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

PROTOCOL FOR LANDSCAPE QUALITY STUDIES

ANALYSIS OF RESULTS
and
MAPPING LANDSCAPE QUALITY

ANALYSIS OF RESULTS

Analysis of the results is the heart of any study for it is here that new knowledge will be discovered and further understanding of human perception of landscapes gained.

The analytical tools used in the analysis of data relates to the objectives of the study. Many of the Australian studies were directed towards measuring scenic quality but not necessarily mapping it. The following summarises the purpose of a range of studies:

  • Zube & Mills (1976) used group correlations to compare the ratings by two groups, landscape students and residents, of the coast near Lorne.
  • Williamson and Chalmers (1982) sought to understand what landscapes were preferred, the visual impact of changes to the landscape and also to verify the Forest Commission’s Visual Management System. They used correlation analysis of relate the ratings with the landscape factors (or dimensions) and multiple linear regression to describe the variance.
  • In their study of Warringah Shire, Lamb and Purcell (1982) sought to determine the relationship between planning controls, building regulations and the character of the coastal area using people’s assessment of that character. They examined the differences in assessments between the groups and arranged the ratings of scenes in descending order. From this they identified key factors in generating high scenic quality. Detailed analysis of the scenes enabled them to describe the effects that future changes would make to scenic quality.
  • In their later study of the influence of vegetation structure on scenic quality, Lamb and Purcell (1990) asked participants to assess the naturalness of scenes and then used factor analysis to identify the factors which influenced the results. The naturalness scores were compared with the standard deviations and variance and naturalness related to the various structural forms of the vegetation.
  • Purcell and Lamb (1994) used groups of Italian and Australian students rate landscapes and urban scenes from each country to assess any differences. Using analysis of variance (ANOVA) they examined the subjects’ ratings of scenes by each country and used cross tabulations to examine the ratings of natural vs built scenes.
  • Preston (2001) measured community ratings of landscape aesthetics and measurements of landscape factors in areas near Brisbane. He used multiple regression analysis to develop a scenic preference model and applied the model to predict the likely rating of different landscapes and land units.
  • Williams and Carey (2002) asked participants to rate scenes of vegetation and also asked ecologists to assess their ecological integrity. They used factor analysis to identify patterns of preference and compared preferences and ecological integrity for each vegetation component. 
  • Wu et al, (2006) used GIS derived data on the physical landscape, a viewshed analysis and community preferences and used multiple linear regression to produce a scenic beauty assessment formula. He then applied the results to mapping landscape quality for the study region of Mornington Peninsula in Victoria.

From this brief review of some of the Australian preference studies it is evident that the methodology of the study and the nature of analysis undertaken relate closely to the purpose of the study. More sophisticated tools and analyses are evident in the more recent studies.

Several of the studies that I have undertaken are directed towards mapping scenic quality for the study region (Lothian, 2005a and 2005b, 2007). The steps taken in analysis are summarized below.

  • The survey results are inspected for strategic bias, i.e. where the participant uses the survey to fulfil their own objectives, for example rating all or most scenes 1 or 10. These surveys are discarded. Strategic bias can be assessed by extracting the means of ratings for each participant and examining the low and high means. Expect to find around 1% of participants employing strategic bias.
  • Incomplete surveys may be discarded and the survey analyse only complete surveys. However recognising the time and effort that participants invested in participating in the survey, surveys which have completed say more than 75% of scenes could be included. Frustration with slow Internet links is often the main reason for drop out, but this will diminish as broadband access becomes more common.
  • The characteristics of the participants are compared with the wider community – State or national, by age, gender, birthplace and education. Inevitably the participants tend to be far better educated and older than the wider community, however most surveys find it very difficult to gain the participation of members of the community who lack interest (Tucker, et al, 2006). If these differences influenced ratings, it would be expected that ratings would vary across, say age groups, or education. Averages of ratings across participant characteristics however generally indicate close similarity. Table 1 and Figure 1 indicate the average ratings for the scenes in five South Australian studies. While there are minor variations, none are sufficient to suggest significant variation in ratings across the respondent characteristics.

Table 1 Average ratings across respondent characteristics in South Australian studies

Tree amenity Coast Barossa River Murray Flinders Ranges
Age
5.55
5.46
5.55
5.73
6.35
6.55
6.48
6.53
5.57
5.50
5.60
5.60
6.03
6.07
6.01
6.07
6.07
6.19
6.37
6.31
Education
5.08
5.80
5.46
5.49
6.49
6.44
6.54
6.56
5.46
5.53
5.68
5.49
6.04
6.02
6.09
5.97
6.16
6.26
6.36
6.32
Gender
5.46
5.53
6.44
6.57
5.57
5.55
6.02
6.05
6.35
6.26
Birthplace
5.52
5.45
6.50
6.52
5.61
5.36
6.07
6.08
6.34
6.09

participant characteristics

Figure 1 Average ratings across respondent characteristics in South Australian studies

  • Chi square tests are conducted to assess the significance of differences between the participants and the community.
  • The analysis aims to uncover as full an understanding as possible of the ratings that have been obtained and to explain these by reference to the landscape factors. Analysis of the ratings commences with the general and moves progressively to the specific. The overall frequency of ratings is determined. Analysis then covers sub-regions or areas, and each of the landscape factors. The scores of landscape factors are compared with the ratings and with other landscape factors and further insights gained.
  • Using the landscape factor scores together with the ratings of scenes, multiple linear regression analysis is used to model the contribution of each landscape factor to the ratings obtained. The model may be tested against the ratings for each scene.

