The use of GIS for one-dimensional modelling of large Dutch rivers


Abstract

The Netherlands have a very intense relationship with the rivers Rhine and Meuse. In the densely populated country a balance has to be found between flood protection, transport, housing, industry, recreation and nature. The recent floods have made it very clear that the rivers still have powerful and threatening natural features. To find a balance between all the uses of the river, one and two dimensional models are being used to evaluate the measures that are foreseen in the river system. In order to keep the one- dimensional models, based on the SOBEK software package, up to date, a GIS based system has been developed to create the models. The system uses data from different sources and in different formats on a 1:5.000 scale. Amongst others, data on altitude, lakes, dikes, flow areas and storage areas and hydraulic roughness are required. Using results from the 2-D model WAQUA, even two dimensional problems are relatively easy to model in a one dimensional model.

The source information is converted into arc-INFO format input for the calculations. Using GRID, an AML based application computes the geometric information for the main section, the bank section and the floodplain. For every calculation a selection is made of the appropriate gridcells. These are used to calculate the flow with and storage width at every level, taking into account the spatial variation of the river system. The calculations for the main section take place on a more detailed scale than for the other parts of the cross section. The output from the application are files that can be used directly by SOBEK. The new way of modelling the river is completed with a system which can be used for a visual check and for presentation of the model. The system gives insight in both the river as a system and the way it is modelled. In combination with other information used by the model, the user gets a quick overview of all the data used in de the model.

The process results in a very robust and consistent way of modelling in which changes are easy to apply in relative little time. After calibration the model results meet the high standards required for flood-prediction calculations. The use of GIS allows for easy updates of the data. The way the model is visualized makes it even for non-hydrologists possible to understand the way the river has been modelled. It allows the Netherlands to keep it's models up to date with regard to the changes that take place in the river system. The current development of a central GIS database for river modelling, and the research on further automatic calculation of specific features will make the method of modelling even more efficient.

Introduction


The Netherlands is the delta of major, busy shipping rivers such as the Rhine branches and the Meuse. The outflow of water, ice and sediment is the main function of the rivers, in order to protect against high water. Recent high water levels have proven that the river claims its right to space from time to time, with all the consequent economic and emotional damage for those involved. The same rivers are used intensively for shipping, potable water extraction and recreation, and form an important ecological axis.

In order to manage their design and utilization properly, one dimensional water movement models are under development, with which the consequences of (proposed) measures on the river can be simulated. These models are based on schematizations in which characteristics and dimensions of the main channel and flood plains of a river are represented as realistically as possible. The development of Geographic Information Systems opens the door to more accurate and flexible schematization techniques. An ArcInfo application has been developed for the one dimensional SOBEK water movement model used by RIZA, in order to rapidly generate unambiguous schematizations from digital geographic information.

This article deals with the schematization technique developed for one dimensional modelling of the major Dutch rivers. Three phases can be roughly distinguished, all three of which will be discussed in this article. The first phase comprises the collection, processing and conversion of data from various sources. Phase two is the translation of this data into profiles readable by SOBEK. The final phase concerns the presentation of the characteristics of the river and the manner in which these are schematized. This allows for visual control of the schematization.


Framework for a SOBEK schematization

The framework for a SOBEK schematization is formed by the way in which the river is divided into a number of one dimensional stretches and consists of a series of consecutive branches and profiles. SOBEK is a one dimensional water movement model which was developed jointly by the RIZA and Delft Hydraulics. Besides hydraulic calculations, it is also possible to apply the model for morphological and water quality calculations. RIZA uses this model for all one dimensional calculations for the Dutch Rhine branches and the Meuse.


branches

A branch is a section of the river and is the basis for a SOBEK schematization. The specific characteristics of a river (section) are recorded in a branch. Calculation points can be given, structural works and monitoring stations indicated and geometric information defined (in the form of profiles). Another possibility is the attribution of certain characteristics such as roughness. An example of how a river can be divided into branches is given in illustration 1. The number and the length of the branches is determined by important system characteristics, comprising forking and merging of river stretches, weirs and important spillways. Through effective gearing of the location of each branch with the river axis and the kilometer values, the location of each characteristic of the river can be clearly recorded in the model.

SOBEK branches

profiles

In a SOBEK schematization, the geometry of a river is described by a series of consecutive cross sections. Each SOBEK profile is as representative as possible a geometric description of a certain river stretch with a fixed length of 500 meter along the river axis, a SOBEK compartment. The calculated profile is situated in the centre of this section. All data in a compartment is used to calculate an average cross section, thus making the profile less sensitive to coincidental deviations in the local geometry.

Right from the lowest level in the channel section to the highest level of the flood plain, the profile contains data on the stream width and storage width. Data is also recorded on the width of various sections (channel, bank and flood plain sections), the average summer dike height, prevention of lakes and the height of the groynes. Illustration 2 gives a schematic representation of a SOBEK profile.

