POTENTIAL OF GIS IN COASTAL BOUNDARIES DETECTION AND PITFALLS IN REPRESENTING THE COAST AS A BOUNDARY

Maria Raffaella Lamacchia (1) , Darius Bartlett

(1) SAT Department, Mediterranean University of Reggio Calabria (IT)
(2) Geography Department, University College of Cork

Introduction
"Boundary" and "coast" are terms linked by a twofold relationship: on the one hand, the coastal zone is unanimously defined as the boundary area between land, sea and air (Carter 1988); on the other, the issue of coastal boundaries definition is addressed by the international literature as one of the most thorny subjects implied by ICZM theories and practices (Cicin-Sain 1993). The definition of coastal boundaries is a tricky task, since it implies the definition of the boundaries of the boundary itself.
A review of the main definitions set up by scientists, practitioners and local communities indicates use of a wide range of boundary types to delimit and manage the coastal zone. Upon closer examination of these, it is possible to confirm Carter's (1988) view that all these definitions move from a common idea: the coast is that boundary region between sea and land made up of that portion of land considerably affected by proximity to the sea, and that part of ocean considerably affected by its proximity to the land.
The illustrated twofold relationship connecting the words 'coast' and 'boundary' in turn leads towards further dual relationships between nature and humanity, and between coastal boundaries and GIS applications. GIS tools are successfully applied by practitioners in terrestrial and maritime boundary delineation, for example where they are used to locate legal and conventional boundaries established by humans, and also where GIS are employed by scientists for detecting the limits of the natural elements of the marine and territorial domain. At the same time, GIS databases and tools show enormous limits in facing the boundary nature and the associated fuzziness of coastal systems.
A number of comparatively recent theories such as Fuzzy Logic, Fractal Theory, and Boundary Objects, along with emerging disciplines such as Landscape Ecology, etc., support us in dealing with the boundary nature of coasts.
The deriving conceptual model of the coast as a boundary is not easy to implement in current geographical information systems, since these are designed to deal almost exclusively with static, well-defined 'objects', exactly located in space with properly defined attributes (Burrough and Frank 1996). At present the coast is generally represented within GIS through one-dimensional or two-dimensional datasets, and in few experimental cases it is represented taking into consideration its third (height) and fourth (temporal change) dimension (Bartlett 2000; Wright and Bartlett 2000). Even in these latter cases, however, coastal representations are forced into the classic, crisp mode of conventional information theory and traditional cartographic methods, since most GIS seems theoretically inadequate to deal with objects with poorly defined boundaries, such as the coast and its elements.
The current paper examines these issues from the specific perspective of the territorial and regional planning professional, and considers the implications of coastal boundary delimitation for the elaboration, communication and implementation of international coastal zone management plans. To illustrate the problems raised, examples are drawn from case studies of the coasts of Bari on the Adriatic coast of southeast Italy, and County Cork on the southwest coast of Ireland.

1. Coastal boundaries

1.1 A multiplicity of coastal boundaries
Although it is universally recognized that the coastal zone is a portion of the earth surface easy to identify at a subjective level, it seems to be one of the most problematic places on Earth to represent formally.
Broadly defined as the geographical place where land sea and air meet, the phrase "Coastal Zone" assumes several meanings and boundaries depending on (i) the physical characteristics and legal systems of the area concerned, (ii) the discipline and cultural background of the observer (iii) and the purpose of the definition or delimitation.
As suggested by Davis (1992) and confirmed by the observation of coastal boundaries adopted within the EU-Demonstration Programme on Integrated Coastal Zone Management 1997-1999 (Humphrey and Burbridge 1999) all coastal boundary delimitation criteria currently adopted can be organized in to 3 main categories: (i) geometrical-linear criteria, (ii) legal-administrative criteria (iii) and ecological-natural system criteria. In some cases a single criterion is adopted, reaching a sole definition of coastal zone extent; while in other cases multiple criteria are adopted and several boundaries are identified according to different objectives.

