Ch2 The Nature of the Landscape

CHAPTER 2

2.1 Introduction

This chapter explores the nature of landscapes by addressing the definition of the term "landscape", discussing landscape planning issues, reviewing landscape assessment research, and identifying important characteristics of perceived landscape. This serves several purposes. First, it provides a rationale for this study by arguing that landscape is a resource with values that need to be reconciled along with other landuses. It demonstrates the need for landscape research and highlights the importance of classification. Lastly, the identification of important characteristics of landscape enables the establishment of a set of criteria that a landscape classification should incorporate.

2.2 Definition

"Landscape as a concept is bedevilled by semantic differences, misunderstanding, and controversies" (Countryside Commission for Scotland, 1970, p.1). Despite this, three common interpretations of the word landscape can be deciphered. Landscape Ecologists and some Landscape Architects use the word as if it is synonymous with the word "environment". In this context it has been defined as "the total spatial and visual entity of human living space, integrating the geosphere with the biosphere and the noospheric man-made artifacts" (Naveh and Lieberman, 1994, p.4). Fabos (1979, p.4) defined this understanding of the term landscape as "a homogeneous segment of the environment (including the surface of the land, the air, and all useful resources) which support all living creatures."

Within physical geography, "landscape" has often been used in relation to the physiographic, geological, and geomorphological features of the earth's crust (Naveh and Lieberman, 1994). In this context the word "landform" or "topography" would be more exact and is used by most physical geographers.

The third meaning of "landscape" is the "environment perceived, especially visually perceived" (Appleton, 1980, p.14), or the appearance of the land (Swaffield, 1991). The Countryside Commission (1970) used the phrase "the spectacle presented by the countryside"(p.2). This is the meaning intended in this thesis. More precisely, it can be defined as the overall impression obtained from viewing the land (environment) from a reasonable distance. Land includes the flora, fauna, cultural developments, surface soil and rock, landform, and water, but not all these can be perceived from a distance. Often this perceptual connotation of landscape is called scenery. The term landscape used in this way can also be found in early literature. A very early reference is the Book of Psalms, where it is used in reference to the beautiful overall view of Jerusalem (Naveh and Lieberman, 1994). The nature of perceived landscapes is discussed in detail in section 0.

2.2.1 Natural versus cultural landscapes

Within landscape assessment, it is common for natural and cultural landscapes to be treated separately (Auckland Regional Authority 1984). Cultural landscapes incorporate human modification and heritage links (Jones, 1991), while natural landscapes focus more on natural components. Although it is very difficult to completely separate the two, as they are very much interlinked, certain landscapes are heavily modified, such as urban areas, while others are not. With natural landscapes, landform and vegetation are important (a point that will be discussed in more depth in section 0), while with urban landscapes, the architectural style and layout are important. It can perhaps be said that with natural landscapes people are orientated towards the attractiveness of the existing landscape, while with urban landscapes people look to see what improvement or renovations can be made (Auckland Regional Authority 1984). Although this study is concerned mainly with natural landscapes, there is an interface between natural landscapes and cultural landscapes, albeit not very well defined, which is important. The degree of human modification in natural landscapes affects the character of natural landscapes. Therefore, the influence of cultural landscapes will be considered in this study by means of a classification of naturalness. The detailed composition and classification of cultural landscapes will not be a part of this study.

2.3 Landscape as a resource and a landuse

Landscapes have been considered a resource by many authors, such as Cloke and Park (1985), and Mitchell (1993). They can be seen as a resource within the classic framework proposed by Zimmermann (1951) which provides a functional interpretation of resources as relevant today as when it was first proposed in 1951 (Mitchell, 1993). Zimmermann argued that parts of the environment are not a resource until they can satisfy human needs and are therefore valued. For example, coal was not a resource until people found utility for it and wanted it. With this interpretation the landscape is a resource since people value landscapes.

It is useful to divide landscape values into three categories - economic, environmental, and ethical (Jackman, 1988). The economic value system is perhaps the most widely recognised, but could be argued to be the least important. There can be no doubt that landscapes have considerable economic value. The obvious example of this value can be seen in the tourism industry. New Zealand's tourism industry is totally dependent on its landscape because it is this that attracts the tourist. Currently, international and domestic tourism in New Zealand is an $8.2 billion per year industry (Statistics New Zealand, 1994). The economic value of landscapes may also be realised in other ways, for instance from the flow of wealthy immigrants attracted partly by the quality of New Zealand's landscapes, or the country's "clean green" marketing image.

The environmental value system is associated with the quality of the environment and the quality of life. Although difficult to measure, this quality of life value is particularly significant because landscapes are all around us and are experienced daily. It is possibly of higher value than the economic value. How happy would New Zealanders be if New Zealand was totally flat, urban, unforested, and landlocked?

The ethical value system is defined as the expression of the culture of the people. It includes society's spiritual or religious beliefs, and may include the perceived relationship that a culture may have with the land. Land ethics, such as the right to own land, and concepts like sustainability are all part of this system. With landscapes, the ethical values can be significant, as demonstrated through experiential research such as Hay (1990). People over time often develop a cultural and spiritual bond with landscapes that they have become familiar with, either through work, leisure, or home environment.

