Vous êtes sur la page 1sur 23

Gary Nobles Analysis

4.0 Analysis

This section is designed to provide a methodology for future analysis of impressions of


circularity within defined landscapes; it provides a series of possible methods for this
depending upon the definition proposed. Rather than concentrating solely on the horizons of
the sites in question the same analysis was conducted upon 100 random points. This was in
order to assess whether this impression of circularity can be defined as part of the general
nature of the containing landscape and thus calculate its significance.

The analysis of the data was conducted within a standard spreadsheet with the data being
imported from text files and SQL queries. The analysis is broken down into an appropriate
structure; firstly the natural character of the landscape needed to be established in order to
ascertain the background population. This aided in the assessment of the data signifying
whether or not the observed points, those of the stone circles, were located at significant points
within the study area when compared with the rest of the landscape. This was achieved
through the use of the Monte-Carlo test.

The results from the questionnaire were analysed to quantify perceptions of circular horizons,
this was attempted using standard deviation and by looking at the graphical representations of
the various attributes, these being: inclination, distance and elevation. These results were then
used in comparison with the standard deviations and graphs of the horizons as viewed from
the Recumbent Stone Circles.

4.1 The Monte-Carlo Test

4.1.1 Establishing the Background Population

A Monte-Carlo test was implemented upon both of the study areas; these consisted of a
1000 random points being scattered throughout each study area, these were constrained to a
rectangle which encompassed all of the observed sites. For Scotland, the southern and western

27
Gary Nobles Analysis

boundaries of this rectangle were allocated by the southern most and western most sites,
whereas the northern and eastern boundary was the coast line. For Ireland a similar approach
was utilised, having the northern and eastern boundaries restricted by the extremities of the
sites and the southern and western edges bound by the coastline. The random points were
created using the GRASS GIS module r.random, a list of these random points are located in
appendix???.

Time of analysis played a large part in this study, it was not known until a vast amount of the
data had been produced that the target of 1000 random points was unattainable. An
approximate time to conduct the r.los is outlined in table 2, values for r.horizon are not
included as this analysis was relatively quick. However its speed did not help as the r.horizon
module had to wait for the r.los module, also r.horizon was only available within a UNIX
environment which was accessible to the author through a network via the internet. r.horizon
in its current state is not easily imported into grass for use within a unix emulator, in this case
cygwin. This also resulted in time being lost having to transfer relatively large ascii files back
and forth. Furthermore this approximate timing is only for computational analysis and does
not take into account the time it took to convert all of these text files which were produced by
r.horizon and calculate the relative cumulative frequencies. The author did make use of shell
scripting, however this took a little time to learn and initial mistakes resulted in the delay of
their implementation. Also over 300 separate analyses were conducted using r.los.

The consequence of this resulted in the need of a reduction of the number of points, the author
aired on the side of caution and initially conducted analysis upon every 100th point in which it
became clear that there was only time to conduct this analysis on 100 points, and therefore a
10% sample of these random points was used. It should also be noted that these points were
chosen by listing the geographic co-ordinates of the points arranged from the northeast to the
southwest and selecting every tenth point. This creates a representative distribution of the
points within this random sample and allows for further analysis at a later date of the
remaining 900 points. In retrospect it would have been more advantageous and conventional to
have selected a stratified sample of these points.

28
Gary Nobles Analysis

Table 3. Typical time to conduct analysis of r.los on three platforms.


Computer Specifications Typical time for one analysis Typical time to analyse 100 points
Server - UNIX OS 4 3/4hours 5 in a day 20 days
Unknown Specification
Desktop Computer – Windows OS 3 1/2hours 7 in a day 15 days
Intel(R) D CPU 3.20 GHz
3.20 GHz, 2.00 GB of Ram

Laptop – Windows OS 24 hours 1 in a day 100 days


Intel(R) Core(TM)2 CPU
T5500 @ 1.66 GHz
980 MHz, 1.00 GB of RAM

At each of these sampled random points the horizon was calculated using the GRASS GIS
modules r.los and r.horizon as outlined in the methodology, the quantified attributes were then
converted into a series of relative cumulative frequency graphs as shown in figures 10-12.
These graphs display the relative cumulative frequency of the random points from which the
associated envelopes were created, this envelope is shown behind the values in purple. These
envelopes were created from taking the minimum and maximum values of the entire set of
random points and constructing an envelope which encapsulates all of the random points. It is
this envelope which is used later in comparison with the relative cumulative frequency values
of the actual sites.

