GIS Based Hydrological Modelling of Alluvial Regions

Using the Example of the Kisalföld (Hungary)

The research was financed by and carried out at the International Institute for Aerospace Survey and Earth Sciences (ITC), Enschede, The Netherlands, with the support of the Water Resources Research Institute (VITUKI), Budapest, Hungary, and the Cartography Department of ELTE

    Problem identification. Objectives and main steps of the research.
    Methodological approach to the comprehensive spatial analysis of the regional hydrology. Data integration and preprocessing in GIS. Integration of remotely sensed data into GIS. Single-variable spatial analysis. Characterization of spatial pattern. Interpolation. Spatial class obtaining methods. Multi-variable spatial analysis: regionalization.
    Topography, geology, climate, cover layer, land use, hydrography, water management.
    Mapping the thickness of the cover layer (spatial analysis, accuracy assessment). Hydraulic conductivity maps (spatial analysis, accuracy assessment).
    Separation of crop-covered and bare surfaces by NDVI slicing (sample area: Szigetköz). Accuracy assessment and extrapolation from the Szigetköz to other subregions.
    Delineation of the hydrological system. Regional water budget (storage, precipitation, surface water in- and outflow, actual evapotranspiration, evaluation of the water budget).
    Quick mapping (deep recharge, maximum evapotranspiration loss from the groundwater, ET surface depth, extinction depth). Regionalization using 1D flow modelling (theory and parameters of Hydrus, determination of the mapping units, calibration and verification, simulation of the unsaturated flow, comparison of the applied regionalization methods).
    River-groundwater contact types. Solution of the governing equation of the confined groundwater flow, calibration, sensitivity analysis. Calculation of the effect of the Danube on the confined part of the Quaternary aquifer. Solution of the governing equation of the phreatic groundwater flow, calibration, sensitivity analysis. Calculation of the effect of the Danube on the phreatic part of the Quaternary aquifer. Spatial extension of the calculations. Validation of the river-influenced zone map by lag correlation analysis.
    Conceptual model of the groundwater system. Coupling Modflow with GIS. Grid design and model construction (horizontal and vertical discretization, generalization of maps, hydraulic conductivity, external source/sink terms, boundary conditions, calibration, sensitivity analysis). Modelling results (analysis of the flux between the surface waters and the groundwater, analysis of the fluxes across the groundwater table).


The study investigated how spatial analysis improves the insight into the hydrological processes of alluvial regions, where surface waters and groundwater are strongly interrelated. The following a priori objectives were formulated:

As can be seen, the study focuses on mainly methodological issues. The Kisalföld, Hungary, (Figure 1) was chosen as the study area because a very rich database is available from this region.

Fig. 1.


Qualitative and quantitative methods have been used in a GIS environment to analyze the water regime, and to describe and simulate hydrological processes in the Kisalföld. The structure of the thesis follows the main steps of this analysis and modelling (see also Page 2 and Figure 2). The discussion focuses on the comprehensive analysis of the hydrology (Figure 2), which leads to the construction of a conceptual model of the investigated area:

Fig. 2.


The methodological conclusions and the results of the hydrological analysis of the Kisalföld are organized in the following groups:


GIS technology supports the comprehensive analysis of the regional hydrology by providing tools for:

The conclusions of the study concerning the above functions of GIS are summarized in the following subsections.


Data integration is one of the most powerful functions of a GIS. This has been intensively used in the construction of a raw data set for the Kisalföld. Data transformations were restricted to geometric and sometimes temporal transformations, and a data model transformation (such as vector to raster conversion) was applied only when a specific application required it in the later steps.
Basically, three different kinds of raw data can be differentiated:

Directly measured hydrological data determine the possible spatial and temporal scale of the analysis. Surrogate quantitative hydrological data and qualitative hydrological data need further processing and analysis to assess hydrological parameters.
Transfer functions are used to deduce values of hydrological parameters from surrogate hydrologic data. An example is shown in Chapter 5, where statistical relations between spectral characteristics and land covers are established to separate the vegetated and the bare surfaces by NDVI slicing. The resulting classes are qualitative data, which were transferred to values of root distribution by using empirical transfer functions.
Flexibility in access to the data is one of the most important factors of data preprocessing, because several different kinds of data and several different methods are used in the comprehensive hydrological analysis. There is no need for a complex database structure because it would result in a data (structure)-driven situation. In fact, a database with established spatial and structural links can be designed only on the basis of the comprehensive analysis of the hydrological system. Thus the applicability of the object-oriented data structure - which is a popular field of the research in GIS - in this case is very limited. An object-oriented approach needs the exact determination of the objects and their relationships, but these may become known only as a result of the hydrological analysis.