Zimmermann explains that resources are dynamic because they become available to people through a combination of increased knowledge, expanding technology, and changing individual and societal objectives. This dynamic is evident with landscapes. There has been a growing awareness of the significance of the aesthetic value of different landscapes. This is evident by the growing number of amenity groups, preservation societies, and general environmental lobbyists (Lowe, 1977). This awareness includes all values - economic, environmental, and ethical, although important documents such as the "Brundtland Report" and "Agenda 21", and legislation such as the Resource Management Act do put emphasis on environmental and ethical values. Not only do landscape characteristics change over time, but "the way humans view and value landscapes changes over time. Therefore, the human "measuring instrument" for observing landscape change is not fixed" (Cary, 1995, p.1). The implication of this for landscape classification is that the classification must be flexible - a criterion given in section 0.

2.4 Landscape planning issues

The impact on the landscape is now a major problem for many development initiatives. The planning process for the construction of the new Bealey Hotel, near Arthur's Pass in New Zealand, took four years to get the necessary planning consent (Brabyn, 1991). The main issue was whether it was preferable to have sporadically developed landscapes or to have development intensified in particular locations, like towns or cities, and leave the rest of the landscape in a relatively natural state. Collier (1991) identified that the impact on the landscape was now the most controversial environmental impact of tourism development in New Zealand. This controversy is likely to continue. Kearsley and Gray (1993) in their review of infrastructure requirements to meet the demands of the increasing tourism industry in New Zealand, drew attention to several new road links that may be needed, such as a direct link between Queenstown and the Milford road. The impact on the landscape is also a problem for the forestry industry. This is evident by the planning required for establishing plantations in the Mackenzie Basin (Boffa Miskell, 1993) and work on the public perception of forestry operations (Kilvert and Hartsough, 1993). This landscape issue is not confined to New Zealand but is a major global problem. Large amounts of landscape planning and research have been undertaken in Great Britain (Countryside Commission, 1988), and in the United States (Itami, 1989) for various local and national government organisations.

Landscape planning issues can be seen as a conflict of landuses, which can also be interpreted as a conflict of values, and a conflict of scales. Landscape is inextricable linked with other landuses, such as preservation, forestry, and farming, because these landuses are a part of the perceived environment. Such landuses can add or detract to the quality of the landscape, depending on whose perception is considered, the landscape context, spatial context, and the landuse in question (Amedeo et al., 1989). Such compromises can be within a particular value system, or between different value systems. The economic gain from commercial forestry could compromise the economic gain from tourist viewing the landscape, similarly it could affect the quality of life and spiritual fulfilment the landscape offers. Landscape perception can, as a landuse itself, cause conflict because landscape perception through tourism can provide economic value, but the facilities to provide for this, such as roads and viewing towers, can generate costs economically, environmentally, and ethically. The values provided by the landscape and different landuses can vary with scale resulting in conflict being even more difficult to resolve. On the west coast of the South Island, the locals may value the indigenous forests more for their timber than their aesthetics. While on a national scale, the forests may be more valued for their aesthetics. The question that then needs to be addressed is, whose values are more important? Whatever the answer, decision making will be more informed if information on the importance of landscape values at all scales is known. Therefore landscape research needs to be conducted at all scales, with appropriate levels of generalisation.

The landscape and spatial context are also an important consideration. With the Bealey Hotel example previously given, the landscape context was mountainous and reasonably natural as it was close to Arthur's Pass National Park. If the landscape context had been different, such as the Canterbury Plains which are flat and developed, it is likely that there would have been considerably less controversy. The Bealey Hotel is also located next to a busy road frequented often by tourists. This spatial context means that the hotel is highly visible. If the hotel had been located in another spatial context that was less visited, this too could have reduced the controversy. However, these alternate locations would probably be unacceptable to the tourism industry.

Throughout the world, planning agencies have been forced to consider landscape values because of statutory laws. In the United States there is the "National Environmental Policy Act 1962", in Norway there is the "Nature and Conservation Act 1970", and in New Zealand there is the Reserves Act (1977), which establishes provisions for the "preservation of representative samples of all classes of natural ecosystems and landscapes..." (section 3 (1) (b)), the Conservation Act, 1987, which gives power to the Department of Conservation to advocate conservation, and the Resource Management Act (RMA), 1991.

In the RMA, landscape values can be considered under the general umbrella of environmental values, which are provided for throughout the Act. However, in several sections specific reference is given to landscape values. For example, in Section 6 - matters of national importance, resource planners need to provide for:

"6(a) The preservation of the natural character of the coastal environment", and

"6(b) The protection of outstanding natural features and landscapes from inappropriate subdivision, use, and development."

Under section 7(c), particular regard to "the maintenance and enhancement of amenity values" is also required.