This envelope provides the necessary background character of the landscape; this is the area in
which one would expect all of the relative cumulative frequency line graphs of the sites to fall
completely within, if this is the case then one would accept the null hypothesis. If the null
hypothesis is accepted then the site or sites are within keeping with the general character of the
surrounding landscape in relation to the nature of the far horizon. However if a site or sites go
beyond the limit set by the envelope then an alterative hypothesis would have to be proposed.

29
Gary Nobles Analysis

Figure 10. Relative Cumulative Frequency Graph of Inclination for the Scottish Random Points

Figure 11. Relative Cumulative Frequency Graph of Distance for the Scottish Random Points

30
Gary Nobles Analysis

Figure 12. Relative Cumulative Frequency Graph of Elevation for the Scottish Random Points

Figure 13. Relative Cumulative Frequency Graph of Inclination for the Irish Random Points

31
Gary Nobles Analysis

Figure 14. Relative Cumulative Frequency Graph of Distance for the Irish Random Points

Figure 15. Relative Cumulative Frequency Graph of Elevation for the Irish Random Points

32
Gary Nobles Analysis

4.1.2 Comparisons with the Scottish Sites

Of all of the possible and probable RSC’s in Northeast Scotland only 52 could be accepted as
definitely belonging to this class of monument, the majority of sites were plotted with an
accuracy of up to 100 meters, although some of these were plotted with an accuracy of up to
10 meters. The sites which were used, along with their accuracy and site number can be
accessed through the database on the accompanying DVD or in a table in appendix.

As seen in the previous subsection (4.1.1) the relative cumulative frequencies of the random
points provide an envelope of the background population, upon which the associated values of
the Scottish sites can be plotted. Figures 16-18 display the values of the observed sites against
the envelopes of the expected background population, the majority of the sites fall well within
the envelopes, however a few sites have values beyond the limit as set by the envelopes, these
will be discussed in due course.

Figure 16. Relative Cumulative Frequency Graph of Inclination for the Scottish Sites

33
Gary Nobles Analysis

Figure 17. Relative Cumulative Frequency Graph of Distance for the Scottish Sites

Figure 18. Relative Cumulative Frequency Graph of Elevation for the Scottish Sites

34
Gary Nobles Analysis

Figure 19. Relative Cumulative Frequency Graph of Elevation for the Scottish Sites

The overall nature of the relative cumulative frequencies can be assessed; the inclination
values of the observed sites are largely skewed to the left of the envelope. This appears to be a
typical bias when compared with the values of the random points, as the envelope is
constructed from a group of values which are also well represented to the left of this envelope.
The second of these envelopes (figure 17) assesses the relative cumulative frequencies of the
sites in relation to their distance to the far horizon; the majority of the sites are fairly uniform
in respect to their distribution within the envelope. This is also the case of the underlying
random values from which the envelope is created, as seen previously in figure 11. The third
graph (figure 18) evaluates relative cumulative frequencies of the elevation of the far horizon
from each individual site. These values are located well within the boundary set by the
minimum threshold. Analysing this further there appears to be three trends within the data set,
these are highlighted in figure 19, in this the turquoise lines highlight those sites which contain
elements of the sea, and this is why the lines begin above zero on the y-axis. The yellow lines
are sites which contain horizons where the majority of their elevations are within 100m of

35
Gary Nobles Analysis

variance, whereas the orange lines indicate sites where their horizons are more wide-ranging
in elevation. This trend would initially appear significant, however when the same
categorisation is applied to the values of the random points three similar trends can be
detected, this is illustrated in figure 20.

Figure 20. Relative Cumulative Frequency Graph of Elevation for the Scottish Random Points with
Trends

This type of analysis has looked at the defined landscape as a whole with out discrimination,
an expansion on this analysis would be to confine the random points which were analysed to
areas with specific topographic attributes. For instance to narrow down these envelopes one
could select areas which are at similar altitudes as the observed sites, alternatively areas with
the same geology or these could be used as a complimentary selection technique. So, for
instance random points could be limited to areas below 500m and located on crystalline
schists.

