GEOLOGIJA 46/2, 329–323, Ljubljana 2003 Digital map databases: No more hiding places for inconsistent geologists! Kristine ASCH Federal Institute for Geosciences and Natural ResourcesStilleweg 2; 30655 Hannover, Germany Key words: digital geological map, GIS Introduction A geological map is without doubt the visual language of geologists (Rudwick, 1976). Given a geological map of anywhere in the world a geologist will be able to share a basic understanding of the disposition of the rocks that the map author depicted. Further, with a little time to interpret the maps and their legends, most geologists could make sense of two maps of adjacent countries, even though the linework and classification systems may not always be the same. Unfortunately computers, GIS and digital databases do not possess such powers of interpretation and deduction. They do not comprehend that polygon X on one map is probably the approximate equivalent of polygon Y on the other. Though systems using fuzzy logic are currently being investigated, most GIS and databases require data to be logically structured and relationships between features and attributes to be explicit and not merely tacit. Using the example of the IGME 5000 project, this paper will explore some of the re- asons for the inconsistency in geological maps and classification systems and illustrate why this poses serious problems for those who wish to construct and use geological GIS across regions and countries. Maps, geologists and the advent of IT Generations of earth scientists (“Geogno-sten” and other geoscientists) have summarized the results of their fieldwork and research in map form (Asch , 2003). The geological map has been the means for “geologists” to record, store and disseminate their knowledge and the results of their investigation of the rocks and unconsolidated deposits of the Earth’s surface. For several hundred years geological maps have been, and still are, “the visual language of geologists” (after Rudwick, 1976). They represent the “ ... knowledge simply of what is where on the Earth surface ...” (Maltman, 1998). Geological maps have always provided for their users basic knowledge about the distribution of natural resources such as ore, water, oil or building stones. They may, al- 330 Kristine Asch beit indirectly, warn about the danger of natural hazards or supply information about suitable sites for land-fill, house-building or tourism. They thus provide the basis for environmental planning and protection and support public policy decisions. Geological maps are the basis for understanding the earth and its processes. In the last quarter of the 20th century, the era of IT arrived and changed the world of geosciences totally and irrevocably. Loudon (2000) points out: “IT influences the way in which scientists investigate the real world, how they are organized, how they communicate, what they know and what they think”. We are just at the dawn of that era. Now many factors that constrained our predecessors no longer exist. Modern computing systems (for example databases, GIS and Internet tools) allow us to store, retrieve and present far more information and knowledge about an area than we could ever display on a 2-dimensional piece of paper. The key point is that we can now separate the storage and recording of information from the means of disseminating it; we are no longer forced to try and serve all purposes with the same “general purpose document”. Using IT we can select the area, change the scale and topographic base, choose the theme, amend the colours and line styles. We can distribute the knowledge in an infinitely variable number of ways, delivering it on paper, on CD ROM, or across the Web and choose a variety of resolutions, qualities and levels of complexity. Increasingly, geologists are now using modelling software to create 3- and 4-dimensional models, allowing users, through a variety of visualisation methods, an insight into the original sci- entist’s interpretation of the Earth below our feet. In many respects the 1:5 Million International Geological Map of Europe and Adjacent Areas (IGME 5000) project is bridging the domains of the traditional paper map and the digital era which have been summarised above. The next sections describe the project and discuss the issues it faces. GIS and paper map: The IGME 5000 Project The 1:5 Million International Geological Map of Europe and Adjacent Areas (IGME 5000) is a major European geological GIS project which is being managed and implemented by the Federal Institute for Geosci-ences and Natural Resources (BGR) under the umbrella of the Commission for the Geological Map of the World (CGMW). It follows a long tradition of the BGR and its predecessors to produce international geo-scientific maps of Europe. The IGME 5000 is a collaborative European project involving to date, 48 participating geological Surveys and is supported by a network of scientific advisors.. Its aims are to develop a Geographic Information System (GIS), underpinned by a geological database, and a printed map providing up-to-date and consistent geological information. The main theme of the project is the pre-Quater-nary geology of the on-shore and, for the first time at this scale, the off-shore areas of Europe (Asch , 2002). Standard procedures, data structure and dictionaries were developed in order to gather, integrate and constrain the necessary spatial and attri- Figure 1. An example of inconsistency at national boundaries from the IGME 5000 project. The differences are notable particularly in regard to geological classification, mapped units and level of detail. Digital map databases: No more hiding places for inconsistent geologists! 331 bute information from the participant organisations. Some Recurring Problems Organising the co-operation of so many participating nations and compiling their input proved to be a considerable information management task. Without doubt the major challenge was coping with the inconsistency of approach by the participants: different interpretations, variable data input, generalisation and drawing quality techniques. It seems that almost every geological survey organisation in Europe has created its own conventions (and sometimes several conventions) to produce traditional paper maps, and now their digital representation within a GIS (a fact subsequently reinforced by a FOREGS census of 29 Geological Surveys (Jackson & Asch, 2002). Significant discrepancies (Asch , 2001) were found in the following items: • geological classification, such as litho-logy and chronostratigraphy, • mapped units (emphasis, number, ...), • topographic base (co-ordinate system, ellipsoid, drainage system, projection), • draft map scale, • level of detail and completeness (especially off-shore), • colours, symbols, • data structures and hierarchies. Not unexpectedly these differences gave rise to discontinuities at the political boundaries - the well known “national boundary faults” (Figure 1), not to mention highlighting the substantial differences between the mapping of onshore and offshore areas. Generic Reasons behind Inconsistencies There may be numerous reasons for the inconsistencies described above, inconsistencies that are repeated within the mapping of most national territories. The amount of data available in areas will vary; different classification schemes have been used; the mapping may be of different ages and advances in the scientific techniques and new data will have occurred. But perhaps the underlying and most fundamental reason is surely that geology is a de- ductive science, and a geological map is the result of the interpretation of often sparse and variable data by individual geologists, each with their own idiosyncratic approaches. Are Standards Important? Does it matter if we have these inconsistencies? After all, given a little time, geologists can usually establish the intended equivalence or otherwise between the “apparently different” rock types on adjacent maps? Given time, they may be able to, but the total effort taken to research and solve these discrepancies in an ad hoc way must consume an enormous amount of time. These variations and the adjustments made to correct them will inevitably also lead to misunderstandings between geologists and make it more difficult to recognise relationships and associations between geological sequences. This will result in obstruction of the progress of cross-border scientific understanding. Further, those without the benefit of geological training will not be able to appreciate or resolve the inconsistencies, a fact which seriously limits the worth of geological maps and databases outside the geological profession. In addition, when the maps are used as the basis for applied products, e.g. geoha-zard or mineral maps, the differences may lead to potentially serious inconsistencies in future risk or resource prediction. In this context should be also considered the need to provide coherent geoscience information for pan-regional or pan-national initiatives, e.g. the European Water Framework Directive (EU, 2000) or Mineral Waste directive initiative (Cliford & Fernandez Fuen-tes, 2002). Last but not least, while geologists may be able to deal with uncertain relationships, computers, GIS and database systems find it extremely difficult, if not impossible. Such systems demand a much more rigorous approach to geometry, data structure and attribution. Thus, the potential benefits of Information Technology, i.e. interoperability, data integration and the ability to share and supply 332 Kristine Asch harmonious information for scientific research to address pan-national geological problems across frontiers, are entirely dependent on the continuity and consistency that standards would bring. References Asch, K. 2001: The IGME 5000: A Means to Pan-European Geological Standards? (or just another local and transient convention?), IAMG Annual Conference (Cancun): http:// www.kgs.ukans.edu/Conferences/IAMG/Sessi-ons/I/asch2.html A s c h , K. 2003: 1:5 Million International Geological Map of Europe and Adjacent Areas: Development and Implementation of a GIS-enabled Concept. – Geol. Jb., Sonderheft, SA3. 133 p.; BGR (Hannover). Cliford, J. & Fernandez Fuentes, I. 2002: May 2002 Report to EFG Council on Brussels’ Office and EC/EU-related Activities; http:// www.icog.es/framestexto/EU%20Re-port%20to%20Council%20May%202002.pdf EU (2000): Directive 2000/60/EC of the European Parliament and of the Council establishing a framework for the Community action in the field of water policy; Official Journal (OJ L 327), Brussels: http://europa.eu.int/comm/environment/ water/water-framework/index_en.html J a c k s o n , I. & A s c h , K. 2002: The Status of Geoscience Mapping in Europe. – Computers and Geosciences, 28, 783–788, (Elsevier), Ottawa. L o u d o n , T. V. 2000: Geoscience after IT; A view of the present and future impact of Information Technology in Geoscience. – Computers and Geosciences, 26, 142 p.; Kidlington (Pergamon). Maltman, A. 1998: Geological Maps – An Introduction. – 2nd Ed., 260 p.; Chichester (John Wiley). R u d w i c k , M. J. S. 1976: The emergence of a visual language of geological science 1760–1840. – History of Science, 14, 149–195; Cambridge. GEOLOGIJA 46/2, 333–338, Ljubljana 2003 GIS-based assessment of aggregates in Carinthia (Austria) Richard BÄK1, Maria HEINRICH2 & Gerhard LETOUZÉ-ZEZULA3 1Amt der Kärntner Landesregierung, UAbt 15GB (Geologie und Bodenschutz), A-9021 Klagenfurt, Flatschacher Straße 70 2 and 3 Geologische Bundesanstalt, FA Rohstoffgeologie, A-1031 Wien, Rasumofskygasse 23 Key words: mineral resources, assessment, GIS Abstract For the last thirty years the Geological Survey of Austria (GBA) has made assessments of surface-near mineral resources putting emphasis on regional aspects of land use, environment and economy. In collaboration with land use experts, GIS-tools have been developed to evaluate the sustainability of mineral deposits, taking into account possible conflicts during it’s exploitation. An example for the province of Upper Austria, representing the main target of the 1990’s, has been shown at 1998’s ICGESA (L e t o u z é et al., 1998). Recent studies have focused on the province of Carinthia, where a Geological Information System has been designed which includes: • A GIS-based geological map “Setting of gravels, sands and clays” for the whole province of Carinthia (1:50.000 ArcINFO® concept). • Assessment, input and evaluation of basic data from archives, boreholes, pits and literature in general, in order to advance knowledge about surface-near mineral resources. Linking digital geological maps to specific data bases allows for the evaluation of mineral resources quality and usability, thereby contributing to a modern planning instrument within the administration of Carinthia. Internet and intranet availability of such data should initiate strategic mineral planning and strengthen sustainable land use in general. Equivalent treating of solid rocks is ongoing task required to complete land-use relevant mineral resources’ assessment in Carinthia. European regionalization efforts and the common future of Austria and Slovenia within the European Union may support at least comparable structuring of geological data. This paper outlines such an example for Carinthia, one of Austria’s boarder provinces to Slovenia. Preface In Austria, each year 42 mio m3 (=75 mio tons) gravels and sand, 1,5 mio tons of clay and 16 mio m3 (=44 mio tons) of hard rocks are produced and used. Nearly the whole amount is used in construction industry. Austria has an average per capita consumption of 15,t tons of such mineral resources (Heinrich, 1995). Carinthia, Austria’s southernmost province, is able to cover the necessities of it’s gravels and sand consumption from it’s own territory. In detail, significant abundance areas in the south-eastern part contrast with shortage areas (due to lacking geological conditions for reasonable accumulation of such sediments) in the western part of the province. Geological setting Most of the gravels, sands and clays in Carinthia are originally sedimentary products of glaciation through alpine ice ages. 334 Richard Bäk, Maria Heinrich & Gerhard Letouzé-Zezula Carinthia during this period was almost completely covered by glaciers, only the easternmost part stayed ice-free. Metamor-phic rocks of the East Alpine Cristalline as well as the paleocoic rocks above the ice level were significantly alterated and contributing high amounts of debris. Remaining deposits of gravels and sands in Carinthia show a complex suite of glacigene, glaci-lacustrine, glaci-fluviatile and fluvio-lacustrine sediments of the Wuerm glacia-tion and its residual stages. GInS – Geological Information System of Carinthia less, hydrogeological research generated a lot of informations about aggregate resources too. The 2001 amendment of the Federal Mining Law put executive power for exploitation of gravels, sand, clay and natural stones into the hands of the provincial governments. In Carinthia, this triggered a complete assessment of aggregates as well as leading-off a GIS-based Geological Information System (GInS) at the regional Geological Survey, designed to fulfil an expert role in risk-, hydro-, mineral resources- and environmental geology. GInS is built up on tow major fundaments, the digital geological map and the digital Geo-archive. For a long period, Carinthia’s supply with aggregates was only matter of singular licensing while planning was entirely left to the industry. The only matter of public concern and research was the conservation of relevant groundwater bodies, which - in some way – implicated negative consequences for the supply with aggregates out of the same geological setting. Neverthe- GIS-based geological map “Setting of gravels, sands and clays” (Figure 1) Out of geological maps of different origin and quality, different scale and purpose, a digital map for the whole province of Carinthia was compiled and processed into a 1:50.000 ArcINFO® concept. This because Figure 1. Setting of non consolidated rocks showing their relevance as concrete and road construction material (red/green/brown: high/average/low relevance) Stability analysis of underground openings for extraction of natural stone 335 the Survey’s official concept of the „Geo-logical Map of Austria 1:50.000” so far only has edited 5 sheets from 34 sheets, which partially or totally cover the province of Carinthia. All polygons of the new digital map are digitally assigned a) the original legend text and b) a hierarchically structured general legend data base. Classification and description of lithology is according to respective mineral resource’s quality. The general legend has turned out to be necessary for combining polygons of different origin and value. This concept should enable the user to generate a map of any scale and any significance, according to his necessities. For the purpose of Carinthia’s land use planners the recommended map points out aggregates of “relevance as concrete and road construction material” and distinguishes between high / average / low relevance. Digital Geo-Archive Carinthia’s Geological Survey had a large amount of analogous data derived from decades of daily assessment which have been structured for a GIS-based digital archive and partially have been put into GInS. Items covered are: General items, garbage water, railroads, chemical hazards, dump sites, land use, power plants, cable ways, scientific Figure 3. Detailed GIS-image of a gravel pit in the vicinity of Bleiburg, Carinthia projects, bike routes, mining sites, disaster damages, groundwater, road building and geothermal heat pumps. Research for archive data may start out of the digital archive (Figure 2) or out of the GIS-application (Figure 3). Lists of search criteria are helpful for data selection. In parts this archive is already linked to existing Intranet facilities of the Carinthian government. Data input is supported by especially designed MS-Access® applications. Relevant documents could be scanned and linked to the archive. The functionality Figure 2. Digital archive at the Geological Survey of Carinthia 336 Richard Bäk, Maria Heinrich & Gerhard Letouzé-Zezula Figure 4. Data input of boreholes for different purpose (highways, power plants, railroads, others) within MS Windows is optimized through ODBC®-facilities. In order to advance knowledge about surface-near mineral resources, data input for the following items had priority: Boreholes (Figure 4) Data from about 4000 wells have been collected from the archive of the Carinthia’s Geological Survey, from the archive of Carinthia’s Bridge Building Department, from the archives of two regional power plant companies, from the archive of the railway company, out of unpublished projects and out of geological literature in general. Open Pits Information about open pits was essential for mineral resource’s quality evaluation. Contributing archives were once again Carinthia’s Geological Survey and the archive of the national Geological Survey, in total the documentation 1250 mostly gravel pits have been evaluated (Figure 5, detail see also Fig. 3). Data input fields are numerous, most of them provide lists of possible input data, others are designed for free textual input. Figure 5. Position of aggregate pits (red/blue: active/inactive) Stability analysis of underground openings for extraction of natural stone 337 Towards an Austrian Mineral Resources’ Plan In October 2000, Austrian Parliament engaged the Ministry for Economy and Labour to develop a National Plan of Mineral Resources. Starting on a generalized level the supply with mineral resources of the entire nation should be outlined verbally, by figures and maps. In a second attempt, supply concepts will be worked out together with the regional governments. Work started in spring of 2002 and is scheduled for five years. Overall target is approaching a sustainable development on the mineral resource’s sector. References H e i n r i c h , M. 1995: Bundesweite Übersicht zum Forschungsstand der Massenrohstoffe Kies, Kiessand, Brecherprodukte und Bruchsteine für das Bauwesen hinsichtlich der Vorkommen der Abbaubetriebe und der Produktion sowie des Verbrauches - Zusammenfassung. - Berichte der Geologischen Bundesanstalt, 31, Wien. L e t o o z é - Z e z u l a , G., K o c i u , A., L i p i a r s k i , P., P f l e i d e r e r , S. & R e i t n e r , H. 1998: GIS-based geodata processing for regional land use planning. - Proceedings of the International Conference on GIS for Earth Science Applications (ICGESA), 121-128. Ljubljana. Moshammer, B., Posch-Trözmuller , G., L i p i a r s k i , P., R e i t n e r , H. & H e i n r i c h , M. 2002: Erfassung des Baurohstoffpotentials in Kärnten Phase 1: Lockergesteine. - Endbericht Bund-Bundesländer-Rohstoffprojekt, Geol. Bun-desanst., Wien. 338 Richard Bäk, Maria Heinrich & Gerhard Letouzé-Zezula GEOLOGIJA 46/2, 339–342, Ljubljana 2003 A new Slovenian digital cartographic standard for geologic map symbolization Milo{ BAVEC & Marko KOMAC Geolo{ki zavod Slovenije, Dimi~eva 14, SI – 1000 Ljubljana, Slovenija E-mail:milos.bavec@geo-zs.si, marko.komac@geo-zs.si Key words: digital geologic standards, geology Abstract Almost four decades have passed since the last “new” graphic standard has been issued by the former Federal Geological Survey of Yugoslavia (1964). Although it was prepared for the project of Basic Geologic Map at the scale of 1:100.000, its use surpassed the primal purpose, and it became broadly used in various graphic representations of geologic information. Through years, however, the standard outdated and have therefore been sporadically upgraded. Constant changes gradually made it unsystematic and inconsistent. An effort was made a couple of years ago to revise it again, and to convert it into digital form but the result provided another conclusive evidence that a total revision of the original is needed. A new digital cartographic standard for geologic map symbolisation is therefore being prepared in Slovenia. The project is run by a team formed at the Geological Survey of Slovenia in close co-operation with contributors from other geologic institutions in Slovenia. The aim of the project is to prepare a consistent and comprehensive set of graphic symbols and rules of representation that would cover the needs of the geologic maps production in scales between 1:10.000 and 1:100.000. The focus of standard’s applicability is, however, the new Geologic Map of Slovenia in scale 1:50.000. Initial requirements Requirements that the new Standard needs to meet are: • uniform graphic appearance of geologic maps, • employability in maps of scales between 1:10.000 to 1:100.000 with employability focused to the new Geologic Map of Slovenia in scale of 1:50.000, • employability in litho/chronostrati-graphic maps as well as in formation maps, • employability in geologic cross-sections and stratigraphic/lithologic columns of adequate scales, • strict systematics, • employability in various application environments, • user-friendliness (valid for the maker as well as for the reader of the product) • applicable in solving local geologic problems yet comparable with global standards. Methodology A work group of 17 experts has been formed to revise existing standards and to prepare proposals by topics. The topics were distributed among group members according to their professional specialities. Each of 340 the group members prepared a Standard proposal on a selected topic given only the rough outlines of the expected joint product. The proposals were then co-ordinated and revised by the whole work group. After all the suggestions and corrections were taken into account, the manuscript proposals were sent to the Geologic Information Centre of the Geological Survey of Slovenia for digitalisation (transfer into the application environment). After the digitalisation is finished, the draft version will be revised again by the work group and published on the Internet for public revision. After the three-month revision period, the work group will consider suggestions, make necessary corrections and finally publish the Standard on the home page of the Geological Survey of Slovenia. The product will stay open for further suggestions through a digital form published aside. Several Standards exist on the “market”, so it has been clear from the very beginning that there was no need for introducing brand new systematics or to apply a completely different approach. The following Standards were used as the basis for the new product: ISO (1974–1989), JUS (2001), USGS-FGDC (2002), OGK 1 (Savezni geolo{ki zavod SFRJ, 1964), the Standard proposal for OGK 2 (Savezni geolo{ki Zavod SFRJ, 1985) and the manuscript proposal for the Slovenian OGK 2 Standard (Premru & Jev{enak, 1996). However, a simple compilation of existing standards proved to be impossible for several reasons. Some of existing standards are inconsistent by their systematics, some are not compatible with the geology of Slovenia, some are too extensive, and the others are too simple. Preparation of the new proposal therefore required plenty of authorial work, in some cases starting completely from scratch. The structure of the standard The Standard consists of three major topical complexes: 1) graphic and alphanumeric symbols, 2) geologic timescale and 3) rules of representation. Graphic and alphanumeric symbols part is the major topical complex of the Standard. It comprises all graphic symbols, hatches and alphanumeric symbols that may Milo{ Bavec & Marko Komac apply on a geologic map with an exception of formation, and chronostratigraphic notations. The complex is divided into following topics: 1) sediments, sedimentary rocks and sedimentary environments, 2) volcanic and volcanoclastic rocks, 3) magmatic rocks, 4) metamorphic rocks, 5) minerals, 6) tectonics and structural geology, 7) palaeontology, 8) geomorphology, 9) hydrogeology, 10) engineering geology, 11) mineral resources, 12) special symbols, 13) natural heritage. The guiding line through the process of preparation was the strict systematics. Namely, there are several examples of existing Standards that have failed in solving the problem of systematics adequately. By comparing them, it became evident that there are three main reasons for a failure: 1) certain geological phenomena are described by two or more symbols, 2) similar or identical symbols are applied to describe different geological phenomena, 3) basic and expanded symbol sets are put together regardless of any hierarchy. The basic, and the expanded symbol sets are strictly divided in the new Standard. Introducing the basic (obligatory) symbol set along with the expanded set means that the mapping geologist will be able to use the Standard regardless of his field of specialisation. Vice-versa, the specialist should find all the symbols needed to conduct any specialised work with the scale of the map/ profile/column taken as a limitation, of course. All symbols are presented in standardised tables and described by a consecutive number, the name, picture, alphanumeric symbol where applicable and a short comment describing the rules of its use. The timescale is the second topical complex. The main idea that followed the preparation of the timescale was to make a local chronostratigraphic division in accordance with global chronostratigraphic/geochrono-logic divisions. There is a “near-perfect” global time scale that is being constantly updated by the International Commission on Stratigraphy (ICS) of the International Union of Geological Sciences (IUGS). However, the specifics of Slovenian geology make the global timescales non-applicable in certain cases. It was concluded that, for various reasons, local particularities have to be considered for the Carboniferous, the Lower A new Slovenian digital cartographic standard for geologic map symbolization 341 Triassic and the Paleogene – Neogene geo-chrons. The main frame, however, stayed completely comparable with the (semi)offi-cial ICS’s timescale. The divisions go down to the stage or to the substage where reasonable. Beside the main timescale, the new Standard will provide links to three additional (informative) scales that are being broadly used for comparison with official chrono-stratigraphic units. These are: the geomagnetic polarity, ?18O and the Alpine Pleistocene morphostratigraphy. In case of the Tertiary and the Quaternary we took the conservative approach. Namely, the ICS ceased to use those two as formal chronostratigraphic units, but we found them both too well-nested in minds of geologists so the decision was made to use them. In addition, the cancellation of Tertiary and Quaternary has not been formalised yet. Rules of representation define the notation rules for chronostratigraphic, and for-mational geologic units. The aim of the standard is to be applicable for all types of general geologic maps so the rules do not exclude any type of such maps. Transformation into the digital form & GIS To transform the analogue data into the digital form and further into the GIS environment in a effective manner, proper standards are needed. These standards have to be strict, practically perfect and upgradable enough, so that absolutely no divergence from the rules is allowed. At this junction of needs, the concept and the GIS, the consistency and applicability of standards work hand in hand. Graphic symbols, hatches and alphanumeric symbols are transformed into the digital form in proper scales, using CAD tools. After the shape, colour and dimensions are confirmed by the author and the review group, the symbols will be introduced into the standardised procedure of the map digitalisation. For the purpose of the onscreen digitalisation of the Basic Geologic Map at the scale of 1:100.000, the CAD application “Geolog” (i.e. Geologist) was developed (Fig. 1). This standardised pro- cedure is used to minimise the analogue-to-digital transformation errors. The operator (digitiser) uses simple, user-friendly menus in which symbols from different topics are listed and shown. With the described procedure, the basic attributes of a specific symbol are entered and can be further used for the linkage with the symbol’s properties and more detailed description in GIS environment. The already adopted protocol, with necessary updates taken into account, will also be used for digitalisation of geologic maps compiled on the basis of the new Standard. The new Standard is supposed to support production of maps in various scales (from 10.000 to 100.000), hence graphic symbols have to be adjusted to the specific scale and can not be simply scaled. Each symbol has to be defined and designed for each specific common scale used. Also colour charts will have to be defined for each of the symbols. For hatches the RGB and CMYK values need to be defined due to different screen and printer/plotter properties. For graphic symbols (objects and lines) 8-bit colour palette is advisable, to avoid unnecessary dithering on devices that can only display 256 colours (Brown & Feringa, 1999). Standard layouts of maps, which will comprise the map itself, the title, the cross-sections, the columns, the legend(s), the design of the scale bar, the north arrow, appendices, and text, will also be defined. With all described bearing in mind, there are several problems that arise during the process: • non-systematic (chaotic) symbols (hatches), • scaling of symbols, • colour vs. black/white print, • variable hatch orientation within a single geologic unit (bedding…), • ironically, the inter-PC compatibility can sometimes pose big problems due to specifics of CE fonts, commonly used for Slovenian. During the development of the new Standard, all problems stated above should be considered. Than again, geologists should participate in the process of the GIS software development more actively, since our needs sometimes differ from the needs of other spatial related sciences. 