Investigations in slope development through landslide activity - concepts, methods and implications for interdisciplinary and interoperable data management

Jochen Schmidt, Holger Gärtner, and Richard Dikau
Department of Geography , University of Bonn, Meckenheimer Allee 166, D-53115 Bonn Germany
E-Mail: jochen.schmidt@uni-bonn.de

Abstract

It is an open research question to which extend landslide activity contributes to landform evolution, especially under moderate humid climatic conditions. In a multidisciplinary research project at the University of Bonn, we are trying to get insight into the process of slope development through mass movements. Research methods include local field investigation and stability analysis, mapping and modelling of landslide susceptibility, geophysical subsurface monitoring, and geomorphometric slope profile analysis. Research aim of the coupled use of these different techniques is to model quantitative measures for sediment transport through landslides on hillslopes in the Bonn area. In this paper, the research approach, methods and a few first results are presented. Special emphasis is given to the data management: The scheme and system used to store and manage the data and analysis results is critically discussed. Additionally an alternative, object-oriented approach is presented.

In quantifying the sediment transport through mass movements over longer time scales, we have to cope with several problems, e.g. (1) unknown boundary conditions in time, (2) discontinuity of the process, (3) different process types, (4) coupling with other (slope) processes. Therefore, a statistical approach, coupled with slope stability analysis is used to estimate the contribution of landslides to slope evolution under variable (i) climatic, (ii) morphometric and (iii) geologic boundary conditions. First, several indices describing mass movements over longer time scales were itentified and selected for the study. Slope profile types are extracted for the study area using several morphometric algorithms. Material properties from laboratory tests are related to geologic units. Several slope stability models based on different approaches are used to model the selected landslide indices under different morphometric and geologic conditions. The models are calibrated using data from several landslides in the Bonn area (see below). First results show the dependency of landslide occurence in the Bonn area on hillslope morphometry and geology. These findings are used in an sensitivity study on the influence of different climatic conditions, which shall lead to an estimation of spatio-temporal landslide activity.
Investigations on sites near Bonn (Germany) are carried out, producing a large amount of field and laboratory data. Supplementary information is available by climate data, geologic maps, topographic maps, DEMs, etc. Additionally, data resulting from interpretation, analysis and modelling of the field data must be handled. In our project, storage, visualization and analysis of these data is realized using GIS (Arc/Info , GRASS ), geotechnical software (GeoDin ), databases (Access, ORACLE ), slope stability programs and several other software products. This shows that landslide investigation is a typical example for the heterogeneity of data and methods used in Geosciences, which necessitates a careful and consistent data management.

Presently, a relational data model and diverse structures, methods and tools to handle and analyze the data are used. This (common) practice complicates the exchange of data, methods and research results. In contrast, an object-oriented approach, developed in cooperation with a project on Open Information Systems is compared with classical concepts. The results show, that object-oriented data modelling can facilitate user access to multiple datasets, support integrated use of different analysis technologies and could aid in the development of standards for exchanging data in multidisciplinary environment.

1. Introduction

In this paper we focus on two aspects of landslide research: (1) The question of contribution of landslides on slope development and (2) the problems and issues of data integration in landslide research and in geosciences. These topics come directly out of two interdisciplinary research projects currently located at the University of Bonn : (1) The Collaborative Research Center (SFB) 350 , working on problems of interaction between Geo-Sysems on different scales, and (2) the project OPALIS , working on object-oriented modelling of geoscientific data. The cooperation between these projects arise from data management problem within the Collaborative Research Center (SFB) 350 .
Therefore, in this paper we focus not only the research problems of slope evolution through landslides, we also like to stress the strong need for data integration and efforts in datamanagement in geosciences, which is, from our point of view, a problem often underestimated in geosciences. For a actual description of our research progress and results please look at http://www.sfb350.uni-bonn.de/Wob/de/view/class58_id8.html and http://slide.giub.uni-bonn.de/~holger/opalis1.html .