The first part of this section has established that the majority of the sites are within the
boundaries of what one would expect, for these sites the null hypothesis can be accepted. That
is that these sites are not situated within locations where the attributes of the far horizon

36
Gary Nobles Analysis

(inclination, distance and elevation) are significantly different to the far horizon at different
locations. However there are a few sites which go beyond the limits of these defined
envelopes. Site 47 exceeds the maximum elevation and is below the minimum values for
distance and site 51 exceeds the envelope in respect to distance and inclination. Site 122 is
also below the minimum expected for distance and both sites 119 and 124 exceed the envelope
for inclination. For these sites the null hypothesis, as previously stated, has to be rejected and
an alternate hypothesis proposed. There is a chance of generating a type two error from these
results; especially as for this test only 100 random samples were used. It has been established
that for a Monte-Carlo test 1000 or more random samples should be generated (Conolly and
Lake 2006:161).

4.1.3 Comparisons with the Irish Sites

The same analysis can be conducted upon the RSC’s from Southwest Ireland; these consist of
56 monuments which are located with an accuracy of 100m and some at an accuracy of 10m.
The sites and their accuracy are located in appendix Z. Figures 21-23 display the values of the
observed sites plotted against the associated envelopes of the background populations, this
uses the same method as utilised in the previous subsection (4.1.2). A few sites exceed the
limits set by the envelopes, these sites will be discussed at the end of this section.

The relative cumulative frequency for inclination can be investigated, as illustrated in figure
14 the values for the sites are skewed to the left of the envelope, this bias is also seen in the
values from which the envelope is constructed, therefore these appear to fall within the overall
character of the surrounding landscape. The second of these envelopes examine the relative
cumulative frequency of the sites in relation to their distance to the far horizon. Their
distribution appears to reflect that of the underlying values which were used to create the
envelope, as seen previously in figure 20 (Section 4.1.1); they both appear uniformly
distributed within the envelope. The third graph (figure 23) evaluates the relative cumulative
frequencies of the elevation of the far horizon from each individual site.

37
Gary Nobles Analysis

Figure 21:

Figure 22:

38
Gary Nobles Analysis

Figure 23:

Figure 24

0.9

0.8
Relative Cumulative Frequency

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Elevation (Km)

39
Gary Nobles Analysis

This reflects a similar result as seen in the underlying values of the envelope, as with the
Scottish sites there appears to be three trends within the data. These trends are highlighted in
figure 24; in this the turquoise lines illustrate the sites whose far horizon falls in the sea. These
are also identified due to the starting position on the graph being above zero on the y-axis. The
yellow lines are sites which contain horizons where the majority of their elevations are within
100m of variance, whereas the orange lines indicate sites where their horizons are more wide-
ranging in elevation. These are not significant trends as the same trend is reflected through
elevation of the horizons of the random sites as illustrated in figure 25.

Figure 25

As acknowledged in relation to the Scottish sites (Section 4.1.2) the random sites were located
irrespective of any other factors, these envelopes could be narrowed or reduced by selecting
random points on areas with similar topographic attributes, for instance areas below 500
meters and with a geology of old red sandstone.

As previously established, the majority of these sites fit within the character of the landscape
with respect to the far horizon. Therefore the null hypothesis can be accepted, this being that

40
Gary Nobles Analysis

these sites are not situated within locations where the attributes of the far horizon (inclination,
distance and elevation) are significantly different to the far horizon at different locations.
However there are a few sites which go beyond the envelopes set out previously. Sites 4, 7, 15
and 26 have inclinations which exceed the maximum values for the envelope. Sites 90 and 91
pass beyond the minimum values for distance and sites 41 and 69 deviate from the envelope in
respect to elevation. The null hypothesis as stated earlier can not be applied to these eight
sites; therefore one must reject the null hypothesis and seek an alternative hypothesis. This
risks the chance of generating a type two error from these results; especially as for this test
only 100 random samples were used. It has been established that for a Monte-Carlo test 1000
or more random samples should be generated (Conolly and Lake 2006:161).

4.2 People’s perception of Horizons

As described in the methodology (Section 3.1) an online interactive questionnaire was


circulated in an attempt to establish whether the perception of a circular horizon is definable
and quantifiable. A total of 78 responses were received, two of which were incomplete and
discarded. Of the remaining 75 responses only 33 took into account the full questionnaire, due
to the quantity of hits on the website and the amount of failed questionnaires this was reduced
from 41 questions to 9, the omitted questions were not vital to this study, however these
omitted responses will be analysed in the latter part of this section. The two forms of this
questionnaire can be seen in appendices? And?

