new banner
 

Home
Services of Scenic Solutions
Completed landscape projects
Papers and Presentations
Beautiful landscapes
Australian and International studies of landscape quality
The science of scenery
Landscape miscellany
Landscape resources
About Dr Andrew Lothian
Contact us
Site map

AUSTRALIAN LANDSCAPE PREFERENCE STUDIES

In contrast to the physical studies, preference studies involve the community in assessing scenes and rating their scenic quality. Because they require people to assess the scenes, preference studies have been regarded as more resource intensive than physical studies. However with the availability of the Internet to gain community aesthetic preferences as well as digital cameras and computer based statistical programs, the major impediments against the use of preference studies have diminished significantly.  

Click on the following topics:

Characteristics of preference studies
Descriptions of preference studies

CHARACTERISTICS OF PREFERENCE STUDIES

Table 1 summarises characteristics of the 31 studies which assessed the community’s scenic quality preferences, including several which combined physical assessment with preferences.

Table 1 Characteristics of Preference Studies

Author
Region
Scenes
Sample size
Preference methods
Rating scale
Landscape dimensions
Multiple reg. R2
Map
Zube &
Mills, 1976
Lorne
Coast
24
101 Lorne
22 USA
Q sort
5 groups
No
Kane, 1976
South
Australia
5 regions
varied
10
Bi polar, Component
Checklist, Marker
Scenes, Factor equation
varied
Factor
Equation
18
-
No
Radford
& Bartlett,
1977
Lower
Hunter
Valley
15
164
Q sort
5 groups
15
Yes
Simson, 1977
Gold Coast
5
51
Semantic differential
5
4
-
No
Dare, 1978
Fleurieu
Peninsula
SA
90?
10
Landform
elements
scored
Elements
Impact (0-2)
multiplied
(-2 to +2) by
contribution
to quality
No
Williamson
& Chalmers,
1979
Bright
Forest
10
97
Ranking,
rating
unstated
Revell,
1981
Fleurieu
Peninsula SA
100
15
Rating
0 - 100
No
Williamson
& Chalmers,
1982
Ovens
Valley,
Victoria
1st
survey
10
2nd
survey
56
1st survey
95
2nd survey
253
1st survey rating,
ranking, list positive
& negativefeatures
2nd survey Q sort, rating, paired comparison, landscape dimensions
1 high –
7 low
29
80%
No
Lamb &
Purcell,
1982
Warringah
Shire,
Nth Sydney
180
97
Rating
0 - 10
-
No
Purcell
& Lamb,
1984
Narrow
Peninsula
Sydney
105
85
Rating
0 – 10
-
No
Correy,
1984
Sydney
Harbor
12
1st survey 40
2nd survey
100
1st survey identify
significantcomponents
2nd survey bi-polar, ranking
2nd survey
1 – 5
ranking
24
No
Lamb &
Purcell,
1990
Native
vegetation
NSW
71
81
Naturalness
score
Scale
rule
130 mm
7
-
No
Prineas & Allen, 1992
Wet Tropics
Qld
80
306
Rating
1 – 15
8
85%
Yes
Brouwer,
1994
Whitsunday
Qld
24
32
Rating, effect & scale of development
1 – 7
-
No
Purcell
et al,
1994
Sydney
Region +
Padua, Italy
24
(12 X 2)
192
(96 X 2)
Rating; Like to  live, work & visit
place
1 – 7
liking of
scene
-
No
Purcell & Lamb, 1998
Eastern
Australia
96
49
Rating
preference
0 – 100
-
No
Lothian,
2000
South
Australia
160
319
Rating
1 – 10
27
-
Yes
Preston,
2001a
Mogill,
Qld
52
210
Ranking, rating polar adjectives,
familiarity
1 - 10
110;
6 used in
model
72%
Yes
Preston,
2001b
Glen Rock,
Qld
21
60
Ranking, rating
evocative scoring,
familiarity
1 - 10
114;
7 used in
model
52%
Yes
Williams & Carey, 2002
SE
Australia
36
(b/w)
1232
Rating,
Interviews of 20%
1 like
- 5 dislike
5
57%
No
Davis,
2003
Adelaide
42
106
Rating. Ecological
integrity rated
1 – 10
No
Lothian,
2004
SA Tree
Amenity
112
454
Rating
1 – 10
9
53%
No
Lothian,
2005a
Coast
SA
138
2200
Rating
1 – 10
7
86%
Yes
Lothian,
2005b
SA Coastal
Development
82
1659
Rating
1 – 10
-
No
Lothian,
2005c
Barossa
Region SA
120
1210
Rating
1 – 10
7
54%
Yes
SEQRESA
Steering
Com; 2005
South
East Qld.
15000
1000
Public
preferences
survey
1 – 10
31
Yes
Raymond & Brown, 2006
Otway
Region,
Victoria
-
1400
12 landscape
values identified
-
12
81%
No
Wu et al,
2006
Mornington
Pen., Victoria
240
unstated
Rating
1 - 7
8
49%
Yes
Lothian,
2007a
R. Murray
SA
120
1673
Rating
1 – 10
8
81%
Yes
Lothian,
2007b
R. Murray
Development SA
80
1259
Bi polar
1 – 9
-
No
Lothian, 2009
Flinders Ranges, SA
127
2422
Rating
1 - 10
8
85%
Yes