1.1 Slope failure and slope development: between process modelling and landform evolution

It is an open research question to which extend landslide activity contributes to landform evolution. Recent investigations towards the problem range from local field investigations, local stability modelling over detailed physical process modelling, simplified physical process modelling and statistical modelling to conceptual, genetic approaches (see Hergarten & Neugebauer, 1999b). The problem of filling the gap between actual process description and longterm geomorphic evolution is obvious like in no other research field (Dikau, 1999), because ``on the geological time span over which slope profile evolve, landslides and other rapid mass movements occur almost instantaneously'' (Kirkby, 1987). This means, the change in time scale requires a change in model description. The discontinouity in mass transport and the formative activity of landslides cannot be transfered in time by methods of simple averaging and bilancing, often used in processes with more continuous character.

Therefore, temporal descriptors like lifetime of landforms, reoccurence interval of events, etc. should be used in discontinuous processes like landsliding (see Brunsden & Thornes, 1979; Cendrero & Dramis, 1996). These descriptors can lead to the determination of frequency and magnitude of landslide activity in space and time (Brunsden & Thornes, 1979; Crozier, 1996a). Predicting frequency and magnitude in landslide processes using actual process descriptions as used in recent stability and movement models is an important issue in landslide research. Estimating temporal descriptors and magnitude / frequency behaviours by triggering thresholds is one possiblity in approaching this problem. The concept of thresholds was often used in modelling landform evolution, because of it´s attractive simplicity (Francis, 1987). The variablity of thresholds in space and especially time (which means under varying boundary conditions) is one crucial problem. These varaibility appear to be related to internal dispositive factors (Crozier, 1996a). However, the utilitity of the concept of variable thresholds in modelling landform evolution using slope stability models was shown in several studies (e.g. Brooks et al., 1999; Casale et al., 1993; van Beek & van Asch, 1999). A related problem, stressed by Palmquist & Bible (1980), is the identification and differentation of external and internal dispositive factors as landslide causes and triggering factors leading to slope failure. They propose to identify important dispositve factors to group landslide occurrence with special statistical characteristics in space and time.

Recently, the workshop ``Process Modelling and Landform Evolution'', held by the Collaborative Research Center (SFB) 350 in Bonn , addressed the problem of unsolved discrepancy between (1) measurement and modelling of actual geomorphic processes and (2) description and modelling of geomorphic landform evolution (Hergarten & Neugebauer, 1999b), which applies especially to landslide processes (see above). The actual modelling approaches use well known field sites and more or less sophisticated models. The evolution models often use simplified field data (e.g. ideal slope profiles or artificial landform topography) and comparatively simple / statistical / conceptual approaches (Cendrero & Dramis, 1996; Hergarten & Neugebauer, 1999a; Kirkby, 1987). Appropriate field data for validation of landform evolution models are undefined and/or missing (Dikau, 1999). Therefore, validation of those models often have to be done by methods like the principle of actualim or the ergodic principle (Brooks et al., 1999; Dikau, 1999; van Beek & van Asch, 1999).
As part of the Collaborative Research Center (SFB) 350 , we are trying to give a contribution in filling the gap between modelling processes and modelling landform evolution in a multidisciplinary research initiative. Our approach is based on combining detailled field investigation, local and regional stability analysis and models of slope evolution, leading to an estimation for contribution of mass movements to slope development (see section 2).

1.2 Landslide investigation and the need for integrated data management

In general, two types of landslide investigation can be distinguished: Local site investigation and regional landslide hazard analysis. In both cases, the quality of the analysis results depends stronlgy on the used input data: In the first case, usually more or less intensive field and laboratory measurement programmes and field monitoring is carried out, the regional approaches usually gather spatial data available for the considered area (as geological maps, DEM´s, etc.). However, in both cases, it is necessary to get as good as possible insight in the morphometry, geology and ground water situation of the site (Dikau et al., 1996). Usually, landslide investigations start with geomorphic and/or geotechnical mapping of the field site, followed by a drilling programme and analysis of the collected samples in the laboratory. The produced maps and digital elevation models can be stored within a GIS environment. Graphical representation of the drilling and sampling results are often carried out by specialized software based on national and international (geotechnical) standards (ISO, DIN, EU) (compare section 4). Various stability and/or hazard models are used to analyze the (interpreted) field and laboratory results with respect to local or regional stability of slopes.