342 Milo{ Bavec & Marko Komac Figure 1. CAD application “Geolog” (Geologist) Discussion References Creation of a new Standard consists, in its basics, mostly of compilation of existing standards. However, considering the specifics of regional geology and any kind of special requirements, a pure compilation is practically impossible, and the authorial approach is needed. Two issues arose along the process of preparation that we have not solved in-full yet. The first is the problem of sytematics and division of symbols according to hierarchical criteria. The second problem, which is in-part specific for Slovenia, is the problem of translations. Beside grammatical issues that demonstrated while translating names of chronostratigraphic units, we are still discussing the decision on whether to derive alphanumeric symbols (two- to three letter) from the original (Greek, English....), from English, or from Slovene. Prior to creating data presentation standards for the whole country, three levels of consistency have to be addressed: consistency of the original survey (standardised data collection), consistency of descriptive information (standardised data model) and consistency of coding (Johnson et al., 1997). The former two can be controlled and “enforced” up to a certain degree, while the first one depends purely on the knowledge and consistency of the field geologist. Brown, A. & Feringa, W. 1999: A colour handbook for GIS users and cartographers. – ITC, Enschede. D i m i t r i j e v i } , M. 1978: Geolo{ko kartiranje. – Izdava~ko-informativni centar studenata, 487 pp., Beograd. Geologic Data Subcomittee, Federal Geographic Data Comittee (FGDC), 1997: Proposed Digital Cartographic Standard for Geologic Map Symbolization. – United States Geological Survey. (http://ncgmp.usgs.gov/fgdc_gds/home.html, 2003) International Comission on Stratigraphy (ICS), 2002: Overview of Global Boundary Stratotype Sections and Points, strat info, strat guides and color codes. – International Union of Geological Sciences. (http://www.micropress.org/stratigra-phy/, 2003) International Organization for Standardization, 1974–1989: Graphical symbols for use on detailed maps, plans and geological cross-sections – Parts 1–6. – International standard ISO 710/1– 6. J o h n s o n , B. R., B r o d a r i c , B. & R a i n e s , G. L. 1997: Draft geologic maps data model: version 4.2. – AASG/USGS Geologic Map Data Model Committee. (http://ncgmp.usgs.gov/ngmdbproject/stan-dards/datamodel/model42.pdf, 2003) Komisija za geolo{ke oznake i simbole, 2001: Jugoslovenski standard (JUS) za gelo{ke oznake i simbole. – Savezni zavod za standardizaciju, Beograd. P r e m r u , U. & J e v { e n a k , B. 1996: Dimenzije standardnih znakov za geolo{ko karto Republike Slovenije v merilu 1:50.000, 1. del, revizija 0 – rokopisno poro~ilo. – 68 pp., Geolo{ki zavod Slovenije. Savezni geolo{ki zavod, 1964: Uputstvo za iz-radu osnovne geolo{ka karte SFRJ. 2. izdanje. – Gra?evinska knjiga, 100 pp, Beograd. GEOLOGIJA 46/2, 343–348, Ljubljana 2003 The Architecture of the Georgia Basin Digital Library: Using geoscientific knowledge in sustainable development B. BRODARIC1, M. JOURNEAY2, S. TALWAR2,3, R. HARRAP4, J. van ULDEN2, R. GRANT5 & S.DENNY2 1Geological Survey of Canada, 234B-615 Booth St., Ottawa, ON K1A 0E9 2Geological Survey of Canada, 605 Robson Street, Suite 101, Vancouver, BC V6B 5J3 3University of B.C., Dept. of Geography, 1984 West Mall, Vancouver, BC V6T 1Z2 4Queens University, Dept. of Geological Sciences, Miller Hall, Kingston ON, K7L 3N6 5University of B.C., SDRI, 1924 West Mall, UBC Vancouver, BC Canada V6T 1Z2 {bmjourne,stalwart,jvanulde,rgrant,sdenny}@NRCan.gc.ca; harrap@geol.queensu.ca Key words: Digital Library, architecture, geoscientific knowledge, GIS Abstract To address societal issues such as sustainable development, geoscience knowledge must be transformed from its conventional packaging. A holistic approach to this transformation requires that (1) geo-information be used to develop indicators such as mineral potential, material availability, groundwater capacity, etc., and that (2) the indicators are combined with socio-economic factors to guide generation of future scenarios for land use, population, urban growth, etc. In this paper we discuss the architecture of a prototype system designed to support these two activities. The system consists of a geoscientific data and knowledge repository that is loosely linked with a future scenario modeling tool (QuestTM). The repository supplies transformed information to the tool, and also provides explanations for the input variables and output results. An example implementation, Georgia Basin Digital Library, is also presented. Introduction It is commonly understood by geoscien-tists that geoscience information has a vital role to play in how significant issues, such as climate change, sustainability, biodiversity, natural hazard prevention and mitigation, are to be addressed. Though the importance of geoscientific information may be self-evident to geoscientists, the information itself is nevertheless generally overlooked by non-geoscientists dealing with those issues. Improving accessibility to geo-scientific information through database construction and web delivery raises the visibility and availability of geoscientific information, but may not increase its use in non-geologic domains where the information is not well understood, largely because the information is complex and primarily aimed at geoscientists rather than at biologists, environmental scientists, etc. Geo-science information must therefore be transformed from its conventional packaging, typically as a map, to modes specifically usable by other disciplines prior to its delivery to those disciplines. This transformation may involve the development of geoscientific indicators that reduce significant geological situations into a single re- 344 B. Brodaric, M. Journeay, S. Talwar, gion-specific value for prediction or assessment purposes, such as estimates of mineral resource potential, building material availability, groundwater capacity, etc. It may further involve the incorporation of such geoscientific indicators into broader models that integrate natural resource indicators with socio-economic factors for purposes of generating region-specific predictions for land use, population growth, urban expansion, etc. In this paper we focus on the latter approach and discuss the architecture of an internet-based system that integrates geological information into natural resource-socio-economic models. The system was developed within the Georgia Basin Futures Project (www.basinfutures.net/), an academic research project carried out at the University of British Columbia (UBC) and the Geological Survey of Canada (GSC). The ultimate goal of the project and resulting system is to better inform decision-making in sustainable development by allowing future scenarios to be developed and inspected. The main notion here is firstly, that using geoscientific information as an additional input will improve the accuracy and utility of future scenarios, and secondly, that exploring the nature and consequences of future scenarios will ultimately lead to better decisions. To achieve these two goals we develop a system that links two main modules: (1) an information and knowledge repository (i.e. ontology + data), called the Georgia Basin Digital Library (Journeay, et al., 2000), that allows relevant information and related knowledge to be stored, accessed and explored, and (2) a modeling tool (QuestTM), from UBC, that uses the repository information to generate future scenarios. These linked modules constitute a simple decision support system for sustainability. The Georgia Basin digital library The knowledge and information repository is structured as a web-based digital library that contains 64 significant geospatial information layers such as land use, transportation, various natural resource layers, etc. It also contains a network of concepts, organized as ontologies and described using stories, to explain what role the layers and R. Harrap, J. van Ulden, R. Ggrant & S. Denny their contents play in the sustainable development of the region. So, the repository not only holds information but also possesses knowledge elements that explain what the information means and how it applies to sustainable development. The repository can thus be viewed as also supplying a knowledge layer to the stored information. Because initial implementation of the repository is focused on the southwestern region of British Columbia, the repository is called the Georgia Basin Digital Library (GBDL). Coupled to the repository is a web-based user interface that enables repository contents to be viewed and explored. The interface also connects the repository contents to other relevant sources of information and knowledge, such as publications in other digital libraries, and news stories from web-based news agencies. It further allows individuals and groups to input their own geospatial sites and descriptions, to foster the input of local knowledge and thereby stimulate dialog on issues of sustainability in the context of a community or region. This user interface is called the Georgia Basin Explorer (GBX). The repository is coupled with the user interface via web-enabled functions. As shown in Figure 1, the system architecture has 3 tiers: information, web-services and presentation. The information tier There are three types of information sources that together comprise the information and knowledge repository: metadata, geospatial layers, and a knowledge-database. The metadata component stores in- Figure 1. The three tier technical architecture of GBDL. The Architecture of the Georgia Basin Digital Library: Using geoscientific knowledge ... 345 formation about users, projects, and information content. The 64 geospatial layers represent different geospatial themes for the Georgia Basin and are stored as ESRI shape files. The knowledge-database maintains the ontologies in a relational database (SQL Server): it consists of concepts, their relationships to each other, and to the metadata and geospatial data. The Figure 2 depicts the main elements of the logical schema used by the knowledge-database, and illustrates how geologic information can be stored in the design (c.f. Brodaric & Hastings, 2002; also Brodaric & Gahegan, 2002). The Service Tier The Service Tier provides a suite of web-based functions that manipulate the contents of the Information Tier. Spatial data is exposed via the WMS service specified by the OpenGis Consortium (www.opengis.org), metadata about spatial layers is exposed via the Z39.50 protocol, and project metadata and the knowledge-database are exposed through a custom-built web service. The service tier thus serves as a bridge from the repository to external web clients who can be either persons or software agents. It also serves as an internal bridge between the repository, metadata and spatial data for operations internal to GBDL such as building and updating the repository. The custom web interface is divided into two sets: low-level foundation services operating on the knowledge-base, spatial data (WMS) and metadata (Z39.50), and high-level presentation services that perform dedicated procedures for specific interface components. The services are currently accessed via http, but we are migrating these to SOAP/WDSL platforms. XML encoding standards are used to pass requests and results between tiers (Figure 3). theme —- - _ *¦ - jpckjjnjt _ 'X formation14* % •rock type 'grapndicrite' conceptual • grano Llthology c pdrphyril ¦» ^ monzogran ;e recrystall zed^., - •** Figure 2. Logical schema for the knowledge-database (c.f. B r o d a r i c & H a s t i n g s , 2002). 346 B. Brodaric, M. Journeay, S. Talwar, R. Harrap, J. van Ulden, R. Ggrant & S. Denny The Presentation Tier The presentation tier (GBX) provides a user interface to the information and knowledge repository—it acts as a client to the services tier, exclusively using the services tier to access the library contents. GBX aims to build awareness and understanding about sustainability amongst municipal and scientific communities in the Georgia Basin by directly engaging those communities using the world-wide-web. GBX operates by using the services tier to access the information and knowledge repository. It has five components: (1) News and Information, which connects specific concepts from the knowledge-database with topically relevant web-based news stories and other pages; (2) Library Collections, which uses specific concepts to search web-based catalogs for maps, images and reports and displays them for viewing; (3) Local Stories, which allows users to annotate geospatial layers with per- Figure 3. Example requests to the web services tier, encoded in XML. sonal geospatial sites and related stories that convey local knowledge about issues of sustainability in the community or region; (4) Ideas and Perspectives, which enables concepts and their story-like descriptions (from the knowledge-database) and related geospatial layers to be explored interactively; and (5) Future Scenarios, which connects with the Quest modeling tool for generating future scenarios for the region. Particularly significant is the Ideas and Perspectives component, as it presents community, academic and NGO (non-govn’t organization) perspectives on sustainability, each organized as a separate ontology in the knowledge-database. Figure 4 displays the Ideas & Perspectives component, which enables the perspectives (ontologies) to be explored by browsing a semantic network of concepts (top left) triggering the display of related maps (bottom left) and story-like documentation (right). Figure 4. Exploring sustainable development concepts in GBDL. The Architecture of the Georgia Basin Digital Library: Using geoscientific knowledge ... 347 Figure 5. Explaining the results of a Quest future scenario in GBDL. The quest scenario modeler Apart from its representation and presentation roles, GBDL also provides dynamic connection to the Quest scenario modeling environment, in which future scenarios, such as land use, economic and demographic projections, are constructed. Quest is a computer simulation that enables people from all walks of life to construct alternative futures for a region and view the trade-offs and consequences of their choices. The scenarios are controlled by input from the user and are driven by an integrated suite of system models that reflect the expert knowledge of more than 35 researchers in the fields of environmental, social and economic health. In this modeling procedure GBDL is designed to serve as an information repository, supplying information to Quest for modeling, and as an explanatory tool, allowing the user to browse input variables and output results, thus helping build a context for understanding. The results of a Quest scenario are uploaded to the GBDL scenario library (Figure 5), where the various input parameters and modeling assumptions can be reviewed and further explored in the Ideas & Perspective module (described above). At the time of this paper, the dynamic interaction between the GBDL and Quest is still under development, with positive initial results. Of broader significance is the fact that geoscientific knowledge is utilized as an input theme and therefore as part of the modeling system, clearly demonstrating the rel- evance of geoscience information in the context of sustainability planning and decision making. In our work to date we have prototyped a sub-model that integrates the potential socio-economic impacts of earthquakes into the Quest modeling framework; we are currently working on developing an integrated surface/groundwater sub-model to help address issues of sustainable yield and aquifer vulnerability in the region. Conclusions The GBDL concept has been prototyped in two geographic regions: Bowen Island and the Georgia Basin (http://georgiabasin.info/ ). The Bowen Island prototype is most mature to date, as it has been used extensively in a grade 10 experiential learning environment where students were asked to develop their own Local Stories. GBDL is just recently coming on-line as part of a larger research project, the Georgia Basin Futures Project (http://www.basinfutures.net/). Our experiences from this project indicate that the combination of knowledge representation techniques, modeling tools, geoscience information and the world-wide-web is practical, informative and useful. We anticipate developing these further, specifically by introducing more geoscientific information into the modeling component, tighter coupling of the modeling and knowledge repository components, and applying the system in different geographic regions. References B r o d a r i c , B. & G a h e g a n , M. 2002: Distinguishing instances and evidence of geographical concepts for geospatial database design. In: Geographic Information Science-2nd Int’l Conference, GIScience 2002, M. Egenhofer and D. Mark (eds.), LNCS 2478, Springer, 22-37, Berlin. B r o d a r i c , B. & H a s t i n g s , J. 2002: An object model for geologic map information. In: Richardson, D., van Oosterom, P. (Eds.), Advances in Spatial Data Handling, 10th International Symposium on Spatial Data Handling, Springer, 55-68, New York. J o u r n e a y , M., R o b i n s o n , J., T a l w a r , S., Walsh, M., Biggs, D., McNaney, K., Kay, B., B r o d a r i c , B. & H a r r a p , R. 2000: The Georgia Basin Digital Library: Infrastructure for a Sustainable Future. Proceedings, GeoCanada 2000, May 29-June 2, Calgary. 348 B. Brodaric, M. Journeay, S. Talwar, R. Harrap, J. van Ulden, R. Ggrant & S. Denny GEOLOGIJA 46/2, 349–360, Ljubljana 2003 Feature Map Classifier – a possible approach to morphological/ geological evaluation of terrain Janez HAFNER Republic of Slovenia, Ministry of Defense, Kardeljeva plo{~ad 25, SI-1000 Ljubljana, Slovenia E-mail:janez.hafner@pub.mo-rs.si Key words: Classification, geology, image processing, neural networks, self organising maps Abstract This paper investigates the practicability of new classification approach to image processing. Landsat TM image of Slovene coastal area was used to perform lithological classification. Standard classification methods, based on statistical principles do not always give satisfactory results. Therefore a variety of new approaches are being tested in order to achieve better accuraccy. One of the most promising fields is artificial intelligence where artificial neural networks (ANNs) have proven to be usefull. An artificial neural network represents a limited analogy of the neural functioning of the biological brain (Sejnowski, Koch and Churchland, 1988). Several ANN methods have been developed to solve classification problems. The one represented in this article is a combination of two methods: Self Organising Maps and Backpropagation network. This kind of ANN, also called Feature Map Classifier, is not the best in sence of accuraccy but has one big advantage in comparison with other ANNs – it is much more transparent. In comparison with standard approach better results were gained especially in more complicated cases, where classes are not linearly separable. The separability of classes is shown to be one of the most important factors. ANN methods tend to give better results as statistical clustering technique especially in cases when classes overlap and are not easily separable. Introduction The morphologic/geologic mapping of a territory is one of the preliminary steps in selecting the optimal highway route. Slovene geologists are under pressure to provide planners with fast solutions, for all intended areas of the extensive highway construction program covering the entire country. Mapping requires extensive fieldwork, the cost and duration of which is directly linked with the feasibility study. Developed is an morphologic/geologic map. It, being derived by the interpretation of field data, is subject to many different influences such as the availability of data, impassability of a territory, it’s overgrowth, and human expertise. The question then remains, whether it is possible to obtain a low cost and relatively quick solution that would be independent of hu- 350 Janez Hafner man subjectivity. The solution consists of two parts. First, there is a problem of input data. The acquisition should be based on already existing or easily derived data. One of the fastest ways is to use remotely sensed data. Nowadays, there are a variety of commercial satellites supplying a multitude of data that can be used in different ways. Highresolution satellite multispectral imagery (LANDSAT, SPOT, Ikonos) is useful for the analysis of vegetation and soil characteristics, whilst radar (RADARSAT, ERS) and stereoscopic pairs (SPOT, Ikonos) imagery can be used for the creation of digital elevation models. In the work presented hereafter, an attempt was made to compensate for the contribution of lithology with the use of the LANDSAT Thematic Mapper image. Besides satellite data, other already existing sources of data were considered, including digital geologic and geodetic maps that were used to derive various derivative data layers. The classical approach to aerophoto imagery was also used to delineate different vegetation cover types. Second, the processing of data should be automatic. This means that all data should be gathered or converted into a digital format, then processed at a later stage by computer. The digital georeferenced data can be considered as digital imagery, whilst the procedural methods can be considered as digital image processing. In order for them to be performed, a knowledge base about examined phenomena, must be designed. Several different designs exist. For instance, fcUSTfÖ* ^v r^ iL SLOVENIA ¦^ä ^•=5^^ \ XJ^ expert systems try to incorporate expert knowledge, in the form of IF-THEN rules. The problem arises, when one makes an attempt to reshape expert know-how to comply with strict rules. Mapping is usually described as a highly intuitive process, which is often very difficult to describe. Therefore, a methodology with self-learning capabilities is needed. In the field of digital image processing, various parametric and non-parametric methods (classifiers) have won wide recognition, and are now widely used. With this work, an attempt was made to investigate the application of a special non-parametric method, inspired by the human brain – artificial neural networks. Study area The study area occupies approximately 50 square kilometres of the Slovene coastal area, near the Italian border (Fig.1). It bears all the morphological, lithological and vegetation properties of the whole coastal region, and can be considered to be representative. The examined territory was first mapped using standard techniques, such as fieldwork and aerophoto interpretation. The mapping produced a morphologic/geologic map, where 7 different area categories were extracted: stable areas, labile areas, landslides, sinkholes, debris, moist areas and erosion zones. In the later course of work, this map was used for the random extraction of learning, and for testing samples. The resulting map is mainly an interpretation of lithological and morphological factors. In this place it should be pointed out that the morphological shape of terrain exibits strong correlation with lithology. The flysch, composed of sandstone and marl, is less resistant to weathering than limestone. As a consequence, clastic sediments are covered with a thick, overgrown, weathering cover and the morphological shape of flysch area is heavely ditched and full of ravines. Areas with limestone are re- ^--------------- Crni Kai region Figure 1. Geographical position of ^rni Kal region. Feature Map Classifier – a possible approach to sistant to mechanical but non-resistant to chemical weathering. Ordinarily, they are seldom found with any significant weathering cover and are marked with a variety of karstic phenomena such as sinkholes and doline. In the mapping (classification) process the problems arise because it is possible that certain areas belong to multiple categories at once. For instance, labile areas and landslides are much alike. Furthermore, erosion zones and debris areas are often subject to sliding. The problem is twofold. First, logic of the model obliges us to select only one category, and second, categories are highly overlapping. Consequently, decisions are made by experts, according to their knowledge and experience. Input data Lithological data Lithological mapping of the entire Slovene coastal area in a scale of 1:5000 took place in 1994. The authors (Ribi~i~ et al, 1994) found 12 different lithological units. For the purposes of this project, they were combined to form 5 units (Fig. 2a): alluvium, limestone, deluvium, and two types of flysch – one with a majority content of sandstone, the other with majority content of marl. Digital elevation model and derivatives A digital elevation model (DEM) with 5 meters of spatial resolution was made using isolines from base topographic plans in a scale of 1:5000. The information derived from a DEM (see Fig. 3 – shaded relief) does not alone contribute heavily to a model, but is very important as a foundation for several DEM based derivatives. The slope data layer is expressed as the change in elevation over a certain distance. In this case, the distance is the size of a pixel. The slope of the terrain is directly connected with weathering, erosion and deposition. It was expected to directly influence in the occurrence of landslides and labile areas. The aspect data layer describes the direction of the slope at each pixel. The aspect data was not expected to have a direct influence on the model, but is important nevertheless for the creation of related statistic layers. morphological/geological evaluation of terrain 351 Both slope and aspect data layers were derived using 1st order derivative filters (gradient operators). By employing a 2nd order derivative filter (Laplacian operator), data layers describing convexity/concavity of landscape morphology were made. Two different filters, using neighbourhood sized 7×7 and 11×11 cells were used to produce two different data layers. The curvature data layer was produced using the 4th polynomial function (ESRI, 1992) to describe curvature of landscape morphology. This calculation uses the neighbourhood sized 3×3 cells and is closely related to convexity/concavity of data layers. The outputs were three data layers: general curvature, planform curvature – curvature perpendicular to the slope direction and profile curvature – and curvature of the surface in the direction of the slope. Curvature, convexity/concavity, slope and aspect data layers are all strongly related to the physical characteristics of a drainage basin and can be used to describe erosion and runoff processes. The slope affects the overall rate of downslope movement, whilst the aspect defines its direction. The profile curvature affects the acceleration and deceleration of flow, therefore influencing erosion and deposition. The planform curvature, together with convexity/concavity, influences convergence and the divergence of flow. The flowlength data layer details the downstream distance along a flow path of the hypothetical rainfall/runoff events. This layer is used with the intent of extracting erosion zones that are typically formed in the upper flow areas of flysch lithology. In calculating the standard deviation for slope and aspect data layers using two different neighbourhood sizes (5×5 and 8×8 cells), four standard deviation data layers were derived. The variability within the aspect data layer is closely connected with ridge/ ravine detection, where slope direction opposes one another. The variability of slope is related to the undulation of the surface – a typical phenomena for unstable areas. Distance to surface waters This layer, made by a calculation within an individual cell, measures the minimum distance to surface waters, and could assist in defining the process of erosion in nonkarst areas. 