2.4.1 Landuse information modelling - the path to conflict resolution

To satisfy the requirements under sections 6, and 7 of the RMA, a comprehensive landscape assessment programme is required. The Resource Management Act, Section 35 -Duty to gather information, monitor, and keep records, makes this explicitly clear. It requires that:

"(2) Every local authority shall monitor -

(a) The state of the whole or any part of the environment of its region or district to the extent that is appropriate to enable the local authority to effectively carry out its functions under this Act;

(b) The suitability and effectiveness of any policy statement or plan for its region or district."

An example of a response to these statutory requirements is the Canterbury Regional Council's (1995) policy regarding landscapes. This is to "protect landscapes' aesthetic values" (p.94), which they intend to do this through information provision and monitoring of trends. A landscape study has already been completed (Canterbury Regional Council, 1993).

The Department of Statistics (now Statistics New Zealand) and Ministry for the Environment (1990) have pushed the concept of State of the Environment Reporting (SER). SER is defined as "the systematic analysis, description, and presentation of credible, scientifically based information on environmental conditions and trends, and their significance to human activity and its effects on the biosphere" (p.12). The main product from this is a national State of the Environment report. Several countries now produce regularly such reports, notably Canada and the Netherlands. The OECD reports five-yearly (OECD, 1991). By comparing SERs over a period, trends can be identified.

To implement SER in NZ, environmental monitoring is required (Ward, 1991). Ward and Beanland (1992) have consequently determined a list of appropriate environmental indicators to be used for monitoring the environment. The affect on aesthetics is listed as an issue, but no indicator is suggested. As will be discussed in section 0, public preference can be used as an indicator of landscape quality using psychophysical assessment. It is important that the indicator used is standardised so that national reports can be aggregated from regional reports.

The United Nations, through its environmental programme, is developing a Global Environmental Monitoring System (GEMS). One of the most important tasks is the harmonization of environmental data so that national and global assessments can be implemented. The data are being stored in digital form and the intention is to establish a central global resource information database (GRID) that can be accessed from every country through computer networks (United Nations Environmental Programme, 1990).

Jackman (1988) proposed a comprehensive framework on which landuse information could be structured for decision makers (refer figure 2.1). The framework groups landuse into ten components. Landscape fits in the category "humans as users". For each component, economic, ecological, and ethical values are determined for five different scales ranging from national to site. It is recognized that each component and value system is connected, but separation is required to determine the linkages between them. From this parametric approach to planning, tradeoffs between different landuses, value systems, and scales can be assessed enabling planners to be proactive, rather than reactive to planning issues. The development of such a model may seem a formidable task, but computers, in particular GIS, are providing useful support for capture, storing, analysing, and retrieving such information. It is a model that planners can work towards for identifying information deficiencies. The Canterbury Regional Council, for example, has a GIS that contains information on most of the ten components. However, information discerning the different value systems at different scales still needs to be developed. Over time, research and the development of planning tools will enable planning authorities to use increasingly sophisticated models of landuse that one day may approach the model proposed by Jackman. The computer aided study for optimizing the location of transmission lines in New Zealand (Electricity Corporation of New Zealand Ltd, 1988) is a good example of a study that approaches this model.

For such parametric models of landuse to work, information on all the different landuses is needed. Jackman (1988) showed that there were major deficiencies in information on landscape values in New Zealand. Compared to the amount of research done on other landuses there is very little assessment of landscapes. Yet landscapes may be of equal or more value to society and more vulnerable than those other landuses. The benefits of research on landscapes may be just as productive for enhancing or maintaining value to society as research in other landuses. This leads to the question: what does landscape research involve?

2.5 Landscape research

Zube, Sell and Taylor (1982) derived a landscape research framework after reviewing 160 landscape articles, covering 20 different research journals. They identified four different research paradigms. These have been labelled expert, psychophysical, cognitive, and experiential. The expert and psychophysical paradigms are concerned with applied landscape assessment and seek to determine whether landscapes are significant or beautiful. They are strongly motivated by the pragmatic concerns of resource planners. The cognitive and experiential are more concerned with theoretical issues, such as the nature of landscapes, why people have preferences for particular landscapes, and the meaning people attach to particular landscapes.

Daniel and Vining (1983) developed a similar framework using five groups - ecological, formal aesthetic, psychophysical, psychological, and phenomenological. The main difference between these frameworks is that the latter has an extra paradigm called "ecological", which has been included within the expert paradigm in

Zube, Sell and Taylor's framework. Otherwise, both frameworks are very similar, except the labels used. The frameworks have been accepted by landscape researchers as a valuable frame of reference for assessing different approaches to landscape assessment and research, and are often cited, for example Dearden (1889).

Steinitz (1993) also provides a useful research framework for addressing landscape issues. With his 25 years experience working with GIS and landscape planning he realized that there was an "overwhelming (and perhaps necessary) structural similarity among the questions asked by and of landscape planners and other environmental design professionals" (p.42). His proposed framework consists of six questions:

1. How should the state of the landscape be described: in context, space and, time?

2. How does the landscape operate? What are the functional and structural relationships among its elements?

3. Is the current landscape functioning well?

4. How might the landscape be altered: by what actions, where and when?

5. What predictable differences might the changes cause?

6. Should the landscape be changed? How is a comparative evaluation of the impacts of the alternative changes to be made?