SCHEMATIC

Working method

The geographic data required for the schematization is derived from various sources. A distinction is made between source data, the rough information which comes from many different sources and is in various formats, and basic data files used to make the actual profiles. Three phases are distinguished in the production of the profiles for SOBEK. In phase one, the source data is collected and converted into basic files. In phase two, these basic files are used to record the SOBEK profiles, with the aid of an ArcInfo application. The third and final phase is the visualization of the river and the way in which this is schematized.


Phase 1: Compilation of basic files

Due to the differences in the nature and origin of the source data, various processes must be carried out to arrive at the required set of basic files. Table 1 gives the basic files used for schematizations of the Rhine and Meuse. This paragraph then gives a brief description of the nature of these files and the way in which they have been compiled.

Table 1: Basic files

The main source of data is the D(igital) T(opographical) F(ile) river. This is a vector oriented digital system which comprises many spatial elements of the river. The file is compiled by the geometricians of the Dept. of Public Works on a scale of 1:5,000. The DTF river also contains elevation data in the form of a Digital Terrain Model (DTM).

Compartment division
Each SOBEK profile is divided into three river sections:

The channel section is defined as that part of the summer bed which always flows with the river. For the Rhine branches, this is estimated to be the zone between the regulating boundaries (lines which connect groyne heads with one another). In the Meuse, which has hardly any groynes, a contour line has been used which corresponds with the flooded summer bed at a discharge of 500 m3/s (inundated for around 45 days per annum).
The bank section of the Rhine branches is defined as the zone between the regulation boundaries and the lines connecting the groyne toes. A fixed width of 5 metres on both banks was used for the Meuse.
The outside of the flood plain section is usually bordered by the main dike. Where there is no such dike, the maximum flood limits of extreme high waters (based on aerial photos) was used.

Limits of sections
Each branch is sub-divided into SOBEK compartments for which representative profiles are calculated. The compartment limits of the channel section are perpendicular to the river axis and are extended to the outside boundary of the flood plain on both banks. Where possible, the compartment limits of the flood plains are perpendicular to the local flow direction. Where possible, great differences in the flood plain characteristics between successive compartments are avoided, when determining the limits of the compartments.

Elevation data on channel section and flood plain
The elevation data of the bed of the channel section and bank section is derived from soundings made in cross lines perpendicular to the river axis. The elevation data of surface levels in the flood plain is derived from the DTF elevation model. The point data from both files are combined to form a file with which a DTM is generated to cover the entire area by means of interpolation. A series of preparatory processes were carried out prior to measurements being taken in the channel section, in order to take account of the great difference in the spatial density of the measurements.

Delimitation of flow conveying and storage areas
The amount of water which can be discharged through the flood plain depends strongly on the available stream width. Discharge rates, calculated with the aid of WAQUA, the two dimensional calculation model, were used to determine the boundary between the stream width of the flood plain and the storage width of the flood plain. The discharge rates are calculated at point locations, input in ArcInfo and then converted to a grid file.

In determining the boundary, the assumption was made that all grid cells in the flood plain with a flow rate below 5 cm/s (in the case of Design High Water discharge) are storage cells. For all other parts, the quadrate of the flow rate was determined for each grid cell, after which the average quadrate rate was calculated for each SOBEK compartment. It has been empirically determined that a quadrate rate which is less than half of the average per compartment is a good criterium for the boundary between flow conveying and storage areas.

At low flow rates, the lakes in the flood plains can still make a considerable contribution to discharge, because they allow for flow over a great depth. Based on the discharge pattern of lakes in the storage part of the flood plain, it has been determined whether they should not be considered discharge areas after all. Thus, a clear picture is gained of variations in the flood plain such as stagnant areas behind dikes, bridges and areas not reached by high water. Illustration 3 gives the effect of various steps on a reach of the IJssel, one of the Rhine branches.

Flow conveying area

Summer dikes
Summer dikes are of great influence on the discharge pattern in the flood plain. Two types of summer dikes are distinguished in the schematization method, namely discharge flow and storage dikes.

Dikes are assigned a height per SOBEK stretch, through the use of data from WAQUA, which determines at which discharge level the area behind a dike begins to flow or to store water. The water height pertaining to this level is the height level attributed to the dike and can therefore deviate from the actual height. The dike segments in successive SOBEK compartments have dike heights attributed to them which follow the drop in the river. This ensures that a diked flood plain which reaches over a number of SOBEK compartment begins to flow evenly at a certain discharge level.
The location of the dikes is derived from the DTF river, the other properties are attributed manually to the dikes.