The first question was designed to identify a general definition for the term circularity, the
following four options were offered:

• Definition 1: like, involving, resembling, shaped like, relating to, or in the form of a
circle.
• Definition 2: traveling, occurring, or recurring in a cycle; or ending in itself;
roundabout; circuitous.
• Definition 3: both

41
Gary Nobles Analysis

• Definition 4: none of the above

The majority agreed with definition one and three with a number of people settling for
definition two, this is illustrated below in figure 26. No one believed that the definitions

Figure 26. Definitions of Circularity and their significance


40
Number of Responses

35
30
25
20
15
10
5
0
1 2 3 4
Definition

proposed here were inadequate for describing the term circularity, therefore one can accept
that the remaining definitions adequately fulfil this term. These results are from a very small
section of the population they were mostly those who are friends of the author, a member of
the computer heritage forum www.antiquist.org or a member of UCL. Furthermore only
persons who had access to a computer and the internet could fill in the questionnaire; therefore
it was impossible to get responses from a full cross section of society.

Table 3.
Panorama
Definition 1 2 3 4 5 6 7 8
Circular 1 16 15 9 15 13 12 13 15
Not Circular 1 12 13 19 13 15 16 15 13
Total 1 28 28 28 28 28 28 28 28
Circular % 1 57.14 53.57 32.14 53.57 46.43 42.86 46.43 53.57
Not Circular % 1 42.86 46.43 67.86 46.43 53.57 57.14 53.57 46.43
Circular 2 11 6 2 5 3 4 4 10
Not Circular 2 1 6 10 7 9 8 8 2
Total 2 12 12 12 12 12 12 12 12
Circular % 2 91.67 50 16.67 41.67 25 33.33 33.33 83.33
Not Circular % 2 8.33 50 83.33 58.33 75 66.67 66.67 16.67
Circular 3 23 14 16 21 14 16 18 24
Not Circular 3 12 21 19 14 21 19 17 11
Total 3 35 35 35 35 35 35 35 35
Circular % 3 65.71 40 45.71 60 40 45.71 51.43 68.57
Not Circular % 3 34.29 60 54.29 40 60 54.29 48.57 31.43

42
Gary Nobles Analysis

Taking into account this low representative sample each definition will be investigated. The
preceding table displays the results of the questionnaire looking at the answers of the
panoramas in reference to the definition which was given. To assess the significance of these
results a Chi-squared test can be employed. A significance level of 0.05 was used with 3
degrees of freedom this supposed that a value of χ2 needed to exceed 3.84146 (Shennan
2004:Table F) to be of significance. The χ2 values, which are located in appendix X, have
varying significance, for the first definition none are statistically significant. The second
definition has three results which are statistically significant and the third definition only has
one significant result. For the results where there are no significant results a heuristic approach
will be attempted.

4.2.1 Resembling a circle

The first definition which was proposed defined circular as resembling a circle, by this
definition if one was stood at a point in the landscape would they expect the resulting
panorama to be represented as a circle on a plan? As there was no significant consensus for
which landscapes applied to this definition this can only be examined through a heuristic
approach. For this standard deviation (S.D.) can be applied, if a horizon was a perfect
geometric circle then the S.D. for inclination, distance and elevation would be zero. This can
be visualised by imagining a cross section through a sphere, the viewing location would be
central. One can imagine the inclination increasing which would result in an increase of
elevation, the distance can also be increased by expanding the sphere; this could be supported
by an analogy with a partially inflated football.
Table 5. S.D. of Attributes of the panoramas ordered by (I+D+E)
Inclination (I) Distance (D) Elevation (E) (I+D+E)
Panorama 2: Scratchy Bottom 2 4.65 170.89 15.87 166.3424
Panorama 4: Scratchy Bottom 1 5.55 182.13 19.96 194.7755
Panorama 1: Lyscome Bottom 3.42 331.09 39.14 310.6553
Panorama 3: Luton Down 0.95 345.99 13.59 356.0576
Panorama 6: Down Farm 1 0.35 3052.05 67.07 3113.565
Panorama 5: Lulworth 2.45 3153.07 29.56 3155.244
Panorama 7: Down Farm 2 0.4 3430.65 68.58 3489.878
Panorama 8: Corfe 0.25 11023.2 75.37 11040.3

43
Gary Nobles Analysis

As shown in table 5. the S.D. of the raw values can provide a guide to the nature of the
variation of the horizon, these cannot be added together to produce an overall variation. This is
due to the distance having possible values from 0m-60,000m, the elevation can have a partial
impact as it ranges from 0m-1000m, but the inclination has virtually no impact as it ranges
from 0˚-180˚. If one normalises the raw data from zero to one, one being the maximum and
zero being the minimum, then what is produced is a relative range. The data is normalised
using equation 1.