These questions summarise the breadth of landscape research that is being implemented. The first two questions are more theoretical and concerned with the nature of landscapes. While the remaining questions are more concerned with landscape assessment, which is discussed in the following section. The nature of landscapes is discussed in section 0.

2.5.1 Landscape assessment

There has been much discussion on methodological issues regarding landscape assessment. In particular, what is the preferred approach - the expert or the psychophysical? The main factor that distinguishes the expert approach is that it is based on the judgement of experts who have been trained in the field of aesthetics. There are two kinds of experts, one that has had training in the field of fine art perspectives, and the other expert who is more ecologically orientated.

The fine art expert uses the formal artistic properties, which have been defined as form, lines, colours, and textures (Bennett, 1985). From these basic elements, landscape architects determine what are called the principle determinants of landscape quality. Daniel and Vining (1983) have called these variety, harmony, unity, and contrast. These determinants vary with different experts. Boffa Miskell (1993) and Ministry of Works and Development (1987) used vividness, intactness, and coherence to determine beauty. In the Boffa Miskell study of the Mackenzie Basin "vividness" was defined as the memorability of the visual expression received from contrasting landscape elements. "Intactness" was defined as the integrity of visual order in the natural and man-built landscape, and the extent to which the landscape is free from visual encroachment. "Coherence" was referred to as the way the landscape "hangs together" and is explicable, particularly in terms of its natural formative processes. Even when definitions are given, as with the Boffa Miskell study, the definitions of such criteria are still ambiguous. What is meant by "hang together" or "integrity of visual order"?

The ecological expert assumes that particular attributes of the landscape, such as naturalness and mountains, are of high value and uses the presence or absence of these attributes to judge beauty. Linton (1970) exemplifies such an approach. These are bold assumptions that are questionable, although they may be proven correct with cognitive research using preference surveys. These attributes are reasonably objectively defined, using parameters such as relative relief, presence of roads, and particular vegetation. They are also familiar concepts to the general public.

The psychophysical approach is based on the work of classical psychophysics, which sought to establish quantitative relationships between physical features of environmental stimuli and human perceptual responses (Daniel and Vining, 1983). Landscape quality is determined by the public as opposed to experts. This approach typically involves two stages. The first stage is the classification of the landscape into reasonable homogeneous classes, which are usually based on aspects of landform, landcover, naturalness, and water (eg. The Auckland Regional Authority study, 1984). The second stage is the evaluation of quality of the different classes within the classification. This is done with a public preference survey and with the classification acting as a frame of reference. With public preference surveys, typically, samples of the public are questioned about how they rank each landscape class for its visual quality. The survey could require a sample of the public to visit landscape sites, but usually photographs of the classes are used as a surrogate, and these photographs are ranked. The psychophysical approach might identify a consensus within society, but it could also show that different groups in society value landscapes differently.

Both the expert approach and the psychophysical approach have their strengths and weaknesses. The expert approach sacrifices reliability (repeatability) for utility and is more concerned with getting the job done. It is the more practical approach but is considered invalid because it claims that quality is inherent in the landscape (Daniel and Vining, 1983). The psychophysical approach, on the other hand, claims that quality is related to both the landscape and the observer, which is consistent with landscape theory. The psychophysical approach achieves a high level of precision and consistency but at the expense of generality and resources. It is the more scientific method as it is more concerned with measurement that is free from the bias of the researcher. Daniel and Vining (1983, p.79) concluded that;

"...no other approach has come so close to meeting the criteria of the ideal assessment system".

The psychophysical approach attempts to combine the cognitive research on the subject (ie. the viewer) with the object (the physical landscape). For scholarly enquiry, factors such as reliability and validity are more important considerations than generality and utility. If GIS can be used to classify landscape character, it is hoped that the cost of the psychophysical approach will be substantially reduced, and that it can be applied readily to a wide range of landscapes.

2.6 The purpose of landscape classification

This study is principally concerned with developing a methodology to classify landscape character. This is an important task that requires specialization because of technical developments in information and its processing.

There is a distinction between landscape classification and landscape description. Classification groups objects into categories, while description does not (Countryside Commission, 1988). Description has traditionally been the common means of communicating about landscapes. It describes a particular landscape in a way that conveys a clear picture. This approach is commonly used by Landscape Architects for analysing a site for a proposed development, or by writers who try to evoke the character of particular landscape in literature. However, description can be an inefficient means of communicating about a resource. If landscapes are similar then description can be repetitious. Descriptions are also difficult to further analyse and evaluate.

Classification is important to science because it provides a frame of reference that enables different researchers to communicate their results effectively. It also helps order and structure what is known (Haines-Young and Petch, 1986). In fact, classification is an important part of cognition (Langridge, 1992). The importance of classification for landscape research is no exception (Countryside Commission, 1970). The psychophysical approach to landscape assessment, as discussed previously, is based on a landscape classification. Without a classification the approach would be of little use. Without a landscape classification, landscape researchers are unable to effectively communicate their discoveries, and as a result a body of theoretical knowledge will be slow to develop.

