the model forecasts of expected water level, used to compute under keel clearances, are complete, accurate, and complemented by a realistic accuracy description (i.e., one that is a function of space and lead time), and desirable that all available observations are exploited in the framework of data assimilation. As shown below, these criteria cannot be met with the operational water level forecasting systems currently available in the Netherlands. Our primary objective is to design a new hydrodynamic model that is properly embedded in the observational network and that does meet all above-mentioned criteria. The model is intended to provide water levels, temperature, and salinity.

Why we can't rely on the current operational systems

The continuous investments in hydrodynamic model development in the Netherlands, well aware of its vulnerability to flooding, are primarily driven by the need for reliable storm surge forecasts. Indeed, massive improvements of the model accuracies have been achieved; the latest storm surge model represents water levels with a root-mean-square error of ~7-8 cm. At the same time, this statistics is to some extent misleading; it is computed after applying an approximate bias correction that originates from the fact that the model is barotropic (2D), i.e., the baroclinic/ steric contribution is missing. Missing the baroclinic forcing has three consequences: i) the modeled water levels are not complete, ii) any interaction between tides, surges, and the baroclinic water level is ignored, and iii) the modeled water levels lack an absolute vertical reference. Although work-arounds have been developed to counteract the above consequences these are not always valid. For example, recent work conducted by some of the project team members as part of the recently finished STW project 12553 “Vertical Reference Frame for the Netherlands Mainland, Wadden Island and Continental Shelf (NEVREF)”, has shown that the misalignment between the density-induced pressure gradients as taken from an existing 3D baroclinic model and the bathymetric contours of our 2D storm surge model causes instabilities and erroneous tilts in the water level that may reach decimeters. Moreover, these work-arounds are hard to operationalize.


Nowhere is the missing baroclinic contribution more critical than at the mouth of the Rotterdam Waterway. At the entrance to the Maasvlakte 2, lenses of fresh-water are ejected every ebb tide. They spread forming ever larger lenses that merge and interact with the lenses emitted on the previous tidal cycles forming the Rhine Region of Freshwater Influence (ROFI). The near field ‘bulge’ and downstream ROFI region is highlighted in the figure (image courtesy Gerben de Boer) and is populated by multiple fronts. It is dynamic, with significant tidal currents modified by baroclinic effects, due to the highly variable temperature and salinity. 2D barotropic models can neither capture this variability nor resolve the strongly sheared counter-rotating tidal currents. The 3D model we will develop and use in this project solves all three aforementioned problems.

However, accurate water level forecasting also requires the assimilation of measurements. In this project, we consider the assimilation of observed water levels from tide gauges and satellite radar altimeters, in-situ temperature and salinity profiles, and remotely sensed sea surface temperature data. Regarding the first data type it should be noted that after adding the baroclinic forcing to the model the full water levels need to be considered, i.e., the bias correction approach no longer applies. This in turn requires that we have to adopt one and the same height datum for both the model and all water level observations, and apply corresponding datum transformations accordingly. For radar altimeter data this involves a simple reduction of the sea surface heights to the (quasi-)geoid, but for tide gauge data this is less trivial as each country has its own height system, and the relations between them are not accurately known. In any case, the adopted height system should be accessible at all tide gauges (including the ones located at islands and offshore platforms) and have centimeter accuracy; even small errors imply large erroneous water fluxes and tilts in the water levels and are a potential source for model instabilities.

Versatile Hydrodynamics


The challenge to realize a unified height system that has the required coverage and accuracy on the one hand and an accurate hydrodynamic model whose vertical reference is defined consistently with this height system on the other hand is the topic of subproject P1. P2 aims to develop the hydrodynamic model with a resolution sufficient to resolve the most important physical processes in the coastal waters including the Rhine ROFI. The detailed input needed to calibrate this model requires innovative strategies for acquiring temperature and salinity profiles; currently such information is very sparsely available. We will develop a novel processing strategy to derive such data from multibeam echosounding (MBES) measurements. In P3, we will develop the assimilation techniques that allow full exploitation of all available observations, and we will design an operational methodology to describe the accuracy of the model forecasts. This, combined with the results of P1 and P2, allows us to build the prototype of a seamless forecasting system for total water depths (total water depth = bathymetry w.r.t. the (quasi-)geoid + forecasted water level) in the Dutch North Sea (also part of P3). The connection among the three subprojects, their key contribution to the development of the forecasting system, and the applications that will be developed in the subprojects are illustrated in the figure.

Subproject P1: Realization of a regional height system

In P1, we will reference both the tide gauge water levels and the hydrodynamic model to a unified height datum. This is a prerequisite to assimilate full water levels and to forecast total water depths, respectively. The designated choice for such a datum, the European Vertical Reference Frame 2007 (EVRF2007), lacks the required coverage and accuracy. The key limitation is that the EVRF2007 is mainly based on spirit leveling, which i) is prone to systematic errors, ii) is expensive, and iii) cannot be used to cross large water bodies. The latter implies that islands and platforms cannot be connected and that the connection of coastal countries, such as the Netherlands to the spirit leveling network used to realize the European Vertical Reference System (EVRS) is intrinsically weak. In P1 we solve these problems by adding model-based hydrodynamic leveling data to spirit leveling data, connecting points separated by water bodies and along the coast. Development of the technique to conduct model-based hydrodynamic leveling is the key novelty of P1.

Contact: Dr. ir. D.C. (Cornelis) Slobbe -

Subproject P2: The Rhine outflow area & MBES-derived salinity/ temperature profiles

For the Netherlands, the region at the mouth of the Rotterdam Waterway is an area of significant economic activity and importance. Here the fresh river water outflow from the Rhine and Meuse rivers, creates the Rhine ROFI (Regions of Freshwater Influence), an area with large spatial and temporal variability in the water column. The key novelty of P2 is twofold: i) to develop a high resolution hydrodynamic model which resolves the fronts and lenses, and ii) to develop a methodology to determine salinity or temperature profiles from an analysis of multibeam echosounding (MBES) data, which then are used to calibrate and validate the hydrodynamic model in the ROFI. This data is indispensable for an accurate prediction of water depths in these areas, as it is not possible to place moorings in the busy shipping lanes. We will use the model to i) compute the MDT needed in P1, ii) understand the processes contributing to it, and iii) determine which conditions (e.g., river discharge, wind, waves) result in extreme low water levels.

Contact: Prof. dr. J.D. (Julie) Pietrzak/ Dr. ir. M. (Mirjam) Snellen -

Subproject P3: A seamless forecast of total water depth

Due to limited predictability, our ability to forecast the total water depth decreases with increasing lead time. On the other hand, our ability to act is reduced with reducing lead time when the event approaches. Ideally, forecasts are updated frequently in real-time to include the latest observations and weather forecasts. Moreover, the forecasts should be seamless, i.e., without any jumps caused by mixing different methodologies conventionally used to compute forecasts over different time horizons. The contribution of this subproject is to develop i) the assimilation techniques that allow full exploitation of available observations, and ii) an operational methodology describing the accuracy of the model forecasts. Together with the contributions from P1 and P2, this enables us to build a prototype of the seamless forecasting system for total water depths in the Dutch North Sea. Building and validating this prototype is also part of P3.

Contact: Prof. dr. ir. M. (Martin) Verlaan -