352 Janez Hafner Vegetation data The coarse vegetation map (Fig. 3b) separating three different categories (dense, medium and non-overgrown areas) was compiled using aerophoto interpretation. resolution (except thermal infrared – 120 meters). The satellite image was, due to coarse spatial resolution and occasional cloud cover, not expected to completely substitute lithological data contribution. LANDSAT TM image Two models, the first using a lithological map, the second using a LANDSAT Thematic Mapper image, were made in order to study the application of a model in cases where no lithological data was accessible. The LANDSAT TM image consists of 7 spectral bands: blue, green, red, near infrared, thermal infrared and 2 mid-infrareds, with an 8-bit radiometric and 30 meter spatial Methodology The modelling of a terrain is primarily a classification process. The input data is clustered into homogeneous and separable groups according to an appropriate measure of similarity. With the intention to improve the classification process, new methods, among them Figure 2. Lithological input data (a), morphological/ geological map obtained by standard mapping methods (b), result of classification using lithological data (c) and result of classification using satellite data (d). Feature Map Classifier – a possible approach to morphological/geological evaluation of terrain 353 Figure 3. Shaded relief (a) and vegetation data layer (b). also the subject of this article – artificial neural networks (ANNs), are being developed. Although similar to the k-NN, they are more efficient and require less data for training. The distribution free nature enables them to join remote sensing and geographic data of multiple type/statistical distributions, in the single classification model. The overall objective of this paper is to investigate and discuss their application in the morphological/geological evaluation of an examined area. An artificial intelligence technique, ANN, inspired by the human brain, has shown promising results and is now considered to be an alternative method in digital image processing. The human brain forms a massive communication network, consisting of billions of nerve cells, known as neurones. In functional terms a neuron can be seen as a processing unit receiving and transmitting electrical impulses. Learning, and the storage of knowledge take place by modifying the conductivity between the neuron connections. ANNs are mathematical models that try to mimic the biologic brain. Artificial neurons, also called processing elements, or PEs, are interconnected through numerical weights that function analogously to conductivity connections. The selected ANN method FMC (Feature Map Classifier) combines great representational and analytical power of Self Organising Maps (SOM) with the classification abilities of Backpropagation (BPG) networks. A backpropagation network can be used as a classifier itself, and as such it remains the most popular ANN classification method, but there is an inherent problem. That is, it is very difficult to give physical meaning to the weights connected to the neurones. The use of an SOM method makes the evaluation of classes and the classifier itself much easier. When confronted with multidimensional data it is often very difficult to determine how the data is structured; therefore, it is desired to reduce the dimensionality. The statistical method usually used in performing this task is factor analysis. However, as with parametric statistical classifiers, there is again a problem of normal distribution and linear relations among variables. To solve this problem a special ANN, Self Organising Map was developed by Tuevo Kohonen (Kohonen, 1984). It reduces a multivariable space into two (sometimes three) dimensions in such a manner that makes it possible for every n-dimen-sional input pattern to occupy its place in a 2-D map. Results Two FMC models were made, the first for the classification using lithological data, whilst the second substituting lithological data with 7 bands of LANDSAT TM image. Both artificial neural networks constituted of an input neural layer (with 20 and 22 neurones), Kohonen’s layer (matrix of 20 × 20 neurones) and an output layer (7 neurones representing 7 categories). After the learning stage, the recall procedure took place to actually perform the classification of the studied area. Results are shown in figure 2. 354 Janez Hafner Classification accuracy The assessment of classification accuracy of multivariate data unfortunately does not reach the ability to produce digital land cover classification. In fact, this problem sometimes precludes the application of automated land cover classification techniques, even when their cost compares favourably with a more traditional means of data collection (Lillesand & Kiefer , 1994). Methods originate from the field of image processing and are often described in expert literature. The most common way to represent the classification accuracy is in the form of an error matrix (Congalton, 1991). Accuracy assessment results Accuracy was assessed using a test sample taken by stratified sampling. This kind of accuracy is not dependent upon categories existent in the examined territory. The solution is a general model that describes the model’s behaviour, not just in this case but also for any other resembling region. Another fact in favour of the general model is the nature of input data. All the data, except the satellite image, is independent of the acquisition time and, therefore, dependant only upon the mode of acquisition. Ideally, where procedures are standardised, input data generation should yield identical or at least similar results. This leads to the conclusion that the model using lithological data is general and the model using satellite images is less so. The error matrixes of both models were used as the basis for an accuracy table (Table 1 and Table 2) generation. The equality of average produced accuracy and overall accuracy is the result of equivalent sample sizes for singular categories. The results show satisfactory accuracy. Significant difficulties arise only in the case of labile and stable areas. Whilst the stable areas in both models show similar proclivity to moist areas, the labile areas in the model with lithological data are mixed with a debris category, and the model using satellite images is mixed with erosion zones. Despite the mixing problem, both of the models are within the borders of serviceability, especially in the feasibility stage of study where one strives to yield a fast and low-cost outcome. The comparison of models (Table 3) shows that the model using satellite images yields a somewhat weaker result than the model using lithological data. This confirms the conjecture of the ability of LANDSATs TM images to substitute the contribution of the lithological data. Table 1. Accuracy table for the classification – usage of lithological data REFEREN CLASSIFIED CORRECT PROD. A. USER. A. KHAT EROS 250 260 197 78,80% 75,77% 71,73% DEBRIS 250 278 223 89,20% 80,22% 76,92% LABIL 250 195 106 42,40% 54,36% 46,75% MOIST 250 301 216 86,40% 71,76% 67,05% LANDS 250 254 207 82,80% 81,50% 78,41% SINKH 250 275 240 96,00% 87,27% 85,15% STABIL 250 186 109 43,60% 58,60% 51,70% AVERAGE: 74,17% 72,78% 68,24% OVERALL KHAT: 69,87% OVERALL ACCURACY: 74,17% Table 2. Accuracy table for the classification – usage of satellite image REFEREN CLASSIFIED CORRECT PROD. A. USER. A. KHAT EROS 250 311 200 80,00% 64,31% 58,36% DEBRIS 250 224 139 55,60% 62,05% 55,73% LABIL 250 216 94 37,60% 43,52% 34,10% MOIST 250 289 208 83,20% 71,97% 67,30% LANDS 250 263 213 85,20% 80,99% 77,82% SINKH 250 278 236 94,40% 84,89% 82,37% STABIL 250 168 91 36,40% 54,17% 46,53% AVERAGE: 67,49% 65,99% 60,31% OVERALL KHAT: 62,07% OVERALL ACCURACY: 67,49% Feature Map Classifier – a possible approach to morphological/geological evaluation of terrain 355 Table 3. Comparison of accuracies ACCURACY LITHOLOGICAL DATA SATELLITE DATA Overall acc. Overall KHAT Average Prod User KHAT 74,17% 68,87% 74,71% 72,78% 68,24% 67,49% 62,07% 67,49% 65,99% 60,31% Another way to evaluate the application of the model is to use expert opinion. In spite of the fact that such judgements are heavily affected by human subjectivity, it can still be used as a quick and approximate method. In this case, expert opinion is in favour of both models. The thick vegetation cover of flysch terrain makes it difficult to investigate. This affected the quality of learning/testing data. The resulting classification for this region, therefore, is evaluated to be even better than the actual input (reference) data. The explanation could lie in the ability of ANNs to reduce the noise of learning data, and thus produce a highly generalised solution. The classification of moist areas is inappropriate. That is why it would be reasonable for this category to be either excluded from the model, or to be joined with the stable area category. The feature map classifier has proven to be a useful tool. High delimitation abilities of input feature space and a capacity to work with distribution free data of various data types makes it superior in comparison with statistical classifiers. ANNs in general are criticised because of nontransparent functioning – the black box effect. Fortunately in the case of FMC, this statement does not apply. While the usual classification of ANNs enables users to perform only one single analytical operation – impact analysis, it is the SOM part of FMC, which is known for its great power to present data in the form of 2D matrices. Impact analysis Impact analysis is a technique where the relevance of each input variable is determined using the leave-one-out method. The effect of disabling each input neuron in turn is determined in terms of its percentage reduction in the accuracy of the classification. Figure 4 and table 4 illustrate the results of impact analyses for both models. The Y coordinate is defined as a portion of accuracy reduction when an input neuron for a certain category is turned off. On the X-axis, all the individual and, in cases where they constitute a logical unit, joint categories (preceded by SUM) are presented. The result for the first model indicates lithology to be the most important input data. Seven bands of LANDSAT TM satellite image do not represent proper compensation. Nevertheless, the accuracy of the second model is affected to a smaller extent than expected. The contribution of lithological data is compensated by the increased importance of other factors, particularly vegetation, ridge/ravine data and wa- n n 1 r J \ If 1 l r r n-, m HI Figure 4. Result of impact analyses. 356 Janez Hafner Table 4. Importance of input variables according to impact analysis LITHOLOGICAL DATA SATELLITE DATA IMPORTANT Lithology Undulation Elevation Ridge/Ravine Vegetation Vegetation Ridge/Ravine Undulation Flowlength Distance to surface waters LESS IMPORTANT Distance to surface waters Slope Flowlength Elevation Satellite image Aspect UNIMPORTANT Aspect Curvature Convexity/Concavity Slope Curvature Convexity/Concavity ter related factors. The examined territory is especially characterised by high lithology-vegetation correlation, bare limestone and highly overgrown flysch. Data on the basis of standard deviation, ridge/ravine and undulation data, shows that the neighbourhood of 8 × 8 cells is more important than the neighbourhood of 5 × 5 cells. The low importance of slope data (Table 4) upon analysis gave an unexpected result. It was expected that the slope would strongly affect the stability of the ground. This presumption however, holds only for flysch areas. An imbricated limestone region is, on the other hand, characterised with almost vertical, but fully stable slopes. An important point to note here is that there still exists a variety of attributes not included in this model. Human intervention especially in the form of irrigation, civil engineering activities, tree cutting and water dams often decisively influence the nature of environment. Competitive layer analysis While the impact analysis gives an impression about the importance of the input variables, questions about their mutual relationship, the separability of output categories and the impact of input variables on a particular category remained unanswered. Self-organising maps, as a part of FMC, offer perhaps the best solution at this point in time, that artificial neural networks are able to produce. The unsupervised learning that takes place in the self-organising phase forms so-called Kohonen’s maps, where bright tones represent characteristic regions, and dark tones uncharacteristic regions. Organising data into groups depends of course upon the purpose we are seeking to achieve. Groups can be organised according to output categories. In this way Kohonen’s maps, describing categories, are made (Fig. 5). Examination of their mutual relationship enables a resemblance/separability study of the classification model to be made. Ideally, each category would occupy its own part of the Kohonen’s map, and would not overlap with any other categories. The importance of input variables can be studied by partitioning data into classes, according to individual variable values. The Kohonen’s maps in this case represent inner-variable groups (Fig. 5). Variables can be differentiated either by their individual class values (vegetation: dense, medium, non-overgrown areas) or by some other kind of arbitrary values (curvature: concave [–?, -1], flat [-1,1], convex [1,?]). The differentiation of inner-variable groups serves as an indicator of variable importance - the better the discrimination, the more important the variable. For instance, the differentiation of variable Slope (Fig. 6) shows that three different groups for slope: <10°, 10° to 20° and >20° exist. The comparison of inner-variable Kohonen’s maps of different variables is used for the study of their relationships, whilst the comparison of inner-variable Kohonen’s maps with output category Kohonen’s maps is used to estimate the influence that the individual inner-variable group has on the particular category (Fig. 5). Visual comparison of Kohonen’s maps is useful only in cases where there is a likeness/ dissimilarity of the two maps. Obviously, in Feature Map Classifier – a possible approach to morphological/geological evaluation of terrain 357 MorfJGeol. categories erosion zones Lithology Undulation vivid 4 no "4 Elevation 3ÜÜ 450 Vivid Moder No High Moder Bare Alluv Lime Fl-s Fl-m Deluv Eros Debris Labil Moist Landsl Sinkh Stab Ridge No 0,00 1,48 1,52 0,76 1,28 0,00 1,18 0,64 0,45 1,48 0,00 0,00 0,78 1,16 0,12 1,45 0,00 1,29 1,34 1,14 1,56 1,59 0,20 0,70 0,94 0,27 Weak 1,48 0,00 0,08 1,51 0,15 1,48 0,30 1,82 1,44 0,00 1,48 1,48 0,40 0,27 1,48 0,13 1,48 0,16 0,32 0,33 0,09 0,28 1,79 1,35 1,56 1,65 Ridge/Ravine 1,52 0,08 0,00 1,45 0,27 1,52 0,28 1,83 1,64 0,08 1,52 1,52 0,36 0,30 1,55 0,16 1,52 0,26 0,33 0,32 0,08 0,18 1,80 1,26 1,51 1,70 Pronounced 0,76 1,51 1,45 0,00 1,76 0,76 1,83 0,37 0,45 1,51 0,76 0,76 1,13 1,81 0,56 1,70 0,76 1,76 1,79 1,84 1,41 1,09 0,52 0,13 0,06 0,47 Elevation <150 1,28 0,15 0,27 1,76 0,00 1,28 0,19 1,73 1,38 0,15 1,28 1,28 0,57 0,11 1,32 0,14 1,28 0,01 0,14 0,20 0,33 0,67 1,61 1,66 1,79 1,54 <300 0,00 1,48 1,52 0,76 1,28 0,00 1,18 0,64 0,45 1,48 0,00 0,00 0,78 1,16 0,12 1,45 0,00 1,29 1,34 1,14 1,56 1,59 0,20 0,70 0,94 0,27 <450 1,18 0,30 0,28 1,83 0,19 1,18 0,00 1,75 1,56 0,30 1,18 1,18 0,46 0,09 1,40 0,15 1,18 0,18 0,22 0,01 0,34 0,61 1,51 1,59 1,83 1,54 >450 0,64 1,82 1,83 0,37 1,73 0,64 1,75 0,00 0,32 1,82 0,64 0,64 1,65 1,79 0,49 1,74 0,64 1,71 1,62 1,74 1,72 1,57 0,21 0,65 0,29 0,26 Undul Vivid 0,45 1,44 1,64 0,45 1,38 0,45 1,56 0,32 0,00 1,44 0,45 0,45 1,22 1,52 0,25 1,46 0,45 1,42 1,45 1,54 1,50 1,49 0,36 0,67 0,50 0,20 Moderate 1,48 0,00 0,08 1,51 0,15 1,48 0,30 1,82 1,44 0,00 1,48 1,48 0,40 0,27 1,48 0,13 1,48 0,16 0,32 0,33 0,09 0,28 1,79 1,35 1,56 1,65 No 0,00 1,48 1,52 0,76 1,28 0,00 1,18 0,64 0,45 1,48 0,00 0,00 0,78 1,16 0,12 1,45 0,00 1,29 1,34 1,14 1,56 1,59 0,20 0,70 0,94 0,27 Veget. Highly overgr 0,00 1,48 1,52 0,76 1,28 0,00 1,18 0,64 0,45 1,48 0,00 0,00 0,78 1,16 0,12 1,45 0,00 1,29 1,34 1,14 1,56 1,59 0,20 0,70 0,94 0,27 Moderate ov. 0,78 0,40 0,36 1,13 0,57 0,78 0,46 1,65 1,22 0,40 0,78 0,78 0,00 0,42 0,94 0,56 0,78 0,59 0,79 0,48 0,51 0,57 1,24 0,80 1,35 1,23 Bare land 1,16 0,27 0,30 1,81 0,11 1,16 0,09 1,79 1,52 0,27 1,16 1,16 0,42 0,00 1,35 0,20 1,16 0,12 0,24 0,09 0,46 0,76 1,53 1,59 1,84 1,58 Lithology Alluvium 0,12 1,48 1,55 0,56 1,32 0,12 1,40 0,49 0,25 1,48 0,12 0,12 0,94 1,35 0,00 1,46 0,12 1,34 1,36 1,34 1,55 1,55 0,26 0,69 0,77 0,10 limestone 1,45 0,13 0,16 1,70 0,14 1,45 0,15 1,74 1,46 0,13 1,45 1,45 0,56 0,20 1,46 0,00 1,45 0,15 0,18 0,16 0,18 0,43 1,76 1,64 1,72 1,53 Flysch-sands. 0,00 1,48 1,52 0,76 1,28 0,00 1,18 0,64 0,45 1,48 0,00 0,00 0,78 1,16 0,12 1,45 0,00 1,29 1,34 1,14 1,56 1,59 0,20 0,70 0,94 0,27 Flysch-marl 1,29 0,16 0,26 1,76 0,01 1,29 0,18 1,71 1,42 0,16 1,29 1,29 0,59 0,12 1,34 0,15 1,29 0,00 0,12 0,20 0,32 0,65 1,60 1,66 1,78 1,55 Deluvium 1,34 0,32 0,33 1,79 0,14 1,34 0,22 1,62 1,45 0,32 1,34 1,34 0,79 0,24 1,36 0,18 1,34 0,12 0,00 0,22 0,40 0,70 1,59 1,75 1,78 1,50 Morph./Geol Erosion zone 1,14 0,33 0,32 1,84 0,20 1,14 0,01 1,74 1,54 0,33 1,14 1,14 0,48 0,09 1,34 0,16 1,14 0,20 0,22 0,00 0,38 0,66 1,49 1,62 1,86 1,48 categories Debris 1,56 0,09 0,08 1,41 0,33 1,56 0,34 1,72 1,50 0,09 1,56 1,56 0,51 0,46 1,55 0,18 1,56 0,32 0,40 0,38 0,00 0,09 1,77 1,26 1,43 1,61 Labile areas 1,59 0,28 0,18 1,09 0,67 1,59 0,61 1,57 1,49 0,28 1,59 1,59 0,57 0,76 1,55 0,43 1,59 0,65 0,70 0,66 0,09 0,00 1,69 0,93 1,11 1,55 Moist areas 0,20 1,79 1,80 0,52 1,61 0,20 1,51 0,21 0,36 1,79 0,20 0,20 1,24 1,53 0,26 1,76 0,20 1,60 1,59 1,49 1,77 1,69 0,00 0,57 0,54 0,24 Landslides 0,70 1,35 1,26 0,13 1,66 0,70 1,59 0,65 0,67 1,35 0,70 0,70 0,80 1,59 0,69 1,64 0,70 1,66 1,75 1,62 1,26 0,93 0,57 0,00 0,21 0,72 Sinkholes 0,94 1,56 1,51 0,06 1,79 0,94 1,83 0,29 0,50 1,56 0,94 0,94 1,35 1,84 0,77 1,72 0,94 1,78 1,78 1,86 1,43 1,11 0,54 0,21 0,00 0,60 Stabile areas 0,27 1,65 1,70 0,47 1,54 0,27 1,54 0,26 0,20 1,65 0,27 0,27 1,23 1,58 0,10 1,53 0,27 1,55 1,50 1,48 1,61 1,55 0,24 0,72 0,60 0,00 CD 1 C SE C -a o 360 Janez Hafner generalisation and already existing data, it serves as an efficient tool to produce fast, low-cost estimations of morphologic/geologic properties. This kind of modelling is, however, constrained for usage in the preliminary stages of such works, and should be in the later course of work, combined with standard fieldwork. References A t k i n s o n , P.M. & T a t n a l l , A.R.L. 1997: Neural networks in remote sensing, International Journal of Remote Sensing, 18(4), 699–709. C o n g a l t o n , R.G. 1991: A review of assessing the accuracy of classifications of remotely sensed data, Remote Sensing of Environment, 37, 35–46. C o r t i j o , F.J. & Perez De La Blanca, N. 1997: A comparative study of some non-parametric spectral classifiers, Applications to problems with high-overlapping training sets, International Journal of Remote Sensing, 18(6), 1259–1275. D e v i j v e r , P. & K i t t l e r , J. 1982: Pattern recognition. A Statistical approach, Prentice Hall International, London. ESRI, 1992. ArcInfo User’s Guide: Cell-based Modeling with GRIDR, 2nd edition, Redlands, CA. F r e e m a n , J.A. & S k a p u r a , D.M. 1991: Neural Networks. Algorithms, Applications and Programming Techniques, Addison Wesley Publishing Company. G a l l a n t , S. 1993: Neural network Learning, MIT Press, Cambridge, Massachusetts. J a n ‘ a , M. & R i b i ~ i ~ , M. 1998: Prediction of landslide occurrence possibilities with spatial decision support system, ICGESA 98. K o h o n e n , T. 1984: Self-Organization and Associative Memory, Springer-Verlag. Komac, M. & Ribi~i~, M. 1998: The application of remote sensing – satellite imagery in engineering geology, ICGESA 98. L i l l e s a n d , T.M. & K i e f e r , R.W. 1994: Remote Sensing and Image Interpretation, 3rd Edition, John Wiley and Sons, New York. P a r k e r , D.B. 1985: Learning Logic, Technical Report TR-47, Centre for Computational Research in Economics and Management Science, MIT, Cambridge, MA. R i b i ~ i ~ , M., ^ e n ~ u r , B., J e r m a n , J. & Buser, I. 1994: Engineering geological map of Slovene coastal area, Geological survey of Slovenia. R u m e l h a r t , D.E., H i n t o n , G.E. & W i l l -iams, R.J. 1986: Learning internal representations by error propagation, In D.E. Rumelhart and J.L. McClelland (Eds.), Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1: Foundations. : 318–362, MIT Press, Cambridge, Massachusetts. Story, M. & Congalton, R.G. 1986: Accuracy assessment: a user’s perspective, Photogram-metric Engineering and Remote Sensing, 52, 397– 399. Werbos, P. 1974: Beyond Regression: New Tools for Prediction and Analysis in the Behavioral Sciences, PhD thesis, Harvard, Cambridge, MA. @ e p i ~ , F., H a f n e r , J., G o o s s e n s , M., R i b i ~ i ~ , M. & [ i n i g o j , J. 1998: Necessary first step in the process of the establishing national geological information system – the Slovenian experience, ICGESA 98. GEOLOGIJA 46/2, 361–366, Ljubljana 2003 Why geologists don’t listen and the public can’t read geological maps1 Ian JACKSON British Geological Survey, Keyworth, Nottingham, UK Key words: geological map, GHASP, GIS Introduction Geology is a science that is not only tremendously interesting and exciting; it is also of fundamental importance to our lives, our environment and our assets? No argument with this is there? This statement is obviously true, right? Well if your answer is yes, then why are the majority of our politicians completely unaware of the financial and human cost of ignoring geological hazards?; why had the guy you met in a bar the other evening never heard of your organisation?; why are geological surveys so poorly funded?; and just in case you are still not convinced, why are you so poorly paid? Geology is not irrelevant, but for decades we geologists have been largely creating products that appeal only to one audience: ourselves. I concede they may be much sought after for colourful wall posters, but have you ever met anyone outside the profession who could understand a conventional geological map, let alone comprehend what that map has to say about risks and resources? And what geological maps are try- ing to reveal of the 3rd dimension remains a mystery to the majority. Hold on, I hear you say, now we are in the 21st Century and things have changed; we’ve got sophisticated new computers with GIS and 3D modelling software, and we can devise all sorts of colourful coverages, dynamic databases and mutating models! OK, but how convinced are you that the public (who, I might add, provocatively, probably pay your salary though their taxes) now understand our message any better? Geological maps and models (digital or analogue) are crucial, but they must not be seen as the end point; they are only a means to an end; and that end must be ensuring our science is understood and meets the needs of our users and not just us. If you work for a geological survey this has to be your over-riding priority. In addition to examining the relevance and perception of geological surveys, this paper will take a critical look at traditional products (and some recent digital ones), review some alternative options and discuss the issues which arise when a geological survey tries to take the products of geological 1 With acknowledgements and apologies to Allan & Barbara Pease, authors of the best-selling book “Why men don’t listen and women can’t read maps” 362 surveying and research out of the sophisticated, but limited, circles of the geoscience cognoscenti (yes, that’s us!) into the real world. Relevant – Yes; Understood and appreciated - No Geological factors are important in disaster mitigation and planning, environmental protection and resource exploitation. An understanding of them is essential in establishing policies for sustainable development and can assist in addressing a range of socio-economic, biodiversity and landscape issues. However, decision–makers, the politicians, planners, financiers, businessmen and the legal profession, often fail to take geology into account, leading to increased financial costs (e.g. badly located construction schemes, inadequate planning for the use of natural resources), reduction in the quality of life of citizens (e.g. radon emission, pollution of water supply) and at worst, loss of life (e.g. landslides). Examples from Great Britain (a relatively geologically stable country) show the majority of politicians and planners seemingly unaware of, for instance, the swelling and shrinking properties of clay or the dissolution of gypsum, and allowing housing development that is inappropriate in terms of both location and design. Roads and car parks have been constructed over landslipped ground causing death and injury. The importance of including geoscience knowledge in the prediction of radon-affected areas is only just being recognized. In GB a lawyer would be deemed as professionally negligent if s/he did not obtain a report into possible coal mining beneath a property prior to purchase. But at the moment there is no compulsion to seek out information on potentially damaging natural hazards and yet the case is equally compelling. The estimate of insured losses due to natural geological instability in GB is approximately 450 million Euros per year. There are more than 40 individual nations in greater Europe and each of these countries has a geological survey organisation (GSO). There exists within each geological survey an enormous wealth of relevant geological data and knowledge. Infor- Ian Jackson mation, that can, for example, help to mitigate the affects of radon, flood-risk and subsidence. But this is a knowledge base that is grossly under-used and it will stay that way until we understand better how to convert it into the products and services that people want. At the beginning of the 21st century, at a time when “the environment” has the highest of profiles, geoscience knowledge should be occupying a more prominent role. But it is not. It is a sad fact that the importance of geology to the environment, and to human health, property and assets is not well understood outside the geological profession. Geoscientists and geological surveys and research institutions must accept a substantial part of the responsibility for this lack of understanding and for the failure to persuade potential users to use the geoscience knowledge base. Traditionally the output of a geoscientist’s work has been complex, technical and academic maps and reports. The quality of the science is not in question but too often that science remains obscure and remote from the end-user and its significance to society and the environment is not obvious to the public, to governments and to commerce. It is worth making clear that this paper is not challenging the absolute necessity of a strong foundation of high quality geoscience research and information. But there is an need to reassess the balance and the traditional focus on “academic” output. There is a need to build products that genuinely meet society’s requirements. These products must be expressed and provided in a way that is meaningful to an audience that does not, for the most part, have geological training. Traditional geological output, such as lithostratigraphical maps, may be perfectly clear to a professional geologist, however, they convey little or nothing to the non-geologist. The various “stratigraphical” schemes and codes that we use, almost without exception, on geological maps, may allow geoscientists to share information, but they are just impenetrable secret codes to other potential users. These users seek straightforward information on the rock types, their physical properties and their hazard or resource potential. They want our knowledge articulated in a way that will help them solve their problems. Why geologists don’t listen and the public can’t read geological maps 363 If we try and understand what the users want, we have never been better equipped to be able to meet it. The availability of inexpensive, powerful and sophisticated IT tools provides all surveys with the facility to provide customised and flexible products based on their unique geoscience knowledge bases. But how well are we doing with this? Not as well as perhaps we might. Instead of helping to disseminate our message to a wider audience, GIS and other software is often only being used to recreate digitally products as equally indecipherable as those we produced by manual methods in the past. There may be another factor in our failure to reach the wider audience – the tension between short-term research/scientific advancement, and reliable long-term survey programmes. Many of the new users of geological survey digital data not only seek information that is intelligible to them; they also expect data that are consistent and available nationally. It is a fact that much of our work in the past has taken the form of local (spatially restricted) research projects, which, however innovative and scientifically stimulating, collectively produce neither consistency, nor extensive geographic cover. In the geological surveys of many countries there is a powerful case for spending a greater proportion of the funding on managing existing datasets more coherently and effectively, converting more legacy data into digital form and making these data consistent; rather than focusing excessively on new research and acquisition. This would put us in a position to be able to exploit our already extensive knowledge bases more fully. While this may be a deeply unpopular strategy amongst some geoscientists, many of our potential may view it differently. Think like a wise man, but communicate in the language of the people? The quotation is from W B Yeats, an Irish writer; perhaps this should be a guiding principle behind our new products and services? Over the last 20 years the British Geological Survey (BGS) has had to progressively increase its earnings from external sources to around 50% of its income, i.e. the BGS grant from the UK Government now covers only half its costs. While this funding model has at times produced a number of problems for the organisation, one obvious benefit has been that, because of the need to earn income, priority and much effort has been devoted to trying to better understand what BGS’ users want, and then to attempt to design and deliver appropriate products and services for them. In business-speak BGS is aspiring to market-pull and not product-push. A variety of products has been developed, some very successful, some less so, but in many of them there is no longer a presumption that the user must have a qualification in geology, the talents to visualise 3D objects from a 2D