The use of preference studies has accelerated: five each in the 1970s, 1980s and 1990s, and fourteen since 2000. The studies have been spread fairly equally between Queensland (6), New South Wales (7) and Victoria (6), but there have been eleven in South Australia, including eight by this author.

Number of scenes

Leaving aside the South East Queensland study which used 15,000 photographs, the mean number of photographs used in preference studies was 77 (SD 59). The minimum was 5, the maximum 240 (Figure 1).

Nos of scenes
Note: Omits SEQRESA Steering Committee, 2005, which used 15,000 photographs
Figure 1 Number of Scenes Used in Preference Studies

Apart from this author, only one other study used scenes from other locations to benchmark the scenes and provide a comparison. The other study was that of the Wet Tropics World Heritage Area in Queensland which used scenes from other WHA areas for comparison purposes (Prineas and Allen, 1992).

Sample size

The sample size of participants ranged from 10 to over 2400 with the mean 531 (SD 705). However this does not convey the whole story. Since the availability of the Internet, large samples have been made economically and practically possible. Figure 2 indicates that nine of the studies have used samples of 1000 and greater and all of these have occurred since 2002. Prior to this, the mean sample was only 127.

sample size

Figure 2 Sample Size of Preference Studies

Preference methods

The methods used to elicit scenic quality preferences of scenes were mainly rating and ranking instruments (Table 2). Ratings against a scale (e.g. 1 – 10) provide an absolute measure of scenic quality (interval scale), whereas rankings which compare one scene to another provide only a relative measure (ordinal scale). Only ratings from one study can be compared with ratings from another study; rankings cannot be compared. Some studies used more than one method.

Table 2 Preference Methods used in Preference Studies

Preference methods Frequency
Rating
22
Ranking & Q sort
8
Likert scale (bi polar)
5
Components checklist
4
Other
8
Total
47

Note: Ratings require the respondent to choose a number on the rating scale representing the scenic quality of the scene; rankings and Q sort require the respondent to place the scene in order relative to other scenes; bi-polar uses a like – dislike continuum; components checklist require the respondent to identify components in the scene. Other tools include scores of naturalness, ecological integrity and familiarity.

Rating scale

A range of rating scales was used with 1 – 10 being the most popular (Table 3). Odd numbered rating scales (e.g. 5, 7) enable the median to be an integer with equal number of ratings on either side. Adopting a common rating scale would better enable comparison of ratings between studies.