Tab. 1 Coastal Boundaries

GEOMORPHOLOGIC BOUNDARIES - coastal plain boudaries;
- continental shelf boundaries;
- continental slope boundaries;
- continental rise boundaries;
- active cliff line;
HYDROGEOLOGICAL BOUNDARIES - Coastline;
- low/high tide shoreline;
- inshore - inshoreline, and offshore;
- low/high tide breakline;
- Backshore - backshoreline,
- Foreshore - foreshoreline;
- Nearshore (Surf zone and Swash zone boundaries);
- Water tables, watersheds;
- Erosion subjected areas boundaries;
- Boundaries of regions alternatively floated and exposed during sea level fluctuations over millennia;

BIOLOGICAL BOUNDARIES		 - Benthic and Pelagic zone boundaries;
- Epipelagic, mesopelagic, batipelagic, abissopelagic zone boundaries;
- Euphotic and aphotic zone boundaries;
- Neritic and oceanic zone boundaries;
ECOLOGICAL BOUNDARIES - Ecosystems boundaries;
- Habitats boundaries;
- Ecotono boundaries;
Sandy coast:
- Beaches: supralittoral zones, littoral zone, surf zone, transition zone, outer turbolent zone boundaries;
- Dunes: foredune, back dunes boundaries;
- Pioneers zone, shrubs zone, scrub-ticket zone, thicket or forest zone boundaries;
Rocky coast:
- Supralittoral zone, littoral zone, infralittoral zone boundaries;
- Low liken zone, aereosaline zone, upperwindswept zone boundaries;
- Eutlittoral zone and littoral fringe boundaries;
Wetlands:
- Salt marsh, mudflats boundaries;
- Low marsh, high marsh boundaries
LEGAL BOUNDARIES
- administrative boundaries (baronies, county, regions) ;
- ownership boundaries (public and private) ;
- Planning zones boundaries;
- Baselines, internal waters, territorial sea, contiguous zone, continental shelf boundaries;
- Exclusive economic zone, exclusive fishing zone boundaries;
- Protected areas boundaries;
- Census zones boundaries;
UTILITARIAN BOUNDARIES - Fishing areas boundaries;
- Hydrocarbons exploitability boundaries;;
- Limit of daily accessibility for tourists ;
- Limit of daily travel to work;
- Land-use boundaries;
PERCEPTION BUNDARIES - Visual boundaries (from the sea and from the land) ;
- Current perception boundaries ;
- Historical perception boundaries (by historical charts)

From an international management perspective the coastal zone concept has widened its geographical coverage, as Coastal Zone Management has evolved from a sectoral to a more integrated approach. Arising from this international debate, the geographical coverage of Coastal Zone Management programmes has widened from a strict focus on the coastal fringe, defined according to administrative (landwards) and jurisdictional (seaward) criteria, or arbitrary criteria (isodistance from a baseline), to a wider area defined according to administrative and ecological criteria. The seawards extent of this new definition covers a much-extended jurisdictional zone, while in the landwards direction it can, and frequently does, cover entire river basins.
A review of the main definitions set up by scientists (geomorphologists, oceanographers, biologists, ecologists, economists etc), pratictioners (politicians, administrators, entrepreneurs, etc.) and local comunities reveal a wide range of boundary types used to define the coastal zone. These include geomorphologic boundaries, hydrogeological boundaries, biological boundaries, ecological boundaries, legal boundaries, utilitarian boundaries, perception boundaries, etc. [cfr tab.1].