representation, locate themselves by grid reference or even be able to read a map! In 1993 BGS developed GHASP (GeoHAzard Susceptibility Package) aimed at the UK insurance industry. It was a simple assessment of geohazard potential, which by utilizing digital mapping and GIS, effectively distilled geological knowledge down to a spreadsheet containing a list of GB post(zip)codes and a potential hazard rating between 1 and 10! GHASP proved to be a considerable success. Subsequently BGS developed ALGI (Address Linked Geological Inventory) for the urban areas of Bristol and London. This was a prototype turnkey system to supply geological information for those involved in property transactions. It used GIS, in combination with an address/ coordinate database and automated report writing scripts to deliver a standard geological report on any specific address. It was a major advance, but its restricted geographic extent and the geoscientific nature of the information provided, limited its usefulness and take-up. Developing GHASP and ALGI provided experience and the basis for a range of products and services that BGS is offering today. In 2002 the GeoReports service was launched (Figure 1 and http://www.bgs.ac.uk/ georeports/home.cfm). This is a full e-commerce service that uses a number of national databases of geohazards, GIS, address-linking, and automated report-writing scripts. It allows customers to select from a variety of report types using postal address or grid reference and then receive the report (in secure PDF format) by email. Report types range from a simple listing of the data BGS 364 Ian Jackson holds for the specified area, to reports describing potential radon risk or natural ground stability hazards in non-technical language. A portion of a report on natural ground stability is included as Figure 2. Note that after assessing the likelihood of any hazard the report first advises the client on what they should do next and only then gives information on the likely cause; the concern of most members of the public is not what may cause the geological hazard, but what they should now do about it. BGS continues to try and understand users’ needs better. This is not an easy task, in part because many users are not really aware of the range of data and potential services a geological survey can offer and often have difficulty articulating their needs. But through one-to-one dialogues, partnerships and user forums (including a regular Parliamentary briefing) our appreciation of the real requirement is slowly growing. In recent months new discussions have taken place with representatives of the insurance companies, the legal profession and the financial sector on the content and design of potential new products they might wish BGS to supply. Additionally, continuing negotiations are taking place with local administrations, transport infrastructure organisations and national environmental and conservation agencies. For these major organisations, Figure 1. “The location entry” screen of the BGS GeoReports ecommerce service which have an ongoing need for geological data, the opportunity of direct and dynamic, customised access to BGS knowledge via Virtual Private Networks and web services is being explored. Working outside the comfort zone? Going beyond the delivery of conventional geological maps and reports and reaching out to a non-traditional user base means facing a new set of problems. If a GSO then charges for these products and services, even if on a non-profit making basis, only to recover costs, then these problems are compounded. The first of the problems is resources; in addition to the costs of defining and developing the products, there are the operational costs of delivery and maintenance. Creating the products may divert staff away from their (perhaps preferred?) core duties of survey and research and cause tensions in the organisation. Running a service which provides information to the public will probably require a help-line or inquiry point to deal with enquiries and complaints. The issue of liability and the risk of being sued for supplying erroneous information is not new, but it does increase considerably, as these new products are going to an audience that Why geologists don’t listen and the public can’t read geological maps 365 «mammal Natural Ground Professional Search Stability Search Results: Important notes The term ’search area’ as used throughout this report means the property extent and a 150m buffer zone. The property extent will be defined using the original details specified by the client This search is concerned with potential ground stability related to NATURAL geological hazards only. It does not search for man-made hazards, such as contaminated land or mining. Searches of coal mining should be carried out via The Coal Authority Mine Reports Service (www.coalminingreports.co.uk/) Question 1 Answer Is significant natural ground instability possible in the area? YES Question 2 Answer How significant could natural ground instability be in the area on a scale of 1 to 4 (low to high)? Level 3 Question 3 Answer What action should be taken? If natural ground instability has been indicated, then this means there is potential in your area for some properties to suffer subsidence damage. However, it does not necessarily mean that your property will be affected, and in order to find out if this is the case or not you, should obtain further advice from a qualified expert, such as a building surveyor. Show them this report and ask them to evaluate the property and its surroundings for any signs of existing subsidence damage as well as advise on the likelihood for subsidence to occur in the future. The notes at the end of this report may be useful in this regard. Note that the type of building and its surroundings (e.g. the presence of trees) are also very important when considering subsidence risk. Many types of properties, particularly newer ones, are very well constructed and unlikely to be affected by subsidence, even in areas of very significant ground movements. Question 4 Answer Which natural geological hazards could be contributing to the ground instability in the area? How much ground instability each hazard may cause is indicated by the Level 1 to 4 in brackets. This corresponds to the (’low’ to ’high’ significance) scale used in Q.2 Clays that can swell when wet and shrink when dry, causing the ground to rise and fall (’Swelling Clays Hazard’) (LEVEL 2) Weak or unstable rocks that could slip downhill on steep slopes (greater than c. 5 degrees) or into excavations (’Landslip Hazard’) (LEVEL 1) Very soft ground that might compress and progressively sink under the weight of a building (’Compressible Ground Hazard’) (LEVEL 3) Figure 2. Part of a sample report on natural ground stability from the BGS GeoReports service is not familiar with the “fuzzy” nature of geological information and may misuse them. The cost of legal advice to make sure the products are properly described and “caveat-ed” must be taken into account, as must the potential cost of legal representation, should someone actually take you to court. National and European directives and statutes may prescribe whether a GSO may provide such services and also what and how they may charge for them (if anything!). The whole issue of charging and pricing policy is complex; should data be licenced or sold outright, how much should be charged, 366 Ian Jackson should the charges differentiate between commercial use and public good use? In the UK the 1998 Competition Act, enforced by the Office of Fair Trading, introduces a further set of rules with which BGS must comply; these relate to operating fairly within the commercial market place,. Intellectual Property Rights (IPR) and copyright are equally complex issues, not only in terms of the protection of data originating in the GSO but also because data from other organisations may have been used in developing the new product (for instance a digital elevation model or mine plan data). Perhaps one of the most difficult issues is the dilemma posed by the problem of “blight”. Geological maps have always contained implied information about hazards and resources that may affect decisions about planning in general and property in particular, but that information has been understood by only a few. When one develops products and services that make that information accessible to and understandable by the general public, suddenly any potentially damaging implications for health and property are there for all to see. It is not difficult for, instance, to envisage the affects on property prices of a GSO releasing information that describes a particular area of a city having a potential risk from subsidence or landslipping. Some will argue that making such information available (information which can only ever be indicative and never site specific or definitive) is irresponsible, others will assert that this is precisely the duty of a responsible public body. Making the potential hazard information available has not increased the actual level of risk, but it has given it a higher profile. It is also true that, however comprehensive the disclaimers or explanations, there is always a possibility that some users will, innocently or otherwise, misinterpret the information. Bringing science to the public is not always easy. GEOLOGIJA 46/2, 367–372, Ljubljana 2003 Geohazard map of the central Slovenia – the mathematical approach to landslide prediction Marko KOMAC Geological Survey of Slovenia, Dimi~eva 14, SI – 1000 Ljubljana, Slovenia E-mail: marko.komac@geo-zs.si Key words: Geology, Landslide prediction, Geomorphology, Multivariate statistics, Analytical Hierarchy Process (AHP), Geohazard map, Slovenia Abstract Issues connected with unwanted natural occurrences, such as landslides, floods or earthquakes, are a source of concern around the world and Slovenia is no exception. Landslides belong to the category of “manageable” natural disasters. Today, we cannot envisage spatial modelling and prediction of various events without the information technology. GIS is also used to analyse the landslide data and satellite images can serve as a support to the ground reconnaissance. Using the methods of univariate statistics, the influences of individual spatial factors on the different landslide types and on landslides generally were tested. Using multivariate statistical methods, the interactions between factors and landslide distribution, and defined the importance of individual factors on the landslide occurrence were tested. Having combined all the spatial data available, several models were developed. Those that produced best results were then used to determine and locate the potentially hazardous areas and to draw the map of landslide risk. The landslide risk-map permitted the assessment of the hazard to the inhabitants and infrastructure (roads) on the tested area. Introduction As a result of the recent natural disasters in Europe, like floods in 2002, the need for a better understanding of natural phenomena has arisen. Independently of whether these events result from human actions or are the work of nature, their prevention or mitigation is an important factor when the preservation of the modern man’s environmental quality is at stake. Hence the need for better understanding of these phenomena, especially when their consequences can be to some measure controlled, like in the case of landslides. For this purpose, the statistical approach to analysing the influence factors and their contribution to landslide occur- rence was chosen (Carrara, 1983; Carrara et al., 1991). For the study area, the central part of Slovenia, west of Ljubljana, was selected (Figure 1), covering approximately 35×35 km. A better understanding of the described relationships should enable a more precise and a more affordable identification of the landslide-prone areas. In order to determine the capacity of an accurate spatial prediction of landslides or landslide-prone areas, several linear prediction models, using AHP (Saaty , 1977), were developed. The applicability of the AHP (Analytical Hierarchy Process) method to landslide prediction has been shown before (Barredo et al., 2000; Mwasi , 2001; N i e et al., 2001). 368 Marko Komac Figure 1. The study area Data acquisition To successful predict landslide occurrence and to produce the map of landslide-prone areas, relevant spatial data are needed. The data needed for this investigation were obtained from several sources. The landslide data were obtained from the landslide database that was constructed at Geological Survey of Slovenia. For the study area, it contains data on 614 landslides. Further, the digital elevation model (DEM) data were obtained from the national 25 m resolution DEM (InSAR DMV 25) (Survey and Mapping Administration, 2000). All the additional data on the terrain morphology (curvature, elevation, slope, aspect, basins, primary slope-units) were derived from the DEM. The Basic Geological Map of Yugoslavia at the scale of 1:100.000 served as a source for the geologic data of the area (Buser et al., 1967; Buser, 1973; Buser, 1986; Grad & Ferjan~i~, 1976). For the land use and the vegetation cover, satellite images from different sources were used and combined, using PCA (Principal Component Analysis) merging method. The multi-spectral part of the satellite data was obtained from the Landsat-5 TM images, and the high-resolution part was obtained from the Resurs-F2 MK-4 images. The topologic map in scale 1:50.000 was used as a source of the surface water net data (Survey and Mapping Administration, 1994). The population density data were obtained from National Office of Spatial Planning et al. (1997) and infrastructure data from Survey and Mapping Administration (2000). Data analysis The aim of the paper was to examine several topics, related to the landslide prediction in the central part of Slovenia, west of Ljubljana. One of the project’s main goals was to study spatial factors that influence the occurrence of landslides, individually and conjointly, and to statistically establish the univariate and multivariate relations with the landslide distribution. A better understanding of the described relationships should enable a more precise and a more affordable identification of the landslide-prone areas. In order to determine the capacity of an accurate spatial prediction of landslides or landslide-prone areas, several linear prediction models, based on various methodologies were developed. Univarate statistical analysis Using methods of univariate statistics, the influences of individual spatial factors on the different landslide types and on landslides generally were tested. For the categorical variables, Kolmogorov-Smirnov and ÷2 test were used, where actual frequency of the landslide occurrence was compared to the expected frequency. Bigger difference represents stronger influence of the observed factor. Continuous variables were also tested with Student’s t test. On the basis of these results the stability characteristics of the individual classes of the observed factors were assessed. The factors that proved to have played an important role are shown in the following table (Table 1). The steepness and the curvature of the slopes, Geohazard map of the central Slovenia – the mathematical approach to landslide prediction 369 Table 1. Factors that play an important role in landslide occurrence (univariate statistical methods) Variable All landslides LS_typel LS_type2 LS_type3 LS_type4 jL K-S K-S j Vh E.O. 0.301 0.282 0.333 0.404 0.200 0.437 Kappa = 0.5609 where Py = Vjkj is the probability of forest existence at point x, b0 is a constant and b1 to bn are coefficients for each explicative variable X] to xn. Results range between zero and one, indicating terrain-forest incompatibility and compatibility, respectively. As with the discriminant analysis method, the potential vegetation model is mapped by assigning to each cell the type of forest with the highest value on the corresponding suitability map. Although it is not actually possible to test which of the models would result in better reforestation (it would mean waiting dozens of years for the forest to grow), we were able to test the validity of the model by analysing the extent to which the results matched the distribution of existing forests. Table 1 shows the error matrices and the Kappa index measurements (Rosenfield & Fitz-patrick-Lins, 1988) for the discriminant analysis and logistic regression models. It can be observed that the potential distribu- 382 Celestino Ordónez Galán, Javier Taboada Castro, Roberto Martínez Alegría López Table 2. Error matrix and Kappa index for the logistic regression model and the weights of evidence method. 1 2 3 4 5 6 E.C. 1 23117 394 480 5510 35 3132 0.2924 2 1870 2888 356 113 333 589 0.5303 3 21 7 12004 93 209 2291 0.1792 4 5185 343 709 8324 49 5116 0.5780 5 333 1024 3311 312 4225 1565 0.6077 6 2367 18 3668 3071 38 10572 0.4643 E.O. 0.2981 0.3821 0.4154 0.522 0.1358 0.5456 Kappa = 0.4802 tion obtained using discriminant analysis provides a better match to the existing situation than the model obtained using the logistic regression method. Inclusion of lithology as a variable in the model. Analysis of correspondences The inclusion of the lithology variable in the models deserves special attention; this is because, as a non-quantitative variable, it receives a different treatment from the other variables. For the logistic regression model it is necessary to define C-1 dummy variables for the C lithology classes present. In our case, this meant adding a further 18 variables to the 6 existing variables. This can be problematic in some cases - for example, when using the Idrisi Geographic Information System (version Idrisi32), which can only manipulate a maximum of 20 variables. For this reason we used the SPSS program (ver- sion 10.1) to calculate the logistic regression model. Combining the six logistic regression models into a single map and assigning to each cell the forest with the greatest value for this cell, we obtained results which were a poor match to the existing forest distribution, as can be seen in the error matrix in Table 4. Felicísimo et al. (2002) use the weights of evidence method to multiply the probability value for each cell by a constant that depends on the different lithologies for this type of forest. The constant is greater than one when the weights are positive, less than one when they are negative and zero when the value is minus infinite (i.e. when a certain kind of lithology is not present in the forest type in question). Table 2 shows the matrix error for this method. As can be seen, the match is better than that of the logistic regression model without the lithology variable. To introduce the lithology variable into the discriminant analysis, real values were Table 3. Contingency table for actual forest type and lithology, with first-dimension lithology scores. FOREST LITHOLOGY 1 2 3 4 5 6 Total Score 1 0 0 2 0 1 1 4 -0.7812 2 0 2 14 0 1 0 17 -0.7801 3 58 0 19 16 0 30 123 0.3276 4 55 3 84 53 3 81 279 0.1712 5 17 23 0 6 1 1 48 -1.3657 6 0 0 0 0 7 0 7 -3.0170 7 2 0 9 0 0 7 18 0.0078 8 10 0 0 0 0 1 0.3870 9 39 9 14 57 0 27 146 0.2005 10 1402 35 638 479 99 837 3490 0.1777 11 556 12 428 505 0 462 1963 0.2911 12 70 0 27 83 2 18 210 0.3644 13 130 0 0 37 0 30 206 0.4353 14 70 190 256 10 118 22 666 -1.4845 15 0 0 0 7 0 1 8 0.5877 16 21 34 19 0 76 20 170 -1.9161 17 12 5 0 0 3 0 20 -1.0161 18 0 0 0 0 11 2 13 -2.4842 19 0 0 0 0 4 0 4 -3.0170 Total 2442 313 1510 1253 326 1549 7393 Developing a suitability model for potential vegetation distribution based on GIS 383 Table 4. Error matrix for model with lithology: discriminant analysis model with scored lithologies combined with correspondence analysis (left); logistic regression (right). 1 2 3 4 5 6 E.C. 1 2 3 4 5 6 E.C. 1 24119 362 5 2957 0 1098 0.155 2 1123 3252 16 19 238 133 0.319 3 9 57 13423 319 1036 3382 0.263 4 5339 409 305 10357 56 4143 0.497 5 70 546 3213 76 3069 700 0.600 6 2275 48 3570 3695 490 13809 0.422 E.O.0.268 0.304 0.346 0.406 0.372 0.406 1 20112 337 5521 9383 2 6124 2880 121 718 3 988 325 10272 232 4 4723 563 2846 6149 5 44 489 486 13 6 944 80 1286 928 E.0. 0.389 0.384 0.499 0.647 150 10891 0.566 423 596 0.735 1543 3209 0.380 1553 6916 0.729 1177 306 0.532 43 1347 0.709 0.759 0.942 0.596 Kappa = 0.5629 assigned to the different lithology classes in accordance with their capacity to discriminate the different types of forest through a correspondence analysis. Correspondence analysis is a way of factoring categorical variables and representing them in a space that reflects their association in two or more dimensions (Greenacre, 1984). It tends to be used when a large number of rows and columns in the contingency table make it very difficult to understand associations between variables, but it can also be used to assign numerical values to categorical variables. The scores for the categories of one variable reflect their capacity to discriminate the other variable. Table 3 shows our contingency table, with an extra column added to record the scores for each lithology and forest type for the first dimension, which explains 68.8% of the variance (second dimension explains 16.3%). From these scores and the contingency table we can draw conclusions in regard to the relationship between both variables. Table 3 shows that lithologies Kappa = 0.23 6 and 19, with very similar negative scores, are only present in forest 5. The potential vegetation map is obtained by assigning first-dimension scores to each lithology variable and using this as a new quantitative variable in the discriminant analysis, together with the other variables (Figure 3). Comparing this error matrix model with the model without the lithology variable, we can observe that the errors are lower for four of the six forests. Of particular note is forest 2, with commission error falling from 50.56% to 31.98%. Conclusions The use of a deductive method supported by statistical methods to create potential vegetation models reduces subjectivity and thus represents an important advance in reforestation project design methods. These models are created using multivari-ate analysis methods, such as discriminant Figure 3. Potential maps including lithology as an independent variable: discriminant analysis model combined with correspondence analysis (left); logistic regression model (right). 384 Celestino Ordónez Galán, Javier Taboada Castro, Roberto Martínez Alegría López analysis and logistic regression, which produce mathematical expressions associating independent variables with the dependent variable (in this case the forest). When compared to the earlier logistic regression method proposed by some researchers, the discriminant analysis method combined with correspondence analysis produces a potential vegetation map that provides a better match to actual forest distribution. Matching results to existing distribution patterns is the only realistic way of corroborating results; the only other possibility would be to wait dozens of years for the forests to grow. The potential vegetation model only uses species which currently exist and in sufficiently large areas to provide a statistically representative sample, it not being possible, obviously, to take into account other poorly represented species or species which have already disappeared. References F e l i c í s i m o , A. M., F r a n c é s , E., F e r n á n -d e z , J.M., G o n z á l e z - D í e z A. & V a r a s J. 2002: “Modeling the Potential Distribution of Forests with a GIS”. – Photogrammetric Engineering and Remote Sensing, 68, 457–461. F i s h e r , R.A., 1936: “The use of multiple measurements in taxonomic problems”. – Annals of Eugenics, 7, 179–188. Fernández-Cepedal G. & Felicísimo, A.M. 1987: “Método de cálculo de la radiación solar incidente en áreas con apantallamiento topo-gráfico”. – Revista de Biología de la Universidad de Oviedo, 5, 109–119. G r e e n a c r e , M. J., 1984: Theory and Applications of Correspondence Analysis. London: Academic Press. G u i s a n , A., J. P. T h e u r i l l a t & F. K i e s -n a t , 1998: “Predicting the potential distribution of plant species in an alpine environment”. – Journal of Vegetation Science, 9, 65–74. Taboada, J., Vaamonde, A., Saave dra, A., O r d ó n e z , C., 2002: “Geostatistical study of the feldspar content and quality of a granite deposit”. – Engineering Geology, 65, 285–292. Jordan, M., Mateu, J. & Boix, A., 1998: “A classification of sediment types based on statistical multivariate techniques”. – Wate, Air and Soil Pollution, 107, 91–104. M a y e r , M., W i l k i n s o n , I., H e i k k i n e n , R., O r n t o f t , T. & M a g i d , E. 1998: “Improved laboratory test selection and enhanced perception of test results as tools for cost-effective medicine”. – Clinical Chemistry and Laboratory Medicine, 36: 683 – 690. M o r r i s o n , D. F., 1976: Multivariate statistical methods, 2nd Ed.: McGraw-Hill, Inc., New York, 415 p. R o s e n f i e l d , G.H. & K. F i t z p a t r i c k -L i n s , 1986: “A Coefficient of Agreement as a Measure of Thematic Classification Accuracy,” – Photogrammetric Engineering and Remote Sensing, 52, 223–227, Skidmore, A. K., 1989. “A comparison of techniques for calculating gradient and aspect from a gridded digital elevation model”. – Int. J. Geographical Information Systems, 3, 323–334. V a n d e R i j t , C.W.C.J., L. H a z e l h o f f & C.W.P.M. Blom, 1996: “Vegetation zonation in a former tidal area: A vegetation-type response model based on DCA and logistic regression using GIS”. – Journal of Vegetation Science, 7, 505–518. GEOLOGIJA 46/2, 385–390, Ljubljana 2003 Application of remote sensing and GIS in Mt. Mangart landslide observation (Slovenia) Kri{tof O[TIR, Tatjana VELJANOVSKI, Toma‘ PODOBNIKAR & Zoran STAN^I^ Scientific Research Centre Slovenian Academy of Sciences and Arts Novi trg 2, SI-1001 Ljubljana, Slovenia E-mail: kristof@zrc-sazu.si Key words: landslides, radar interferometry, digital elevation models, satellite images, Mt. Mangart, Slovenia Abstract On 17 November 2000 a major landslide occurred on the slopes of Mount Mangart in the Upper Poso~je region, Slovenia, as a direct consequence of extreme rainfall and assortment of several inconvenient circumstances. A research group was established immediately after the event to find possible causes of the landslide and monitor its consequences. As a part of these attempts also remote sensing and integration of remotely sensed data to GIS was used. In the paper usefulness of satellite images as one of the most convenient data source in natural hazard observation is demonstrated. Satellite images were acquired within the “Space and Major Disaster” Charter, started just a few weeks before the event by the European Space Agency, the Centre National d’Etudes Spatiales and the Canadian Space Agency. Advanced image processing was performed carefully to analyze various aspects of the event. Before and after radar images were used to detect soil moisture and to observe the changes in water runoff. Optical images together with DEM were used for GIS analysis of areas affected by the slide. Land use maps, generated from processed imagery, proved to be highly useful for damage estimation. Introduction The use of remote sensing is becoming increasingly frequent in environmental studies. In the 1970s and 1980s satellite images were mostly used in simple interpretations or as a map background (Merifield & Lama r 1975, R i b & Liang 1978). However, more recently there are almost no serious environmental studies that do not include advanced image processing and analysis. Remote sensing has been successfully applied to forest fires detection, flood monitoring, deforestation studies, co-seismic displace- ment monitoring, pollution tracking in the atmosphere and the sea, weather devastation observation, pollution prevention, desertification and erosion observation and many more (ESA 2001, Cracknell 2000, Sabins 1997, Dixon 1995). One of the most important applications of satellite technology can be found in the case of natural disasters, where satellite images can be used to provide advance warning for specific hazardous events (G e n s & G e n -deren 1996, Gu o et al. 2001), to monitor the concerned, or for a quick evaluation of the damage and therefore support the deci- 386 Kri{tof O{tir, Tatjana Veljanovski, Toma‘ Podobnikar & Zoran Stan~i~ sion-making process in the rescue operations. Satellite and airborne imagery alone can offer an efficient contribution to natural resource management. Still, the most promising seems to be the application of remote sensing in combination with geographical information systems. In the paper the use of remote sensing and geographical information systems in the Mount Mangart landslide observation is presented. A description of the “Space and Major Disasters” Charter is given, and details on image interpretation and analysis are described. Special attention is given to data integration and GIS modelling performed within the Mount Mangart landslide case study. At the end some general remarks and guidelines are presented. The Mount Mangart landslide Following several weeks of heavy rainfall, a major landslide occurred on the slopes of Mount Mangart in North-western Slovenia in the night between 16 and 17 November 2000. The landslide hit the village of Log pod Mangartom, claimed seven dead and caused immense damage. After weeks of continuous rain on 15 November 2000, a mass of morainic material and slope gravel moved down to the Predelica gorge, blocked the water flow of Mangart stream and stopped there for several hours. One day later, in the early morning of 17 November 2000, a major landslide Figure 1. Location of the Mount Mangart landslide. occurred on the slopes of Mount Mangart (Figure 1). The landslide rested for several hours and became saturated from the waters of the Mangart stream supplemented by the heavy rain. This, together with the local dynamics, caused the ground material to become “liquefied”. Within a few hours the slide was transformed into a debris flow – a fast moving mixture of water, soil and other material. It is estimated that about 1,000,000 m3 of various material flowed downwards along the bed of the Mangart stream, hitting the village of Log