Table 3 Rating scales in preference studies

Rating scale Frequency
0 – 10
2
1 – 10
11
1 – 5
2
1 – 7
4
1 – 9
1
1 – 15
1
1 – 100
2
1 – 130
1
Total
24

Regarding the use of the zero as the baseline, it is difficult to conceptualise the appearance of a landscape of zero value – i.e. the complete absence of aesthetic appeal. A flat, featureless plain which might be regarded as of minimal scenic quality has some appeal as attested by papers on the Canadian prairies (Evernden, 1983) and Great Plains of the US (Cook & Cable, 1995). Even the Hay plain, the archetypal boring Australian landscape, has intrinsic visual appeal.

Go to top

Landscape characteristics

In seventeen studies, the landscape characteristics present in the landscape which may explain its aesthetic quality were evaluated. The landscape characteristics included naturalness, diversity, vegetation height and density, landform characteristics and the presence of water. The number of landscape characteristics varied from 5 to 114 (Figure 3) however, without the two studies with more than 100 dimensions, the average was 14. In these two studies with more than 100 characteristics (Preston, 2001a and b) the actual number of characteristics used in modeling was only six and seven.

landscape factors

Figure 3 Number of landscape characteristics in studies

The landscape characteristics used in the preference studies totaled 193. The majority of these, 86% were feature specific (e.g. trees, buildings, water) and 14% were generalized characteristics such as naturalness, diversity, harmony or human impacts. Vegetation, trees and forests dominated with water-related, buildings and structures, and agricultural features following (Table 4).

Table 4 Nature of landscape characteristics

Landscape characteristics Frequency
Vegetation, trees, forests
58
Water, sea, coast, rivers
35
Buildings, structures, roads
33
Agriculture
15
Terrain and landforms
7
Naturalness & diversity
6
Transitory elements
5
Formalist qualities
14
Viewpoint
20
Total
193
Note: Some items were double-counted

Table 5 details the variations found in the vegetation, water and artefact landscape characteristics. Vegetation was dominated by its physical characteristics (e.g. height, form, density) followed by type (e.g. eucalypt, alpine, pine). Health of vegetation was well covered.  Water characteristics were dominated by descriptions of the water bodies followed by their interface with the land. Buildings dominated as artefacts. Other urban uses included residences, commerce and industry.

Table 5 Vegetation, water and artefact landscape characteristics

Vegetation characteristics
Nos.
Water characteristics
Nos.
Artefact characteristics
Nos.
Vegetation characteristics
19
Water, sea, river
17
Buildings
11
Type of vegetation
15
Beach, shore, edge
9
Other urban uses
8
Health/regeneration/regrowth
9
Waves, movement
3
Other non urban uses
9
Forest
4
Other
6
Roads
2
Trees with roads
4
Total
35
Powerlines
2
Trees
3
 
Quarries
1
Vegetation
4
 
Total
33
Total
58
 
 

The main purpose of defining landscape characteristics by these studies was to enable the use of multiple linear regression analysis and other statistical tools to model the relationship between these dimensions and scenic quality. The correlation coefficient, R2, was used as the measure of how well the model explained the scenic quality.  An R2 of say, 0.8 explained 80% of the variance. Figure 4 compares the number of landscape characteristics and R2 and indicates there to be no relationship (y = 0.0003x + 0.67, R2 = 0.00001). Indeed a higher R2 was found by some studies using only six factors as studies which used twelve. The careful selection and scoring of factors is critical to achieve a high R2.

ls factors & R2
Note: Excludes Williamson & Chalmers, 1982 which used 29 dimensions and achieved an R2 of 80%.
Figure 4 Relationship between landscape characteristics and R2

Maps of landscape quality

Ten of the 29 preference studies produced a map of landscape quality, including four by this author. The first map was produced by Radford and Bartlett in 1977 of the Lower Hunter Valley. Queensland and South Australia scored most of the maps with four in each state (Table 6).