1.2 Potential of GIS in Coastal Boundaries Detection
Some coastal boundaries can be easily managed by means of GIS and Remote Sensing tools, both where the objective is to locate legal and conventional boundaries set up by humans in order to establish some rights, and also where the objective is to demarcate boundaries of any natural element.
In the first case the ability of GIS tools to perform spatial analysis based on geometrical rules is exalted: marine boundaries in fact are delimited, not demarcated, and generally there is no physical evidence, but only mathematical evidence, of these boundaries left behind (Fowler and Treml 2001). For instance the United Nations Convention on the Law of the Sea (UNCLOS) has defined a number of marine boundaries, including the baseline (art.7) and the maritime zone that are claimed from it, the Territorial Sea (art. 2), the Contiguous Zone (art. 33), the Continental Shelf (art. 76), and the Exclusive Economic Zone (art. 55-57). GIS techniques have been successfully applied to delineating such boundaries, for example the GIS-based Global Maritime Boundaries Database set up and distributed by Verdian MRJ Technology Solutions (www.marittimeboundaries.com), and the Ocean Planning Information System (OPIS) developed for the south-eastern U.S. by NOAA's Coastal Services Center (www.csc.noaa.gov/opis). [fig.1]

Fig. 1 Marine boundaries according to the United Nation Convention on the Low of the sea (UNCLOS)

In the second case the capability of GIS to retrieve information from existing geodatabases, from satellite images and from GPS, proves useful for the identification of natural elements and coastal ecosystems boundaries. For this reason an increasing number of coastal zone management projects make use of georeferenced databases and GIS software in order to define the project area and to manage the project itself. (Doody 1995).

2. Coast as boundary

2.1 Conceptual models
A careful examination of a range of coastal zone definitions set up by scientists, practitioners and local communities show a minimum common denominator: them all are based on the idea that the coast is a boundary region, a transition zone between sea and land where land is considerably affected by its proximity to the sea and vice versa (Carter 1988), a region characterised by the fuzziness of its external and internal boundaries.
Some comparatively recent theories such as Fuzzy Logic, Fractal Theory, Boundary Objects, and some emerging disciplines such as Landscape Ecology, help us in exploring theoretical and practical consequences for coastal database modelling and the implementation of coastal zone management plans. Fuzzy Logic theories (Zadeh 1965, 1987) help us in grasping the structural indeterminateness of boundaries between spatial regions; Boundary Objects and the related concept of Trading Zones (Chrisman 1999) represent attempts at circumventing the problems of defining objects that are inherently subject to multiple interpretations by "agreeing to differ"; Fractal theories (Mandelbrot 1983, Milne 1988, 1991, DeCola 1989, Wiens and Milne 1990), help us in assessing boundary length and shape, and in understanding the spatial and temporal dependences of such measurements; while Landscape Ecology offers considerations on environmental boundaries (Forman 1995, Forman and Moore 1992) that allow us to reflect on the structure, functions and quantitative aspects of the boundaries between landscape elements.
According to Landscape Ecology introducing energy causes a system, such as a coastal landscape, to become spatially heterogeneous in one of two ways: (i) as a mosaic, and (ii) as a gradient. In a mosaic, landscape elements (tiles or tesserae) are distinctly distinguishable, and a "core" and one or more "edges" characterize each element. Edges of two adjacent tesserae are divided by a border and together make up the boundary or the boundary zone. According to this model, in most cases, the coastline represents the border between the marine edge and the terrestrial one, these edges together make up the boundary, which is the coastal zone. In a gradient, landscape elements gradually change, giving rise to landscape heterogeneity through a continuum of fluctuations. There are no real boundaries in such a representation, but instead a sole transition zone characterised by the constant increase/decrease of a factor according to a distance. In some cases the coastal zone can be described according to this model, such as along coastal wetlands, where the wholly marine changes imperceptibly to become wholly terrestrial, and it is not possible to mark any clear border line between the two (Forman 1995). [fig. 2]

Fig. 2 Spatial relationship of boundary, border, edges and transition zones. (Forman 1995, modified)