Equation 1: Calculating the Normalised Value


Value – Minimum Value
Normalised Value =
Range

Once normalised the S.D. can be taken of the inclination, distance and elevation values, these
are seen in table 6 below, these values can be added together to give a number between 0 and
3, the lower this number to zero the more circular the horizon is. That is to say that the more
circular when compared to a geometric circle. From here on in it is referred to as an index of
circularity or the circularity index. This treats each attribute, these being inclination, distance

Table 6. S.D. of normalized values ordered by Circularity Index


Inclination Distance Elevation Circularity Index
(I) (D) (E) (I+D+E)
Panorama 8: Corfe 0.12 0.28 0.27 0.68
Panorama 4: Lulworth 0.18 0.32 0.23 0.73
Panorama 3: Scratchy Bottom 2 0.28 0.25 0.22 0.75
Panorama 5: Luton Down 0.23 0.27 0.27 0.77
Panorama 1: Lyscome Bottom 0.33 0.20 0.31 0.85
Panorama 2: Scratchy Bottom 1 0.28 0.34 0.25 0.86
Panorama 6: Down Farm 1 0.20 0.35 0.32 0.87
Panorama 7: Down Farm 2 0.19 0.35 0.34 0.87

and elevation, as having an equal weighting of its influence upon the perception of circularity
of the horizon. With further research if this is not found to be the case one could weight these
accordingly. The circularity index can range from 0-1.5, this is a result of the attributes, these
can have a maximum S.D. of 0.5 as they are normalised between 0 and 1. The lower this index
the more likely the horizon is circular, whereas the closer this value is to 1.5 the more likely
the horizon is not circular.

44
Gary Nobles Analysis

This methodology has been applied to the RSC’s, for Scotland the circularity index ranges
from 0.35 to 0.98 with an average of 0.67. This means that none can be defined as definitely
having or not having a circular far horizon. It is better described as being on a sliding scale
between circular and not circular. Therefore the results can be placed in four groups created
from dividing the scale into four equal sections.

Table 7. Scottish sites


Quartile 1 2 3 4
Division Values 0-0.375 0.3751-0.75 0.751-1.125 1.1251-1
Circular horizon Highly Likely Likely Unlikely Highly Unlikely
Number of Sites 1 39 10 0

This shows that with these arbitrary thresholds only one site, Whitehill (66), is highly likely to
have a circular far horizon. The same methodology can be applied to the Irish sites, the
circularity index ranges from 0.27 to 1.43 with an average of 0.69. These have a wider range
in respect to the circular index than the Scottish sites with three sites, Gortanimill (19),
Reanascreena (29), Coolaclevane (75), being highly likely to have a circular far horizon.

Table 8. Irish Sites


Quartile 1 2 3 4
Division Values 0-0.375 0.375-0.75 0.75-1.125 1.125-1
Circular horizon Highly Likely Likely Unlikely Highly Unlikely
Number of Sites 3 36 15 1

These thresholds were used merely as an indicator as there were no significant results from the
questionnaire (Section 4.2.0). Adding a weighting to the far horizon’s attributes could alter
these results, to apply a weighting one would have to visit all or at least a sample of the sites.
Through phenomenological investigation one could then define which sites are contained
within a circular far horizon and then define to what proportion each of the three attributes
influence the observer.

4.2.2 Continuous Horizon

45
Gary Nobles Analysis

The second of the definitions, as defined by the questionnaire, was where the far horizon
forms a complete circuit, ending in itself or occurring in a cycle. Only three results from the
questionnaire were of significance, panorama‘s 1, 3 and 8, of these panorama’s 1 and 8 were
seen as having a circular far horizon but panorama 3 did not. This seems harder to analyse,
without the results from the questionnaire one would add the number of parcels of land on
which the far horizon segments fall. A continuous horizon would have one unbroken horizon
line; this could be of varying inclination, distance and elevation. Table 9 displays the number
of parcels of visible land which contains a horizon segment, the number of horizon parcels
was then divided by the mean distance to the horizon.