Table 6 Location of maps produced from preference studies

Author Region
Radford & Bartlett, 1977
Lower Hunter Valley, NSW
Prineas and Allen, 1992
Wet Tropics Qld
Lothian 2000
South Australia
Preston, 2001a
Mogill, Qld
Preston, 2001b
Glen Rock, Qld
Lothian, 2005
Coast, SA
Lothian, 2005
Barossa region, SA
SEQRESA Steering Committee, 2005
South East Queensland
Wu, et al, 2006
Mornington Peninsula, Victoria
Lothian, 2007
River Murray, SA
Lothian, 2009
Flinders Ranges. SA

Having examined the overall characteristics of preference studies, the following section details some of the studies.

Go to top

DESCRIPTIONS OF PREFERENCE STUDIES

An early Australian study using preferences was in 1976 in South Australia where the Nature Preservation Committee of the National Trust engaged Dr Phillip Kane from California to develop a method of landscape assessment (Kane, 1976, 1981). The study is detailed here as it illustrates the complexities that can be injected into an otherwise simple method in an attempt to rate scenes.

Kane developed and tested four methods:

  • bi-polar semantic differential scale – emotional responses to the whole scene
  • component checklist – emotional responses to parts of the scene
  • set of marker scenes – record and ‘objective” appraisal of the whole scene
  • equation which “objectively” appraises landscape components

 
Table 7 classifies the methods by landscape and appraisal.

Table 7 Summary chart of technique developed for the National Trust by Dr Phillip Kane, 1976

  Whole landscape Landscape components
Emotional response
BIPOLAR SCALES
(Semantic differentials)
Scoring: use of mean response of at least 10 people
COMPONENT CHECKLIST
Scoring: Use of mean response of at least 10 people
Application: must be used on-site
Visual appraisal
MARKER SCENES
Scoring: Use of mean response of at least 10 people
Application: best results with colour slides
FACTOR EQUATION
Scoring: use of equation
Application: measurements taken by an individual from slides or prints.

All four methods provided scores within range 0 – 100 (least – most attractive). Each view or site was evaluated by all four methods, providing an independent set of responses. The conversion of the evaluations using these tools to ratings relied on complex calculations which are summarized below.

Bipolar scores Used 14 adjective pairs and scored each out of 10. Averages for all respondents were multiplied by a weighting factor to produce a score which could range from 49 to 343, and which was divided by 3.42 to yield a score out of 100.

Components checklist Scored each component on a scale: -2 to +2 (bad to excellent).

  1. Total number of items checked by each observer (= N).
  2. Scores for each component added (= S).
  3. Count up total number of items scored +1 and +2 by each observer (= P).
  4. Checklist score for each observer (I) = 10(S)/N + √P where 10(S)/N is a measure of component quality, and the factor √P is a positive measure of landscape diversity.
  5. View’s total score (T) by summing individual scores (I) and then divided by number of observers (B). View score not reliable unless evaluated on-site by minimum 10 observers.
  6. Convert score T to base of 100 points.

Example:

  Observer
  1 2 3 4 5 6 7 8 9 10
1. N
15
17
12
23
19
15
13
16
13
18
2. S
21
20
16
15
29
21
16
19
21
33
3. P
14
15
14
16
18
14
11
13
11
17
4. I
17.7
15.7
17.0
10.5
19.5
17.7
15.6
15.5
19.4
22.4

7. Score T = ∑I/B = sum of Iandscapes/number of observers = 17.1
8. Checklist score, C = 2.2(Y + 20) = 82

Factor equation

  1. Enter slide dimensions – width, height, area, sky are, landscape area, area & percentage of the landscape of land cover, water, ground types, land uses, human impacts, background (>0.4 km), area above observer. Sky area should be no more than 25%.
  2. Calculate scores using percentage recorded and some ratios of these values - e.g. woody/total vegetation, woody/water. Total out of 80.

Marker scenes
Included coastal and non-coastal scenes. Observers scored these and average scores for all observers were obtained.

The method involved a small group of people traveling through five regions and recording their assessments of views (Table 8).