In this sense, with boundaries present in a mosaic but not in a gradient, boundary and gradient are mutually exclusive patterns or concepts, but in nature they coexist in the same landscape. In many cases, changes in the character of a landscape appear, when observed from a certain distance, to imply existence of boundaries but, when these same changes are observed at a shorter distance, they are seen to become transition zones instead. Looking at a coastal zone from an aircraft window it is possible to note that boundaries distinguishable from a certain altitude are different from the ones distinguishable from a lower - or a higher - altitude: from an altitude of few meters the landscape seems dominated by fine scale boundaries, such us boundaries between lots, but when gradually the aircraft gains height lots boundaries disappear and we recognise instead broad scale boundaries such us the ones between land uses, geological forms or climatic zones.
The coastal zone can thus be interpretated as a boundary both at the fine scale and at a broader scale: it corresponds in the first case to the intertidal strip, while in the second case to the continental shelf seaward and the coastal plain landwards.
This means that in a coastal landscape gradients and mosaics, abrupt boundaries and transition zones, all coexist. Similarly fuzzy theory describes a grey world characterised by frayed contours, where all the things change fluently and where all is a matter of measure and depends on the observation scale (Kosko 1993).
Kosko (1993) explains how the boundary between the hand and fingers appear as a transition zone, a gradient - the hand gradually becomes finger-, while the skin around a finger appears more similar to an abrupt boundary -on one side of it there are all the finger tissues, on the other side all the air molecules- but at a finer observation the gradualness of this change appears evident: some molecules belong to the finger, some others to the air, but there are also molecules belonging in a certain measure to the finger and in a certain measure to the air, and at the same time belonging neither to the finger nor to the air. Similarly the change between land and sea -the coastline- is normally regarded as a boundary but at a finer observation its nature of transition zone appear evident.
According to fuzzy logic theory, abrupt boundaries and transition zones are simply the theoretical extreme limits of a range of boundaries that are more or less sudden: all is a matter of measure and all is to a certain degree fuzzy, vague and indefinite; abrupt boundaries and perfect transition zones do not exist, each boundary is characterised by a certain degree of fuzziness, they are extreme limits present only in theoretical interpretative models
In order to explain this concept Kosko (1993) use the example of a man eating an apple: the man while eating the apple transforms it into a not-apple; mouthful after mouthful the apple will be less and less apple. Similarly if we imagine sailing - as the Titanic did- from the port of Cobh in Cork Harbour towards New York, the more the ship navigated away from the harbour the less it is sailing in the coastal zone and the more it is sailing in the ocean. The voyage from the coast towards the not-coast is describable, such as the eating process that transform the apple into a not-apple, as a linear change.

Fig.3 Coast's boundary-line and fuzzy-curve.

Fig. 4 Fuzzy curves slopes of land, sea and coastal zone.

The fuzziness is not a matter of ignorance, a more detailed knowledge of an issue do not reduce its fuzziness: the information that an house has been built more than 400 metres away from the coastline, do not allow us to assert if it lies on the costal zone or not. If we gain some more precise information, knowing that the house has been built at a distance of 450 meters from the sea, the increased accuracy of the information does not help us in answering the question; we simply know that our house is "more coastal" than another house built at a distance of 500 meters from the sea. This is because the coastal zone is a fuzzy region, a Boundary object inherently subject to multiple interpretations. (Chrisman 1999)
The Galasso Law in Italy, similarly to the Law for the Protection of Nature in Denmark, states that houses cannot be built within a distance of 300 meters from the sea. These laws draw a line between the coast and the rest of the land (not-coast) on the scale of the distance from the sea; some other laws in other states put this line at a different distance: for instance the French Loi Littoral draws it at 100 meters while the Greek law 2344/40 puts it at 50 meters. In some other countries this line is drawn according to an altitude criterion, or within the immediate visual influence of the sea, and it is not easy to agree upon which is the most correct. For this reason an approach based on Fuzzy logic would not draw a line but a curve between coast and not-coast. [fig. 3]
Fuzzy logic theories recognize the existence of concepts and regions that are more or less fuzzy, and measure the degree of fuzziness of them through the steepness of this curve. The boundary between the sea and the land is steeper than the boundary between the coastal zone and the in-land. It may be technically possible, but very difficult, to establish the precise location of the boundary between sea and land (coastline) at any given point in time, and much more difficult, virtually impossible, establish the location of the boundary of the portion of land affected by his proximity to the sea and the portion of ocean affected by his proximity to the land (coastal zone boundary). [fig. 4]
Similarly the science of landscape ecology recognizes the existence of more or less "hard" boundaries, and distinguishes curvilinear from straight boundaries: (i) curvilinear boundaries result from natural processes, contain concave and convex surfaces, may include tiny patches of one ecosystem type in the other, and induce considerable interaction or movements between adjacent landscapes; (ii) straight boundaries, very common in human-imprinted areas, contain no lobes, coves or tiny patches, and function as a barrier for adjacent landscapes interactions (Forman 1995).
An urban coast, and specifically a dock, represents a clear example of a hard boundary, while a saltmarsh can be considered an example of a soft one. The different degree of softness depends both on geomorphologic factors - coasts on convergent margins are frequently more indented than coasts on diverging margins - and on the degree of artificialization of the site - human action, in fact, tends to cleanly divide the water from the land, theorically in order to protect the second from the invasions of the first, but practically often obtaining the opposite result. [fig 5]