To derive a landscape classification one must first consider the specific purposes to which such a classification is to be used. Objects can be classified in many ways. Objects are assigned to classes according to the characteristics that they have in common, but even the simplest objects have many features that could be used for this purpose. A simple table can be classified by its size, colour, design, style, etc. The choice of classification criteria is related to purpose (Langridge, 1992).

It is intended that the landscape classification developed in this study be used for landscape research and planning. The information demanded by planners was discussed in section 0. This includes the relative qualities (values) of different landscapes and the tradeoff in landscape quality associated with landuse change. If it is quality that is the focus of planning, then the landscape classes within a classification must distinguish this. This is not saying that the classification should identify quality. It is hoped that by dividing the landscape into homogeneous character classes, then the quality is also being divided into homogeneous classes. This may or may not be the case. It is possible that someone may perceive differently the quality of two landscapes that are reasonably identical in character, but located in separate areas. For example, people may value the area where they live more than another area that is similar in character, but far away from where they live. However, it is fair to say that the perception of landscape quality is dominated by landscape character and that this should therefore form the basis of a landscape classification.

2.7 Manual landscape classification

Conventional methods for classifying landscape character have relied on manual techniques (eg. Auckland Regional Authority, 1984), whereby maps, photographs (aerial and ground based) and field observations are used. Usually the classification process has not been explicitly stated. This is because of the large amount of information that needs to be analysed. Practitioners have avoided using computational and quantifiable methods and have instead used subjective approaches that Densem (1980, p.8) calls, "gut reaction". Without the use of GIS and detailed databases this has really been the only feasible option.

The manual approach, if done properly by operating with strict definitions of attributes measured on maps or in the field, becomes incredibly labourious and tedious. For example, relative relief can be used to differentiate between a mountain and a hill, but relative relief is dependent on scale. So it is necessary to use relative relief within a certain area, say a grid cell, examine a topographic map of the whole study area with a grid template, and then within each grid cell calculate the relative relief. This is time consuming and may still give the wrong answer as a flat area on top of a mountain may occupy an entire grid cell and thus get classified as flat when it should be classified as mountainous. Possible solutions are to consider the neighbouring grids cells as well, but this would make the process even slower. Alternatively use a larger spaced grid, but then accuracy will be lost. As of yet, no universal classification of landscape character exists that has clearly defined definitions. This is probably because the technology to implement these definitions has not existed until recently.

Manual techniques instead have generally used ambiguous definition involving more intuitive methods, whereby, for example, a contour map is viewed and areas that appear to be mountainous are defined as mountainous and areas that appear hilly are defined as hilly, or areas that look natural defined as natural and areas that look forested defined as forest. The problem with such an approach is that it would be impossible for other researchers to replicate it exactly, as the definitions of the classes would not be known. The inconsistency that results would mean that two independent studies could not be compared. This means that such landscape classifications can only be used as a frame of reference for the particular study where they were developed. For example, if a landscape classification was completed for the North Island of New Zealand by one person and a classification of the South Island was completed by another person independently using different definitions, it would then not be possible to say, based solely from these two classifications, that the South Island is more mountainous than the North Island.

Even when all the attributes have been mapped for the study area, compositions of these attributes need to be considered. Manual overlays are often used but this can be a time consuming task with many problems. It is often necessary to weed out unnecessary classes, or rework the generalisation process to cut down on the number of classes. With the Auckland Regional Authority (ARA) (1984) classification, 85 classes were derived. Yet, with the combination of different attributes used, clearly more classes would have been identified. How this reworking was done was not reported. Again, explicitness was compromised for expediency.

Another problem with the manual approach has been the high costs involved in time and resources. If it is necessary to do field observations of every landscape then this is going to cost a considerable amount. Because of these high costs only sporadic areas have been classified in New Zealand.

Landscape classification requires identifying areal units or enclosed areas that aid analysis. With manual classification the options available for this have been (Robinson et al., 1976):

1. the character tract (usually based on macro landforms),

2. the viewpoint and its associated visual envelope (viewshed),

3. the grid square, and

4. the whole study area.

With GIS, a new areal analysis option is available that is not practical to implement manually. This is the focal neighbourhood function, which opens new possibilities for landscape classification, and will be discussed in detail in the following chapter.

With GIS and a computer automated approach, many problems encountered with manual classification can be considerably reduced. Computers force one to define exactly what procedures are being followed, there is a high degree of consistency between applications, different information layers can be easily integrated, the results can be easily displayed and used for further analysis, and all this can probably be done at less cost than the manual approach. A fuller discussion of this is given in the following chapter. It remains possible that the manual approach, by using field observations and intuitive methods, can capture more of the subtleties of landscapes than an automated GIS approach. Therefore, a comparison of the two approaches is required.

