DETAILS


COLUMNS


CONTRIBUTIONS


ARCHIVE



New Visualization Techniques

Vol.34 No.1 February 2000
ACM SIGGRAPH


Application of VisAD in Hydrological Modeling and Simulation



Ugo Taddei, Olaf David, Christian Michl
Friedrich - Schiller - University Jena

Abstract

For visualization of distributed hydrological phenomena, a multidimensional visualization tool such as VisAD embeds sophisticated visualization techniques in hydrological models. Heterogeneous hydrological properties can be visualized draped on top of a digital elevation model (DEM), in order to model parametrization or results. This article discusses how the VisAD library was applied to hydrology related information in the headwater catchments of the Schmücker Graben and the Steinbach utilizing DEMViewer.

Introduction

Visualization plays an important role in simulation model development and application. The distributed variability of physiographic properties contributing to the river basin’s hydrological budget needs to be reflected in a suitable visual representation. Sophisticated tools are required to visualize distributed entities. These entities have to be accessed relative to the relief of a given basin. Distributed hydrological models take into account a spatial distribution of factors controlling the hydrologic dynamic. In order to be able to analyze, interpret and compare spatial-related information from simulation and observation, software tools are needed which allow "ad hoc" assessment of this information.

VisAD [2] is a scientific visualization library written in and for the Java architecture [3] to provide visualization of data using a flexible tool that assembles and arranges data sets in multiple dimensions. The application of VisAD in distributed hydrological modeling focuses on data models that enable the visualization of properties in a given location in a digital elevation model. With this goal in mind, software tools based on VisAD were developed to direct and simplify the application of VisAD in a hydrological sense. The DEMViewer is one result of this work, and is described in the Tool Application section. In the Background Hydrology section, the modeling background is described as the driving factor for the creation of the visualization applications.


Figure 1: Sub catchment classification visualization (with vertical exaggeration of 2x).


Figure 2: Visualization of (a) linear flow algorithm and (b) multiple area flow algorithm.

Background Hydrology

Drainage basins must be seen as a heterogeneous assembly of distributed entities having specific physiographic properties and precipitation input, each contributing differently to the basin’s hydrological cycle and discharge output. Therefore the modeling of these basins has to account for this heterogeneity if it can lay any claim to simulating the physical processes realistically. This involves the use of physical laws and the establishment of empirical relationships from field studies. In order to fulfill these process-oriented conditions, the study area must be well observed (in the sense of the determining hydrological processes), and the model must be flexible enough to deal with the "real world" conditions. Therefore, the headwater catchments of the Schmücker Graben and the Steinbach were selected. Both catchments are located on the northern slope of the ridge of the Thüringer Wald in central Germany and have been well studied since 1958. The Schmücker Graben drainage basin covers about 2.92 km2. with a stream length of 2.5 km and an average slope of 3.5 degrees. The Steinbach catchment is less than half that size, with an area of about 1.32 km2, a stream length of 1.5 km and an average slope of 6.0 degrees. At the end of the 1930s, more than 70 percent of the wooded areas in both catchments were lost to heavy winds followed by a Scolytidae catastrophe. Under the upgrowing forest, discharge, precipitation, groundwater level and soil moisture, as well as meteorological data, were collected. In the 1950s and 1960s, both catchments were research basins for water balance studies. Since then, they have been a part of the state measurement station network.

Based on the time series and the former research, two distributed and physically based precipitation runoff models PRMS (Precipitation Runoff Modeling System) [4] and TOPMODEL (TOPographic modeling system) [1] were applied to simulate the hydrological dynamic of the system over the last 40 years, using a daily time step. The deterministic HRU (Hydrological Response Unit) and the stochastic REA (Representative Element Area) concepts account in both models for the distributed character by aggregating regions with topo-pedo-geological associations and a common land use controlling their unique hydrological dynamic. A major factor in both concepts is elevation information, which is available in the form of high quality DEMs, and their derivatives slope, aspect, curvature, flow direction and flow accumulation which are then used to quantify the relief charasterics. An example of 16 different relief classes, delineated as subcatchments and accounting for the aspect, is shown in Figure 1. Additionally the wet and gushy catchment parts following the narrow valley floors, as well as upper depressions, are mapped (dark blue). These relief positions are the determinant for the runoff production distinguishing overland flow, interflow and groundwater flow in both models (PRMS and TOPMODEL).

A further evaluation of different flow direction algorithms and their derived flow accumulation was achieved by applying the DEMViewer. The results are shown in Figures 2a and 2b. Whereas a linear flow algorithm allowing only the steepest flow path for the water flow tends to produce parallel water patterns along the slopes (see Figure 2a), a multiple flow algorithm will account for an areal flow accumulation (see Figure 2b). Comparing both calculations to the observed basin characteristic (dark blue pattern in Figure 1), the multiple flow direction algorithm represents the hydrology and therefore the process of runoff generation much more efficiently in middle mountain areas like the described catchment.

As the application shows, the DEMViewer is an effective tool to judge the areal distribution of model parameters conceptually and test it against hard data if that is available.

Tool Application

Clearly, we had a need for custom-made applications that could extend the functionality of the already available programs while also introducing some new features, such as interactive and remote manipulation for modeling support (the latter being very important in the Internet times in which we live!) The use of a third-party visualization library is also justified by the need to create an open visualization package for the Object Modeling System (OMS). OMS is software for hydrological model development currently under development in our department. OMS grew out of other modular/object oriented systems and is written in Java.