Spatial analysis highlights the spatial characteristics of the variables.
The spatial characteristics of a variable can be analyzed from a data set if it was obtained by a sufficiently dense observation network. In the case of the cover layer thickness data of the Kisalföld, the sufficiency of the network was evaluated by analysing the effect of different network densities on the representation of the subregional spatial variances. The different network densities were generated by stratified random resampling from the original data set. It could be concluded that the density of the observation network sufficiently represents the overall variance of the cover layer thickness in the investigated area. The analysis of the subregional variances highlighted differences in the spatial characteristics of the cover layers of the Danube and the Rába alluvial fans, the two fans were, therefore, tackled separately in the later steps.
Density of networks also may be described qualitatively. A qualitative description of the drainage densities was used in the calibration of the groundwater flow model of the Kisalföld.
Variogram analysis (or the analysis of any other spatial continuity functions) describes the spatial continuity of a variable. Spatial continuity is related to the range of a variogram. If the range is relatively small compared with the mean distance between the observation points, the spatial continuity of the variable cannot be represented properly, as was found in the case of the thickness and the hydraulic conductivity of the cover layer in the Kisalföld.
Moving window statistics highlights the spatial anomalies of a variable by calculating the local descriptive statistics, such as the local means or the coefficients of variation. Moving window statistics proved the presence of a proportional effect in the cover layer thickness data of the Kisalföld.
When using a raster data model, the resolution of the map of a continuous spatial variable (i.e. the grid size) is determined by the permissible largest gridding residual, and the largest local gradient of the variable as it is described in Chapter 4. Densification of a raster map beyond the optimal resolution - which might be a temptation when large computer capacities are available - does not result in a better representation of the variable.
Problems occur when maps of different optimal resolutions have to be used together in the further analysis. The maps have to be transformed to a common scale using proper operators for the generalization, e.g. majority operators are more meaningful for categorical coverage maps (e.g. as were used in the generalization of the land use map for groundwater flow simulation in Chapter 9) than the arithmetic mean.


Analytical and visualization functions of GIS were used to describe the hydrological system and to analyze the processes (Chapter 6). The regional water budget calculated from hydrometric data contained higher uncertainty due to unavoidable inaccuracies in the measurements (e.g. river discharge) than the magnitude of some important subregional and boundary fluxes of the regional flow system.
Regionalization is the most complex spatial analysis method. It can be described as the derivation of one categorical spatial variable from several other spatial variables. A systematic approach to regionalization proposed by Simmers was modified and applied in a GIS environment to analyze subregional fluxes across the groundwater table and between the rivers and the groundwater.
Surface recharge and evapotranspiration loss from the groundwater are the fluxes across the groundwater table. They cannot be measured directly on a regional scale, therefore they have to be deduced from other variables (Chapter 7).
It is relatively easy to implement empirical transfer functions. The disadvantage of such empirical functions is that they are not based on objective laws of the nature; thus their accuracy depends on subjective factors.
Physically-based numerical methods are usually more complex than the empirical transfer functions. For example, a one-dimensional variably saturated flow model (Hydrus) was used for the simulation of the recharge/evapotranspiration processes in the unsaturated zone. Quasi- steady-state simulation of the flow was carried out for mapping units formed on the basis of sediment layering types and land use. Such solutions suffer from the scale effect. Point measurements have to be used for the calibration of the hydraulic parameters of the flow medium. In the case of the Kisalföld, the hydraulic parameters of four sediment types were determined on the basis of the data of soil moisture measurement sites: clay, loam, sand and gravel. Such calibration may reflect the local variance of the hydraulic parameters, which is then erroneously extrapolated to the whole area. Similar difficulties may occur in the determination of other parameters such as the root distribution which determines the water uptake by the plant. Vegetation related factors are handled in Hydrus in a simple way, which may also result in conceptual errors.
It can be concluded that the recently available transfer functions - simple or complex makes no difference - cannot describe the processes in the unsaturated zone on a regional scale with high accuracy. It is difficult to determine the spatial variability of the process from point measurements. Up-scaling of unsaturated flow, determined on a local scale, is therefore highly uncertain.
The groundwater heads and flow vary strongly in the zones influenced by the rivers. An analytical solution of the flow equation was implemented in a GIS environment to assess the effects of floods on the groundwater heads (Chapter 8). The result highlights the spatial variability of the river effects on the groundwater heads. Two major simplifications were applied in the model: one was related to the geometry of the flow system, and the other was related to the homogeneity of the medium. Large discrepancies could be found in places where the simplifications did not describe the system properly. Parameter determination problems also occurred, similar to the ones shown in the surface recharge modelling.


