Video Showing A Landscape Genesis

Video Showing A Landscape Genesis

Boehler, W., Scherer, Y., Siebold
Institute for Spatial Information and Surveying Technology
FH Mainz, University of Applied Sciences, Holzstrasse 36, 55116 Mainz, Germany
Tel: +49-6131-2628-27, Fax +49-6131-2628-15
Email: i3mainz@geoinform.fh-mainz.de

Abstract
A video was produced showing the landscape genesis of the 'Pellenz', an area in the western part of Germany, which was formed by various volcanic events. A general project description was given at the Onuma Workshop of ISPRS (Boehler, Scherer, Siebold, 1999). Together with a presentation of the final film product, a more detailed description of the mathematical modeling of the volcanic events, such as the development of maars and scoria cones as well as pyroclastic and lava flows can now be presented at the Ayutthaya International Workshop. Information from topographic and geological maps was used together with data from mining and exploration to create digital terrain models of the landscape for the last 300.000 years. For the volcanic events, the topographic changes for every vertex of the used raster DEM had to be modeled for every increment, i.e. for every one of the many thousand single frames of the video. Simultaneously, the changes affecting the texture overlays had to be modeled accordingly.

The Landscape
A video showing the landscape genesis of the Pellenz area in the last 300.000 years has been developed at i3mainz recently. In a first period of volcanic activities, between about 250.000 and 200.000 B.P., many maars and scoria cones with lava flows changed the appearance of this landscape. A long period without volcanic activities followed where the landscape was changed only by erosion and loess deposits. About 13.000 B.P., a huge volcanic explosion of the Mt. St. Helens type occurred. An estimated 8 cubic kilometers of material was thrown out within a few days, causing heavy fall-out. Several pyroclastic surges, consisting of hot gases and ashes, crashed down into the neighboring valleys. The area and its geological history is well investigated and published (Meyer 1994, Schmincke 1988, Ippach 1999, 2000).

The Video Project
The basic requirements for digital landscape modeling were DEMs and texture overlays for every single frame. Those were animated, rendered and combined for a digital video showing the landscape development without any gap for the last 300.000 years. A general outline of the project has been presented at the Onuma Workshop of ISPRS, Commission V (Boehler, Scherer, Siebold 1999). Meanwhile, a film has been produced, combining the digital video with film sequences of comparable volcanic events in the last decades. A sound track was added, too. The film is displayed several times a day at the Rauschermuehle Information Center in the Pellenz area and will also be shown in the poster session of this workshop. Figure 1 gives a general idea of the whole process. No off-the-shelf software is available for modeling the build-up of volcanic cones and maars or the flow of lava and pyroclastic surges. Own software was developed using C as programming language.

For landscape animation, a complete digital terrain model (DTM) is precondition. It consists of a digital elevation model (DEM) describing the landscape's morphology and textural information illustrating the landscapes appearance, depending on vegetation (depending again on climate). Therefore, solutions had to be found and developed for modeling both of the components (DEM and image data).

The following sections describe the modeling process (according to the leftmost column in Fig. 1) for maars and volcano cones as well as for lava and pyroclastic flows in more detail.

Figure 1 Production workflow for the whole project. Modeling, animation and rendering was accomplished at i3mainz, the final film was produced by 'Institut fuer den wissenschaftlichen Film', Goettingen.

Dem Modeling

Initial DEM
A 40m raster DEM is available from the local land surveying authorities. The same raster was used for the whole project. So any removal or addition of material had to be modeled changing the elevation values of the 40 m raster vertices.

Before any modeling of volcanic events could be started, the initial DEM of the Rhenish Massif had to be reconstructed as it appeared most likely 300.000 years ago. This was accomplished starting from a DEM around the year 1900 which was computed from vectorized contour lines from the first topographic maps (scaled 1:25.000) available for the area. The present DEM could not be used because the topography of the landscape was considerably changed in the 20th century by intensive mining of pumice and basalt with industrial methods.

To obtain the initial elevation model, all volcanic deposits had to be removed mathematically. Since the thickness of the deposits is well known for most regions from mining activities or investigative drilling, the elevation changes for the DEM raster points could be interpolated and considered. For some areas, where geological data were insufficient, contour lines of the initial landscape were drawn manually by geologists and again converted into DEM raster information.

Maars and Volcanic Cones
Quite often, maars are the initial stage for the build-up of a volcanic cone. They result from explosions being triggered by underground water coming into contact with hot magma. In contrast to volcanic cones, maars consist of a low circular embankment only that surrounds a deep crater, often filled with water. Nevertheless, volcanic cones and maars can be modeled in the same way. So, if the term 'volcano' is used in the following, it refers to both, volcanoes and maars.