Landslide investigation therefore is a typical example of a complex geomorphic / geotechnic investigation. It inheritates the need for multiple types of data and geocomputational modelling and analysis techniques (see section 3). Finding methods, integrating different datasets and techniques is an important issue. This may not only assist geoscientists in concentrating more on their work, but can also help in (interdisciplinary) exchanging of data. New information systems allowing integrated access to various types of data and therefore new kinds of queries, could not only simplify our work, but also extend available information sources. Additionally, data storage within these types of database management systems facilitate the analysis process and connect data in a consistent and recoverable (well documented) way (see section 4). The project OPALIS is an research initiative of geoscientists and computer scientists at the University of Bonn, working on the development of a framework for multidisciplinary data integration and data management.

2. Methodology

Our research aim is to get insight in the process of slope development through landslide processes for the Bonn area. As indicated in the previous section, this research topic contains several problems:

1.: The boundary conditions, which control landslide activity are unknown or only poorly known in a temporal (e.g. paleo surfaces) and spatial (e.g. soil parameters) context.
2.: In timescales of slope evolution, landslide activity is a episodic process.
3.: Landslide processes are coupled with other slope processes.
4.: Landslides control the occurence of subsequent landslides (often leading to several generations of landslides).
5.: Process internals of landslide processes are only poorly known.
6.: Dating of landslides in a representative way according to their spatio-temporal distribution is very difficult (see points 3 and 4). So, the easiest method to analyze slope development through mass movement, the determination of size and date of all previous landslides for a given area is hardly possible or impossible.

Therefore the following statements can be drawn:

It is not possible to model the development of one specfic slope in detail, because of unknown or only unpresice or poor known boundary conditions and poor models.
Even if it would be possible to model a specfic slope, it is not known, whether the model results are representative in a regional context (or only a spatial singularity). One slope is part of a system of landforms and cannot be seen isolated in terms of landform evolution (Dikau, 1999).
That means, modelling slope development can only be done in a statistical way. This modelling process can be seen as ``averaging'' of landslide activity for a slope type over a period of time (which should be long enough). This process has to consider the discontinuity of landslide processes and should include temporal descriptors (see section 1 and table 1). Therefore, it maybe possible to a certain extend (accepting a statistical uncertainty) to simulate the evolution of a typical slope under typical boundary conditions for a specific area.
The reliability of these model results depends strongly (1) how representative my spatial slope model describes a typical situation in the investigated area (which means the definition of ``typical''), (2) if it is possible to determine and describe its position and function in the related assemblage of landforms, and (3) on the quality of the used physical and mathematical model (which means: 'How good describes the specific model the real process?').

**Table 1:** Some indices of mass movements for longer timescales relevant for slope development.
index (for time interval T)	symbol/relation
number of events	n
life time of gravitational forms	T_l
frequency (or return interval)	T_w=T/n
maximum area per event	a_max
average area per event	a_ave
maximum volume (or mass) per event	v_max bzw. m_max
average volume (or mass) per event	v_ave bzw. m_ave
runout distance	l_ave, l_max
runout height difference	h_ave, h_max
average energy per event	E_ave=m_ave g h_ave
total area involved in landslides	A_M(!= n a_ave)
percentage of area involved	I_A=A_M/A
total moved volume	V_M=n v_ave
total energy	E_M
average movement rate	Q_ave=V_M/T
maximum movement rate	Q_max=v_max/T_w
relation to other geomorphic processes	e.g.
e.g. weathering, fluvial transport	T_w/T_l

Modelling slope evolution in time can therefore not be done by investigating one specific slope, it is also related to a spatial problem. Stability models are often poor in terms of modelling a single landslide (also because of poor parameter knowledge). However, if these models use a general valid description of the physical relationships and processes, they should be good enough to model a typical slope system in a statistical sense. ´Modelling in a statistical sense´ thereby means to simulate measures of landslide processes within a slope system relevant for slope development (see table 1 and compare Cendrero & Dramis (1996), Crozier (1973), Crozier (1996b), Crozier (1996a) and Dikau et al. (1996)).

So we can redefine the introductory sentence of this section: We looking for a quantification of mass transport of landslides by modelling landslide measures for longer periods of time, using physically based models for typical (spatial and temporal) situations of the research area. Based on the these statements a research strategy was developed, investigating slope development by mass movement by a coupled local and regional approach using multiple techniques (figure 1). Important in this approach is to get a good knowledge (databasis) of the site (slope) and to gather as most as possible additional information available for the specific site and the site environment.