pod Mangartom, and finally flowing into the So~a river. Both landslides were most probably influenced by the specific geological composition of the ground, the considerable seismo-logical activity of the nearby area and the intense rainfall. The mountain ridge west of Mount Mangart is composed of massive Upper Triassic carbonate that is in areas interrupted by clastic rocks, and some poorly permeable Carnian calc stoneware. In the Pleistocene, over the stepped bedrock, poorly permeable grounding glacial sediments rich with silt were deposited over the dolomite gravel. The bedrock of the landslide, represented by a block of poorly permeable carbonate-clastic succession, is situated between the fault-bounded blocks of massive and bedded dolomite. A direct triggering mechanism of the landslide and consequentially of the development of the debris flow was the intense rainfall. The landslide scar in the upper part Application of remote sensing and GIS in Mount Mangart landslide observation (Slovenia) 387 of the slope exposed a cliff in the bedrock topography, probably produced by faulting. Considering the geological situation in the area, it seems that the fundamental trigger for the landslide was the poorly permeable bedrock combined with the extreme weather situation. Low permeability of the bedrock caused the concentration of water in diamicts and thus a rapid increase in material-rich water tension. The end of this process caused a rapid “liquefaction” of the first landslide material into an immense flow with almost no solidity at all. Satellite image interpretation Shortly after the disaster a group of professionals was established in order to monitor the slide and propose solutions for its stabilisation. As the area was dangerous and further slides could occur at any time, the group relied on remote sensing techniques, both airborne and spaceborne. The actions to obtain and process satellite imagery started a few days after the landslide when the European Space Agency was contacted, and afterwards a request was made to the “Space and Major Disasters” Charter. The Charter was initiated following the UNISPACE III conference held in Vienna, Austria, in July 1999, by the European Space Agency (ESA) and Centre National d’Etudes Spatiales (CNES). The Canadian Space Agency (CSA), Indian Space Research Organisation (ISRO), and US National Oceanic and Atmospheric Administration (NOAA) joined the initiative later on. The Charter aims at providing a unified system of data acquisition and delivery to those affected by natural or man-made disasters. It was declared formally operational on 1 November 2000, less than three weeks before the events on Mount Mangart, and the landslide discussed in this paper was actually the first time it was activated. After the problems which were to be analysed were defined, a plan of action was proposed by ESA and the Scientific Research Centre of the Slovenian Academy of Sciences and Arts. It was immediately submitted to the various space agencies for tasking satellites. In total 13 satellite images from 1992 to 2000 were utilised: • five ERS (both ERS-1 and 2), • two RADARSAT, • four SPOT (two panchromatic and two multispectral), and • two Landsat images. In the analysis, an additional layer – a digital elevation model of Slovenia, produced using radar interferometry from ERS images and advanced modelling – was also used. The first post event image, an ERS-2 scene, was acquired a week after the landslide. This was followed by two further acquisitions, the SPOT and RADARSAT images made during the second week. The images were supplemented by archive data taken under approximately the same conditions. All the necessary data and were distributed by mail as soon as possible. Nevertheless it took almost a month to gather all the necessary images. What suggests that in such cases electronic distribution would be highly desired and needed. After the images were received a visual inspection was made. The landslide was detected directly or indirectly in the images made after the event: ERS-2 (24 November 2000), RADARSAT (1 December 2000) and SPOT (29 November 2000). Visual inspection was followed by geo-coding and image interpretation. All scenes were georeferenced to the national system – that is the Gauss-Kreuger projection on the Bessel ellipsoid. Georeferenced satellite images were integrated into a GIS system, together with other already available referenced data (Landsat images, digital elevation model, etc.). Within the project ERS images were used in two ways – to produce a digital elevation model and to observe the land properties at the time of the landslide. A digital elevation model for the area under investigation was made in the beginning of 2000, mainly to test the usability of ERS data in rough terrain and to support the observation of co-seismic activity after the 12 April 1998 earthquake (O{tir & Stan~i~ 1999, O{tir 2000). In the area mentioned seven ERS-1 and 2 scenes were used from both the ascending and descending orbit. Partial elevation models and other height data sources, such as contour lines and a coarse digital elevation model with a resolution of 100 m, were used to produce a final digital elevation model InSAR DEM 25 (O{tir 2000, Podobnikar et al. 388 Kri{tof O{tir, Tatjana Veljanovski, Toma‘ Podobnikar & Zoran Stan~i~ 2000). The model has a resolution cell of 25 m; its overall accuracy is approximately 8 m, from better than 2 m in plains to more than 10 m in the mountains. Contemporary ERS images were used to observe land properties, mostly humidity in the time of landslide. RADARSAT images, obtained in the frame of the Charter, offer very high spatial resolution (fine beam mode). They provided clearer results than ERS, despite the fact that the relief in the area of Mount Mangart is very steep and therefore causes severe problems to all radar satellites (layover and shadows) and considerably limits their use. The humidity observed on the RADARSAT image map is not as extensive as in the case of the ERS data. The reason for this lies in the fact that the second RADARSAT image was taken several days after the ERS image and that there was no significant rainfall in the meantime. The interpretation of SPOT imagery gave a more detailed insight into the consequences of the disaster. Two panchromatic (21 August 2000 and 29 November 2000) and two multispectral (19 August 2000 and 29 November 2000) SPOT scenes were used to detect the landslide and to evaluate its impact on the natural environment. Figure 2 shows the scene acquired after the landslide. One can clearly see how the landslide changed the valley of Log pod Mangartom. The interpre- Figure 2. SPOT satellite map of landslide area (image was acquired on 29 November 2000). tation of SPOT images allowed us to obtain the most accurate information on the slide location and compare the situation before and after the event. However, as a consequence of the very low sun position in November (shadows were emphasised) the interpretation of SPOT data was not straightforward. In addition to the shadows the November image (Figure 2) contained snow in higher areas and the August image included some clouds. Remote sensing data integration and analysis Image interpretation can offer useful information; however, it is often used merely as a data source for the GIS analysis. Therefore all available satellite images have been integrated within a geographical database, together with the digital elevation model and land use map. Initially, the exact location of the landslide and its direct area of influence were determined. Due to the high spatial and spectral resolution of the SPOT satellite images (panchromatic and multispectral) acquired on 29 November 2000, these images were used to isolate both areas. The estimated total area of the landslide, i.e. the area of the slipped land, is 25.7 hectares. The additional area of destruction in the valley is therefore estimated to be 50.1 hectares, summing to the total direct impact area of 75.8 hectares. As described before, a digital elevation model InSAR DEM 25 was produced for the area using ERS satellite images with inter-ferometric processing (Figure 3). From the elevations also a slope map was produced. Average elevation, slope and terrain orientation were computed for the landslide and its impact area; the results are listed in Table 1. The landslide occurred at an average elevation of almost 1400 m, at a very steep slope (24%) facing south-east (161°). The standard deviations for both slope and orientation are small, showing that the landslide area is very homogenous. On the other hand the impact area lies much lower, on average at approximately 800 m. It is also modesty inclined (19%) and oriented to the south-west (224°). The impact area is rather heterogeneous, with standard deviations from two to more than three times larger than that for the landslide. Application of remote sensing and GIS in Mount Mangart landslide observation (Slovenia) 389 Table 2. Land use categories in respect to the landslide and its impact area. Figure 3. Digital elevation model of landslide area (InSAR DEM 25) produced from ERS images with radar interferometry and advanced modelling. Table 1. Elevation, slope and orientation of the landslide and its impact area. Landslide Impact area Elevation (m) Average 1386 824 STD 109 243 Slope (%) Average 24 19 STD 6 12 Orientation (°) Average 161 224 STD 25 83 Aside the digital elevation model, land use is amongst the most important natural environment variables. The land use map for the area of the landslide was produced from a combination of Landsat and SPOT images. Classical supervised image classification method has been used in order to obtain land use (Sabins 1997). The land categories were divided into ten classes: urban, built-up, individual houses, coniferous forest, deciduous forest, mixed forest, bushes, water, agricultural, and open. Additionally advanced post-classification techniques – such as elevation modelling and forest mixing – were also used. The estimated thematic accuracy of the produced land use map is approximately 90%. A detailed analysis of the changes in the environment was carried out. Table 2 and Landslide area Impact area Class ha % ha % Urban 0.0 0% 0.0 0% Build-up 0.0 0% 1.7 3% Individual houses 0.0 0% 1.9 4% Coniferous forest 1.0 4% 10.1 20% Deciduous forest 18.6 72% 5.8 12% Mixed forest 2.2 9% 5.4 11% Bushes 0.3 1% 3.6 7% Water 0.0 0% 4.0 8% Agricultural 0.0 0% 9.4 19% Open 3.6 14% 8.2 16% Total 25.7 100% 50.1 100% Figure 4: Land use classes destroyed by the landslide. Figure 4 show areas that were destroyed by the slide in respect to land use. The landslide directly destroyed forests and a small amount of open areas, while other classes were not present. The impact area was more heterogeneous – forests covered almost half of it, but there was also a notable quantity of built-up land, individual houses and agricultural land. Conclusions The disaster below Mount Mangart is a classical case used to show the value of satellite remote sensing. The landslide happened in late November 2000 after several weeks of heavy rainfall and had such extent that it can be clearly detected with the available satellite sensors. SPOT optical images offered a good illustration of the situation and could be compared with the archived data in order to evaluate the damage. Multi-spectral optical data was supplemented with radar images, acquired on four dates before and after the event. Due to the rough terrain, it was hard to directly detect the landslide and its consequences on radar imagery; 390 Kri{tof O{tir, Tatjana Veljanovski, Toma‘ Podobnikar & Zoran Stan~i~ however, the high humidity in the area could be observed even several days after the event. To evaluate the landslide consequences a detailed GIS analysis of the available satellite images and other data was made. The landslide has been identified on several post event images, most notably on the SPOT panchromatic image, which was used to outline both the landslide and its impact area. The total damage area was estimated to be almost 76 hectares – 26 hectares representing the surface of the landslide and 50 hectares the impact area. The landslide occurred on steep south-east facing slopes, at an average elevation of approximately 1400 m. With respect to slope, elevation and orientation the area affected in the valley was lower and more heterogeneous. The evaluation of land use showed that the landslide occurred mainly in areas covered by deciduous forest (almost three quarters of its surface). The impact zone was again more heterogeneous, half of it being covered with forests. There was also significant damage in agricultural land and built-up areas. The Mount Mangart landslide study has proven the value of remote sensing technology for monitoring natural disasters and it has in particular proved the usefulness of the “Space and Major Disasters” Charter. It has shown that remote sensing can be used to estimate the damage and under suitable conditions also in rescue operations. In rescue operations the processing speed is critical and near real time data distribution is needed. In the case of damage estimation the processing speed is less important than the accuracy and quality of results. It has been proven, that remote sensing enables mapping and analysing topographic and land cover changes caused by a catastrophic event within a considerably short period of time. We also believe that with advanced simulations it can be used to determine hazardous areas and predict the triggering conditions. Satellite remote sensing may therefore be one of the most important steps in the development of an early hazard warning system. Acknowledgements This project would not be possible without the “Space and Major Disasters” Charter. The authors would like to thank the European Space Agency, the Centre National d’Etudes Spatiales and the Canadian Space Agency for starting the Charter and providing most of the satellite images used in this project. Jerome Bequignon of the European Space Agency helped us to define the plan of action and Professor Bojan Majes of the Faculty of Civil and Geodetic Engineering invited us to contribute to his expert group. Adrijan Ko{ir of the Scientific Research Centre of the Slovenian Academy of Sciences and Arts provided valuable interpretation on geological characteristics of the area. References C r a c k n e l l , A.P., N e w c o m b e , S.K., Black, A.F. & Kirby , N.E. 2001: The ABDMAP (Algal Bloom Detection, Monitoring and Prediction) Concerted Action. Int. Jour. Remote Sensing, 22, 205–247. Dixon, T.H. 1995: SAR Interferometry and Surface Change Detection. RSMAS Technical Report TR 95-003, University of Miami, Rosenstiel School of Marine and Athmospheric Science, USA. ESA, 2002, Earth Observation, Earthnet Online. http://earth.esa.int/ Gens, R. & Genderen, J.L. Van 1996: SAR interferometry – issues, techniques, applications. International Journal of Remote Sensing, 17, 1803–1835. G e n s , R. 1998: Quality assessment of SAR interferometric data (Enschede: International Institute for Aerospace Survey and Earth Sciences). G u o , Z., Hu , G., & Q i a n , S. 2001: Spatial Detection Technology Applied on Earthquake’s Independent Forecast. 22nd Asian Conference on Remote Sensing, 5 – 10 November 2001 (Singapore: Asian Association on Remote Sensing), 295– 299. http://www.crisp.nus.edu.sg/~acrs2001/pdf/ 192Guo.pdf Merified, P.M. & Lamar, D.L. 1975: Active and inactive faults in southern California viewed from Skylab. TM X-58168, vol. 1, NASA, USA. O{tir, K. 2000: Analysis of the influence of radar interferogram combination on digital elevation and movement models accuracy (Ljubljana: University of Ljubljana). (Ph.D. thesis in Slovenian) O { t i r , K. & S t a n ~ i ~ , Z. 1999: Interferomet-ric generation of DEM for mobile telephone network planning. “Fringe 99” Workshop on ERS SAR Interferometry, 10 – 12 November 1999 (Liege: European Space Agency). P o d o b n i k a r , T., S t a n ~ i ~ , Z. & O { t i r , K. 2000: Data integration for the DTM production. International Co-operation and Technology Transfer, proceedings of the Workshop, 2 - 5 February 2000 (Ljubljana: Institute of Geodesy, Cartography and Photogrammetry), pp. 134–139. R i b , H.T. & L i a n g , T. 1978: Recognition and identification. Landslides - analyses and control, edited by R.L. Schuster & R.J. Krizek (Washington DC: National Academy of Sciences) pp. 34–69. S a b i n s , F.F. 1997: Remote sensing: principles and interpretation, (New York: Freeman). GEOLOGIJA 46/2, 391–396, Ljubljana 2003 Extracting NDVI temporal profiles of vegetation types in the Ri‘ana spring catchment area from NOAA-AVHRR data using linear mixture model Mitja JAN@A Geolo{ki zavod Slovenije, Dimi~eva 14, SI – 1000 Ljubljana, Slovenija E-mail: mitja.janza@geo-zs.si Key words: remote sensing, linear mixture model, NOAA-AVHRR, NDVI, NDVI temporal profiles, Ri`ana spring, Slovenia Abstract This paper presents methodology that was used to derive NDVI temporal profiles of the vegetation types in the Ri‘ana spring catchment area. In the methodology the Landsat TM image was used for the classification of the area into vegetation cover classes and to determine proportions of classes within AVHRR pixels. According to the influence of the vegetation types on the water cycling process in the catchment area, six vegetation classes were defined (deciduous forest, grass, agricultural areas, shrub, coniferous forest and areas with no/sparse vegetation). This data and the NDVI data derived from AVHRR satellite images was then used in the linear mixture modelling that was applied to estimate the mean NDVI value of each vegetation class. The resulting temporal NDVI profiles of vegetation cover classes, with exception of class with no/sparse vegetation, are in general in agreement with the observed vegetation characteristics area. Introduction Recharge of the aquifer is a dynamic process that depends on many factors. For accurate modelling of this process data is required that is distributed in time and space. One of the very important data are vegetation characteristics. Remote sensing data is a potentially very useful source of information on the state and development of a long range of vegetation and hydrological parameters (Sandholt et al., 1999). For vegetation monitoring from satellite platforms are common used vegetation indices. One of the most widely used is normalized difference vegetation index NDVI = (NIR-IR)/(NIR+R) (Rouse et al., 1973). For describing spatial variability of vegetation in the catchment areas Landsat TM (30 m nominal spatial resolution) provides in general adequate data. But to monitor temporal dynamic of vegetation development Landsat TM, with its temporal resolution of 16 days, is often not sufficient. Especially if we have in mind that atmospheric conditions (cloudiness) at the time of the satellite overpass can make interpretation of satellite images impossible. On the other hand NOAA-AVHRR satellite system provides fine temporal resolution (daily frequency). But its coarse spatial resolution (nominal 1,1 km at nadir) is a huge limitation in many applications. An ideal satellite system for observation of vegetation changes in catchment area would have spatial resolution of Landsat TM and temporal resolution of NOAA-AVHRR satellite system. 392 Mitja Jan`a An approach towards this kind of system is a linear mixture model. This paper presents methodology that was used to derive NDVI temporal profiles of vegetation types in the Ri‘ana spring cath-ment area from Landsat TM and AVHRR images. Landsat TM image was used for classification of the area into vegetation classes and to determine proportions of the classes within AVHRR pixels. Linear mixture model based on multiple linear regression was then applied on the set of 21 AVHRR images to estimate the mean NDVI value of each vegetation class. Study area: The Ri‘ana spring catchment area Ri‘ana spring is the most important water resource for water supply of the costal area of Slovenia. Its cathment area covers nearly 240 km2 (Figure 1). The terrain is mainly hilly, with altitude ranging from 70 m to 1028 m. Very complex karst aquifer system was developed in limestones that cover most of the catchment area. Minor part of the area consists of flish sediments. According to the used classification most of the area is covered by deciduous forest, following by grass, agricultural areas, shrub, coniferous forest and areas with no or sparse vegetation (Table Table 1. Vegetation cover classes and theirs cover portions in the stuy area vegetation class cover portion (%) 1). R1 no/sparse vegetation 12 R2 grass 19 R3 agricultural areas 18 R4 decidious forest 32 R5 coniferous forest 10 R6 shrub 19 Satellite data In the study one Landsat-5 TM image ((c) ESA, Eurimage, ZRC SAZU, 1992) taken on 18.8. 1992 and set of 21 images AVHRR (Advanced Very High Resolution Radiometers) -LAC images (NOAA Satellite Active Archive, http://www.saa.noaa.gov) were used. This set of images was selected from all available images for the studied period of two years (1992, 1993). Cloud contaminated images were rejected and in order to minimize scan angles effects only images where the study area is within 30o scan angle were selected. On the AVHRR images only Sun angle correction and correction of panoramic distortion were applied. Neither atmospheric nor topographic corrections were performed on the used AVHRR and Landsat TM satellite images. Radiometric calibration was performed on all images and for the first two bands of NOAA 11-AVHRR images Figure 1. Position of the study area, NOAA-11 AVHRR image, colour composition – RGB: Bands 1,2,4 (left) and Landsat TM image, colour composition – RGB: Bands 1,2,3 (right). Both images ware taken on 18.8.1992 Etracting NDVI temporal profiles of vegetation types in the Ri‘ana spring catchment ... 393 Figure 2. Classified study area non-linear correction that accounts for sensor degradation was applied (R a o & Chen , 1994). All AVHRR images were coregistered to a Transverse Mercator projection (as the Landsat-TM image). Second order polynomial transformation and nearest neighbor resampling method were used. A georefe-rencing of AVHRR image was based mainly on the referenced points selected along the coastline. Vegetation cover classification Vegetation cover classification was carried out by supervised classification based on maximum likelihood classifier. For the reference data CORINE Land Cover (Ho -~evar et al., 2001) was used that covers part of the study area. According to the influence of the vegetation types on the water cycling process in the catchment area, six vegetation classes were defined. For each class more training areas (spectral classes) were selected (altogether 62) in order to incorporate variability within the class. After the classification smoothing with majority filter (3x3) was applied. The distribution of vegetation classes is shown in Figure 2. Spatial degradation of the fine resolution image Used methodology of preparation of data for linear mixture model is based on the procedures presented by Oleson et al. (1995). From the classified map a set of single class maps was determined. For each single-class map, a digital number (DN) 1 was assigned to each pixel that contains corresponding cover type. All other pixels were assigned a DN of 0. Each single class map was then degraded to a spatial resolution approximating that of the AVHRR. With regards on the AVHRR point spread function (PTF), which defines the characteristics of the image of a point source formed by an optical system, convolution with a Gaussian filter was applied. This is similar approach to that used by Moreno & Melia, (1994), who modeled AVHRR PTF at nadir with two-dimensional Gaussian distribution. The standard deviation (?) of the Gaussian distribution was defined with the expression (O l e s o n et al., 1995): AVHRR_ pixel TM pixel 2.8ct 394 Mitja Jan`a A series of new (filtered) class maps was created this way. Value of each pixel on these maps corresponds to the vegetation cover class proportion within an AVHRR spatial resolution pixel. These portions have been used in linear mixture model (as independent variables). Linear mixture model Linear mixture modelling considers that the pixel’s radiance results from the linear combination of the radiances of the elements composing the pixel multiplied by their respective proportion within the pixel. These base elements of the landscape are named endmembers (Kerdiles & Grondona, 1995). This assumption should be strictly true only for the original bands. However, studies of Kardiles and Gordona (1995) showed that linear combination of NDVI values implies only very minor inaccuracies. The linear mixture model in this study was formulated as: NDVI, = /n ¦NDVI11+... + flk-NDVIlk + e, NDVI, = fA ¦ NDVInl +... + frik.NDVI„k+en The system of n equations corresponds to the number of pixels and k (independent or regressor variables) to the number of vegetation cover classes. Using more general symbology, this set of equation can be expressed in matrix notation as 7 = Xß + L and Where Y is a (n x 1) vector of observation, X is a (n x k) matrix of the levels of the independent variables, ß is a (k x 1) vector of regression coefficients and ? is a (n x 1) vector of random errors. Multiple linear regression model was used to solve the set of equations. The least squares estimate of ß is (Montgomery & Runger, 1994): ß=(X'X)-1X'y and the fitted model in the matrix notation is 9 = 4 Results The mean NDVI temporal profiles of vegetation cover classes derived from AVHRR images are presented in Figure 3. Temporal profiles for two years show phases of vegetation cycle that are in agreement with known vegetation characteristics in the study area. In general low values of NDVI at the beginning and at the end of the year in the winter are shown. In the spring a green-up is observed. Very fast increase of NDVI values, that starts to increase in the middle of April and reaches the climax at the end of May and beginning of June. In the next phase in the summer period NDVI profile is relatively flat with gentle decrease until the end of August or beginning of September. After that phase in the autumn period NDVI profile drops fast and reaches low value, characteristic for the winter period, around the middle of November. Extracted NDVI profiles for individual vegetation cover type show evidently separated profile of R4 (deciduous forest) that has far highest NDVI values in the spring and summer. Les evident characteristics of the others NDVI vegetation cover type profiles are: • R5 (coniferous forest) profile has the lowest amplitude and relatively the highest value in the winter period, • R2 (grass) profile has in average the lowest value. J2 Jn. X ¦ xn ¦ x2k , ß = "A" • Xnk _ A J and ? = Y Etracting NDVI temporal profiles of vegetation types in the Ri‘ana spring catchment ... 395 • R6 (shrub) profile has relatively high amplitude similar to the R4 profile. Exception is the profile of R1 (no/sparse vegetation) vegetation type that doesn’t follow the characteristics of the other profiles. In general it has the lowest value, but very high amplitude, which makes it very difficult for interpretation. The reason for this is most probable low cover portion (2%) of that vegetation cover type in the study area. Conclusion mentioned criteria. Other vegetation classes’ temporal profiles, with the exception of R1 profile, are of mixed quality and less separable but still in general in agreement with observation. Temporal profile of R1 that covers a very small part (2 %) of the study area is difficult to interpret and doesn’t agree with expected results. This fact shows inability of linear mixture model to extract subpixel information for the vegetation classes that are not well represented in terms of cover portion within the study area. The study confirmed the potential of using linear mixture model for extraction of NDVI temporal profiles of vegetation cover types in Ri‘ana catchment area and pointed out some characteristics of the used method. Results are in general in agreement with the observed vegetation development cycle in study area. Study shows the ability of extracting NDVI temporal profiles for vegetation cover types that are spectrally separated and well represented in the mean of cover portion, which confirmed conclusions of O l e a s o n et al. (1995). In this study it is the vegetation cover type R4 that satisfied References (c) ESA, Eurimage, ZRC SAZU, 1992: Landsat – 5 TM satellite image taken on taken on 18.8. 1992. Kerdiles, H. & Grondona, M.O. 1995: NOAA-AVHRR NDVI decomposition and subpixel classification using linear mixing in Argentina Pampa. - Int. J. Remote Sensing, 16(7), 1303-1325. H o ~ e v a r , M., K o b l e r , A., V r { ~ a j , B., P o l j a k , M. & K u { a r , B. 2001: Corine karta rabe tal in pokrovnosti Slovenije = Corine land cover phare project Slovenia: Podprojekt: Fotointerpretacija in rezultati: zaklju~no poro~ilo. - Gozdarski in{titut Slovenije, 83 pp., Ljubljana 5 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 I I fr OOt- CM CM t- (O II I I I I CO lf>- 1992 1993 Figure 3. The mean NDVI temporal profiles of vegetation cover classes 396 Mitja Jan`a Montgomery , D. C. & Runger , G. C. 1994: Applied Statistics and Probability for Engineers.