2.8 Landscape classification criteria

The classification of landscapes has many problems. One of the main problems is that researchers involved in landscape classification have originated from many different professions and do not classify landscapes from the perspective of the general public. The classifications they produced were often too detailed, especially regarding information relating to their original profession. This problem is also compounded by the fact that "landscape" is an ambiguous term and people have different interpretations of what should be in a landscape classification. As a result many classifications that claim to be landscape classification are not landscape classifications using the definition of landscape adopted in this thesis. The Protected Natural Areas (PNA) program is an example of this. Here the protection of representative landscapes was an objective and a landscape assessment was required (Myers, et al., 1987), but a detailed geomorphological and botanical classification was developed. The landscape classification produced by Canterbury Regional Council (1993) also falls far short of being a landscape classification, since a general land inventory was produced rather than a classification based on the appearance of the land. It is therefore necessary to establish criteria for a landscape classification. This should consist of general criteria that apply to all classifications, and specific criteria that apply only to landscape classification. Specific criteria will be listed in the following section after the important characteristics of landscape are discussed.

General classification criteria are listed below. They have been adapted from Rhind and Hudson (1980) who have used them as criteria for a landuse classification.

(i) The classes must be exhaustive and mutually exclusive, i.e. all geographical individuals must be classified, but no individual must fall into more than one class.

(ii) It has to be easily understood and applied.

(iii) It has to produce repeatable results that are independent of the researcher.

(iv) It has to be hierarchical, to cope with needs at different levels of resolution in different areas.

(v) It has to be sufficiently flexible for new interests and tasks to be met from a modified, rather than a completely new, classification.

(vi) It must incorporate some recognition of seasonal or other cyclical changes.

In order for a classification to be repeatable by different researchers, it is necessary that the classification be totally explicit. This does not necessarily require the process to be quantitative, however, this is usually the most efficient means of being explicit. If a classification process has been automated with computers then this criterion will be met.

Regarding criteria (vi), elements of seasonality can be incorporated into a landscape classification by ensuring that within a class the whole class changes similarly through the seasons. For example, low land plains and a highland plateau need to be in different classes as these areas change differently with the seasons because of the differences in altitude. A plateau may get covered in snow in the winter, while a lowland plain may not. This is generally the case in New Zealand but may not be the case in colder or warmer parts of the world. It is therefore necessary to restrict the domain of the classification being developed in this study to New Zealand.

Besides these general criteria a landscape classification must also incorporate the nature of landscapes. This is needed for a landscape classification to be useful for psychophysical research, and for the identification of quality - an important use of a classification. Haines-Young and Petch (1986) say that classification needs to be undertaken in the context of theory. This begs the question, what is the current theory behind the nature of landscapes?

2.9 The important characteristics of landscape

The definition of landscape in section 0 focuses the intent of this thesis, but leaves unanswered questions about the important characteristics of landscape. What are the important components of landscape, and what does viewing the land entail?

The environment is different to the appearance of the land (landscape). The environment contains entities that are not commonly perceived, while landscape is restricted to objects above a certain scale. The perception of landscape is dominated by the larger, more visible entities. Many components of the environment can be seen if one cares to look closely, for example small fauna such as lizards, snails, and hedgehogs. However, these cannot be seen from a reasonable distance, and would not contribute significantly to the overall spectacle of the landscape. The definition of landscape in this thesis excludes such visually insignificant components. If, however, there were a sufficient number of a particular small component that together could be seen from a distance, then this group could be part of the landscape. A blade of grass on its own is not a part of the landscape, but a paddock is likely to be. Landscape does not include small isolated patches of mosses, lichens, and small shrubs as they are unlikely to contribute significantly to the overall spectacle of the countryside. Often the surface rock and soil are not a part of the landscape as they are hidden from view, usually by vegetation, or buildings. So, for a landscape classification, it is not necessary to do a detailed assessment of soil, geology, entomology, grasses, or forest undercover. The notion that "landscapes are perceived from a distance" is useful for indicating the minimum degree of generalisation involved.

Is the weather, or the polluted air a part of the landscape? Or is the sun setting on the horizon a part of the landscape? Certainly, weather, air pollution, and the position of the sun affect the view of the land. However, they are highly variable components that are normally held constant in landscape studies, as in the ARA (1984) landscape study. Consequently, they are not generally included in a landscape classification, so will not be considered in this study.

To discuss the important characteristics of landscapes, it is necessary to have a means for conceptualising landscapes. As discussed in section 0, there are two lines of thought by experts about how landscape components can be conceptualised -either the fine arts or ecological perspective. If two areas are similar in physical components, then there is a high probability that they should be similar in terms of form, colour, lines and texture. It therefore should not make too much difference which approaches are used for identifying landscapes. However, there is likely to be a difference resulting from implementation. The major problem with using the formal artistic approach is the lack of definition about how these artistic principles can be defined. The approach is usually very intuitive and not explicit. This study uses physical components because they can be described explicitly. Common language also suggests that this is how most people conceptualise landscapes, for example mountainous, forested, and coastal. As will is discussed in this section, there has also been research to suggest which of these physical components are important for landscape classification.