Step 1:: Local field survey and slope stability modelling is used to learn as much as possible of the process behaviour in the site environment, which means it´s dependency to triggering factors (e.g. thresholds of groundwater level and precipitation) and specific geologic and morphologic boundary conditions. These results are used in the following modelling steps.
Step 2:: Combining the local data and the regional data available is used to estimate how representative the local situation is in a regional context. This stage of comparison is carried out by morphometric analysis, comparing geologic and morphologic situation (see figure 5).
Step 3:: Morphometric slope profile analysis and geologic surveying is used to extract typical slope situations. These typical slope situations are analyzed with respect to their stability using slope stability models in combination with the model results from the local investigations (compare step 1).
Step 4:: Regional stability analysis, based on the infinite slope model is carried out using available site parameters and areal data (e.g. DEMs, Geological mappings, etc.).
Step 5:: In addition, movement models and simplified slope evolution models are applied, extending the choice of simulated landslide indices (especially runout parameters, compare table 1).
Step 6:: The different model results are used for cross check and model validations.
Step 7:: Climate data avialable for the last tens of years and climate proxi-data shall be used to estimate the modeled slope failure under different climate conditions. The applied model types are therefore extrapolated in time using this data.

These modelling efforts shall lead to an estimation of landslide parameters as described in table 1 under various (but ``real'' or ``typical'') morphometric, climatic and geologic boundary conditions.

**Figure 1:** Overview of the methodological approaches to analyze the relation between mass movement and slope development.
$\begin{figure} \begin{center} \leavevmode \epsfxsize 15 cm \epsfbox {geocomp99papermethod.eps} \end{center}\end{figure}$

3. Research area and data collection

3.1 Morphometry and geology of the Bonn area

**Figure 2:** Topography of the Bonn area. Areas of high landslide susceptibility are predominantely the small insected valleys (Katzenlochbachtal, Melbtal) in the horst structure of the Kottenforst and the mountainous area east of the river rhine, the Siebengebirge. Two field sites, the *Dollendorfer Hardt* and the *Melbtal* were chosen for detailed investigations. **Click on the these two sites to see a simple sketch of the typical morphometric / geologic situation.** Data source: Mapped and digitized by D. Kirschhausen, V. Schmanke and U. Hardenbicker
$\begin{figure} \begin{center} \leavevmode \epsfxsize 14 cm \epsfbox {geocomp99paperarea.eps} \end{center}\end{figure}$

Previous investigations and several problems during the construction of roads and buildings indicated serious instability problems and many old landslides on the slopes in the Bonn area (Grunert & Hardenbicker, 1991; Grunert & Hardenbicker, 1993; Hardenbicker, 1993). Most of them were interpreted as holocene mass displacements, whereof several could be dated in the 20st century (Hardenbicker, 1991). However, most of the recent landslides were triggered by construction activities. Landslide susceptibility in the Bonn area is mostly influenced by specific geologic situations (Grunert & Schmanke, 1997). Bonn is located in the so called ``Niederrheinische Bucht'' at the border between the Rheinische Schiefergebirge and the subsidence area of the Lower Rhine region. The site experienced a highly active and complicated geologic history, especially during the Tertiary. Uplift of the Rheinische Schiefergebirge and subsidence of the Lower Rhine region caused erosion and deposition processes, active tectonic movement and Tertiary and Pleistocene vulcanic activity. Therefore, west of the Rhine, today terrace sediments are found above a series of Tertiary layers (clay, sand and gravel) and a Devonian base layer (compare figure 2 and figure 3). Pleistocene tectonic processes uplifted the Kottenforst as a horst with steep slopes to the Low Terrace of the Rhine. Pleistocene and Holocene fluvial processes dissected the plateau of the Kottenforst and created several small valleys (Godesbachtal, Melbtal, Katzenlochbachtal) which are often donwncutted to the Devonian base level. The mountainous area of the Siebengebirge is located east of the Rhine. The Siebengebirge is formed by Tertiary vulcanic and subvulcanic activity. Nowadays, eroded basaltic intrusions form the peaks of the Siebengebirge. The slopes are covered with vulcanic ashes and with similar tertiary sediments as on the west side. The valley floors often reach the Devonian base layer. Both, on the slopes of the Kottenforst and on the hillslopes in the Siebengebirge, landslides are found varying in size and age (Grunert & Schmanke, 1997).