Groundwater modelling is the key to the quantitative description of the hydrological system of alluvial areas. The most important GIS functions which support groundwater modelling are summarized in Table 1. In reality, the steps of the modelling procedure cannot always be separated from each other (e.g. all preprocessing steps need data collection), but for sake of clarity they are listed one-by-one in the table.
The most important point in coupling a GIS and a hydrological simulation code is that the analytical tools of the GIS must have direct access to the input and output data structures of the simulation code. This can be solved by data transfer programs. Although the coupling is more a technical than a scientific issue, some specific functions have to be added to the GIS software to support hydrological analysis. These functions may help the determination of the model geometry, and may provide a framework for routinely repeated operations such as visualization, statistical calculations etc. In this sense, the frequently applied pre

  • and postprocessors of different computer codes for hydrological modelling may be considered as special GIS software, although they often have limitations in spatial analysis.

    Table 1 GIS functions in the different steps of groundwater modelling
    Modelling stepGIS function
    PreprocessingData collectionData integration: retrieval from external databases, import from different file formats, merging maps
    Identification of area to be modelled (topographic map - geometric background) Digitizing
    Raster vector and vector raster conversions
    Mesh design (model code specific step)Interactive editing of segment
    and polygon maps
    Coordinate transformation
    Map calculations
    Creation of surfaces (continuous spatial variables, e.g. top and bottom of a layer)DEM generation
    Map calculations, neighbourhood calculations,
    Regionalization (categorical spatial variables), modelling of subprocessesMap calculations
    Spatial statistics
    Reconstruction of 3D objects (e.g. aquifers)Spatial statistics
    Creation of and data retrieval from 3D data sets
    Visualization of 3D surfaces, sections
    Transfer of assembled data to the model code (input file generation)Automation by programming capabilities (batch files, macros, modelling language) for the geometric control of input data and file format conversion
    Calibration, sensitivity analysis, verification, predictionEvaluation of the resultVisualization, overlaying
    Retrieval of values from maps, statistics of a whole map, statistics of selected areas of one or several maps
    Map calculation
    Section editing from 2D and 3D data
    Time series analysis
    Animation of the results of consecutive runs
    Animation of the results of consecutive time steps
    Modification of parametersVisualization, overlaying
    Editing of map values
    Map calculation
    Scenario analysisVisualization, overlaying
    Retrieval of values from maps, statistics of a whole map, statistics of selected areas of one or several maps
    Animation of the results of consecutive time steps


    The Kisalföld (Small Plain) region in Hungary is of alluvial origin. It is located in the centre of a down-warping tectonic basin which was filled up gradually in the Tertiary by fine- grained marine and lacustrine deposits with some embedded sand layers. Sandy and gravelly fluvial deposits covered them in the Quaternary, reaching 700 m thickness at the fastest down- warping part of the basin. The central part of the Kisalföld is referred to as Győr Basin. The alluvium in the Győr Basin contains one of the largest freshwater reserves of Central Europe.
    The Danube is the largest river of the area. Due to the construction of a hydropower station, the largest part of the discharge of the Danube has been diverted into an isolated side channel in 1992. Since the groundwater and the surface waters are in strong contact in the Kisalföld, the diversion of the river effects the groundwater heads. The present study examined the interaction between the subsystems of the hydrological regime of the Győr Basin, and determined the effects of the river diversion on the groundwater heads and the fluxes between the hydrological subsystems.
    The results of the comprehensive analysis of the hydrological system of the Kisalföld were the basis for constructing the groundwater flow model of the region. Model parameter values were derived using the methods summarized above.
    A priori calibration targets were set by determining the acceptable statistical and spatial distribution of the errors. In addition to the usual quantitative methods (e.g. comparison of simulated heads to measured heads), some qualitative methods were also used in the calibration of the groundwater model:

    The calibration results were compared with the maps created by the recharge/evapotranspiration regionalization methods (Chapter 7). It was found that the regionalization methods resulted in an overestimate of those parameters, which have an effect on the simulated loss by evapotranspiration from the groundwater.
    Some discrepancies were also found between the results of the regionalization of the river- influenced zones (Chapter 8) and the calibrated groundwater model. For example, the reach of the Moson-Danube above Mosonmagyaróvár was originally considered as an isolated river bed, but the model required a considerable recharge to the groundwater from this reach.
    Subregions - termed as water budget units - were formed to calculate the subregional fluxes. These units were delineated on the basis of drainage densities and the widths of the river-influenced zones. This subdivision of the Győr Basin provided a basis for a detailed analysis of the interactions of the different parts of the hydrological system. Such a subdivision can be extremely useful in regional planning.
    Two scenarios were analyzed: the -initial- scenario which simulated the mean annual groundwater heads and fluxes of the period 1979-89, and the -diverted Danube- scenario, which simulated the mean annual situation after the diversion of the Danube. The analysis of the scenarios showed that the rivers play a more important role in the recharge to the groundwater than the surface recharge, especially on the Danube alluvial fan. It was found that a considerable drop of the annual mean groundwater heads occurs in the northern and middle part of the Szigetköz if no artificial recharge measures are applied. The loss in the available groundwater for the vegetation would occur in the flood plain forests, and in the middle part of the Szigetköz.
    Because of its equidistant grid of 1000 m spacing, the developed model describes only the regional processes. This resolution does not allow the simulation of the local processes, but this regional model may serve as a starting point for modelling local flow and transport processes in the future.


    The main objective of the study was to improve existing methods for obtaining insight into the hydrology of alluvial areas on the basis of spatial analysis in a GIS environment.
    On the one hand, evidence has been provided in this study of the important effects of local variations of the hydrological variables, e.g. hydraulic properties of sediment types, cover layer thickness. A part of these local variations cannot be described because of practical and theoretical limitations:

    The conclusion is that deterministic up-scaling is not a proper approach for the modelling of regional hydrological processes.
    On the other hand, it was demonstrated that it is possible to simulate regional hydrological processes by deterministic groundwater flow models in alluvial areas. The effects of the unknown local variations can be eliminated by the empirical calibration of the model. Approaching hydrological processes from regional to local scale is the most appropriate way, i.e. down-scaling is the best approach in hydrological analysis.
    Deterministic handling of error propagation is not applicable in hydrological analysis when down-scaling is applied; thus sensitivity analysis is required at every step to identify those parameters which have the most significant influence on the final results. Increasing the accuracy of the determination of the values of these parameters improves most efficiently the modelling result.
    Whenever possible, the results need to be checked by an independent method or at least by an independent data set. If this cross-check (validation) is not possible, the sensitivity analysis is of extreme importance.
    On a more general level, it can be concluded that protocols for hydrological analysis - like the one suggested by Simmers for regionalization - help us to understand the information flow through the steps of the procedure in general, but do not provide a robust guide to solve each problem. Instead of suggesting another protocol, the existing methods may be improved by identifying the -bottle necks- of the information flow. The bottle necks are those steps (and variables) which strongly effect the final results. It was found that such a bottle neck occurs in the first steps of the analysis, namely in the reconstruction of the spatial hydrological variables from the measured data. The spatial characteristics of the variables, therefore, have to be analyzed and the optimal representation method (e.g. optimal interpolation or classification method) has to be determined. This is the field where GIS technology can have the most important contribution to hydrological analysis.

    Vissza a magyar térképészet kezdőoldalára!