Table 9.
Panorama Number of Horizon Parcels Mean Distance Number of parcels/Mean Distance
Panorama 1 4 728.55 0.0055
Panorama 2 3 385.53 0.0078
Panorama 3 5 601.65 0.0083
Panorama4 2 384.13 0.0052
Panorama 5 10 3204.43 0.0031
Panorama 6 19 5540.43 0.0034
Panorama 7 19 5807.01 0.0033
Panorama 8 38 30677.52 0.0012

This method does not appear to be useful; even though it appears that panoramas 1 and 8 have
a relatively low number of parcels per mean distance and panorama 3, which is not circular,
has the highest relative value. This does not seem a very robust method, the mean distance was
chosen as it appears that the distance to these parcels can alter a person’s perception of the
horizon. This is most likely due to object clarity, in this case the objects are the horizon
segments and it is the point of the perceptual interaction between two of these segments which
determines whether the horizon is circular. If one or more of these segments are relatively
close to the perceiver and the remaining segments are relatively far away then this is unlikely
to be seen as a complete circuit. Whereas if the points of perceptual interaction are all at a
distance which is beyond the recognition limit of the perceiver then these segments would
appear to be combined into the same segment and thus perceived as circular.

46
Gary Nobles Analysis

At this point it appears necessary to diverge to identify general factors which can alter ones
perception of these perceived nodes. Firstly these points or nodes at which two horizon lines
converge are artificial, they do not exist in the environment, they are created ultimately by the
action of the retina and its ability to define edges through lateral inhibition. From this they can
be defined as points or areas of contrast by the perceiver, these areas can be enhanced or
masked by features in the landscape. For instance a horizon covered in trees can mask
horizons which are at relatively similar distances from the observer. Whereas if one of these
horizons is relatively distant then the scaling effect of the trees can aid in the enhancement of
the contrast between the horizon segments, thus an imaginary nodal point can be applied to
that location. This point as just discussed has also been referred to by the author as the point of
perceptual interaction.

For analytical purposes this poses a problem, at what distance does this perception of two
horizon segments become indistinguishable? Also at what distance can a second horizon
segment be from the first and still be perceive as a single horizon? These relationships are
illustrated in figure 27. The main methodological issues have just been posed but there are also
computational issues, the output from r.horizon provides a binary raster where the horizon
cells are classified accordingly. Firstly a horizon segment would have to be identified; this
could be achieved through converting the binary raster into a series of vectors. The end nodes
of these vectors could then be joined together to produce a secondary vector map, these
vectors could then be measured. Alternatively and in the authors opinion a more advantageous
method, this would be to stay within a raster environment and calculate the distance between
each raster cell and it’s neighbouring cell. A further possible problem would be that a nearest
neighbour approach would not work as the nearest cell may not be the next cell in the horizon.
To avoid this each raster of the horizon could be numbered from the northern cell clockwise,
possibly using the azimuth generated through r.horizon. Unfortunately a partially manual

47
Gary Nobles Analysis

Figure 27: The point of


perceptual interaction
Point
of
perce
ptual
Dist 2nd
ance Horizon
betw Segment
1st een
Horizo
n
Segme
Dist
ance
to
Hori

Viewing point

calculation of this is too consuming on resources to be attempted within this study; however
this theoretical methodology needs to be tested to see its relevance. This appears to be as far as
this analysis can go without further methodological experimentation.

3.2.3 Putting it all together

The final of these definitions is a synergy between the first two definitions (Sections 4.2.1 and
4.2.2). These define circular as resembling a circle and having a continuous horizon, that is a
far horizon which is perceived as being unbroken through 360˚. In this section one would try

48
Gary Nobles Analysis

to define to what influence does each of these two factors have: can a continuous far horizon
override the need of it resembling a geometric circle? Also: can a segmented far horizon be
perceived as circular if it resembles a geometric circle? It is this balance of the two definitions
which needs to be established, as the results from the second definition were unable to be fully
analysed this limits what can be inferred in this section.

A possible methodology for this would be to take the circularity index from section 4.2.1 and
combine this with the result from section 4.2.2. Rather than adding these values together one
could compare them through a statement, for instance a circular far horizon could be where the
circularity index is below 0.75 and a number derived from the second definition is below or
above another threshold. This appears to be as far as this section can be taken within this study
without further investigation. These three definitions have vast differences upon the
methodology to be applied; it shows that mathematical calculations can be used to quantify the
degree of circularity of the horizon. However to what degree these results reflect an
impression of circularity is still unclear, without a further questionnaire or phenomenological
study.

49

Vous aimerez peut-être aussi