**Figure 3:** The Bonn area: Geologic situation. Data source: Mapped and digitized by D. Kirschhausen, V. Schmanke and U. Hardenbicker
$\begin{figure} \begin{center} \leavevmode \epsfxsize 14 cm \epsfbox {geocomp99papergeol.eps} \end{center}\end{figure}$

3.2 Field sites and data collection

We chose two field sites, which are representative for our research area in morphometric and geologic terms (see above). The first site is a vulcanic peak (Dollendorfer Hardt), located in the Siebengebirge (compare figure 2). There are several mass movements at the side slopes of this peak, including a larger landslide (affected area: 30,000 m²), which we chose for a detailled instrumentation. The second site is the Melbtal, a small valley west of the river rhine, dissected in the Kottenforst plateau (see figure 2). A series of landslides are located at the valley sides (area ranging from 300 to 8,000 m²). The youngest slide (1988) damaged a cemetery and lead to expensive expertises and foundation constructions.

As described in the previous section, our research approach included different methodologies. Recently, several detailed investigations of the Dollendorfer Hardt (mappings, drillings, laboratory works, geophysic subsurface surveys, groundwater monitoring) have been carried out (figure 4). Different mappings (pedologic, geologic, morphologic), drillings (core drilling, vane test) and results of laboratory tests (physical and mechanical soil parameters) are now available. Groundwater and movement monitoring is carried out. Supplementary information is available by maps (either in analog or digital formats) available by official institutions. These include topographic, geologic and pedologic maps, DEMs in different resolutions, and various analogeous maps. Moreover, surface and subsurface information (field and laboratory tests) is available by projects and consultancies working in the same area.

**Figure 4:** Mapping and drillings from field site *Dollendorfer Hardt*. **Click on the drilling sites to see some results of stratigraphical layer interpretations.**
$\begin{figure} \begin{center} \leavevmode \epsfxsize 14 cm \epsfbox {geocomp99paperres.eps} \end{center}\end{figure}$

Although we are presently just in the stage of analyzing the first field data, first statments about the different characteristics of the two field sites can be given: At the field site Dollendorfer Hardt, landslide occurence depends on the existance of steep slopes and the geologic layer of vulcanic tuffs. Long (up to 500 m) and steep slopes formed by the vulcanic and tectonic actvity and the comparatively weak tuffs in high slope positions lead to larger landslides as in the Melbtal (longer runout distances). The Melbtal is characterized by lower slope length (up to 200 m) and hillslopes, which are more gentle. It is dominated by small and shallow landslides mainly influenced by the existence of heterogeneous Tertiary sediments. An additional factor is undercutting through the stream (Engelsbach).

**Figure 5:** Data collection and data analysis tasks.
$\begin{figure} \begin{center} \leavevmode \epsfxsize 14 cm \epsfbox {geocomp99paperdata.eps} \end{center}\end{figure}$

From our point of view integrated and interdisciplinary use of multiple research methods and techniques can lead to a ´best insight´ into the study area, with respect to the studied problem. The combination of different approaches in mapping, surveying and monitoring different parameters reveals a useful combination of information. For example, the geomorphologic, geologic, pedologic and soil mechanic description of the underlying material of a slope involved in landsliding, can help to get a better understanding in the geologic structure and development of the slope. Geophysical surveying techniques can aid in extrapolating point information (like drilling results) and therefore be useful in regional analysis approaches. However, working in a multi-method and multidisciplinary environment means also the difficulty of integrating different sights on the same object. Therefore, the produced data and the used methods to capture, store and analyze the data often are heterogeneous to incompatible.