Fig. 5 Hard and soft boundaries.

2.2 Quantitative aspects
On the basis of the explained conceptual models it is possible to identify relevant dimensions in the study of the coast as a boundary; they are (i) the length of the border and, from it, the degree of curvilinearity and the fractal dimension; (ii) the width of the edges; (iii) the height or differential between adjacent ecosystems; (iv) and the rhythm of change. The study of these measure appear relevant for coastal database modelling and GIS-based data handling.
Measure the length of a boundary such as the coast is a thorny task. First of all we have to decide which is the border line to measure. In fact, as a boundary (or set of boundaries), the coastline changes continuously as it is battered by forces on opposing sides (waves, tides, epicyclical sea level fluctuations, etc.), therefore it is virtually impossible to establish its absolute position at any given point in time. Instead established practice is to be use an approximation to this line, generalized at a certain spatial and temporal scale, choosing from a number of potential candidate "coastlines", including the mean sea level line (MLS), the higher/lower equinoctal water tides line (HAT/LAT), the mean high/low water tide line (MHWT/MLWT) and a lot of other recognised water levels.

Fig. 6 Coastlines data overlay in central Apulia and south Ireland.

A simple exercise of overlay of some different coast lines, mapped along Cork Harbour in Ireland and the north coast of Bari in Italy, clearly show how different cartographers have drawn different coastlines, and how this divergence of approach leads to greater potential error and uncertainty (Bartlett and Bruce, 2002) on an Atlantic coast, that is impacted by wide tidal fluctuations (Ireland), than on a Mediterranean coast, affected by smaller tidal fluctuations. [fig. 6]
The difference between these lines depends also on the different drawing scale: it is impossible to measure the absolute length of curvilinear boundaries, it is possible only in the case of straight ones. Several tourist guides, simplicisticly, state the length of a country or a region coastline as if it was an absolute measure. In reality, the pioneering studies of fractals by Maldebrot (1967) established that if the scale of measurement of the coast is progressively increased the length obtained for the coastline concerned increases progressively towards infinity as ever-increasing levels of detail are considered. Thus the measured length of a section of coast depends on the length of the ruler or the scale of measurement (Maldebrot 1983). A 1 km long ruler measures only major coves and peninsulas, a 1-meter stick, measuring each shoreline rock, results in a longer coastline, and a 1mm ruler that traces around each sand grain produces a longer coastline still (Forman 1995). [fig. 7]

Fig 7. Increase in coastline length according to measurement scale.

Tab. 2 Coastline length of European member staes according to CORINE Coastal Erosion database (scale 1:100.000 and scale 1:1.000.000).