Input data To model the build-up of a volcanic cone, first of all we need the extension of the volcano (outer limit of its projection to the horizontal plane). This borderline polygon is digitized from a geological map, spacing the vertices individually to record the shape as close as necessary. Secondly, the location of the center of the eruption and its altitude are also taken from the geological map. As a third input, we need one single profile representing a typical cross-section (from the center to the outer edge) of a standard cone (cf. Fig. 2). Finally, it has to be decided how many frames the video sequence of the whole event shall comprise. All input data have to be made available as ASCII files in an ESRI® compatible file format since ARC/INFO®, Release 7.2.1, was used for partial solutions (vectorizing scanned maps, building polygon topology or generating and interpolating DEMs) wherever possible.

Modeling To model the intermediate locations of the cone's borderlines during built-up, a linear interpolation is carried out between each vertex of the borderline polygon and the center point, both known from the input data (Fig. 3). The increment depends on the number of frames assigned for the build-up of the cone. The series of concentric polygons documenting the horizontal growth is saved as one single ESRI file.

As a next step, the standard profile is transformed horizontally between the center point and each vertex of the present borderline polygon (dotted lines in Fig. 3). As a result, we obtain an individual profile for every vertex of the border line. All profiles originate in the center and run outwards in a star-shaped figure.


Figure 2 Standard profile (1-2-3-4) for a typical	Figure 3 Interpolation of volcano growth polygons cone (1 = center point). based on vertices of borderline polygon of maximum extension.

The vertical extent is computed as the difference between the altitude of the highest elevation of the volcano taken from the 1900 map and the lowest DEM vertex of the pre-volcano terrain within the horizontal final extent of the volcano. This difference is divided by the number of frames assigned to the event. Thus, we have the increment the volcano grows along the vertical axis during each frame. A linear interpolation by time can be performed using this increment. All profiles are now scaled in the vertical as well. Thus, the maximum vertical extent of every profile of a frame is identical and represents the current height of the volcano for this frame. Consequently, we have to deal with two different scale factors for the horizontal and the vertical transformation of every profile.

In the next step, the volcanic cone has to be prepared for integration into the digital (raster) elevation model. So far, the volcano is just described by a number of profiles. As mentioned above, the raster width used in the project is 40 m; the program could handle any other width, too, however. Affected are all points of the grid that lie within the box defined by the extreme points of the volcano's horizontal extent (of course, the same coordinate system for all data is a precondition). For every grid point G inside this box the two neighboring profiles of the star-shaped figure have to be found (Fig. 4). They are used to compute a new profile running through the grid point under consideration. The distances of profile points 2, 3 and 4 from center point 1 on the new profile are interpolated using the angles between this profile and the neighbor profiles. The height of the grid point can finally be interpolated on the new profile.

When this procedure is repeated for all grid points within the chosen box, a grid based height model of the volcano is available for the box area. These heights now have to be added to the pre-volcano elevations of the general elevation model to obtain new absolute heights. If the old terrain was very uneven, however, the computed cone would be distorted accordingly. Therefore, the volcano grid heights always are added to the horizontal plane defined by the lowest DEM-vertex within the volcano's extent, instead.

All steps described above have to be repeated for each volcano or maar and for every single frame. At the end of processing, we obtain a sequence of elevation models describing the genesis of the maars and volcanoes in this period. It has to be considered, too, that in some cases several volcanic events take place at the same time at different locations.

Figure 4 Interpolation of the height of a grid point (G) from heights
of neighboring profile points (see text).

Lava and pyroclastic flows
Although the geological facts behind their genesis are entirely different, both manifestations of volcanic flow deposits can be modeled in a very similar way.

Input data. The outline of a flow deposit has to be digitized from the geological map in a way similar to the procedure described for volcano cones above. In addition, a line must be specified describing the directions of progression. Again, the number of frames has to be chosen in order to have a value for the increments needed in the computation. Finally, the thickness of the flow must be known at any location. These values derive from digital height models describing the surface of the deposits. If those are not available with sufficient quantity and quality, a constant average thickness has to be used. For reasons already mentioned, all input-data (outlines and DEMs) have been generated as or converted into ARC/INFO® formats and provided as ASCII-files.

Modeling. In order to model the progression of the flow, its front part is moved ahead in increments (Fig. 5). To give the movement a smooth appearance, many front lines have to be defined. Since it would be too tedious to digitize all these lines manually, a program was developed to define these front lines as arcs between the left and right outline, considering also the flow directions given with the input data. A polygon, enclosing the area the flow has moved ahead since the last frame, is generated using the current and the last front line arcs, and labeled according to the frame number. Problems arise in sharp curves where the front line arcs may intersect (Fig. 6). This requires some additional editing (e.g. with ARCEDIT®) to get topologically correct polygons.