-John Wiley & Sons, 531-623, New York. Moreno, J.F. & Melia, J. 1994: An Optimum Interpolation Method Applied to the Resampling of NOAA AVHRR Data. - IEEE Transactions on Geoscience and Remote Sensing, 32(1), 131-151. NOAA Satellite Active Archive 2002, http:// www.saa.noaa.gov. O l e s o n , K.W., S a r l i n , S., G a r r i s o n , J., S m i t h , S., P r i v e t t e , J.L. & Emery, W.J. 1995: Unmixing Multiple Land-Cover Type Reflectances from Coarse Spatial Resolution Satellite Data. -Remote Sensing of Environment, 54, 98-112. Rao , C.R. N. & Chen, J. 1994: Post-Launch Calibration of the Visible and Near-Infrared chan- nels of the Advanced Very High Resolution Radiometer on NOAA-7, -9, and -11 Spacecraft, NOAA Technical Report NESDIS 78. - U.S. Department of Commerce, 22 pp., Washington. R o u s e , J.W., H a a s , R.H., S h e l l , J.A. & D e e r i n g , D.W. 1973: Monitoring vegetation systems in the great plains with ERTS. In: Third ERTS Symposium, NASA SP-351. - NASA, 1, 309-317, Washington. S a n d h o l t , I., A n d e r s e n , J, D y b k j a e r , G., Lo, M., R a s m u s s e n , K., R e f s g a a r d , J. C. & Hogh-Jensen, K. 1999: Use of remote sensing data in distributed hydrological models: applications in the Senegal River basin. - Geografisk Tidsskrift (Danish Journal of Geography), 99, 47-57. GEOLOGIJA 46/2, 397–404, Ljubljana 2003 New general engineering geological map of Slovenia Mihael RIBI^I^1, Jasna [INIGOJ2 & Marko KOMAC2 1Gradbeni in{titut ZRMK d.o.o., Dimi~eva 12, 1000 Ljubljana, Slovenia 2Geolo{ki zavod Slovenije Dimi~eva 14, Slovenia Key words: Engineering Geological map, Slovenia, GIS Klju~ne besede: In‘enirskogeolo{ka karta, Slovenija, GIS Abstract The new lithostratigraphic map of the entire Slovenia (in the scale of 1:250000) created by using the GIS method enabled the production of its derivative – engineering geological map (EG map). The goal of creating this map was to define the general engineering geological characteristics of rocks and soils that will be used for the general review of engineering geological conditions in Slovenia. The map also enables the planing of general interventions in Slovenia. The EG map was created by using the GIS method for merging the lithology units of Slovenia according to EG characteristics on three levels. The first one is the basic separation into soils, soft rocks and rocks. The second level is a more detailed separation on the basis of their origin and the third one on the basis of the composition, rock strength and particle size ranges. The first basic GIS layer determined the EG units merged with the database, giving the spatial and description data for each unit. The basic data for each unit was stored in the GIS-database (serial number, the connection to the lithology unit, the name, short description, comprehensive description, the occurrence in Slovenia). The EG units were also stored in the database (the description of EG units, geotechnical characteristics, the foundation conditions, seismic characteristics). The map was further detailed by the creation of informational layers derived from the map. In this manner the map of rock strength, the map of possible land sliding, the map of weathering cover thickness estimation and the erosion map were produced. The GIS modelling method was used for the creation of these maps. For example, the map of possible land sliding was created regarding these informational layers: lithology structure, the thickness of weathering cover, the slope inclination and the hydrogeological conditions. Kratka vsebina Na novo izdelana geolo{ka karta v GIS-u merila 1 : 250.000 teritorija Slovenije je omogo~ila tudi izdelavo izpeljanke – in‘enirskogeolo{ke karte. Cilj izdelave in‘enirsko-geolo{ke karte je opredeliti splo{ne in‘enirskogeolo{ke lastnosti hribin in zemljin, ki bodo slu‘ili za generalni uvid v in‘enirskogeolo{ke razmere Slovenije. Poleg tega in‘enirsko-geolo{ka karta omogo~a planiranje posegov v prostor v dr‘avnem merilu. In‘enirskogeolo{ka karta je bila izdelana tako, da so bile litolo{ke enote Slovenije s pomo~jo GIS tehnologije med seboj zdru‘ene po in‘enirskogeolo{kih lastnostih v treh nivojih. Prvi nivo je osnovna delitev v zemljine, polhribine in hribine, drugi ‘e detajlnej{i po na~inu nastanka ter tretji po sestavi, trdnosti in zrnavosti. Tako so bile na osnovnem informacijskem sloju opredeljene in‘enirskogeolo{ke enote, kateremu je bila pridru‘ena baza podatkov, ki je za posamezno enoto podajala prostorske in opisne podatke. Za vsako enoto so tako bili v GIS-bazi shranjeni osnovni podatki (zaporedna {tevilka, povezava na litolo{ko enoto, ime, kratek opis obse‘nej{i opis, raz{irjanje v Sloveniji) in in‘enirskogeolo{ke lastnosti (opis in‘enirskogeolo{kih lastnosti, geotehni~ne lastnosti, pogoji temeljenja, seizmi~ne lastnosti). Pri nadaljevanju dela je bila in‘enirskogeolo{ka karta {e detajlirana z izdelavo iz nje izpeljanih informacijskih slojev, ki so izrazili eno izmed pomembnih in‘enirskogeolo{kih zna~ilnosti. Tako so nastali {e karta trdnosti kamnin, karta podvr‘enosti plazenju, karta ocene debeline preperinskega pokrova in karta seizmi~-nih lastnosti tal. Za izdelavo teh kart je bilo uporabljeno GIS modeliranje. Tako je npr. karta podvr‘enosti plazenju nastala z upo{tevanjem naslednjih informacijskih slojev: lito-lo{ke zgradbe, debelina preperine, nagib terena in hidrogeolo{ke razmere. 398 Mihael Ribi~i~, Jasna [inigoj & Marko Komac Introduction The production of the new lithostrati-graphic map (Buser, 1999) in the scale of 1:250,000, dividing in great detail the Slovenian territory according to the lithological characteristic of its structure, also enabled the creation of an engineering geological map of the same scale as its upgrade. To this purpose, the lithological units were merged with regard to their relative engineering geological properties. In the preparation of the engineering geological map, two criteria were primarily used. The first one was the classification of the material composing the Slovenian territory into soils, soft rocks and rocks. The geomecha-nical characteristics of rock and its sensitivity to weathering greatly depends on its maturity and lithification. The second decisive criterion was the content of small clay fraction in rock structure. Rocks composed of clay as well as silt fraction are more susceptible to landsliding and other destructive processes. In joining the rocks according to their similar engineering geological properties, it was necessary to take into account that the Slovenian territory is geologically very complex. A single lithologically homogenous rock is very rare. Most frequently, there is an alternation of different lithological variants, or the prevailing rock comes with inclusions, layers or veins of other rocks. This is the reason why it is not always possible to stick to the classifications set up in the extensive literature. The purpose of engineering geology as a practical science is to offer an engineering geological map as an answer to a certain problem appearing in spatial development or in the preservation of the environment connected with such activities. The general engineering geological map, like this one, thus only presents the generalised engineering geological characteristics of an area. However, general engineering geological maps can also be produced for specific purposes. In such a case, rocks are categorised according to their engineering geological properties that are important for obtaining the answer sought. This part of the task, is the second step in the production of the engineering geological map of Slovenia. The processing of engineering geological data in the GIS environment In the lithostratigraphic map, the 112 lithological units are represented by 4651 separated polygons. On the basis of the key which is described in more detail in the following chapter, each polygon was reclassi-fied into new classes, indicating the engineering geological properties of rocks. The first part of the table (for soils), which was used for the reclassification from the lithostratigraphic map to the engineering geological map, is shown below: Tab.1. Reclassification of the lithostratigraphic map to the engineering geological map ACAD_ ID ELEV no. EG mark Decimal Class. DESCRIPTION clay (Quaternary) brown clay, terra rossa and loam (Quaternary and Pliocene) clay and weathered material with chert (Quaternary and Pliocene) clay, peat (marsh sediments - Quaternary) clay, silt and weathered peat (marsh and lake sediments – Quaternary) clayey silt (continental and marsh loess – Quaternary) alluvium (pebble, sand, silt and clay – Quaternary) fluvial loose sediments in terraces (pebble, sand, silt and clay – Quaternary)0 diluvium (mainly clay with pieces of various rocks – Quaternary) talus (Quaternary) alluvial fan (gravel, pebble and silt – Quaternary) moraines – tuff (Quaternary – Pleistocene) clay, clayey silt with pebbles of flint and silicate rocks (Pliocene and Pleistocene) clay, silt and sand (Pliocene) sandy marl, clay and small pebbles (Lower Pliocene) sand and clay (Upper Miocene and Lower Pliocene) clayey marl, sand, pebble and clay (Upper Miocene) flint pebble, sand and silt (Upper Pliocene) pebble, and sandy clay (Middle Pliocene) mine tailings (anthropogenic recent sediments) 2 1 ZEM-R 111 13 1 ZEM-R 111 14 1 ZEM-R 111 7 2 ZEM-R 112 8 2 ZEM-R 112 9 2 ZEM-R 112 1 3 ZEM-R 113 10 3 ZEM-R 113 5 4 ZEM-P 121 3 5 ZEM-P 122 4 6 ZEM-P 123 12 6 ZEM-P 123 15 7 ZEM-K 131 20 7 ZEM-K 131 19 8 ZEM-K 132 21 8 ZEM-K 132 22 8 ZEM-K 132 16 9 ZEM-K 133 18 9 ZEM-K 133 6 10 ZEM-A 141 New general engineering geological map of Slovenia 399 Tab. 2. Frequency of appearance and the area that it covers in kilometres EG – mark Description Frequency of appearance Area (km2) ZEM-R soil (alluvium) 533 ZEM-P soil (on slope) 305 ZEM-K soil (rocks with soil properties) 203 ZEM-A soil (anthopogenic) 4 POL soft rocks 303 KLA clastic rocks 782 KAR carbonate rocks 2093 MET metamorphic rocks 158 MAG magmatic rocks 232 3696 601 1113 28 1559 2991 8920 759 694 Each lithostratigraphic element, numbered by ACAD_ELEV, corresponds to a ID number according to the engineering geological map. In addition, the engineering geological unit obtained in this way is classified into the basic engineering geological class with regard to its engineering geological properties, i.e. obtains the appropriate decimal mark. Thus, in the table above, the engineering geological mark (EG mark) ZEM-R, means an engineering geological unit classified among soils (ZEM), alluvium deposits (mark R). The decimal classification 111, which has three levels, indicates that the engineering geological unit belongs among soils (first number), alluvium deposits (second number) and that it predominantly consists of clay (third number). The lithostratigraphic elements are divided into 9 classes with regard to their basic engineering geological characteristics. The following table gives the incidence for each class and the surface that it covers in kilometres. The brief description of the logical structure serving as the basis for the preparation of an engineering geological map The basic engineering geological map determining the general engineering geological characteristic of the Slovenian territory is based on the key below. The key distinguishes between soils, soft rocks and rock (level 1). The soils are further divided into alluvium soils (fluvial and stream alluvia), slope soils (diluvia, proluvia, slope alluvial fans and talus), rocks with soil properties and anthropogenic soils (man-made fills of large surfaces). Soft rocks have already been partially lithified, but their humidity, firmness and other geomechanical properties are still too low for them to be classified among rocks. Thus, they represent a class of their own. Rocks are divided into clastic, carbonate, metamorphic and magmatic rocks (level 2). At the third level (level 3), the material is divided into three groups: geotechnically least appropriate, medium-appropriate heterogeneous material and geotechnically most resistant material. When there is an alternation of geotechnically different materials, the criterion for classification is the prevailing material. Each lithological unit connected with ID AcadElev according to the original table is then classified by its engineering geological properties into the engineering geological class defined by the indication of ID no. (the serial number of the engineering geological group), engineering geological mark (generally classifying the material according to its engineering geological properties) and Dec.Cl. (decimal division of materials into classes), like it is shown above. Description of engineering geological units The engineering geological map comes with general and detailed descriptions of the engineering geological characteristics. The general description of an engineering geological unit contains the following information: A. NAME OF UNIT B. LITHOLOGICAL AND EG DESCRIPTION OF THE ROCK C. INCIDENCE IN SLOVENIA D. CHARACTERISTIC TERRAIN MORPHOLOGY E. DESCRIPTION OF THE STRUCTURAL DISCONTINUITIES OF THE ROCK 400 Mihael Ribi~i~, Jasna [inigoj & Marko Komac Tab. 3. The logical structure and the basis for the preparation of an engineering geological map BASIC CLASSIFICATION Level 1 Level 2 Level 3 EG mark Dec.Cl. ALLUVIUM predominantly clayey soils ZEM-R 111 SOILS marsh, lake soils (clay, silt, peat) ZEM-R 112 (and terrace alternation of different soils (pebble, sand, clay, etc.) ZEM-R 113 sed.) pebble and sandy pebble ZEM-R 114 clayey – diluvial, proluvial ZEM-P 121 (ZEM) SLOPE SOILS gravely (with a clayey component) ZEM-P 122 gravely (predominantly thick fraction), moraines ZEM-P 123 ROCKS WITH clayey ZEM-K 131 SOIL alternation of fine and coarse grain soils ZEM-K 132 PROP. pebbly ZEM-K 133 ANTHROPO- mine trailings – gangues ZEM-A 141 GENIC mounds, soil barriers ZEM-A 142 SOILS deposits of urban and other wastes ZEM-A 143 clayey, marly POL 201 clayey, marly and limestone POL 202 alternation of different materials (marl, sand, sandstone, conglomerate pebble, clay etc.) POL 203 conglomerate with possible soil inclusions POL 205 (slaty) claystones with inclusions of other rocks KLA 301 marl and sandstone (flysch) with inclusions of ROCKS other rocks KLA 302 sandstones and conglomerates with inclusions of other rocks KLA 303 stratified and cliff limestones KAR 401 flat limestones KAR 402 limestones and dolomites KAR 403 ROCKS CARBONATES dolomites KAR 404 limestones with marls KAR 405 limestones with inclusions of other rocks KAR 406 limestone conglomerates and breccia KAR 407 METAMORPHIC phyllites, schists and slate MET 501 ROCK amphibolite and gneiss MET 502 MAGMATIC ROCK diabase and other magmatic rocks with tuff MAG 601 amphibolites, serpentinites, diaphthorites MAG 602 tonalite, dacite, granodiorite MAG 603 F. WEATHERING G. WEATHERING COVER H. EROSION I. TERRAIN STABILITY AND LANDSLIDE INCIDENCE J. SUSCEPTIBILITY TO ROCKFALLS K. HYDROGEOLOGICAL PROPERTIES L. SEISMIC SENSITIVITY M.CONSTRUCTION CONDITIONS A detailed description of each engineering geological unit was also made. Part of the description for soils is given below as an example: Soils – alluvium soils (ZEM-R) 111 predominantly clayey soils 112 marsh, lake soils (clay, silt, peat) 113 alternation of different soils (pebble, sand, clay, etc.) 114 pebble and sandy pebble According to the EG classification, fluvial and stream alluvia are divided into four sub-units (111, 112, 113 and 114). The first includes sediments (of Quaternary or Pliocene age), mostly composed of clayey soils (111). It also includes terra rossa. They can be found in the basins of karst sinkholes, primarily in Dolenjska, at the margins of large basins, like the Drava and Mura basins, and in smaller patches also elsewhere in Slovenia. They form a flat or slightly undulating terrain. They are susceptible to erosion along waterways. They are impermeable to water and act as an insulator. Interference with them may be problematic due to their low bearing capacity and possible large differential subsidence. Deep slope and embankments require protective measures in order to ensure the stability of the excavation walls. If they New general engineering geological map of Slovenia 401 are thick, they are appropriate for waste deposits. In case of an earthquake, a considerable increase in the seismic impact is expected. Upgrading of the engineering geological map The next step in the preparation of the general assessment of the engineering geological properties of rock in the Slovenian territory was the creation of maps showing certain important engineering geological characteristics: Thus, the following maps were derived from the basic engineering geological map: – the map of rock classification according to rock strength properties, – the map of rock classification according to stability or susceptibility to landsliding, – the map with the assessment of the weathering cover thickness. In the preparation of the above maps by means of GIS, other information layers were also used. Thus, the map of stability also took into account the following as input information layers: – lithology – the map with the assessment of the weathering cover thickness – the hydrogeological map of Slovenia – DEM (Digital Elevation Model) Tab. 4. Weighting factors of information layers Influence factor Percent of influence lithology 20% weathering cover 40% slope inclination 30% hydrogeology 10% We determined the influence factors for each information layer. For the stability map, they were the following: The basic input data for the production of the derived maps were obtained by making an assessment of a certain engineering geological property for each lithostratigraphic unit, like shown in the following table and the keys attached: Derived maps from the basic engineering geological map are shown below: Conclusion The general engineering geological map in the scale of 1:250,000 was first used in searching for the location for the low radioactive waste deposit in Slovenia. Otherwise, it is not especially significant in construction and other local spatial development, however, it becomes important in spatial planning in a wider area and in understanding the engineering geological characteristics of the Slovenian territory. Tab. 5. Assessment of an engineering geological properties Acad Elev. ID no. EG mark Dec. class. Weathering DESCRIPTION cover – soil Rock Stability/ Erosion strength lithology 2 1 ZEM-R 111 clay (Quaternary) 1 12 2 13 1 ZEM-R 111 brown clay, terra rossa and loam (Quaternary and Pliocene) 1 122 14 1 ZEM-R 111 clay and weathered material with chert (Quaternary and Pliocene) 1 121 7 2 ZEM-R 112 clay, peat (marsh sediments – Quaternary) 1 121 8 2 ZEM-R 112 clay, silt and weathered peat (marsh and lake sediments – Quaternary) 1 121 9 2 ZEM-R 112 clayey silt (continental and marsh loess – Quaternary) 1 122 1 3 ZEM-R 113 alluvium (pebble, sand, silt and clay – Quaternary) 1 113 10 3 ZEM-R 113 fluvial loose sediments in terraces (pebble, sand, silt and clay – Quaternary) 2 113 402 Mihael Ribi~i~, Jasna [inigoj & Marko Komac Weathering cover – soil (LEGEND) 1 Soil, clayey, silty with weathered material properties 2 Soil, pebbly (gravely) with weathered material properties 3 Very thick and thick weathering cover 4 Weathering cover of medium thickness 5 Thin weathering cover Erosion (LEGEND) 1 highly erodable rocks 2 moderately erodable rocks 3 poorly erodable rocks Rock strength properties (LEGEND) 1 cohesionless soils 2 cohesive soils 3 soft rocks 4 soft and medium-hard rocks 5 hard 6 very hard rocks Stability (LEGEND) 1 very high possibility of the landslide appearance 2 high possibility of the landslide appearance 3 medium possibility of thelandslide appearance 4 moderate possibility of the landslide appearance 5 very low possibility of the landslide appearance Fig.1. Weathering cover map New general engineering geological map of Slovenia 403 Fig.2. Erosion map Fig.3. Rock strength properties map 404 Mihael Ribi~i~, Jasna [inigoj & Marko Komac Fig.4. Stability map References B u s e r , S. 1999: Lithostratigraphic Map of Slovenia in the Scale 1:250.000, Geological Survey Slovenia, Ljubljana. Urbanc, J., Komac, M., Poljak , M. & R i b i ~ i ~ , M. 1999: Processing of digital geological space data for Agency RAO – Hydrological, Tectonic and Engineering Geology Map, Geological Survey Slovenia, Ljubljana. GEOLOGIJA 46/2, 405–412, Ljubljana 2003 Geotechnical and seismic microzonation map of the Bovec region Mihael RIBI^I^ Gradbeni In{titut ZRMK d.o.o. Dimi~eva 12, 1000 Ljubljana, Slovenija Key words: earthquake, geotechnical map, map of seismic microzonation, GIS, post-earthquake restoration, Poso~je, Bovec, Slovenia Klju~ne besede: potres, geotehni~na karta, karta seizmi~ne mikrorajonizacije, GIS, popotresna obnova, Poso~je, Bovec, Slovenija Abstract In 1998, the area of Upper Poso~je in the north-west of Slovenia experienced the strongest earthquake in the 20th century in the Slovenian territory. There were no casualties, however, 4200 houses and other building were damaged. The Slovenian Government adopted an extensive plan of post-earthquake restoration, which was almost fully completed by 2003. In place of 160 buildings that suffered too much damage to be repaired new ones were constructed. A geotechnical map of wider Bovec area was produced to be used for planning, location selection and determination of foundation conditions. The geotechnical map was prepared on the basis of the existing geological map, which was additionally reviewed and supplemented on the field. This was added by the geotechnical field research data, including an overview of the existing documents on the foundation construction in the area concerned, engineering geological mapping and drilling of 20 boreholes in areas where the data on ground composition was insufficient. The geotechnical map was supplemented with GIS databases of the damage to buildings and the nature. For buildings for which foundation conditions were determined during restoration, a special database was additionally created. The data collected was also used for the preparation of the seismic microzonation map, which served as the basis for the static designing of seismically safe construction. Kratka vsebina Leta 1998 je bil na obmo~ju Gornjega Poso~ja v severnozahodni Sloveniji najmo~nej{i potres v dvajsetem stoletju na obmo~ju teritorija Slovenije. Smrtnih ‘rtev ni bilo, bilo pa je po{kodovano 4200 hi{ in drugih objektov. Vlada Slovenije je sprejela obse‘en plan popotresne sanacije, ki je bil do leta 2003 skoraj v popolnosti zaklju~en. Namesto 160 objektov, ki so bili preve~ po{kodovani, da bi jih bilo mogo~e popraviti, so se zgradili novi. Za planiranje, izbor lokacij in dolo~itev pogojev temeljenja je bila izdelana geotehni~na karta {ir{ega obmo~ja mesta Bovec. Geotehni~na karta je bila izdelana na osnovi obstoje~e Geolo{ke karte, ki pa je bila dodatno na terenu preverjena in dopolnjena. K temu so bili pridru‘eni podatki geotehni~nih raziskav na terenu, ki so zajemali pregled obstoje~e dokumentacije o izvajanju temeljenja na obravnavanem obmo~ju, in‘enirskogeolo{ko kartiranje in vrtanje 20 vrtin na obmo~jih, kjer je primanjkovalo podatkov o sestavi tal. Postopek izdelave Geotehni~ne karte je bil naslednji. Najprej je bila digitalizirana geolo{ka karta. Geolo{ke enote na karti so bile nadalje zdru‘ene ali deljene v in‘enirskogeolo{ke enote, glede na geomehanske lastnosti tal. Drugi pomemben vhodni podatek za izdelavo Geotehni~ne karte so bili podatki o po{kodbah objektov zaradi potresa. V alpskem svetu, kjer ni objektov so bile uporabljene ugtovljene po{kodbe, ki so nastale v naravi zaradi potresa. Izdelana je bila karta po{kodb, ki je v GIS aplikaciji zdru‘evala lokacije po{kodovanih objektov z bazo popisa po{kodb. Narejena je bila analiza velikosti po{kodb v odvisnosti od sestave tal. Na osnovi korelacije med stopnjo velikosti po{kodb in sestave tal so bili za in‘enirskogeolo{ke enote dodatno opredeljene geomehanske lastnosti tal. Pri tem so bili posebno pomembni podatki o obmo~jih, kjer teren gradijo slabo nosilna tla, ki so se ob potresu prikazala kot obmo~ja z najte‘jimi po{kodbami na hi{ah. Kon~no so bile in‘enirskogeolo{ke enote s sorodnimi lastnostmi zdru‘ene v nov sloj po podobnih geomehanskih lastnostih. Za vsako tako dobljeno zdru‘eno in‘enirskogeolo{ko enoto posebej so bile dolo~eni pogoji temeljenja. Rezultati GIS obdelave so bili pregledno prikazani v izrisih in izpisih: Karta velikosti stopnje po{kodb na objektih in v naravi, Geolo{ka karta, In‘enirskogeolo{ka karta Tabela pogojev temeljenja za in‘enirskogeolo{ke enote in Karta seizmi~ne mikrorajonizacije. 406 Mihael Ribi~i~ Introduction For the presentation and processing of all data the GIS technology was used. The inter-disciplinary data collected was primarily used for three purposes: – the analysis of earthquake impact, – the preparation of the basis for the restoration works on the damaged buildings, and – the monitoring of restoration. tions and geomechanical ground characteristics. The seismic and geotechnical analyses together were used for the preparation of a new seismic microzonation map. The analyses of earthquake impact were conducted for seismic and geotechnical purposes. Seismologists used the data gathered to determine the seismic parameters of the earthquake (depth of earthquake, type of earthquake, definition of the tectonic structure in relation with the earthquake, seismic intensity, etc.). The analysis of the geological-geotechni-cal data enabled the correlation of the impact of the damage to the nature and the buildings with the local geological condi- Maps and Databases The data were organised in three groups: – NATURAL CHARACTERISTICS OR NATURAL CONDITIONS, – EARTHQUAKE IMPACT DATA, AND – RESTORATION DATA (GEOTECHNI-CAL PART). The spatial data were shown on digital maps, while the descriptive data were given in databases. The connections between the graphical representations and databases were made by means of ID numbers or identifiers. In order to determine the natural characteristics, we amended the existing geological map of the Bovec Basin in the scale of 1:10,000, and we also used a more general Fig 1. Section of Geological map of Upper Poso~je in scale 1: 25,000. Geotechnical and seismic microzonation map of the Bovec region 407 Fig. 2. Geotechnical map of Bovec basin Fig.3. Seismic microzonation map of Bovec basin 408 Mihael Ribi~i~ map of the wider area in the scale of 1:25,000, added by the data on the probe boreholes and shallow excavations. The procedure leading to the elaboration of the geotechnical map and seismic micro-zonation map as the final products was the following. First, the geological map was di-gitalised. Further, the geological units on the map were joined or divided to geological-engineering units according to the geo-mechanical and seismic characteristics of the ground. Another input data important for the preparation of both maps were the data on the damage to buildings due to the earthquake. In the Alpine region, where there are no buildings, the determined damage that the earthquake caused to the nature was used. In the preparation of both maps, much aid was provided by the map of damage, joining the locations of the damaged buildings and the database of damage inventory. An analysis of the extent of damage in dependence on the ground composition was made. On the basis of the correlation between the level of damage and the ground composition, the geomechanical and seismic properties of the ground were additionally determined for the geological-engineering units. Here, the especially important data were those referring to the low bearing capacity ground areas which the earthquake revealed as the areas with the worst damage to houses. At the end, the geological-engineering units with related properties were joined into a new layer according to their similar geomechanical or seismic characteristics. For each joined geological-engineering unit obtained in this way, the foundation conditions and the increase in the seismic level due to ground composition were determined. Each map was added by a database describing the data captured. DATABASES TO MAPS database to the geological map database to the engineering geological map database to the geotechnical map database to the sesmic microzonation map The attributes of the database to the geological map and the attributes of the key to the geotechnical map are given as examples: The structure of the descriptive data base to the geological map: Geological units (polygon) – geological unit identifier – description of the mapped geological unit – age of rock – type of rock Layer stike and dip (point) – layer strike and dip identifier – type of layer strike and dip – dip (angle) – strike (angle) Geological boundaries (lines) – geological boundary identifier – type of geological boundary Structure units (lines) – structure unit identifier – type of structure unit ? axis of a normal large fold ? axis of a toppled fold ? axis of a metre fold ? axis of a covered fold ? axis of a plunging fold ? fault ? thrust o cracks Large landslide (polygon) Rockfall – older (point) Rockfall appearing during the earthquake (point) Isolines of equal thickness of quaternary sediments (line) The maps were added by special databases formed within the data capture on the field: FIELDWORK DATABASES database of damage to buildings database of damage to the nature database of boreholes database of probe shafts database of geotechnical foundation conditions The first database contained the inventory of the damage to buildings and the sec- Geotechnical and seismic microzonation map of the Bovec region 409 The database of the geotechnical key contained the following attributes. The right column shows descriptions for the geotechnical unit chosen as examples of the data contained in the database: Attribute Example ROCK CLASSIFICATION ROCK FORMATION ROCK DESCRIPTION MORPHOLOGY PHYSICO-GEOLOGICAL PHENOMENA WEATHERED MATERIAL thickness type USCS EXCAVATION CATEGORY weathered material rock ASSESSMENT OF FOUNDATION AND CONSTRUCTION CONDITIONS description bearing capacity allowed adequacy assessment groundwater slope inclination applicability for building in cohesionless soil; slope sediments (moraine and scree) glacier sediments till (loose moraine) appears as scree of poorly-rounded boulders of limestone gentle to medium dip of slopes subject to strong slope erosion; landslides on steeper slopes 0.5 to 1.5 m clayed gravel to clay with pebbles GC – CL II III ground of medium bearing capacity; requiring careful location selection and foundation 200 to 250 kN/m2 less adequate; where possible, on larger area permeable to water; temporary groundwater above impermeable layers 1 : 2 conditionally applicable The database of the damage to buildings: ATTRIBUTE Example ID NO. 1745 TYPE residential buildings ADDRESS BOVEC, TRENTA, TRENTA 63 OWNER RUDOLF ANA YEAR OF CONSTRUCTION 1941 YEAR OF RESTORATION 1987 NO. OF FLOORS ground floor + 1 FOUNDATION no foundation WALLS stone CEILING wooden ROOF wooden ROOFING other DESCRIPTION OF DAMAGE The structure of the building has suffered much damage, partly due to the earthquake and partly because of its poor state and inappropriate construction method. CATEGORY 4 DAMAGE LEVEL 65.0 