The landscape is derived from an interaction between aspects of the environment (landscape components), and human perception processes. Landscape perception depends not only on these physical landscape components, which may be extremely diverse, interrelated, and complex, but also on the values, experience, and social-cultural conditioning of the observers (Dearden, 1989). Jones (1991) describes this complexity as the "elusive reality of landscape"(p. 229) and adds, that in the past, the lack of recognition that landscapes are both a physical reality and a social or cultural construct has led to an "academic battlefield", with different disciplines and schools concentrating on either the physical landscapes or on the observer.

Identification of the important characteristics of landscapes therefore depends on whose perspective is considered. "Beauty is in the eye of the beholder". One of the most important points to consider is whether the observer is attached or disengaged from the landscape (Bourassa, 1991). For example, an urbanite is likely to perceive a farm differently from the farmer, and a farmer is likely to perceive a city differently from an urbanite. Linked to this, is the observer's familiarity with the landscape. For example, a Geomorphologist who has had training in detecting subtleties in the landscape that give clues to the formation of landforms may perceive an interesting diversity in a landscape, while somebody who has not had this training may think that the same landscape is quite monotonous. According to Appleton (1975a), humans have innate preferences for particular landscapes that are common to us all. This has been suggested to be a savanna type landscape that offers both refuge and prospect, because this would have been an important consideration when humans lived as hunters and gatherers.

Since it is the public's perception of landscape quality that is important to planners, then it is the public's perception of landscape character that should be considered in a landscape classification, not the perceptions of specialized scientists (Zube, 1984a). Specialists are a small minority. The total value that they derive from landscapes would be far less than the total value that the general public derive from a landscape. It could also be argued that for many scientists it is not the landscape but the scientific information that is of value.

The difficult questions that then arise are, what is the general public, and what is the nature of their perception? The general public is not a homogeneous group, but as a notion it eliminates extreme perceptions and is therefore a useful notion. A classification should be based on the lowest common denominator within this group. For example, not all people perceive the difference between exotic and indigenous vegetation (international tourists may be included in this group), but many members of the public do, therefore the distinction between exotic and indigenous vegetation is necessary. This lowest common dominator level is not always obvious. Therefore, a classification should be hierarchical, in terms of detail, to cope with a range of different perceptions. Observer perspectives can also be considered using fuzzy set theory. This is discussed in section 5.4.2.

So what are the main components of landscapes from a general public's perspective? These are assumed to be landform, vegetation, naturalness, and water. This assumption is substantiated by a branch of cognitive landscape research, known as content category identification (Amedeo, Pit, and Zube, 1989), concerned with identifying important components of the landscape that explain differences in how the public perceive quality. It is generally understood that people when determining scenic quality, organize sets of landscape components into classes or categories. Categorization of landscapes is a mental process that proceeds when people actively discriminate among landscapes. How people categorize landscape is important information for determining the appropriate attributes to use for a landscape classification. An ultimate goal of content category identification research is to eventually delineate a landscape classification system resulting from public perceptual assessments, rather than professional judgements of scenic value (Amedeo, Pit, and Zube, 1989). Content category identification studies generally determine features perceived as having negative impacts on quality, and those that have positive impacts. This research typically involves using some form of statistical analysis, for example Q-sorting (Amedeo et al., 1989). Often preferences for different photographs of landscape are determined by asking samples of the public to rank them. Many attributes of the landscapes in the photographs are identified and quantified, and related to the preference of the photographs with some form of regression analysis. Pomeroy, FitzGibbon and Green (1989) used personal construct theory, the repertory grid, and multidimensional scaling to ascertain these attributes.

The Countryside Commission of Scotland (1988) reviewed research involved with identifying those physical attributes of the landscape that determine quality and presented a summary table (refer to Table 2.1). From this, there are some attributes that are consistently identified as determining quality. Man made structures have been identified in all the studies, especially as having a negative effect on quality. This suggests that the degree of naturalness be an important component. Vegetation is identified in five of the six studies, and water features, such as sea, lake and streams have been identified in four out of the six studies. Landform is, surprisingly, only identified in two of the studies. Zube, Sell and Taylor (1982) also identified relative relief, landuse diversity, water, and naturalness, as determinants of quality in their review of this type of research. A recent study by Amedeo et al. (1989) also identified aspects of vegetation, landuse, influence of water, and topography. In this thesis, landform, vegetation, naturalness, and water will be called landscape attributes. The characteristics of these attributes, such as forests, hills, lakes will be called components.

For New Zealand landscapes, there has been no content category research per se. Since landscape quality is specific to cultures, it may be unwise to apply the above results to New Zealand. However, the four attributes identified above have been used for classifying landscapes in Auckland (ARA, 1984), and have been used in various other landscape studies (Mosley, 1989, Kliskey and Kearsley, 1993, and Fairweather and Swaffield, 1994). The components also relate well to common language used in New Zealand to describe landscapes. It appears that these attributes have become fairly standard for describing landscapes.