1:100.000 1:milion
Belgium 97.669,478 71.841,586
Germany 1.863.759,656 1.585.966,249
Denmark 4.488.425,404 3.779.641,433
Spain 6.566.522,153 5.305.636,629
France 7.205.180,882 4.738.704,125
Greece 5.332.934,330 4.596.365,770
Ireland 5.147.530,109 3.841.864,583
Italy 7.408.792,354 6.187.972,030
Netherlands 861.005,147 761.169,666
Portugal 923.551,536 843.490,142
U. K. 15.910.977,428 11.697.812,207
Tot. 55.806.348,480 43.410.464,420

For instance the CORINE Coastal Erosion database at 1:100.000 scale measures the European coastline as more than 55.000 km long, while the same data generalized at the scale of 1:1.000.000, measure a total European coast length of only 43.000 km (http://dataservice.eea.eu.int/ dataservice/metadetails.asp?id=236 ), [Tab. 2]
Moreover, the mathematical proprieties of any segment of a fractal curve are similar to those of the fractal curve as a whole; that is, the statistical distribution of their projections and embayments are the same, irrespective of the scale at which they are measured (Kapraff 1986). Similarly the coastline is composed of coves and lobes at any scale and, although they are not identical copies at each scale, nevertheless it can be considered as a clear example of a fractal curve. From the measure of the length, we can derive the measure of the degree of curvilinearity or "boundary density" that is the total length of boundaries per unit area (Patton 1975, Taylor 1977, Gardner et al. 1991). This dimension can be calculated through the "space filling property". According to this parameter a straight coastline has a dimension D=1, and D increases progressively with the squiggliness of the boundary. When the boundary is so convoluted that it virtually covers the area, D approaches 2, the fractal dimension of a surface. [fig 8]

Fig 8. Lines with fractal dimention D= 1, D=1.5, D=2 and coast stretck appraching such measures.

The width of coastal edges -that is the width of the coastal zone as a whole- is the second relevant dimension. In fact despite the first-sight appearance of the coast as an essentially linear element, in practice it also has a width: maritime influences are felt a considerable distance inland, while terrestrial influences may be felt far off-shore (Bartlett 1993).This width of impact changes considerably from coast to coast. Mininni (2001) for instance points out the major width of the Ionian coast compared to the Adriatic one along the Apulian peninsula. Generally coasts that are more exposed to the kinetic energy of wind, waves and currents have a wider edge, and sandy coasts have an wider edge than rocky ones. Human actions also influence this dimension: by cutting coastal woodlands, building coastal defences and breakwaters, etc. humankind has over many centuries altered the width of coastal areas.
As with edge length, edge width also depends on how it is measured. Landscape Ecology suggests measuring this width by choosing a variable of interest and measuring its properties at intervals along a transect running from the coastline towards the interior (Brandt & Agger 1984).
The coastal zone width (or penetration distance) extends from the coastline border to the point on a transect at which there is no significant change in the variable as we proceed further inland. Connecting these points determined along several transects, it is possible to define coastal zone extent. However, when this method applied taking into consideration different variables (i.e. salinity, emerged land and seabed vegetation cover, population density, species richness, etc.) and using different interval widths along the transects (scale factor), it leads towards very different results.
For instance measuring coastal width according to dispersed house density along a stretch of the central Apulian coast of Italy shows how, proceeding from the sea inland, the density first increases, reaches a maximum value some 100 meters inland, steeply decreases for 4 km and then continues to decrease less steeply in the following 30 km. In this case the choice of a measurement interval that is either too big or too small might influence whether the critical distances of 100 metres and 4 km are correctly identified (Lamacchia 2001b) [fig. 9].