VisAD seems to satisfy the basic conditions. It is also written in Java, so the embedding of remote data in applications is easy, and it is a powerful visualization library permitting interactive manipulation of data. Furthermore, VisAD allows the creation of objects reflecting the mathematical structure of meta-data, making the creation of animations a trivial issue. VisAD also enables the creation of 2D graphs, and because of the way in which it structures its data, it is relatively easy to create different types of graphs without having to change the data structure itself. One is free to experiment with different types of data representation and therefore free to use the representation that best fits one’s needs. Our application spectrum and use varies broadly - from simple 2D graphs to animation of four dimensional data (i.e., an attribute such as precipitation over a 3D surface).



Figure 3: (a) DEMViewer and (b) TOPIndex in DEMViewer.

Figure 4: (a) DEM made of points (top); (b) block view of the catchment (middle); and (c) iso-lines (bottom).

Digital Elevation Model Viewer

The Digital Elevation Model Viewer (DEMViewer) was first developed to test the suitability of VisAD for hydrological issues. DEMViewer is a program (Figure 3a) for viewing digital elevation models previously created using a Geographical Information System (GIS). With DEMViewer, one can visualize the elevation model in three dimensions, and drape some geographical attribute, such as a topographic index (Figure 3b), over the elevation model. The three-dimensional display provided by VisAD allows the rotation and translation of the scene as well as zooming in and out. The program itself allows for the representation of the DEM in different modes, such as a continuous surface, a surface made of isolines (VisAD generates the lines out of the elevations points) or in block mode (Figure 4).

The user can choose between the elevation and two other attributes for coloring the surface. It is also possible to drape a JPEG or GIF image, such as satellite image, over the DEM, thus creating a realistic scene. Choosing between some standard color tables is possible, as is the loading of a predefined one (including legend), and control of the vertical exaggeration. Some further graphic control is provided by VisAD, including drawing of scales, texture mapping, point mode, as well as point size and line width (for iso-lines). Despite the fact that DEMViewer is meant to display the terrain model in the traditional way (with latitude, longitude and elevation mapped to x-, y- and z-axis, respectively, and the attribute mapped to color), nothing prevents the user from mapping the attribute to the z-axis and either elevation or other attribute to color.

Conclusion

VisAD as an open scientific visualization library is well suited for hydrological application. Due to sophisticated conceptual methodology using meta-data description and technical implementation on top of OpenGL and Java/Java3D, the broad application in distributed hydrological simulation model development is possible. DEMViewer is an implementation based on VisAD with a focus on interactive application in hydrological modeling, where distributed information of a catchment is mapped on a digital elevation model. The generated interactive views are valuable for the modeler for an assessment of useful distributed parameter sets as input for hydrological models.

There are some extensions of DEMViewer available. The AnimationViewer animates the temporal variability of a spatial attribute. The user basically has the same controls available in DEMViewer, plus the animation controls provided by VisAD. It is also possible to map time to the z-coordinate, thus creating a display of stacked up images. The ProfileObserver allows the user to interactively extract a profile out of a DEM. By means of a slide bar, the user can select where the profile is to be taken. The profile is then automatically drawn on a 2D graph and its place on the DEM is shown on the 3D display. The user can also predefine points along which a profile is to be taken. The ModelViewer was written to display data from simulation and observation to allow the user to compare them. The user is presented with three 2D displays, each of which shows the daily development of a hydrological quantity (such as precipitation, discharge, total dissolved salt, electrical conductivity, etc.) along a year, which can also be interactively chosen. ModelViewer is also meant to be coupled with a water quality database, which is also under development in our department. Presently, the main emphasis is pointed to the direct integration of the underlying VisAD data model into hydrological models to operate on VisAD data structures directly.

Acknowledgments

Many thanks to Bill Hibbard for the very productive collaboration concerning the development of VisAD for the needs of DEMViewer.

References

  1.  Beven, K., R. Lamb, P. Quinn, R. Romanowicz and J. Freer. "TOPMODEL," Computer Models of Watershed Hydrology, V.J. Singh, editor, Water Resources Publications, pp. 627-668, 1995.
  2.  Hibbard, W. "VisAD: Connecting people to computations and people to people," Computer Graphics 32(3), pp. 10-12, 1998.
  3.  Kramer, D. The Java Platform, a white paper, JavaSoft, Mountain View, California, 1996.
  4.  Leavesley, G. H. and L. G. Stannard. "The precipitation-runoff modelling system," Computer Models of Watershed Hydrology, V.J. Singh, editor, Water Resources Publications, pp. 281-310, 1995.
  5.  Michl, C. "Prozeßorientierte Modellierung des Wasserhaushaltes zweier Quelleinzugsgebiete im Thüringer Wald," Ph.D. thesis, Friedrich-Schiller-University Jena, 1999.




Olaf David is Assistant Professor at the Department of Geoinformatics at the Friedrich-Schiller-University Jena. His major research interests are focused on component integration into modeling systems and frameworks for application in hydrology.

Christian Michl was a Junior Scientist at the University of Jena from 1995-1999, where his research focused on physically based hydrological models and their application in small headwater catchments as a water management tool. Currently, he is in charge of software development for hydrological data at Bundesanstalt für Wasserbau, a government institution.

Ugo Taddei is currently working on his M.Sc. degree at the University of Jena.

Ugo Taddei, Olaf David and Christian Michl
Friedrich - Schiller - University Jena
Löbdergraben 32, 07743
Jena, Germany

Website

The copyright of articles and images printed remains with the author unless otherwise indicated.