Landscapes cannot be classified by simply dividing the land into areas that reflect boundaries between different landscape components, because it is these boundaries that are important characteristics of landscapes (Jackson, 1984, and Robinson et al., 1976). The interaction of components is sometimes more important than the components themselves (Arthur et al., 1977). An analysis of landscape paintings in the Canterbury region showed a clear preference by painters for landscapes that were compositions of plains and mountains (Canterbury Regional Council, 1993). A landscape classification therefore needs to incorporate this juxtaposition of boundaries between different components of the landscape in order to distinguish landscape quality. This consideration makes landscape classification particularly difficult, but if it is not included the classification is of little use for researching perceived landscapes. In one respect landscape is a generalisation of the environment, because only the large visual entities are perceived. However, when compositions of these generalised entities are considered, virtually millions of different combinations become apparent. The perceived environment therefore becomes very complex. There can be no doubt that the perceived environment (landscape) is different to the environment that Landscape Ecologists study. Completely different types of classifications are therefore required.

The geometrical perspective from which landscape components are perceived is also important in determining landscape quality (Higuchi, 1988). For example, a view of plains from a high point, such as the top of a hill, is very different from a view from a point that is at the same height as the plains. The field of view will obviously be greater from the higher vantage point. It has also been suggested that the geometric perspective of the observer is dynamic in the sense that landscape is viewed from a multiple of points, surrounds the observer, and is experienced from movement and exploration (Zube, Sell, and Taylor, 1982). For example, when someone looks from the top of a hill into a valley, they may have just spent the last hour driving through that valley. This experience will be with them when they view that valley from above and will affect the way that landscape is perceived. Many components that cannot be seen from the top of the hill, perhaps because of visual obstruction, but had been seen previously from exploration, will still be a part of the perceived landscape. Such memories may be visually disturbing, such as a rubbish dump, or perhaps more pleasing. When people view a landscape from a point, it is not just the area that is directly visible that gives an impression, but also peripheral information that has been previously experienced (Zube, Sell, and Taylor, 1982).

Landscapes have also been identified as being perceived through multiple senses, and not just sight (Zube, Sell, and Taylor, 1982). Porteous (1990) refers to the "smellscape" and "soundscape". However, sight is considered the dominant sense that provides information on landscapes. Porteous (1990) says that "it yields more than 80 percent of our knowledge of the external world" (p.4). It is necessary to simplify landscapes to visual information to enable the study of landscapes to be feasible. This simplification can be further justified by the assumption that the other sensory inputs are likely to vary consistently with sight.

The above discussion shows that there is some consensus on the important characteristics of landscapes. These should be incorporated in a list of landscape classification criteria. It should be noted that landscape character classification is concerned with the stimulus properties of the landscape, not the outcome of landscape perception, which may be meanings, actions, or values. The following criteria need to be adhered to in order to incorporate the important characteristics of landscapes:

1. The classification should incorporate landform, vegetation, naturalness, and water.

2. The classes should be based on the general public's perception of the above attributes.

3. The classes should be based on an overall impression of the above attributes in an area from a distance, and involve generalisation and composition.

4. The classes should recognize that landscapes surround and are experienced from a multiple of geometrical perspectives that can be obtained from movement and exploration.

The list of specific criteria may appear to be over simplified considering the complex nature of landscapes. Many questions are left unanswered - what is the exact nature of the landscape components and their relationship with each other, and what is the exact nature of the observer? If a greater number of more detailed specific criteria were stated, their validity would be questionable. This is because they would not have been substantiated by research and/or there would not be the consensus among researchers. More research is required. However, to do this effectively requires a landscape classification so that researchers can communicate their results. The dilemma is that it is necessary to know the important characteristics of landscapes to classify, yet to know the important characteristics of landscape it is necessary to classify. A classification needs to be developed with what information is available, and then reassessed as information is forthcoming. It is in this way that classifications evolve. Our understanding of landscapes is at a superficial level - "theoretical vacuum" (Appleton, 1975b, p.2), and needs to be based on many assumptions. Despite these criteria being an over simplification, if a classification can meet all of them, then significant advancement will have been made. As of yet, this has not been achieved using manual techniques.

The specific landscape classification criteria listed above and the general classification criteria given in section 0 focus the issues that this thesis will address. It is worthwhile stressing that the following considerations will not be addressed in this thesis:

1. Cultural landscapes per se;

2. Entities that are not highly visible such as underlying soils and geology, and micro landforms;

3. Highly variable factors such as the weather, atmospheric pollution, and the position of the sun;

4. Non visual senses such as sound, smell, taste, and touch.

The classification criteria will be used later in the thesis for discussing the validity of an automated landscape classification that will be developed. These criteria are only a minimum that a landscape classification must meet to be valid. For a landscape classification to be valid, it also needs to be verified by many independent researchers that have used it in research. Since the criteria include the need to incorporate seasonal variation, then a classification will only be valid for the particular climatic region that it was designed for. The classification being researched in this study is intended for New Zealand conditions. Although the components of each of the four attributes are not part of the criteria, suitable components will be defined and discussed in chapters 4 and 5.