Content uploaded by Priji Balakrishnan
Author content
All content in this area was uploaded by Priji Balakrishnan on Sep 29, 2016
Content may be subject to copyright.
MEASURING LIGHT THROUGH TREES FOR DAYLIGHT SIMULATIONS: A
PHOTOGRAPHIC AND PHOTOMETRIC METHOD
Priji Balakrishnan & J. Alstan Jakubiec
Singapore University of Technology and Design, Singapore
ABSTRACT
Trees play a significant role in influencing daylight
availability inside and outside buildings. They temper,
scatter and transmit light subsequently reducing the
availability or acting as a passive source of daylight.
Current daylighting simulation practices either avoid
modelling trees or model them as cones, spheres or
cylinders with an assumed reflectance value. Trees are
complex in their shape and—depending on crown
density and clumping nature—their optical properties
change considerably. In order to predict the effect of
trees on daylight, researchers need to first measure and
quantify this effect. Hence, in this paper the authors
propose a low-cost method employing high dynamic
range photography and automated image processing to
measure two variables of the tree crown: gap
percentage and transmittance percentage. These
measured variables can be used in daylight simulation
platforms such as Radiance to geometrically model the
crown of a tree and specify its optical properties
INTRODUCTION
Trees are microclimate modifiers. They enhance view
and provide thermal and visual comfort particularly in
warm, sunny climates. Trees also play an important
role in daylight availability within buildings as they
temper, scatter and transmit light.
A characterization study of trees by Federal and
Catarina (2012) showed that though trees attenuate
considerable amount of light, reducing luminance
levels into spaces, they also reflect both sunlight and
skylight resulting in an impact on the neighbouring
buildings. In northern latitudes, where daylight is a
scarcity, trees are considered an obstruction,
increasing the use of electrical lighting in buildings. In
tropical climates trees are considered a suitable
strategy to prevent glare and excessive daylight into
spaces.
Hence in daylight design strategies, trees can be
chosen based on the kind of influence required. In
situations where summer shading is significant, trees
such as the common lime can be planted which
attenuates about 85–86% of light (transmits
approximately 15%), and where more daylighting is
required, trees like ash or birch would be appropriate
choices since they have lower attenuation during
winter 35–40% (Wilkinson, Yates, & McKennan,
1991). Tropical trees like evergreen conifers have an
attenuation percentage of 90–95%; they transmit only
about 5–10% of light through (Tregenza & Wilson,
2011).The influence and effect of trees for daylight
design has been set aside for its difficulty in being
incorporated into daylight calculations. Trees are
complex objects to model given its variability in
shape, colour, crown density, reflectivity and
transmissivity.
Current modelling practices of trees in daylight
simulation platforms simplify trees as cones, spheres
or cylinders with an assumed reflectance of 20%
(Sadeghi & Mistrick, 2015), ignoring the
transmittance property of trees as well as the
variations in shadow patterns different species of trees
produce.
In order to predict the various effects of light through
trees in daylight simulations, their optical properties
need to be measured; therefore, in this work the
authors look at various methods that have been
established to measure and model trees. A new low-
cost method of measurement is proposed employing
hemispherical high dynamic range (HDR)
photography and automated image processing to
measure two variables—gap percentage and
transmittance percentage—that influence the optical
properties of transmittance and reflectance of trees.
BACKGROUND
Measuring light through trees in existing studies are
based predominantly on two methods: photographic
and photometric measurements. Light meters such as
luminance, illuminance or irradiance meters are used
for photometric methods whereas a technique called
hemispherical photography is applied in photographic
measurements.
The photometric methods used in studies by
McKennan (1988); Wilkinson et al. (1991) involved
taking discrete measurements of luminance in rows
across the crown and the sky immediately on either
side of it. As the measurements were taken manually
and at about 40 points across the crown, the
procedures were performed five times to reduce the
occurrence of errors. The study calculated the
attenuation percentage for forty-two tree crowns in
both summer and winter to rank different species
based on their light attenuation characteristics.
Photometric methods can be time consuming. As a
result, the use of photographic methods became
common as it allowed for much larger sample size
collection in lesser time than required by photometric
methods. A comparison of both methods by Wilkinson
(1991) found that there was no significant measured
difference in the light attenuation of tree crowns using
either of them.