Y 403860 X 139619 ond one the inventory of the damage to the nature. In order to connect the database concerning the inventory of the damage to buildings with the abovementioned maps, we used the national house records, which contain the basic information on buildings, in particular spatial co-ordinates. The database of the damage to the nature included the damage to the nature found after the earthquake, with the information being gathered by mapping in the field: ATTRIBUTE Example 3 ID No. of the phenomena Name Mali Leme‘ – [ija Time of triggering 12.4.1998 Place Municipality Y co-ordinate X co-ordinate Surveyed by Description and extension of the large rockfall phenomenon LEPENA BOVEC 398130 128020 Begu{, Ko~evar 410 Mihael Ribi~i~ The research on the field involved a large number of boreholes and probe shafts made next to the buildings. The following are two examples of records contained in the database of boreholes and shafts. The database of boreholes: ning and in the restoration or construction of new substitute buildings. The geotechni-cal map was used for envisaging the foundation conditions for building. ATTRIBUTE Example ID No. of borehole DEPTH BOREHOLE PLACE DATE PROCESSED BY 9 20 m G-3 @i~nica August 1998 M. Bavec subreport – inventory of the borehole: depth AC Description of soil 0.2 humus 1.8 CI-CH brown firm clay of intermediate to high plasticity 2.1 GC brown clayey gravel 9.1 borehole compacted lime breccia 10.5 boulder of limestone 10.6 boulder of sandstone 14 grey marl (flysch) The database of shafts: Building ID No. 17 Place BOVEC Street TRG GOLOBARSKIH @RTEV House No. 16 Owner Lili~ L., Sivec F. Depth 0.0 – 0.60 man-made mound (GP, CL) 0.6 – 1.00 slightly silty gravel (GP) Bearing capacity 200 Foundation The building is founded shallowly, on strip footing 0.6 m under the height of the terrain, in incoherent soil Structure The foundation structure is made of 0.6-m strip footing, constructed of a composition of large rock singlets, poorly bound with concrete or fine sand Assessment The building has shallow foundations. The allowed bearing capacity of the ground corresponds to the freezing criterion. The foundation structure is of poor quality. During the restoration, the geomechanical foundation conditions were determined for all new buildings. The above database of shafts was used for recording the data on the foundation conditions at individual locations. The results and applicability of GIS in post-earthquake restoration The results of the collected information on the geological structure and of the seismic and geotechnical conditions, produced by means of GIS technology (ArcInfo software) were useful during the whole period of restoration, both in the construction plan- During the restoration, the seismic micro-zonation map served as a means of determining the basic seismic level and the impact of the ground composition on its increase or decrease, which is the basis for an expert in statics to be able to design seismically safe buildings. We also created a GIS application which produced the foundation conditions and the seismic properties of the ground from the geotechnical and seismic map of the area of a selected building, i.e. of the location defined by the Y and X co-ordinates. Besides, it was possible to make many useful analyses by means of GIS. Let us only present one of them. The chart below shows the number of damaged buildings in depen- Geotechnical and seismic microzonation map of the Bovec region 411 30,0 UD a 25,0 s 9 i« 0s 20,0 o C k. - 15,0 v L 10,0 s 5,0 0,0 %, V %L ^ X ^ > % % % dence on the ground composition. It can be seen that the percentage of damaged buildings is the highest on the ground with the poorest geotechnical properties. One can conclude that the use of GIS in post-earthquake research and restoration works proved to be successful, since at the beginning it required a clear and long-term concept of approach, and during the work it enabled a quick supply of information, much of which would have otherwise needed long processing, while with GIS it was immediately accessible. REFERENCES V i d r i h , R. & R i b i ~ i ~ , M. 1994: “Influence of Earthquakes on Rock-falls and Landslides in Slovenia”, First Slovenian meeting on Landslides, Idrija, Slovenia, pp. 33–46. R i b i ~ i ~ , M. & V i d r i h , R. 1998: “Earthquake-triggered Landslides and Rockfalls”, Ujma 12, Ljubljana, Slovenia, pp 95–106. R i b i ~ i ~ , M. & V i d r i h , R. 1999a. “Earthquake-triggered Landslades and Rockfalls during Earthquake on April 12, 1998 in Posocje”, Third Slovenian Conference on Landslides, Rogla, Slovenia, pp. 40. Godec, M., V i d r i h , R. & Ribi~i~ , M. 1999b: “The engineering-geological structure of Posocje and damage to buildings. International Conference on Earthquake Hazard and Risk in the Mediterranean Region, Nicosia, North Cyprus, 18–22 October, Near East University, pp.228. V i d r i h , R., R i b i ~ i ~ , M. & L a p a j n e , J. 1999c: “Earthquake on 12.april, 1998 in Posocje (Slovenia) – Phenomena Occuring in Nature During the Earthquake in the Alpine Region”, International Conference on Mountain Natural Hazards, Grenoble, France, 19 pp. V i d r i h , R. & R i b i ~ i ~ , M. 1999d: “Slope Failure Effects in Rocks during the April 12, 1998 Posocje Earthquake and Implications for the European Macroseismic Scale (EMS-98)”, Geologija, Vol. 41, 365–410, Ljubljana. R i b i ~ i ~ , M., V i d r i h , R. & [ i n i g o j , J. 2000: The influence of the geological condition on the increase of the earthquake effects (earthquake on April 12, 1998; Poso~je, Slovenia). V: International conference on geotechnical and geological engineering, 19–24 November 2000, Melbourne, Australia. GeoEng2000 : an International conference on geotechnical & geological engineering, 19–24 November 2000, Melbourne, Australia. Vol. 2, Extended abstracts. Lancaster; Basel: Technomic Publishing Company, cop. 2000, str. 562. V i d r i h , R., R i b i ~ i ~ , M. & S u h a d o l c, P. 2001: Seismogeological effects on rocks during the 12 April 1998 upper So~a Territory earthquake (NW Slovenia). Tectonophysics (Amst.). [Print ed.], vol. 330, no. 3/4, str. 153–175. JCR IF (2000): 1.393; SE, x: 1.321 (16/45), Geochemistry & Geophysics 412 Mihael Ribi~i~ GEOLOGIJA 46/2, 413–418, Ljubljana 2003 Calculation of the moving landslide masses volume from air images Mihael RIBI^I^ Gradbeni in{titut ZRMK d.o.o., Dimi~eva 12, 1000 Ljubljana, Slovenija Key words: landslide, remote sensing, GIS, aerial photography, landslide volume calculation, Slano blato, Slovenia Klju~ne besede: plaz, daljinsko zaznavanje, GIS, letalsko snemanje, izra~un volumna plazu, Slano blato, Slovenija Abstract The landslide Slano blato is of great dimensions, longer than 1 km and wider than 300 m. The movements of 10 m per day mostly happen in heavy rainy seasons and afterwards calm down, while the landslide progresses for a few 100 meters at the time. Because of the size and the inaccessibility of the landslide, common surveying was not possible. So the observation of the sliding masses movements was only possible by successive photography from a plane. For this purpose we carried out two special photo shoots with a special plane equipped for remote sensing. The existent snap shots from regular cyclic remote sensing prior to landsliding were also applied. On the basis of the snaps, TIN meshwork was created for each photo shoot separately. Geodetic maps in 1:2000 scale with contour lines 1 m apart were also produced for this purpose. It is important to know the volume of the moving masses so we can determine which measures are significant for stopping the landsliding (mudflow) that threatens the village of Lokavec. As we had available data of the area size before landsliding, we could easily calculate the mass volume sliding by cross-sectioning the area of the two conditions at different times of aerial photo shoots. The problem in the calculation was the landsliding masses that joined the mudflow from the ground. These were the masses from the previous older slidings, known to had happened at least twice – 100 and 200 years ago. The volume of the old landslide was estimated with the help of geological evaluation which we used to access the depth of the slope base. Geological evaluation was partially based on field drilling and partially on presumption of the depth of weathering cover. The landslide depth data were interpreted with the help of two longitudinal sections and transverse sections each 25 m apart. We put the surface lines from each remote sensing on each cross-section and calculated the volumes of the landslide between two cross-sections. By means of this procedure we assessed that the volume of all the sliding masses was 684.000 m3 (April, 2001). With regard to this and other parallel results we determined that we should stop the sliding before it gets to the village by draining and pushing the masses aside. Part of the masses was impossible to withhold on the slope (between 300.000 m3 and 400.000 m3), so it was removed by means of vehicles to the deposit. It was confirmed that the calculation of the volumes with the help of remote sensing is a very suitable method for large landslides, but will only give the right results by detailed geological interpretation of landsliding. Kratka vsebina Plaz Slano blato je izrednih dimenzij, dalj{i od 1 km in {irine do 300 m. Premiki na njem velikosti do nekaj deset metrov na dan se dogajajo po de‘evnih razdobjih in se zopet umirijo, ko plaz napreduje ve~ sto metrov. Ker s klasi~nimi geodetskimi meritvami zaradi 414 Mihael Ribi~i~ velikosti plazu in nedostopnosti ni bilo mogo~e spremljati premikov plaze~ih se mas, so bila dogajanja na plazu spremljana s pomo~jo zaporednih slikanj iz letala. V ta namen sta bili izvedeni dve posebni snemanji s posebnim letalom za daljinsko opazovanje, uporabljeni pa so bili tudi obstoje~i posnetki rednega cikli~nega snemanja {e predno se je plaz spro‘il. Na osnovi posnetkov so bili izdelani TIN in GRID povr{ine za vsako snemanje posebej in geodetske karte povr{ine v merilu 1:2.000 z izohipsami na 1 m. Za odlo~itev, kateri so nujni ukrepi za prepre~evanje premikanja plazu, ki ogro‘a zaselek Lokavec, ki je neposredno pod plazom, je pomemben podatek, kolik{en je volumen premikajo~ih se mas. Ker je bila na razpolago prvotna povr{ina terena pred plazenjem, se je izra~un volumnov mas, ki so se “razlile” kot blatni tok preko prvotne povr{ine, dolo~il enostavno s presekom povr{in dveh stanj povr{ine v razli~nih ~asih letalskega slikanja. Problem so predstavljale tiste plaze~e se mase, ki so se vklju~ile v blatni tok iz podlage. To so bile mase od starih plazenj, saj je poznano, da je bil plaz aktiven najmanj ‘e dvakrat - pred dvesto in sto leti. Volumen stare plazine smo ocenili s pomo~jo geolo{ke ocene, s katero smo dolo~ili globino do hribinske podlage.Geolo{ka ocena je temeljila deloma na terenskih vrtanjih, deloma pa na predpostavkah o debelini preperinskega sloja. Podatki o debelini plazu so bili interpretirani s pomo~jo dveh vzdol‘nih profilov in pre~nih prerezov na razdaljah po 25 m. V vsak prerez so bile prene{ene linije povr{in posameznih snemanj, dobljene s presekom med ploskvami povr{in in vertikalno ravnino prereza ter interpretirana linija podlage.. Volumni plaze~e se mase med dvema profiloma so bili dolo~eni s produktom polovice vsote plo{~in preseka plazu med zaporednima presekoma in razdalje med presekoma. Po tem postopku je bil dolo~en celotni volumen gibajo~ih se mas, ki je zna{al 684.000 m3 (maja 2001). Ob upo{tevanju teh in drugi vzporednih rezultatov se je pokazalo, da je za prepre~itev prodora plaze~ih se mas do vasi Lokavec treba ~im ve~ mas zadr‘ati z osu{evanjem in odrivanjem na boke na plazu, tiste mase po oceni v koli~ini med 300.000 do 400.000 m3 , ki ni mogo~e zadr‘ati na pobo~ju, pa odvoziti na deponijo. Pokazalo se je, da je izra~un volumnov s pomo~jo letalskega slikanja za velike plazove zelo primerna metoda, ki pa da rezultate {ele ob podrobni geolo{ki interpretaciji dogajanj na plazu. Introduction It is very probably in connection with the global climatic changes which, apart from dry periods, also brought extremely rainy seasons in recent years that four very large landslides have occurred in Slovenia after 2000, such as had not been observed for decades before that. Due to several reasons, we decided to obtain geodetic bases by aerial photography. These reasons were: – inaccessibility of landslide bodies, – large dimensions of landslides, – requirements for quick acquisition of geodetic bases (implementation of urgent restoration measures), – large displacements of soil masses in short time periods, – comparison of the state before and after the triggering of landsliding, – calculation of the volumes of the moving landslide masses. We made air images of all four landslides on the same flight. One of these landslides was Slano blato, treated in this paper. For landslides of extraordinary dimensions, like Slano blato, the calculation of volumes (volumes of moving masses, volumes to the base and potential volumes of new sliding) is important because it provides a basis on which answers of better quality can be given to the following questions: • Is a final landslide restoration possible and sensible at all? • Is it possible to restore it by regrouping the sliding masses? • Is it sensible to remove a part of the landslide material? • What masses may endanger the settled area under the landslide? • What volumes of masses may move at the same time? • What is the most probable sliding prognosis? Only the calculation of the volume of potential and moving masses of the landslide material provides the basic answers specifying to which extent the works on a landslide would contribute to its stabilisation. The calculation also shows approximately what funds will be needed for landslide restoration. Some basic data on the landslide: • Location: Above Lokavec at Ajdov{~ina in Primorska • Date of • triggering: 18 November 2000 • Surface of the • landslide: ? 20 ha, length: ? 1270 m • Largest width: ? 250 m, between 360 and 660 above sea level • Largest progress: ? 90 m/day Calculation of the moving landslide masses volume from air images 415 • Rock in the base: flysch (marl and sandstone) • Composition of • the sliding mass: weathered flysch – clayey gravely soil • Age of the • landslide: first reference dating 200 years ago • Landslide • restoration: first performed in 1903, last- ing for 17 years. In order to calculate the volumes of moving masses, aerial photographs were used as follows: 1. The original surface was taken from aerial photographs taken in 1998, before sliding, 2. The second shoot was carried out at the end of November 2000, 3. The third shoot was carried out in mid April 2001. Preparation of geodetic bases and calculation of volumes between different states of surfaces The preparation of the geodetic bases for the calculation of the volumes was carried out by the Geodetic Institute of Slovenia, which has an aeroplane for aerial photography and all the software needed for data processing. The hardware used was: • aerial photography equipment on the aeroplane with a RC30 aerial metric camera, • analytical photogrammetric instrument Adam Promap, • GPS receivers, • PCs, • electronic theodolite LEICA TCR 307. The data gathered were processed by means of the following software: • Adam System Software, • AutoCAD, • QuickSurf, • Archos, • KarTop, • Polar. The tests at comparative points showed that the error in the determination of the heights does not exceed two metres, which provided a satisfactory precision for the calculation of the volumes, in particular since it is known that landslide displacements of several metres in a short period are no exception. The results of the geodetic processing which, in addition to aerial photography, also included the determination of new pho-togrammetric reference points by GPS measurements on the field as well as classical geodetic survey photography and the inventory of all buildings by entering house numbers, were the following products: • a topographic map of the original state before landsliding of 1998 in the scale of 1:2000, • a topographic map of the landslide state in November 2000 in the scale of 1:2000, • a topographic map of the landslide state in May 2001 in the scale of 1:2000, • TIN1998, • TIN2000, • TIN2001. As examples of geodetic processing, the picture below presents TIN98 states before landsliding (yellow colour) and TIN2001 meshwork of movements in the upper part of the landslide (blue colour): Fig. 1. TIN98 states and TIN2001 meshwork (upper part of the landslide Slano blato) In the first step, we tried to calculate the changes in the volume over time due to the expansion of the landslide by simply determining the volume between two TIN mesh-works. The checking of the data obtained showed unsatisfactory results, because the volume of the change in surfaces between two states did not provide a satisfactory answer about the sliding masses actually involved in landsliding. Consequently, we decided to calculate the volumes of the sliding masses in a more time-consuming way according to the profile method explained below. 416 Mihael Ribi~i~ Calculation of the volumes of sliding masses The calculation of volumes is a long and complex procedure which can practically not be performed by hand. The calculations were made by means of a computer with the software applications AutoCAD 2000, QuickSurf and Microsoft Excel. In order to transfer the data between AutoCAD and Excel, short programs were written in Visual Basic. First, QuickSurf was used for constructing the planes of surfaces for each aerial shoot. These planes were cross-sectioned transversely against the slope at distances of approximately 25 m (right picture – presentation of the upper part of the landslide). Fig.2. Presentation of the upper part of the landslide We obtained three lines of states for each shoot. The fourth line, representing the depth of landsliding, was constructed for each case by means of the data gathered by probe wells and the interpretation of the geological structure, digitalised and transferred to the other three lines. The result was 51 transverse sections, with three of them being shown below as examples: The procedure of calculating the volumes was the following: • For each cross-section, the surface from the reference height was first determined for each of the three landslide states at different times and for the base by means of the Boundary and Area commands in AutoCAD. • The data for the surfaces calculated in this way were transferred to Excel, where the calculation was continued. • The surfaces were mutually subtracted for each case. The surfaces November 2000 and April 2001 were subtracted from the original surface (1998). We also searched for the difference between the state in April 2001 and the interpreted base. A negative difference between surfaces means that masses were carried away from the cross-section area, while a positive difference means that they were brought from elsewhere. • The volume between two cross-sections was calculated by halving the sum of both surfaces and multiplying it with the distance between the two cross-sections: 36 LEGEND - Original state of year 1998 State in November 2000 State in April 2001 Surface of rupture 37 Fig.3. Transverse cross-sections Calculation of the moving landslide masses volume from air images 417 Tab.1. The results of the volume calculation Volumes (m3) 00-98 01-98 01-00 01-Pod 1. upper part of the landslide and Slano blato -35,876 -61,190 -25,313 144,800 2. upper channel -12,714 -25,716 -13,002 42,742 3. “Blatno jezero” 53,431 63,229 9,798 292,092 4. lower channel 459 99,404 98,944 174,403 5. area above the waterfall 0 10,274 10,274 24,344 TOTAL 5,300 86,001 80,702 678,381 • The results obtained were further used for drawing charts and for various summary tables and calculations. Vk: Pk+Pk+l*dM pk k- cross-section pk+1 k+1- cross-section dkk+1 distance between k-and k+1- cross-sections Vk volume of k- area between two cross-sections For easier understanding, the whole landslide was divided into five typical sections. The first “upper part of the landslide and Slano blato (1)” is the area of landsliding, from where all the sliding masses originate. In the area of “upper channel (2)”, these masses moving downhill become wet and turn into a mud mass, which continues its way as a mudflow. After long periods of heavy raining, the mud-flow moves several hundred metres down- wards, until it loses its energy within some days and stops for a few months. In the area of “Blatno jezero (3)” (“Mud Lake”), it spreads, thus producing a secondary accumulation of stagnating mud. When “Blatno jezero (3)” is full, the mud runs out along the “lower channel (4)” and starts to accumulate in the “area above the waterfall (5)”. The calculations showed that 680.000 m3 of material was involved in sliding until April 2001. Each movement of the sliding masses includes new amounts of landslide material, partly also from the base, like the remnants of old landsliding. The basic question is what are the total masses that may get involved in landsliding over the long run. This is shown in the chart below, which presents the changes in the volume along the landslide. The chart and Volume changing along landslide Slano blato upper part and Slano blato distance (m) 2000-1998 ¦2001-1998 • 2001-bedrock Fig.4. Volume changing along landslide Slano blato 418 the table above show that there are still potential large sliding masses in the base in the areas of the upper part of the landslide and of “Blatno jezero”. The “upper part” and “Blatno jezero” still contain around 150.000 m3 and 290.000 m3, respectively, and along the whole landslide there is around 680.000 m3 of landslide material. Landslide restoration The results of volume calculations of landslide masses indicate that successful landslide restoration is possible. Nevertheless, restoration will be time-consuming and financially demanding, taking several years. The above volume analysis show that restoration measures should include: • prevention of the sliding mass from becoming wet (draining), • regrouping of the soil masses from the central landslide area to its sides with the aim of decreasing the sliding amount, • carrying away of the sliding material in the amount of 300.000 m3 to 400.000 m3 to a deposit area. The volume analysis also indicated that over time increasing amounts of mud masses get involved in landsliding. Consequently, any delay in restoration measures results in a more difficult and expensive restoration. The calculation of the movement until November 2000 thus showed that landsliding included ?50.000 m3, while in April 2001 ?170.000 m3 of material was already sliding. Approximately ?150.000 m3 of unstable masses thus still remained in the upper part of the landslide. One-third to half of this Mihael Ribi~i~ material is already sliding. There are some additional sliding masses in unstable sides and the right part of the landslide. Taking into account these and other parallel results, it turned out that in order to prevent the sliding masses from reaching the village of Lokavec as much as possible of the sliding masses should be retained in the upper and central parts of the landslide by draining and pushing the masses to the sides of the landslide, while the masses that cannot be retained on the slope – assessed to between 300.000 m3 and 400.000 m3 – should be carried away to a deposit area. Conclusion It was shown that the calculation of volumes by means of aerial photography was a very appropriate method for large landslides which, however, only produces results after a detailed geological interpretation of the events in the landslide. If a long-term extensive landslide restoration is not carried out, in a few years, the whole of this mass will also activate and begin to move downhill. In this case, the mudflow would reach the village of Lokavec. References Ko~evar , M. & Ribi~i~, M. 2001: Landslide Slano blato, MEGRA (Annual meeting of civil engineers from Slovenia), Gornja Radgona, Slovenia. R i b i ~ i ~ , M. 2001: Landslide Slano blato – volume calculation, [uklje days (Geomechanic conference), Maribor. GEOLOGIJA 46/2, 419–424, Ljubljana 2003 Landslide mapping with the GIS Mihael RIBI^I^ Gradbeni in{titut ZRMK d.o.o., Dimi~eva 12, 1000 Ljubljana, Slovenija Key words: landslide, morphology, GIS, landslide mapping, landslide map, landslide features, landslide description Klju~ne besede: plaz, morfologija, GIS, kartiranje plazov, karta plazov, lastnosti plazov, opis plazov Abstract Any new technology allows an improved approach to the processing of expert data. The article describes how to make the recording of the data of engineering geological landslide mapping of the best possible quality by exploiting the possibilities offered by GIS. All input landslide data acquired during mapping, called landslide elements, were classified, as customary in GIS, into three groups: point, line and polygon elements. A list of the typical landslide elements noted in landslide survey was made for each group. Each element came with a description of its origin and appearance, the surveying method and the graphical representation in GIS. Additionally, a presentation of the landslide in three layers was introduced, historically reviewing the landslide. The first layer represented the past shapes before sliding, the second layer represented the recent signs of sliding and the third one (for old landslides) represented the shapes that arise after sliding. The graphical landslide data were connected with the attributes kept in the descriptive landslide database. The described approach involving the use of GIS considerably improves the quality of data capture in engineering geological landslide mapping. An example of a landslide map prepared according to the described system is attached to the article. Kratka vsebina Vsaka nova tehnologija omogo~a izbolj{an pristop k obdelavi strokovnih podatkov. V ~lanku opisujem, kako ~im bolj kvalitetno shraniti podatke in‘enirskogeolo{kega kartiranja plazu z izrabo mo‘nosti, ki jih nudi GIS. Vse vhodne podatke o plazu, pridobljene pri kartiranju, ki jih imenujem elementi plazenja, sem razvrstil, kot je obi~ajno za GIS, v tri skupine: to~kovni, linijski in poligonski elementi. Za vsako grupo sem izdelal seznam tipi~nih elementov plazenja, ki nastopajo pri pregledu plazu. Za vsak element podajam opis nastanka in pojavljanja, na~in zajemanja na terenu in grafi~ni prikaz v GIS-u. Poleg tega uvajam prikaz plazu v treh slojih, ki plaz zgodovinsko obravnavajo. Na prvem sloju so prikazane reliktne oblike pred za~etkom plazenja, na drugem sve‘i znaki in na tretjem (za stare plazove) dogajanja po kon~anju plazenja. Grafi~ne podatke o plazu sem povezal z atributi shranjeni v opisni bazi podatkov o plazu. Opisani pristop z uporabo GIS-a zelo izbolj{a kvaliteto zajema podatkov pri in‘enirskogeolo{kem kartiranju plazu. K ~lanku prilagam primer izdelane karte plazu po opisanem sistemu. Introduction aim is not only to make data capture as appropriate for the GIS technology as pos-New technology, which GIS no longer is sible, but to also improve the professional actually, also enables feedback. Those who quality through a different aspect enabled master it can tackle the professional tasks by GIS. that have been customary to date in a differ- Many experts have dealt with the use of ent way, from another aspect. In this way, GIS for landslides, so that there is a vast new professional possibilities appear. The literature on various approaches and tech- 420 niques. I have myself frequently used GIS in landslide mapping and thus discovered certain views and considerations that were new to me and which I hope will be at least partly new and useful to somebody else. My goal was not to develop a new system, but to create a useful method of landslide mapping. Landslide presentation in GIS One of the usual techniques in GIS is that when a new spatial problem is being worked on each of the influential factors is presented in its own information layer. When mapping fossil landslides with more or less blurred shapes of sliding, one observes the geological and morphological forms that are or are not the signs of landsliding occurring in the past. My work showed that the one of the most appropriate approach was the method in which information layers were divided into four groups: 1. original forms appearing before land-sliding, 2. forms as the consequence of land-sliding, 3. forms appearing after landsliding, 4. forms due to human activities. The first group includes all morphological forms already existing before landsliding, like remnants of terraces, slope levelling, streambeds and similar features. The second group of information layers contains the forms appearing as the consequence of displacements during landsliding, which are later described in detail. The third group involves the forms that blurred or highlighted the shapes of landsliding, for instance the action of erosion on a landslide which subsequently eroded part of the landslide. Finally, the fourth group contains forms occurring during any landslide restoration or other human activities interfering with the surface of the area concerned. All morphological forms for which it is not clear whether they arose due to land-sliding are inserted in the information landslide layers. When, after the field processing and digitalisation of data, the information landslide layers are drawn separately, the previously hidden integral image of the landslide is clearly revealed. This is, naturally, Mihael Ribi~i~ only true of very large landslides which occurred in the geological history, but not of the smaller and more recent ones, most of whose characteristics can already be determined during mapping. However, I also use for these landslides the division to the forms before, during and after landsliding, because this makes their presentation much clearer. Sometimes such an approach shows that there have been several landsliding stages. In this case, the landsliding shapes are divided into information layers of individual landsliding stages. For a final determination whether a landslide is an old one, the morphological signs must be added by the geological signs of soil composition, pointing to a possible