Lakes
In SOBEK, all forms of water which are not part of the river channel section are treated as lakes. Lakes which are openly connected to the main channel are presumed to fulfil a storage function as from the lowest level distinguishable in the channel section. Those lakes lying in the stream part of the flood plain are presumed to participate in the discharge flow over an extra depth of two metres.

Roughness
Besides the use of GIS for translation of the river geometry into SOBEK profiles, other information for SOBEK is determined with the aid of geographical information. A clear example of this is the roughness in the flood plain, which is mainly determined by the types of vegetation (ecotopes) present there. For this purpose, the surface area of the various ecotopes in the stream part of the flood plain is determined for each stretch. Two extremes are applied to determine the roughness per stretch. These values have been used as the upper and lower limits in the calibration of the model.


Phase 2: Calculations


In the second phase, the basic files compiled are applied to calculate SOBEK profiles. An AML based application is used, which works within the GRID module of ARC-INFO. The use of grids facilitates quick execution of complex selections of various files and calculations. Due to the data in the channel section being more densely available than that in the flood plains, the calculations in the channel section are based on a smaller cell size than those in the flood plain. The output consists of ASCII output files which are directly readable by SOBEK.

The calculation of profiles takes place by selecting those grid cells which meet the right conditions. The surface area of these cells is determined, and then their width is calculated by dividing their surface area by the length of the section. An example of such a selection could be: all cells in compartment 28, in the flood plain, which are part of the stream width, are not behind a dike and have a surface level under 20.00 m + sea level.

The calculations can be subdivided into calculations for the channel section, the bank section and calculations for the flood plain.


Channel section
In order to restrict the calculation time required, the channel section has been divided into five elevation levels. This was possible on the condition that the combined volume stored under the various levels is equal to the total volume that can be stored in the channel section. In order to meet this condition, the channel section is divided into four equilateral trapeziums for which the elevation level, the stream width and the storage width are calculated.

The river bed can vary considerably in height. For the purpose of schematization, it is important to determine a representative bed level per SOBEK compartment because otherwise the bed level between successive compartments will be irregular. This will result in instability of the model and is also undesirable for morphological calculations. The minimum level in the channel section has therefore been determined as the value exceeded by 90% of the grid cells.

Lakes which are openly connected to the river are presumed to fulfil a storage function as from the lowest schematized level. The width of these lakes is added to the storage width in the channel section.

Bank section
The bank section is schematized by two levels, which give the lowest and highest levels of the section. If the SOBEK compartment contains groynes, the discharge flow in the bank section will be hindered until the groynes are under water. Storage increases up until this level, while the discharge flow increases as the level of the average groyne height is exceeded.

Flood plain section
The flood plain section is schematized in a variable number of elevation levels until the highest level of the DTM has been reached. Stream widths and storage widths are determined for all elevation levels distinguished.

An important level in the schematization is the height of the summer dikes. Parts of the flood plain behind summer dikes only contribute to discharge flow and storage once the height of the dike is exceeded. At least one dike height is input in each SOBEK profile. If there are a number of dikes, e.g. on both the left and right hand banks), the dike height used is the one which has most influence on the discharge flow. At this level, the surface area and average surface level elevation of the area behind dikes is determined and included in the profile. If the water breaches a dike, the area does not contribute to the flow at once. The stream width of the area behind a dike is therefore gradually added to the schematization as a function of this change in width. Those lakes lying in the stream part of the flood plain are presumed to participate in the discharge flow over an extra depth of two metres in the schematization.


Phase 3: presentation and visual control


In the past, the schematization of a one-dimensional water movement model was difficult to understand. The one- dimensional model cannot distinguish between elements on the left and right hand banks. Rivers such as the Meuse and IJssel in particular, where a number of complex discharge patterns occur, such as cut corners and channels perpendicular to the river axis, often allow for personal interpretations during schematization, which are difficult to follow at a later date.

With the introduction of geographic information and the new method of schematization, a system has been developed which enables unambiguous visualization of the river, the method of schematization and the model input. An added advantage is that a rough estimate of the effect of certain measures on a location can be made very quickly. Illustration 4 shows the description of profile 32 on the Meuse.


SOBEK profile information
The figure has been compiled by using the available basic data to give a birds' eye view of the SOBEK section. This gives a clear picture of the river and relevant topographical elements such as dikes and lakes. Changes in characteristics of successive SOBEK compartments are also visualized.

Using the same type of grid selections as in the profile calculation, the maximum width of all elements involved in the schematization is determined. Selections are made per bank this time, so that the share of both banks in the flow and storage becomes clear. These widths are presented both in table form and graphically in the form of relative shares of the total width. They are given in the central figure.

The bottom part of the figure gives the SOBEK profile and the location of the profile in the river branch.
This information is not isolated. Each profile requires a description containing supplementary information such as the location of bridges, monitoring points etc. Also when there is a known relationship between water level and discharge at a particular point, this information is given on an opposite page so that all available information on the river stretch in question is available at a glance.