4. Data handling and data modelling: issues for data integration

Several tools and software packages are used to store, analyze and visualize the collected data as well as the results of interpreting the data (e.g. geological layer constructions, shear surface reconstruction). Presently the following scheme is used (figure 6):

Maps and other georeferenced data (e.g. locations of mappings, drillings, geophysical profiles, etc.) has to be stored in the GIS (GRASS and Arc/Info ). GIS have the capability to store data (as names, height information, dates, etc.) and the related coordinates. Map overlay, map production and analysis and morphometric operations are supported by most GIS.
Meta information, drilling data, laboratory results are stored in database structures (Microsoft Access / ORACLE ). Databases allow easy data retrieval, consistency checks and avoiding of data redundancy. That means, a direct connection of field data (e.g. drillings) and laboratory data is possible and the danger of data loss is minimized.
Specific layer interpretations, geophysical data and special field and laboratory results are handled by specific software packages (GeoDin , EnvronMon , etc.). Data from monitoring devices, analysis of geophysical and specific laboratory experiments often require the use of special software (in our case e.g. movement and groundwater measurements, geophysical data, geomechanical laboratory tests).
Additional information (e.g. consultancy reports, data available from external sources, photos) are stored using special data formats.

**Figure 6:** Structure of methods and tools presently used to store primary data
$\begin{figure} \begin{center} \leavevmode \epsfxsize 14 cm \epsfbox {geocomp99papertools.eps} \end{center}\end{figure}$

On top this data scheme and geocomputational concept, several analysis tasks are performed (compare section 2), producing secondary data. These include (1) interpretations of geologic structures on local and regional scales, (2) interpolating three dimensional sediment bodies and landslide bodies, (3) deriving effective soil parameters and (4) slope stability analysis and regional hazard analysis. Until now, the storage of these produced secondary data is only partly realized. Especially the handling of 3D information is a problem to be solved. Therefore, several problems and disadvantages of the used concept to store the data can be identified:

Storage and analysis of spatial data is presently achieved in GIS mostly by two dimensional data structures. Geomorphological data mostly include the third and fourth dimension through subsurface and athmospheric quantities. Although there are some developments towards 3D/4D-GIS (Hamre, 1994; Kidner & Jones, 1994; Raper, 1989), most of the standard systems do not support 3D/4D information (Some more information on our view on present GIS and future issues can be found at http://slide.giub.uni-bonn.de/~jochen/fragments/gisissues.html ).
Meta information (e.g. map descriptions) and complex textual information (e.g. drilling information) often cannot not be stored within a GIS. Therefore GIS had to be coupled to databases.
Various standards in describing and capturing surface and subsurface data (e.g. mappings used in soil sciences, geology and geography) could lead on the one hand to data redundancy, on the other hand to the opportunity of an integrated view on a phenomena through complementary information. However, the technical problem of integrating these different data types still remains unsolved.
Often, geoscientists are forced to use special software packages are used to handle and store specific field and laboratory data. These data sources are not connected other data, which leads to data redundancy and problems in recovering data.
Process modelling tools are often realized by specialized software which has to be coupled to the existing data storage scheme taking more or less effort in data bridging.
In our project we had access to external data sources from related disciplines and previous projects. The integration of these information is often difficult because of various documentation schemes or weak meta data. Therefore these data types are usually stored using native structures, which are not related to other data.

Based on these statements, the conclusion can be drawn, that the development of an homogeneous data model and geocomputational concept in landslide research and, more general, in geosciences is is normaly impossible. That means, we do have a variety of more or less incompatible tools to store and analyze the data, which means lots of time consuming efforts in data bridging and save data recovering. Moreover, the danger of data loss is high, e.g. if a project changes or finishes, if staff changes or if data is transfered from/to external positions (problem of ``data cemetery''). Improvements of this current problems could be reached by more efforts in integrating different data types, improvements of present standards and development of new standards (especially for data documentation), and integrating different standards.

The project OPALIS aims to support interoperability of various data sources by providing uniform access to heterogeneous and distributed sources. For this purpose, we use the Unified Modelling Language to represent the data sets through object models. Aim is to reflect heterogeneous geoscientific data sets and standards by object oriented models. Based on this models, a system will be provided allowing integrated access to the modelled data sources (compare Bergmann et al., 1998). This system will be realized through a OQS (OPALIS QUERY SYSTEM) allowing new integrated queries (see below). Part of this initiative is the landslide database for the Bonn area. Figure 7 shows a simplified OO-Model of several data sets of our landslide database (compare section 3) including different types of mappings, drilling results, stratigraphic layer interpretations, geophysical data and others.