Fig 9 Changes in spreading housing density according to the distance from the coastline, in Central Apulia. [Lamacchia et al 2001, modified]

A similar method applied within the project LaCoast (Land cover changes in Coastal zones) based on the observation of change in Corine land cover percentages, has led to the suggestion that the European coastal area may be defined as having a width of 10 km (Perdigão and Christensen 2000). [fig. 10]

Fig 10 Percentage of surface occupied by different land cover class (Corine Land Cover -first level) according to the distance from the coast.

The height of a boundary or differential between two adjacent ecosystems is the third relevant dimension. It depends on the three-dimensional structure of the border, or on the diversity between the two adjacent ecosystems. In the case of the sea-land boundary, the separation between submerged and emerged land is the condition for the existence of the coastline itself. The steepness of this difference in level also influences the width of the coast. Along rocky coasts this difference in level is concentrated near the coastline, while along sandy coasts it is distributed over a wider area.
The fourth relevant dimension is the speed or rhythm of change. Constant cross and long-shore movements of matter, energy and information at all spatial and temporal scales characterize any coastal zone. As argued before, the absolute position of the land-water interface is difficult to determine: it is constantly changing and, furthermore, each small section changes its position independently of its neighbour, so that one wave may be advancing up the shore at one point, while a neighbouring one is in the process of receding temporally. Superimposed upon these small-scale changes are larger-scale ones such as the daily tidal cycle, annual sediment cycle or the epoch-making sea level rise and fall (Bartlett 1993). Since these changes operate on very different scales both spatially and temporally, it has also been suggested that temporal aspects of these changes exhibit aspects of self-similarity and fractality (Bartlett 1993).
The space-time principle states that most short-duration changes affect a small area, and most long term changes affect a large area: small ripples along a sandy shore last minutes, while big headlands, bays, estuaries, need centuries to take shape (Viles and Spencer 1995). And this in turn suggests the possibility to measure the temporal fractal dimension of a coast as its variation per unit time.

3. Pitfalls in Representing the Coast as a Boundary

3.1 Current coastal data representation models
The complexity implied by the conceptual model of the coast as boundary is not easy to implement in Geographical Information Systems by means of current coastal data models, and it is in performing thorny but fundamental tasks as this that Coastal GIS show their main limits and pitfalls.
Despite general awareness that the coastal system has width, many coastal GIS applications adopt data models based on a simple one-dimensional line, both in complete isolation from the other phenomena (one-dimensional abstract model) and located in multi-dimensional feature space (one-dimensional topologic model), reducing the complexity of the coastal boundary to its borderline, the coastline. In the first case length is the only geometric property stored, in the second case convolution and meandering of the line are regarded as significant and stored, but both models are in practice extremely limited for coastal data handling. [fig.11]

Fig. 11 one-dimentionalabstract data model and one-dimentional topologic model of the Cork harbour and Bari distrect coastlines. [elaboration based on theGISCO-Corine Coastal Erosion Inventory database]