Table 2 summarises the most comprehensive model derived for the Coastal Viewscapes Project in South Australia. The model contains 10 characteristics (or factors) and with an R2 of 0.86 explains 86% of the variance in the data. Use of this model to predict landscape quality involves scoring each of the characteristics (area of beach, quality of beach etc) which can be onerous.

Table 2 Coast project: 10 landscape factors

Method Enter
Factors
Area, height, indentation, quality, diversity, naturalness, tranquility, waves, seaweed, steepness
R2
0.86
Equation
Y = 1.356 + 0.519 Tranquillity + 0.359 Quality + 0.255 Naturalness + 0.26 Area + 0.079 Height + 0.187 Diversity + 0.181 Indentation – 0.141 Seaweed – 0.056 Waves + 0.113 Steepness
Significance
F = 78.12, df 10, 1127, p = 0.000

Area = area of beach
Height = flat – high terrain
Indentation = indentation of the water/land edge
Naturalness = the appearance of naturalness in the scene
Diversity = diversity (busyness) in the scene from its land form, land cover, land use etc
Tranquility = sense of tranquility – awe inspiring perceived in the scene
Quality = quality of any beach in the scene
Seaweed = amount on beach
Waves = extent of wave action
Steepness = steepness of landform

Table 3 and Figure 2 show how this task may be simplified by selecting a fewer number of factors. Use of just one factor resulted a R2 of 0.68, thus explaining 68% of the variance. If this is considered too low, then just three factors resulted in the explanation of 80%. Adding further factors lifts the explanation further but with diminishing returns as shown in Figure 11. Thus a fair measure of the landscape quality of a coastal scene in South Australia could be derived by scoring (out of 5) the sense of tranquility or awe it inspires, and the extentand quality of the beach.

Table 3 Model factors and correlation coefficient

Factors
No factors
R2
Tranquility
1
0.681
Tranquility, Quality
2
0.762
Tranquility, Quality, Area
3
0.803
Tranquility, Quality, Area, Indentation
4
0.824
Tranquility, Quality, Area, Indentation, Seaweed
5
0.838
Tranquility, Quality, Area, Indentation, Seaweed, Steepness
6
0.848
Tranquility, Quality, Area, Indentation, Seaweed, Steepness, Naturalness
7
0.856
Tranquility, Quality, Area, Indentation, Seaweed, Steepness, Naturalness, height
8
0.857
Tranquility, Quality, Area, Indentation, Seaweed, Steepness, Naturalness, Height, Diversity
9
0.859
Tranquility, Quality, Area, Indentation, Seaweed, Steepness, Naturalness, Height, Diversity, Waves
10
0.86

multregfactors

Figure 2 Relationship of correlation coefficient with number of factors

Go to top

MAPPING LANDSCAPE QUALITY

The principal application considered here is for mapping scenic quality. Apart from the author, only two Australian studies mapped scenic quality:

  • Radford and Bartlett (1978) carried out an early study to map scenic quality in the Hunter Valley. Using preferences they overlaid maps of different gradings (water, trees, terrain) and then produced a composite score using a weighted additive model. The dominant score in each are was used as the rating for the area and mapped.
  • Wu et al (2006) used GIS to map scenic quality. The scenic assessment model was applied to all viewpoints and the scenic quality ratings interpolated using three mathematical approaches. This highly sophisticated method mapped scenic quality for the Mornington Peninsula.

The method that I have used in mapping scenic quality is based on the ratings obtained and the understanding gained through the analysis of the ratings and landscape scores. It involves firstly determining the appropriate ratings for each part of the study region, and secondly, entering this on maps of the study region.

Determining the ratings involves close inspection of the characteristics of each area using the full range of resources available including:

  • All the photographs taken of the region
  • Aerial oblique photographs which may be available
  • Google Earth™ images
  • The models derived for the region
  • Ratings of the landscape units
  • Maps of the region

Thresholds for each level on the 1 – 10 scale, as appropriate to the study region, should be defined (Prineas & Allen, 1992).

Basic to the application of the ratings derived through the survey is the principle of equivalence, that the rating of a scene of given characteristics may be applied to another scene of similar characteristics. Thus a rating of say, mallee vegetation, in one region may be applied to similar vegetation in another region. It is the characteristics present in the scene which determine the rating, not their location.

Mapping scenic quality is generally undertaken in conjunction with GIS specialists in interpreting and applying results.

Figure 3 contains six landscape quality maps by the author.

South Australia ls
South Australia

Yorke Pen
Part of the coast of South Australia (Gulfs)

Barossa
Barossa valley and region

victor Harbor
Victor Harbor

River Murray
River Murray, Lakes and Coorong

Flinders Ranges landscape
Flinders Ranges

Lake District landscape quality map
Lake District

Figure 3 Landscape quality maps by Dr Andrew Lothian and Scenic Solutions

Go to top

 
 

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