**Figure 7:** Simplified object-oriented data model of data used in our landslide project: mappings, drillings, layer descriptions, geophysical exploration, etc. (and the according meta information and geometrical attributes).
$\begin{figure} \begin{center} \leavevmode \epsfxsize 14 cm \epsfbox {geocomp99paperoo.eps} \end{center}\end{figure}$

Based on this model, a conceptual query component was applied to show the capabillities of our data model approach. If a query is applied, questioning e.g. a specific geologic term (tol = stratigraphic layer ``basaltic tuff''), the OQS (OPALIS QUERY SYSTEM) can be enabled to extend the query to related namings using the modelled standards. This means the search for ``tol'' will include data sets with different namings for this layer according to domain specific standards. In case of the available drillings and/or mappings these can be e.g. pedologic, geologic and soil mechanic layer descriptions. Therefore the query will deliver as an integrated answer all information from these datasets according to the queried stratigraphic record. Compared to classic systems, this OQS enables the client to query distributed and heterogeneous datasets, rather than using different queries on various systems and therefore simplifies (easy access) and improves (direct access to multiple sources) geoscientific work.

**Figure 8:** Example query of the OQS (OPALIS QUERY SYSTEM) on the modelled data and standards. These types of integrated queries give the opportunity to restore multiple information from distributed systems (given, that they are integrated within OQS). Therefore, they strongly support gescientific work (which normaly uses divere data sources) (see text for description).
$\begin{figure} \begin{center} \leavevmode \epsfxsize 14 cm \epsfbox {geocomp99paperrequest.eps} \end{center}\end{figure}$

5. Conclusions

The research question, to which extend landslide activity contributes to slope evolution over longer periods of time is only partly solved in quantitative terms. Although there are some important concepts (frequeny - magnitude, variable thresholds etc.) (Kirkby, 1987), there are only few links between small to regional scale process modelling and conceptual to statistical landform evolution models (Dikau, 1999). We propose to use available stability, hazard and mass movement models and available field data to model statistic parameters of landslides in a spatio - temporal context as a method to fill this gap. The identification of typical situations in terms of the dispositive factors and their relation to thresholds in triggering factors can be a useful possiblity to handle the problem of missing field evidence. The definition of typical field situations with respect to the site environment has to be done using multiple environmental informations. Therefore multisciplinary investigation techniques and results should be used, which requires handling of heterogeneous data sources and modelling techniques.

Handling diverse data sources in combination with the use of various analysis and modelling tools often lead to the problems of (1) weak integration of data sets, (2) weak documentation of data sets and data sources and (3) missing links between the data. This leads to the risk of loosing data or missinterpreting data. Therefore we argue, that data integration is a important research task in landslide reseach and -- more general -- in geosciences. Object oriented modelling techniques can be used to model a diverse data structure in a recoverable and integrative way. This approach could lead to new types of information systems facilitating the integration of multiple data structures and analysis methods. Moreover, data integration can enforce a necessary documentation of specific data handling and data representation in a multidisciplinary environment.

Acknowledgements

The research presented in this paper have been carried out within the research projects Collaborative Research Center (SFB) 350 and OPALIS , which were supported by the Deutsche Forschungsgemeinschaft (German Research Council) , Bonn . We are grateful to V. Schmanke for providing some of the digital maps of Bonn .

References

Bergmann, A., Gärtner, H., Breunig, M., Cremers, A. & Dikau, R. (1998), `Design and first steps in the development of opalis', International Society for Photogrammetry and Remote Sensing 32(4), 52-58. Proceedings of the ISPRS Commision IV Symposium ``GIS -- Between Visions and Applications''.

Brooks, S. M., Anderson, M. G., Ennion, T. & Wilkinson, P. (1999), Exploring the potential for physically- based models and contemporary slope processes to examine the causes of holocene mass movement, in Hergarten & Neugebauer (1999b), pp. 205-230. Poceedings from SFB350 - Workshop ''Process Modelling and Landform Evolution'', 1997.

Brunsden, D. & Thornes, J. B. (1979), `Landscape sensivity and change', Transactions, Institut of British Geographers. New Series 4(4), 463-484.