Table 8 Assessment of scenes by Kane’s methods

Region Number of scenes Number of observers
Flinders Ranges
16
12
Kangaroo Island
4
10
Southern Eyre Peninsula
8
10
Lake Eyre to Innamincka
-
-
Riverland
13
10 - 13

Applying the methods gave participants experience in their effectiveness. The Nature Preservation Committee concluded (McBriar, 1977) that:

  • Marker scenes sometimes gave anomalous results
  • The factor equation method proved troublesome being dependent on photographs which were often deficient, and involved a lengthy process.
  • Although results were expressed in percentages, it was not possible to reach 100%. The highest figure exceeded 85 and generally the higher scenic quality averaged 70 or higher.

The rejection of the marker scene and factor equation methods resulted in only the bipolar and components checklist being accepted, both of which relied on emotional response (Table 7).

The Committee decided on the following thresholds:

  • 70+ landscapes were worth Classified status
  • 60 – 69 were worth Recorded status
  • <60 noted

The Committee proposed 26 scenes as Classified and 13 as Recorded. The methods could only be used to assess views from specific locations and were not suitable for mapping regional landscape quality.

The small number of participants resulted in a large sample error (31%) which significantly affected the validity of the results. In his efforts to make objective the ass of scenic quality, Kane unnecessarily complicated what should be a simple process.

Dare (1978) followed Kane’s (1976) study and classified landscape character for a portion of Fleurieu Peninsula in South Australia. He used ten assessors to view slides and scored elements. The impact (0 – 2) of each element was multiplied by their contribution to scenic quality (-2 to +2) and scores summed for each landform. A landscape quality map was not completed.

In a follow up study to studies by Kane (1976) and Dare (1978) for the National Trust, Revell (1981) assessed the entire Fleurieu Peninsula. He defined ten tracts, selected ten slides for each tract covering landforms, water, vegetation and man-made forms both close-up and from a distance, involved ten participants and used Kane’s bi-polar scale method. He then had five people score the remaining photographs in comparison with the representative photographs and scaled out of 100. Unfortunately Revell did not translate the scores into a map of scenic quality. However, his approach had some of the key elements of a workable methodology. It used photographs and a rating scale but had far too few participants.

Ms McBriar of the National Trust summarized Revell’s study thus:

Grant Revell has made interesting advances on earlier work; additionally, he has explored further the production of a landscape map, but the ideal still eludes us… (Revell, 1981).

Radford and Bartlett (1977) used a process similar to Q sort in which 164 participants sorted 15 photographs into five groups, from very positive to very negative. The results were used to map scenic quality. Although the number of photographs was small, the approach was sound and the reasonably large number of participants would have provided fairly valid results.

Another early study using preferences was by Williamson and Chalmers (1982) in northeast Victoria. The study was used to validate the assumptions about preferences in the Visual Management System which was being developed by the Victorian Forests Commission. The methodology (Figure 5) used several techniques to gain rankings and ratings of photographs of the region. A 1 – 7 scale was used, unusual in that it was the only study which reversed the normal low-high sequence (i.e. 1 was high, 7 was low). The findings of the study were considered to validate the assumptions of the Visual Management System although it found the VMS placed too much emphasis on landscape variety and insufficient emphasis on naturalism.

Williamson & Chalmers
Source: Williamson and Chalmers, 1982
Figure 5 North East Victoria scenic quality methodology

Correy (1984) used 12 photographs of Sydney Harbor and 40 participants to assess significant components and then a second group of 100 participants used a bi-polar semantic scale with 24 variables to rate the scenes from most liked to least liked. The study found that people’s preferences ranged widely.

Prineas and Allen (1992) used a preferences approach to measure and map scenic quality in the Queensland Wet Tropics World Heritage Area (Figure 6).