A point-in-polygon test is now performed in every frame for each polygon labeled with the current frame number. If a DEM grid elevation point is located inside the last increment of a lava flow movement, its elevation will be replaced by the corresponding elevation of the flow's surface. Pyroclastic flows, consis-ting of gases to a large extent, settle when coming to a halt. This effect was modeled using a multiple thickness during the propagation phase and reducing heights to the final surface location afterwards.

Modeling of volcanoes/maars and flow deposits is integrated into one single program. Thus, it can be performed in one step, if all necessary input data are available.


Figure 5 Polygons for the increments of a flow propagation.	Figure 6 Intersection of arcs of front lines

Texture Modelling

General Aspects
Proceeding in a similar way as with DEM modeling, an initial texture image was produced for the first frame of the video. This image was adapted for all following frames according to the changes occurring. Different textures symbolize different vegetation or deposit types. Climatic changes, occurring quite frequently in the time concerned, can thus be conveyed by the appropriate vegetation patterns, such as forest or prairie.

In the modeling procedure, a segmentation is carried out according to the movement of volcanic material, and an (untextured) color code is introduced for every class of texture. For example, red symbolizes a lava flow, green a volcano or blue a volcanic maar. These colors are used to mark the location of an event in the geo-referenced background image preliminarily (see untextured areas in Fig. 7). Prior to the following animation procedure (cf. Fig. 1) the codes are replaced automatically by more realistic textures (produced based on natural patterns using Corel Photo-Paint®) by means of a script program. The texture files are just another input source for animation in Kinetix 3D Studio MAX®.

Figure 7 Segmented areas (untextured)
to mark areas where texture has to be changed.

Maars and volcanic cones
The same set of polygons that was created for DEM modeling (Fig. 3) can be used for the segmentation of the present volcano cone area of every frame. A point-in-polygon test is performed for the points of the image matrix of each frame and color codes are placed accordingly into the geo-referenced image. While processing consecutive frames, all data are kept in main memory and saved for every frame. Many different events occurring at the same time but at different locations can thus be processed simultaneously.

Lava and pyroclastic flows
The procedure for texture overlays for the flows is basically the same as for the maars and volcanoes: Those color code values of the image matrix appearing inside the area covered by the flow are changed to their new values. Cooling-out of lava after sedimentation is considered as well by change of its color.

Final Film Production
When the modeling procedures had been completed, all DEMs and images were animated with 3D Studio MAX® R2.5 and tested with different variations of lights, cameras and atmospheric effects (smoke, vapor, fall-out, haze, eruption cloud) to get a good realistic appearance. After rendering, many thousand single frames could be combined to show the virtual landscape genesis of the Pellenz area in a digital video. For further details on animation and rendering see Boehler, Scherer, Siebold, 1999. The material was used by the "Institut fuer den wissenschaftlichen Film" (Institute for the Scientific Film), Goettingen, Germany, to produce the final film product mentioned above.

Conclusions
Using only simple linear interpolations, it is possible to model complex volcanic mechanisms and the resulting changes in landscape appearance. DEMs and textural overlays can be created for any stage of development, thus accomplishing a virtual visualization of very complex processes on the basis of scientific knowledge.

Figure 8 Sequences from the video. Left: Build-up of a volcanic cone and subsequent lava flow.
Center: Eruption of a St. Helens type volcano (note fall-out). Right: Clash of two pyroclastic surges.

Acknowledgements
The project was planned and carried out in close cooperation with Roemisch-Germanisches Zentralmuseum, Mainz, and predominantly funded by 'Stiftung-Rheinland-Pfalz fuer Innovation', the innovation foundation of the State of Rhineland-Palatinate. Geologists P. Ippach and E. Harms, both from Vulkanpark GmbH, Mayen, accompanied the project with expert advice. At our own institute, G. Heinz was an advisor on many difficult subjects.

References

Boehler, W., Scherer, Y., Siebold, M.: Visualization of a Landscape Genesis. Internatioanl Archives of Photogrammetry and Remote Sensing, Vol. XXXII, Part 5-3W12, pp. 9-14, Onuma 1999.
Ippach, P., 1999, 2000. Personal communication.
Meyer, W., 1994. Geologie der Eifel. E. Schweizerbart'sche Verlagsbuchhandlung (Naegele u. Obermiller), Stuttgart.
Schmincke, H.-U., 1988. Vulkane im Laacher See-Gebiet. Ihre Entstehung und heutige Bedeutung. Doris Bode Verlag GmbH, Haltern.