Casale, R., Fantechi, R. & Flageollet, J., eds (1993), Temporal occurence and forecasting of landslides in the European Community, Final report, Vol. I and II of the EC Programme EPOCH, Contract no.90 0025.

Cendrero, A. & Dramis, F. (1996), `The contribution of landslides to landscape evolution in europe', Geomorphology 15,191-211.

Crozier, M. J. (1973), `Techniques for morphometric analysis of landslips', Zeitschrift für Geomorphologie 17(1), 78-101.

Crozier, M. J. (1996a), `Magnitude/frequency issues in landslide hazard assessment', Heidelberger Geographische Arbeiten 104, 221-236.

Crozier, M. J. (1996b), `Runout behaviour of shallow, rapid earthflows', Zeitschrift für Geomorphologie, Supplement Band 105, 35-48.

Dikau, R. (1999), The need for field evidence in modelling landform evolution, in Hergarten & Neugebauer (1999b), pp. 3-12. Proceedings from SFB350 - Workshop ''Process Modelling and Landform Evolution'', 1997.

Dikau, R., Brunsden, D., Schrott, L. & Ibsen, M.-L., eds (1996), Landslide recognition, Vol. 5 of Publication (International Association of Geomorphologists), Wiley.

Francis, S. C. (1987), Slope development through the threshold slope concept, in M. G. Anderson & K. S. Richards, eds, `Slope stability', Wiley, pp. 601-624.

Grunert, J. & Hardenbicker, U. (1991), `Hangrutschungen im Bonner Raum - ihre Genese und Kartierung für Planungszwecke', Zeitschrift für Geomorphologie, Supplement Band 89, 35-48.

Grunert, J. & Hardenbicker, U. (1993), `Gravitative Hangabtragung im Bonner Raum während des Holozäns', Würzburger Geographische Arbeiten 87, 325-338.

Grunert, J. & Schmanke, V. (1997), `Hangstabilität im Südwesten Bonns', Geographische Rundschau 49(10), 584-590.

Hamre, T. (1994), An object-oriented model for measured and derived data varying in 3d space and time, in T. C. Waugh & R. G. Healey, eds, `Advances in GIS Research Proceedings', Vol. 2, pp. 868-881. Sixth International Symposium on Spatial Data Handling.

Hardenbicker, U. (1991), `Verbreitung und Chronologie der Hangrutschungen im Bonn er Raum', Arbeiten zur Rheinischen Landeskunde 60, 9-18.

Hardenbicker, U. (1993), `Neue Ergebnisse der Hangrutschungsforschung im Bonner Raum (Südrand der Niederrheinischen Bucht)', Berliner Geographische Abhandlungen 78, 155-174.

Hergarten, S. & Neugebauer, H. J. (1999a), Self-organized criticality in landsliding processes, in Process modelling and landform evolution (Hergarten & Neugebauer, 1999b), pp. 231-249. Proceedings from SFB350 - Workshop ''Process Modelling and Landform Evolution'', 1997.

Hergarten, S. & Neugebauer, H. J., eds (1999b), Process Modelling and Landform Evolution, Vol. 78 of Lecture Notes in Earth Sciences, Springer. Proceedings from SFB350 - Workshop ''Process Modelling and Landform Evolution'', 1997.

A decuctive object-oriented gis for handling multiple representations, in T. C. Waugh & R. G. Healey, eds, `Advances in GIS Research Proceedings', Vol. 2, pp. 882-900. Sixth International Symposium on Spatial Data Handling.

Kirkby, M. J. (1987), General models of long-term slope evolution through mass movement, in M. G. Anderson & K. S. Richards, eds, `Slope stability', Wiley, pp. 359-379.

Palmquist, R. C. & Bible, G. (1980), `Conceptual modelling of landslide distribution in time and space', Bulletin of the International Association of Engineering Geology 21, 178-186.

Raper, J., ed. (1989), Three dimensional applications in Geographical Information Systems, Taylor&Francis.

van Beek, L. P. H. & van Asch, T. W. J. (1999), A combined conceptual model for the effects of fissure-induced infiltration on slope stability, in Hergarten & Neugebauer (1999b), pp. 147-167. Proceedings from SFB350 - Workshop ''Process Modelling and Landform Evolution'', 1997.