According to Bartlett (1993) these models raise and leave unsolved some issue such as: (i) which line has to be selected to represent the shore among the candidate ones (i.e. coastline, backshoreline, foreshoreline, shoreline, etc.); (ii) how to determine the appropriate level of detail for this line, since fractal characteristics of the coastline make it inherently impossible to include every detail, and many coastal management problems require data and concepts at the synoptic level to be integrated with those that match more site-specific scales of resolution (Bartlett et al. 1992); (iii) how to efficiently record variations in attributes along the line, allowing layer integration and avoiding duplication and redundancy of data; (iv) how to integrate data recorded at positions some distance removed from the nominal coastline, especially in the case of heavily intended coasts where it is not possible to unequivocally extrapolate their position to the nominal coastline.
An example of one-dimentional coastal GIS application is the GISCO Database set up by the European Comunity within the CORINE Coastal Erosion Project.
A second type of data model most frequently encountered in coastal GIS applications is the bidimentional model of data representation, both in the exact objects version (vectorial model) and in the continuous fields one (tesseral or raster model).
Both versions leave unsolved the issue of fixing the appropriate area to take into consideration, which is the issue of coastal width assessment. Moreover, each version presents some strengths and weaknesses. In the first version (the vector model) the focus in on boundaries and the interior space bounded by these lines is implied; in the second version (the raster model), by contrast, the primary emphasis is on areas and the boundaries between areas are implied (Berry 1987). The exact object model tends to be seen as more appropriate for representing landscape mosaics, while the continuous field model is preferred for representing landscapes characterized by gradients, such as most of coastal landscapes
An example of bi-dimensional GIS application is the dataset produced by the European Community within the LaCoast (Land cover changes in Coastal zones) project; in this case using as baseline the nominal line chosen for the one-dimensional GISCO database, a coastal strip 10 km deep has been represented (Perdigão and Christensen 2000).
Other types of data model that might be harnessed for coastal GIS applications include three-dimensional and four-dimensional (spatio-temporal) ones (for discussion see Wright and Bartlett, 2000, and especially the chapters by Li, Raper and Varma). Although the coast has a relevant vertical structure, it is very difficult to handle the third spatial dimension in coastal GIS, both because of shortcomings in GIS software, which commonly handles the third dimension as an attribute rather than a real spatial dimension, and because of the shortage of sufficiently-detailed elevation data relating to most parts of the world's coastal zones.

Fig 12 Land cover changes between 1975 and 1990, along Cork coast as presented in the LaCoast (Land cover changes in COASTal zones) project. [Perdigão e Christensen 2000].

Similarlys most current GIS software lacks appropriate methods of representing temporal changes, and coastal GIS applications are severely handicapped and restricted in representing shoreline dynamics. The LaCoast project aimed to assess European coastal landscape changes between 1975 and 1990, for instance, but was based on a bi-dimensional data model where "changes" are regarded as attributes. [fig. 12]

3.2 Indeterminate data models for indeterminate boundary objects
Since the coastal zone is a boundary region dominated by poorly-determined inherently fuzzy boundaries and relatively imprecise demarcations, all the data models analyzed till now show a general limitation: they are designed to deal almost exclusively with static, well-defined 'objects', exactly located in space with properly defined attributes (Burrough and Frank 1996).
At the present time almost all GIS depend on hard Boolean logic and sharply delineated entities and relationships, since investigation and development of alternative data models and spatial data processing techniques are still at a theoric stage with few experimental applications. In contrast, inexact geographical objects, such as coastal landscapes, require inexact data models (Burrough 1996) based on statistical probability and fuzzy reasoning.

Fig. 13 Fuzziness of Cork harbour coastline, resulting from the overlay of different databases of the area.

Frank (1996) attempts to address this question staring from the point of view of experimental realism (Couclelis 1992). According to this theory there are two different environments for our experience, namely small-scale and geographic or large-scale space. Our daily experience with handling small objects in small-scale space (i.e. apples, stones) leads to a conceptualisation of objects with well-defined boundaries. This contrasts with the direct experience of large-scale space, such as coastal landscape, which leads to a different conceptual structure of space that is obtained for the most par without dividing large-scale space into delimited objects. He suggests thatwhen we handle geographical large-scale elements - such as during the construction of a coastal database - our experience with small prototypical objects with well-determined boundaries is metaphorically translated to conceive of the large scale space situation, typically by creating conceptual objects with more or less clear boundaries in the landscape. This process of metaphorical translation of small-scale object characteristics to large-scale ones is what commonly happens when a landscape has to be represented in a cartographic/pictorial space, or in computer storage, and is a main reason for the prevalence of objects with sharp boundaries in coastal Gis.

Conclusions
GIS offers very powerful tools for coastal boundary detection and identification. At the same time, however, many significant shortcomings remain with regard to representing and, to a certain degree visualising, the boundary nature of the coast. Without wishing to diminish the many achievements and advances in coastal GIS of the past decade or more, it remains the case that new conceptual approaches, probably linked to new data structures, are still required in order to more faithfully and effectively represent coastal systems in computer storage.

References