Model results


The schematizations of the Meuse and Rhine branches gained using this method apparently yield good results. The SOBEK model for the Meuse makes it possible to reproduce the water levels measured at the monitoring stations with an accuracy of 0.05 m. The roughness of the channel section is hereby equal over large stretches and remains within realistic physical values, as does that of the flood plain.

The DTF river is not ready for the Rhine branches, so that large-scale information has been used for the schematization. However, at high discharge rates the margin is once again only a maximum 0.10 metres between the measured and calculated water levels. Larger differences between measured and calculated values can mainly be traced back to errors in the schematization (particularly discharge flow and dike heights) and effects of the bottom limit of the model (Van der Veen et al., 1996).



New schematizations and techniques


The sketched method has already been applied up to km 96.5 of the Meuse and to the Rhine branches. Schematization of the Meuse up to km 225 will be completed by the end of 97. In 1998, the Rhine branches will be re-schematized, by which time the DTB river for the Rhine branches will be available. Based on geographic information of a different nature, a schematization of the German Rhine is being made from Andernach (km 614) to Lobith (km 862).

At present, research is underway into the establishment of a central GIS database for 1-D and 2-D river models. This database will allow 1-D and 2-D modelling to be based on exactly the same basic data. Realization of this database signifies the next major step in the standardization of river modelling. It will also be possible to implement measures in the river system in the database in order to make a new schematization. The database will be ready in mid '97.

In one-dimensional modelling, an important assumption is that the drops in the summer bed and flood plain are equal to one another. The boundaries of the floodplain sections are still determined by hand at present. Studies are underway whether the water level field of the 2-D model can be used to determine iso-water level lines which comply with the boundaries in the channel section. Similarly, research is underway as to whether the drop in the channel section can be used to determine the stream dike height in successive SOBEK sections.



Conclusions


The use of geographic data for one-dimensional modelling of the large rivers has great advantages in terms of the consistence, accuracy and production speed of schematizations.

The use of one set of source data for both WAQUA and SOBEK makes it possible to very accurately determine the boundaries of the compartments and discharge flow for SOBEK. The results of the schematization are very consistent and solid and allow for calibration whereby parameters remain within realistic physical limits.

The elevation of the channel section, and the discharge flow and dike heights in the flood plain are apparently important factors for the course of the stream. These aspects are included effectively in the presented method of schematization, as are local variations in the river. The use of ecotopes for determination of the roughness also makes it possible to univocally schematize nature development plans.

Despite the fact that every attempt was made to automate the steps of this process, schematization must not take place fully automatically.Hhuman input (expert judgement) is essential in each phase. However, a great advantage is that the level of personal interpretation is limited, so that different schematizations can be mutually compared.



Literature


Pakes, U., R. Van der Veen and J. van Essen
SOBEK schematisatie Maas, gebruik en presentatie van ruimtelijke gegevens
(SOBEK schematization of the Meuse, use and preparation of spatial data)
RWS-RIZA 1997, in prep.

Schutte, L and R. van der Veen
Bepalen ruwheden winterbed op basis van voorkomende vegetatie
(Determination of winter bed roughness according to prevailing vegetation)
RWS-RIZA, working document 96.177x

Veen, R van der, U. Pakes, J. van Essen, L. Schutte
Calibratie SOBEK Rijntakken
(Calibration of SOBEK Rhine branches)
RWS-RIZA 1996, memo in prep.

Zeeman, M
Schematisatie voor het SOBEK model Nederlandse Rijntakken, rapportage GIS analyses
(Schematization of the SOBEK model for the Dutch Rhine branches, report on GIS analyses)
RWS-RIZA/Dir. Gelderland, Geodan Geodesie, 1994

Zeeman, M
Schematisatie voor het SOBEK model Grens- en Plassenmaas, rapportage GIS analyses
(Schematization for the SOBEK model for the Grensmaas and the waters adjacent to the Meuse, report on GIS analyses)
RWS-RIZA/Geodan Geodesie, 1996

Ubo Pakes(1), Rolf van der Veen(1), Menno Zeeman(2), Johan van Essen(1), Bertus Schutte(1)
(1) Institute for Inland Water Management and Wastewater Treatment (RIZA),
Rivers department P.O. box 9072,
6800 ED Arnhem,
The Netherlands
tel +31 26 3688 911
fax +31 26 3688 678
email: u.pakes@riza.rws.minvenw.nl


(2) Geodan Geodesie b.v.
Overtoom 60-IV
1054 HK Amsterdam,
The Netherlands
tel. +31 20 6125 073
fax. +31 20 6163 848
email: zeeman@geodesie.geodan.nl