Prineas
Source: Prineas and Allen, 1992
Figure 6 Wet Tropics WHA scenic assessment methodology

Over 300 landscape units were defined and a representative sample of 80 selected for photography. Oblique aerial photographs from a helicopter were used due to the inaccessibility of the area, a method criticized by Ramsay (1992) as missing the scenic or aesthetic values of the forest environment. An attempt had been made to include scenes below the forest canopy but was not completed. Criteria were set for the photographs and, unusually, a wide angle 28 mm lens used. The survey comprised 80 scenes from the region and 10 scenes from other WHA areas or elsewhere in Australia for comparison purposes. A sample of 306 random passer-bys used a self-guided survey with a rating scale of 1 – 15 (low - high).

Using multiple linear regression analysis, eight landscape dimensions were found significant: rivers & streams, coral reef, coastline, human disturbances, boundary interferences, water bodies, dense forests, and slopes 25º – 40º. Only 9 of the 307 land units, mostly coastal ranges, were found to exceed the WHA threshold of 11, a figure based on the score for Kakadu. The use of the 15 point rating scale is curious and could have provided difficulties for participants. The study is the only Australian landscape study which used aerial oblique photographs.

In Queensland, Preston (2001) carried out two studies of areas near Brisbane. He used a five stage process commencing with a scenic preferences survey in which over 200 participants rated 52 photographs (Mogill study) and 21 photographs (Glen Rock study) (Figure 7). The results were used to identify 110 attributes based on which a scenic preference model was developed, however only 14 attributes were used to predict emotional response and six attributes in the scenic preference model. This model was used to map scenic preferences. GIS was used to map visual exposure – the frequency of view from various locations. Scenic amenity was mapped combining the scenic preferences and the visual exposure. The method was sophisticated and thorough.

Preston
Source: Preston, 2001
Figure 7 South East Queensland methodology for assessing scenic amenity

The scenic amenity map integrated the preferences with visual exposure to “identify the relative contribution made by different places in the landscape to the collective community appreciation of scenery.” (Preston, 2001, 6).

Preston’s methodology was applied in the largest scenic quality study ever undertaken in Australia, the South East Queensland Regional Scenic Amenity Study (SEQRSASSC, 2005). The study used over 15,000 photographs and interviewed 1000 participants. The public preferences survey identified characteristics of views that influenced scenic preferences and based on this, maps of highly preferred scenery were prepared. The scenic preferences were combined with maps of visibility to map scenic amenity on a 1 – 10 scale.

Williams and Cary (2002) used black and white photographs of five vegetation types from Victorian rural areas. They used 36 photographs and a five point rating scale. The survey was mailed to 3000 respondents (half urban, half rural); the large number involved probably justifying the use of black and white photographs. The study identified a higher preference for eucalypts than non-eucalypts with perceived naturalness being a key influence. Interestingly, no relationship between ecological quality and preferences was detected. Of concern was the use of black and white photographs which emphasises the formalist qualities of line, tone, texture etc while losing the qualities and realism that colour imparts. Black and white photographs also tend to produce more extreme responses than colour photographs (Shuttleworth, 1980).

In a series of six studies, the author has measured and mapped scenic quality in various areas of South Australia (Lothian, 2000, 2004, 2005a and b, 2007, 2009). The methodology (Figure 8) involves photographing the region, classifying landscape units and selecting around 120 sample photographs to represent the area plus 30 South Australian scenes to benchmark the ratings at a State level, having a large number of participants rate the photographs on a 1 – 10 rating scale (low - high), having a small number of participants also score landscape dimensions on a 1 – 5 scale, and analyzing the results, including with multiple regression. Maps of landscape quality have been produced for the South Australia, the entire coast, the Barossa region, the River Murray region and the Flinders Ranges. The scenic amenity provided by large remnant trees, the visual impact of wind farms and developments in coastal and River Murray localities have also been assessed.

Lothian
Source: Lothian, 2008
Figure 8 South Australian landscape quality assessment methodology

The methodology used in the Regional Forest Agreement process was described earlier. These studies used a combination of the physical and preference methods, the later derived from extant cultural information sources.

Go to top

 
 

© 2009-14 Scenic Solutions   Privacy   Terms of Service   Site Map   Contact us