landsliding. Such signs are not treated in this paper. The second basic characteristic of using GIS is that the graphical elements, in this case landslide elements, are presented as points, lines, polygons or bodies. The point elements may be sources, boreholes or measurements on a landslide. The line elements are all main scarps and cracks on a landslide. In time, the line elements of a recent landslide become blurred and rounded, and in old landslides they appear as ridges or steps. Then, it is frequently more appropriate to present them as plane units by polygons. Plane units are the typical forms of different landslide areas occurring during the movement of the sliding masses down the slope. The typical plane forms are flat or undulating planes of characteristic shapes. It depends on the rock in the base and its susceptibility to sliding how fast the initially distinct landsliding forms will disappear and become less and less recognisable. Another factor influencing the recognisability of landsliding is the size of the landslide and the dimensions of the displacements of the sliding masses. Very large landslides with extensive displacements remould the terrain to such an extent that the landslide shapes may remain visible even for several hundred years. Only an assembly of several typical morphological forms with appropriate geological signs constitutes a reliable proof that there was a landslide in a certain place, while individual, although characteristic, forms could have occurred during other natural or anthropological events. The point data are not important for landslide identification by itself. In the GIS pre- Landslide mapping with the GIS 421 sentation, they are used for marking water spring, measurement points, etc. In detailed mapping, point elements are used for the measurements of the morphological form characteristics, like the inclination of the terrain at a certain point. Below, point measurements connected with the line and polygon morphological forms are described in more detail. Line morphological forms For landslide identification, line data as the remnants of landslide scraps and cracks are more important. In precise measurements, certain line elements can also be presented as polygons. The most typical line forms are shown in the following pictures of the characteristic landslide profile and ground plans: Main scarp Minor scarp Right and left flank Transverse cracks Longitudinal cracks Fig. 1. Landslide terms (C r u d e n & V a r n e s , 1996) In the process of ageing, fresh cracks become increasingly blurred, the weathering rounds them, and erosion turns longitudinal cracks into ditches. Thus, with old landslides, the following forms are found instead of fresh cracks: The main scarp is usually rounded and followed by a steep plane, ending in a more or less levelled terrain. Appearance in the nature (sketch): Ulli Presentation in GIS: ^r v V- A longitudinal ridge (left and right) is a remnant of the lateral flank in areas where the material on the sides was piled-up over the edge of the landslide. It is more frequent in the lower part of the landslide. Appearance in the nature (sketch): Presentation in GIS: A longitudinal ditch (left and right) is also a remnant of the lateral landslide flank (left or right) which, after the slide, became lower than the displaced surrounding. It is often deepened by line stream erosion. Appearance in the nature (sketch): Presentation in GIS: Oblong concavity appears in the form of a ridge which usually crosses the landslide transversely. Concavities may be of different dimensions - heights and widths. Ordinarily, they are the consequence of a step in the landslide whose edge later became blurred or plastic deformation of the sliding material. Concavity may also appear in a landslide when the landslide material with plastic behaviour gets compressed, raising the terrain. Appearance in the nature (sketch): Presentation in GIS: ) M M (V I V) V J V J r Oblong convexity usually appears when the landslide is moving, with the material opening up and deforms plastically (tensile stress). It may also be formed as the consequence of an uneven shape in the surface of rapture. Appearance in the nature (sketch): Presentation in GIS: ft r 422 Mihael Ribi~i~ A step is a sharp transition of the terrain from a gentler slope to a steeper one. It differs from a concavity by the sudden change in the inclination. The form of a step is similar to the edge of a terrace. In the field, both inclinations are measured and the point values of the measurements are given. During landsliding, a step may also appear as a sliding surface in the very body of the landslide. One of the opportunities for it to form is also when the rock in the base on which the material is sliding locally changes its direction from a gentler inclination to a steeper one. Appearance in the nature (sketch): Presentation in GIS: In the mapping of line morphological forms, the line shapes that probably appeared due to sliding are added by the line shapes that had existed before sliding (displaced paths, stired terraces, boundaries of changes in the vegetation, etc.) and those that formed after sliding (erosion ditches, new waterways, etc.). As far as the forms appearing after sliding are concerned, we are interested in the extent to which they have covered up the signs of sliding. The natural signs are primarily connected with erosion, frequently with the carrying away of the landslide material, and the anthropological signs are mostly connected with farming, afforestation, remoulding of the terrain by a bulldozer, etc. Plane morphological forms The most reliable evidence of landsliding in the past can be obtained by analysing plane forms. When mapping or identifying old landslides, I found many other characteristic plane shapes, with some of the most frequent being presented below: The main sliding surface appears under the crown and represents the slide of material along the main scarp. This movement is usually the largest one in the body of a landslide. With fossil landslides, the main sliding surface is one of the most distinct features showing that sliding has occurred in the past. In old landslide identification, a distinct lack of soil volume is recorded in this part. Appearance in the nature (sketch): Presentation in GIS: An undulating surface as a very frequent sign of old landsliding appears when the landslide material has plastic properties. When moving down the slope, the sliding material becomes compressed or expands, with various irregularities in the form of the landslide material and the wet masses in the landslide body finding their expression in plastic deformations of the material. Appearance in the nature (sketch): Presentation in GIS: <:mj A horizontal or an inclined plane, on the other hand, normally appears in areas where the sliding is even and regular. It may be a remnant of original levelling before sliding. Appearance in the nature (sketch): Presentation in GIS: A concavity is a round or elliptical depression appearing in a landslide. It often contains water or becomes marshy. It is formed on the upper part of the landslide as a lack of the material that slid or due to the uneven movement of the landslide, with the sliding mass being of such a material that can get plastically deformed. Appearance in the nature (sketch): Presentation in GIS: Landslide mapping with the GIS 423 A convexity is round or elliptical and looks like a small hill on the landslide. It points to a local accumulation of masses in the landslide, primarily in its toe. Appearance in the nature (sketch): Presentation in GIS: The zone of accumulation appears in the area where the shear resistance of the ground increases largely, thus stopping the sliding masses, although they are still under pressure of the higher sliding masses. Due to the slowing down of the landslide, the pressures in the landslide directed towards the slope result in the accumulation of material and even in the rising of the surface in the area of zone of accumulation. The latter are typical of the toe of the landslide. Appearance in the nature (sketch): Presentation in GIS: Landslide measurements When determining the characteristics of line and plane units mostly of an active landslide, it is sensible to also perform point measurements, which provide additional information on the characteristics of the forms: Main scarp, minor scarp or new scarp: Height of vertical displacement - presentation: 36 .... ID No. of point, —> direction of movement, -0.35 vertical movement in meters Right and left flank: Height of vertical displacement (upwards or downwards), horizontal displacement, expansion to the sides – presentation: Transverse cracks and longitudinal cracks: Openness and depth of cracks – presentation: Similarly it can be presented also next landslide elements (units): Longitudinal ridge: height of ridge Longitudinal ditch: depth of ditch Step: inclination of terrain above and under the step Main sliding surface: inclination of the sliding surface Undulating surface: distance between two concavities or convexities Inclined plane: inclination of the inclined plane Convexity: depth of convexity Concavity: height of concavity Zone of height and superposing accumulation: of the zone of concavity Conclusion When mapping old landslides, one no longer encounters fresh signs of landsliding, primarily expressed as lines, but blurred consequences of sliding which mainly appear as plane morphological forms. According to the procedure described in the paper, GIS is used to present these forms in separate information layers, thus supplying new information on landsliding, which is shown in the following picture. 424 Mihael Ribi~i~ ENGINEERING GEOLOGY MAP OF LANDSLIDE smer trase daljnovoda References WP/WLI UNESCO, 1993: Multilingual Landslide Glossary. Bi Tech Publishers, Richmond, British Columbia, Canada. R i b i ~ i ~ , M., B u s e r , I. & H o b l a j , R. 1994: Digital Attribute/Tabular Database of the Landslides in Slovenia for Field Cupturing of Data. First Slovenian Conference on Landslides, Idrija, 17. in 18. november 1994. Idrija: Rudnik ‘ivega srebra, 1994, 139–153. Cruden, D.M. & V arnes, D.J. 1996: Landslide types and processes. In Landslides – Inves- tigation and Mitigation, Transportation Research Board Special Report No. 247 (A.T. Turner &. R.L. Schuster ed.), National Academy, Press, Washington DC, 36–75. R o b i n F e l l R., H u n g r O. L e r o u e i l , S. & R i e m e r W. 2000: Engineering of the Stability of Natural Slopes, and Cuts and Fills in Soil (Keynote Lecture), V: International conference on geo-technical and geological engineering, 19–24 November 2000, Melbourne, Australia. Lancaster; Basel: Technomic Publishing Company, cop. 2000. C a r r a r a , A., C a r d i n a l i , M., G u z z e t t i , F. & Reichenbach, P. 1996: GIS-Based Techniques for Mapping Landslide Hazard, Bologna. GEOLOGIJA 46/2, 425–428, Ljubljana 2003 GESTCO GIS and DSS – A GIS solution to assist with decision making for the geological storage of CO2 from fossil fuel combustion Nichola SMITH1, Frank KEPPEL2 & Sam HOLLOWAY3 1British Geological Survey Murchison House West Mains Road Edinburgh EH9 3LA 2Netherlands Institute of Applied Geoscience TNO - National Geological Survey P.O.Box 80015(Princetonlaan 6) 3508 TA Utrecht 3British Geological Survey Keyworth Nottingham NG12 5GG Key words: GESTCO GIS, DSS, CO2, geological storage, GIS Abstract This project aims to determine whether the storage of CO2 underground, such as is taking place at the Sleipner West Gas Field, North Sea, can become a practical industrial solution to major CO2 emissions into the atmosphere from large point sources such as power plants. If this is a practical proposition it could make an impact on the enhanced greenhouse effect caused by man emitting CO2 into the atmosphere. As part of the project a dedicated Geographical Information System (GIS) and a Decision Support System (DSS) have been developed. The GIS enables the user to view and analyse the large amounts of data collected, whilst the DSS enables emission – source – storage scenarios to be planned and cost evaluated. A webGIS was also set up to enable the project partners to view the progress of data collection and to assist with data checking. Introduction Following the Kyoto climate conference in 1997 a consortium of 8 European national geologic surveys launched a project in 2000, spanning 3 years, which has studied the technical and economical feasibility of wide-scale application of CO2 storage in the subsurface. This EU project was entitled “European Potential for Geological Storage of Carbon Dioxide from Fossil Fuel Combustion” (acronym GESTCO). The EU Kyoto objective implies a reduction of 8% (relative to 1990) of the greenhouse gas emissions. This amounts to a re- duction of approximately 600 million tonnes per year of CO2 between 2008 and 2012. Power generation has the largest individual contribution of CO2 emission and this amounted to 950 million tonnes in 1990. As nearly all fossil fuel power generation occurs at major facilities there is potential for CO2 capture and sequestration. The GESTCO project has aimed to make a major contribution to the possibilities of reducing CO2 emissions into the atmosphere by investigating whether geological storage of CO2, as is taking place at the Sleipner West Gas field, is a viable method capable of wide scale application. The GESTCO project 426 Figure 1. View of UK CO2 sources in the GIS aims to provide documentation and data to show that for emission sources in key selected areas there is sufficient geological storage capacity. A large amount of data has been collected from the participating countries (Belgium, Denmark, France, Germany, Greece, Netherlands, Norway and the UK) for use in the GIS and the DSS. An inventory of major CO2 sources has been made and this data will be combined in the GIS with information on potential underground CO2 sinks and potential CO2 transport routes. Four main types of underground storage sites have been investigated, these being onshore/offshore saline aquifers, low enthalpy geothermal reservoirs, deep methane-bearing coal beds and abandoned coal and salt mines, and exhausted or near exhausted oil and gas fields. The participating countries have also researched several case studies. The DSS, developed through customisation of ESRI’s ArcMap® using VBA, provides the tools for evaluation and comparison of the costs and economic risks of realistic combinations of CO2 emission sources, transport possibilities and storage capacities for various scenarios input by the user, it takes into account all cost relevant parameters for sequestration, transport and storage of the CO2. GESTCO GIS The objective for the GESTCO GIS was to produce a Geographical Information System that would incorporate the wide range of data provided by the project partners and Nichola Smith, Frank Keppel & Sam Holloway Figure 2. View of Hydrocarbon Field sinks in the North Sea allow the partners and end-users meaningful access to the data. The GIS allows users to simultaneously view one or more layers of data including the location of the CO2 sources and possible CO2 sinks, it will also enable the user to perform extensive on screen analysis on all the available data. Geoscience datasets included in the GIS comprise aquifer injection points and aquifer area location, hydrocarbon field injection points and hydrocarbon field locations, coal mines, coal field and coal field injection points as well as the locations of the CO2 sources, existing pipelines and pipeline terminals. Many other datasets have also been provided to enhance the capabilities and information held within the GIS, for example geological, tectonic zone and ecosystem data. CO2 Sources The CO2 sources database was built by EcoFys from data provided by the project partners. The database incorporates a large amount of data including information on the location, emission and sector (power, chemical etc). The data is then converted into shapefile format for visualisation within the GIS as a point dataset with scale rendering to give users an immediate view of the size of emissions. CO2 storage (sinks) datasets These datasets, which include the aquifer injection points, hydrocarbon field injection points and coal field injection points were collated from data provided by each partner. The data incorporates information on the storage capacity for CO2, depth, pressure GESTCO GIS and DSS – A GIS solution to assist and porosity of the sink. The hydrocarbon field injection points database was built by TNO whilst the other datasets were provided as shapefiles by each partner and merged into single datasets by the British Geological Survey (BGS). To provide access to additional information, held within the websites of the Geological Surveys involved, along with other websites, links to external websites have been set upwithin the GIS. The GIS, which has been developed using ESRI’s ArcGIS®8.2 software, uses ArcMap, whilst the datasets, which were initially provided in shapefile or Excel format, are stored within a personal geodatabase which uses Microsoft Access. The personal geodatabase enables the storage of all the datasets in a single location which makes transfer of the GIS data from one location to another much easier. This is a very important requirement for the GIS, as well as the DSS, as it is necessary to ensure the systems are easily transferable to the project partners and the end users on completion of the project. To assist in the ease of this transfer process the GIS has also been set up using relative pathnames which ensures that the GIS will always pick up the location of the datasets. There has been some customisation of the GIS to allow users, who are unfamiliar with the GIS environment, to use the system more effectively. This customisation has taken place using ESRI ArcObjects within the VBA environment. The main customisation has been to develop a selection tool that will allow users to select from within the datasets based on the CO2 emissions or CO2 storage capacity. This tool also allows the user to save their selection as a new shapefile should they wish to keep it for further analysis. Copyright information is also a feature of the GIS. Users must agree to abide by the copyright of the data before the GIS will open fully and there is also the ability to access the copyright information from within the GIS should users wish to read it again. Case study data Many case studies have been carried out for the project and the data from these has been included in the GIS. As this data has been provided in many different formats and is specific to particular case studies this data with decision making for the geological ... 427 has not been merged into single datasets as with the general GIS datasets. There are many maps and diagrams that have been provided for the case studies, as it is highly useful to be able to view such maps, diagrams and seismic profiles, from within the GIS, hyperlinks have been set up. This enables the user to click on a feature with the hyperlink tool and view any maps or documentation associated to the feature. GESTCO WebGIS It was decided that the best way to allow the project partners and end-users to monitor the progress of the data collection was to set up a web-based GIS system. The GESTCO webGIS was developed using ESRI’s ArcIMS® software, which allows the easy dissemination of GIS data over the internet. The webGIS does not have the full functionality of the GESTCO GIS, however it does allow users to view the datasets on screen and perform simple queries on the data. The webGIS also became a very useful resource towards the end of the project when it was used by the project partners to do the final checks on the data they had provided in the preceding 3 years. GESTCO DSS As part of the project The Netherlands Institute of Applied Geoscience TNO, one of the GESTCO participants has developed a decision support system. This DSS calculates costs and economic risks of realistic combinations of CO2 emission sources, transport possibilities and storage capacities for each of the selected areas.The DSS is founded on ArcView®8.2 extended with Spatial Analyst. The end user interfaces with ArcView®8.2 and defines a removal scenario by selecting a CO2 source and a storage location (sink). After scenario composition, Spatial Analyst will determine the least costly transport route. For this ArcView®8.2 is fed with data which expresses costs related to pipeline construction; costs determined by aspects like land use, elevation, artificial and natural barriers, existing pipeline corridors are added in grid format so that Spatial Analyst can take these into account when searching for the optimal route. 428 Once the scenario is completed with an optimal routing between sources and sink, calculation models will kick in and evaluate remaining technical and economical aspects of the problem definition: the costs for CO2 separation at the source is calculated, the size of the CO2 flow in time from source(s) to sink is used to calculate the needed dimensions of pipelines and the number of compression stations along the route. Storage models will evaluate the chosen sink on volumetrics (pore volume, compressibility, sweep efficiency) and injectivity behavior (fluid mobility, injection rate, number of needed wells). These calculation models are implemented outside ArcView®8.2 and coupled as Dynamic Link Libraries. When all calculations are finished, the results are gathered within ArcView®8.2 and the whole scenario evaluation will be presented to the end user in numbers and graphs. And, of course, the chosen route is geographically mapped. The end user will get an answer on whether it is technically possible to separate, transport and store an amount of CO2 over time and how much such a scenario will cost. Conclusions This project has enabled the development of two highly useful systems that should prove invaluable in the decision making process with regards to possible CO2 sequestration. The data collected is a valuable resource and the GIS provides the best interface for accessing and viewing the data. The DSS has been the vehicle that comprises many of the geoscientific and economical study results that were gathered during the GESTCO project. Although it should be viewed upon as a prototype, it is already being used in other projects. Several assumptions and simplifications were made for the sake of implementation. As with many DSS systems the Nichola Smith, Frank Keppel & Sam Holloway Figure 3. Gestco DSS results of a scenario run will not result in deadly accurate figures. The results should be applied as selection criterion for many different scenario runs. The DSS aims however to provide insight in the power of costs that are at hand when dealing with CO2 sequestration. It is the intention to continue to develop and maintain these systems within future projects relating to CO2 sequestration. Acknowledgments The authors would like to acknowledge the project co-workers from GEUS (Geological Survey of Denmark), BGR (Federal Institute of Geoscience and Natural Resources Germany), BGS (British Geological Survey), BRGM (Geological Survey of France), GSB (Geological Survey of Belgium), IGME (Institute of Geology and Mineral Exploration Greece), NGU (Geological Survey of Norway), NITG-TNO (Geological Survey of The Netherlands) and EcoFys Environment and Energy for the work done to provide the data included in the GIS and DSS. This paper was produced with the kind permission of the Director of the British Geological Survey. GEOLOGIJA 46/2, 429–434, Ljubljana 2003 SMART oilfield GIS: Application of GIS for economic and environmental monitoring of oil and gas fields A. TCHISTIAKOV, A. T. GANZEVELD & J. F. KEPPEL Netherlands Institute of Applied Geoscience TNO, Princetonlaan 6, 3584CB Utrecht, The Netherlands; a.tchistiakov@.nitg.tno.nl Key words: SMART, oilfield application, GIS Abstract TNO-NITG has recently developed an extensive exploration and production (E&P) database system, thereby providing a practical and very cost-efficient alternative to the systems existing on the market. Two different approaches were taken: one using licensed software with built-in components, and another using open source software. In this article the merits of both approaches are discussed. Introduction With ever growing possibilities in data gathering, processing speed and storage capacity, the amount of information that can be derived from oil or gas field data has grown enormously over the past decades. A few vendors developed software systems capable handling these complex data streams and their relationships. However, due to the complexity of matters to deal with, they are expensive with respect to the costs of their license and maintenance. At the same time these systems suffer from the 80-20 syndrome: only 20% of the functionality is used while the remaining 80% contributes to the total product costs. This jeopardizes the effectiveness of the users’ software investments. TNO-NITG, being the National Geological Survey of the Netherlands, has developed an E&P database and GIS application that can compete with this off-the-shelf software in functionality and speed. At the same time TNO-NITG ‘s system is cost-effective with respect to development costs, maintenance and customisation. It is a flexible, scalable solution for managing a wide range of information on exploration and production activities of a company, as well as environmental monitoring data. Two national oil companies have already been using this new E&P database system. System Design The distinguishable features of the systems are the open source and modular structure. Thanks to these, the system is highly customisable to meet the requirements of a particular client and it is easy for the customer to maintain the software. The E&P system includes three main modules: • E&P Data Manager (the core database management module) • E&P Reporter (intelligent tool for data mining and flexible reporting) • E&P Spatial Modeller (E&P GIS) 430 A. Tchistiakov, A. T. Ganzeveld & J. F. Keppel Figure 1. E&P Data Manager E&P Data Manager The E&P Data Manager (Figure 1) module includes: • E&P DATABASE management system with POSC-compliant database model • E&P FORMS for data input and analysis • E&P REPORTS for creating standard reports of company activities The most important component of the E&P Database management system is the TNO-NITG’s E&P data model, which comprises all aspects of the E&P enterprise. The TNO-NITG model is based on a subset of the Petrotechnical Open Software Corporation’s (POSC) Epicentre data model, which allows users to store and extract all forms of data and metadata related to E&P: seismic, petrophysical, geological, reservoir engineering, well, borehole, facility, pipeline, rock and fluid sample, and field data. An added benefit is that the database can be easily integrated with other software. Users can thus get hold of any relevant E&P data when performing supplementary analyses or drafting reports on particular topics. The types of data that might be of interest for such purposes are statistics or facts and figures about hydrocarbon production and contouring. The system also incorporates an authorisation function with respect to users. They are assigned specific roles for specific groups of data. This distributes the responsibility for the import and quality control of the groups of data among various users. The E&P database manager is responsible for the referential and application data, and therefore also for the user roles E&P Reporter TNO-NITG has extended the E&P database’s functionality by incorporating Oracle Discoverer into the system. Being an intelligent tool for data mining and flexible reporting, the new E&P module is of particular interest to geo-scientists and production engineers involved in the analysis of oil and gas field performance (Figures 2, 3,4). The module gives the E&P database a number of additional advantages relative to other oilfield management systems. For example, a new report can be created via dialogues similar to those in Windows Explorer. This makes it quite simple for an oil specialist to create a new report tailored to personal requirements and spares the user from having to learn the complex relational data- SMART oilfield GIS: Application of GIS for economic and environmental monitoring of ... 431 Figure 2. Analysis of well geology Figure 3. Netto sand thickness calculations per well Figure 4. Monthly Field Production Report Figure 5. Well productivity analysis by means of E&P Reporter base structure. Once a report has been produced it can be shared with other colleagues. As Discoverer reports have an Excel-like interface, their contents and layout can be easily customised by users, including field geologists and production engineers. The extremely flexible report construction tools allow for in-depth data mining that employs the end-user’s professional expertise. Moreover, the use of dynamically updated graphics significantly simplifies the analysis of production and geological data. All in all, the module has proved itself to be an effective tool for data quality control. Finally, the module can export the E&P data to a large number of external formats (incl. ASCII, Excel and HTML) enabling more sophisticated oilfield analysis using advanced computer simulators (Figure 5 & 6). E&P Spatial Modeller (Oilfield GIS) The GIS module stores information about oil and gas fields as a collection of thematic layers that can be linked together by geography. It is able to visualise the geo-refer-enced data stored in the E&P database and automatically update the information each time while opening a new session. Moreover the module enables selecting geographically (interactively on the screen) and updating the environmental and field operation data whenever it is necessary (Figure 7). The module allows mapping both technical and environmental data as well as studying relationships between contamination and 432 A. Tchistiakov, A. T. Ganzeveld & J. F. Keppel ^MjN^nri^lir^^ ¦ ˇ 9 ? .* ; T 1 • S \Ä ' V Ö © ¦ © w w # *QQ - w » * o -^ lIj • , rQ * -r • • ™^ .ui- r| aii-aj (¦-« j.1 ;j±C ¦ ¦ | 11 Fl J "M^l —' , 11 ¦*•¦_¦ L- ¦ HUUU , L'.bI in liiJUUu '.¦"¦¦. ¦ ¦¦"-"¦" " '• i ¦ „. | -...