The procedure used by McKennan (1995) to collect
data using a photographic method involved capturing
profile photographs of trees to produce a high contrast
monochrome printed image of a dark tree against a
light sky. The print is overlaid with an acetate sheet of
dots at 5mm interval. Attenuation percentage of the
tree is calculated by counting, via visual inspection,
the total number of dots within the boundary and the
dots coinciding with branches, twigs or leaves of the
crown.
Unlike in building science, measuring light through
trees is a mature and active area of research in plant
sciences like forestry and agriculture. In plant
sciences, quantifying the amount of light through trees
helps determine the Leaf Area Index (LAI). LAI is a
dimensionless quantity used to describe the surface
area of leaves that determine the transpiration and
photosynthetic characteristics of trees on local and
global scales. In theory, it is the total leaf area of a tree
by its horizontally projected canopy area.
One of the most commonly used photographic method
in plant sciences is hemispherical (fish-eye) canopy
photography which is a technique of taking
photographs using wide angle lenses (with viewing
angles typically approaching 1800) looking upward
from within a canopy or downward from above a
canopy (Rich, 1990).
Hemispherical photographs are analysed to determine
the LAI of tree canopy by converting it into a binary
image of black and white pixels. The challenge of
conversion is to determine optimal brightness to
threshold the image in order to distinguish leaf pixels
(black pixels) from sky pixels (white pixels). Several
software packages have also been developed to
threshold hemispherical photographs and determine
tree canopies’ LAI (Jonckheere et al., 2004).The input
in most of them are under-canopy photographs with
only the sky as the background. This makes it easier
to define boundary of the canopy and separate it from
the background for thresholding and analyses. In case
of crown profile photographs of tree canopy, there
may be other objects in the background apart from the
sky, such as buildings, other trees etc. Therefore, an
additional step is required to selectively draw a
boundary of the canopy to separate it from the
background before further thresholding or analyses.
Measuring light through crown profiles of trees is
necessary in building science to evaluate its effects on
daylight availability through windows and other
vertical openings. SideLook (Nobis, 2005) is a
software tool developed to analyse vertical vegetation
of grassland communities using automatic
thresholding, an algorithm to find the optimum
threshold value to isolate the vegetation from its
background during image analysis (Nobis &
Hunziker, 2005). This algorithm is implemented and
explained in the methodology section of the paper.
A study by Al-Sallal and Al-Rais (2013) proposes
hemispherical canopy photography to measure
variables called gap fraction and leaf area density for
use in mathematical models to compute lighting
interception of a tree canopy. This model is then used
to geometrically construct a tree for use in daylight
simulation. The setback of the method for calculating
daylight availability in buildings is that it tests and
analyses for light passing through under canopy of
trees which is different from light passing through tree
crown profiles.
The aim of this paper is to propose and describe a
photographic method—Hemispherical HDR
Photography—to measure light through crown
profiles of trees and translate this information for use
in simulation models.
Hemispherical HDR photography is a technique of
utilizing HDR photography with fish eye lenses. In
HDR photography, multiple exposure photographs are
taken to capture wide luminance variations in a scene.
These photographs are then combined into a single
HDR image saved as Radiance RGBE format (*.hdr
or *.pic) (Ward, 2015) in which each pixel value
corresponds to its physical quantity of luminance
(Inanici, 2006).
This paper details steps of taking, processing and
analysing HDR photographs of tree crown profiles
carried out as a pilot study. The outputs of this method
are two variables—gap and transmittance
percentage—that describe optical properties of tree
crowns which can be used to build a model and assign
an appropriate material for a tree crown in daylight
simulation platforms like Radiance. Three-
dimensional tree models are created based on the
measurements, and a short study is performed using
Radiance in order to illustrate the impacts of detailed
tree information in lighting analysis.
METHODOLOGY
Image Capturing – Taking HDR Photos
The pilot study was conducted in open ground where
the crown of the tree could be photographed with only
sky as background. A total of six sets of HDR
photographs were taken from positions facing the tree
as shown in Figure 1. For each of these positions, there
is a photograph taken from the opposite side facing the
sky. Four sets (1,2,3,4) were taken from four sides of
the tree, and two sets (5,6) were taken from under the
crown (Figure 1).
Table 1 lists the equipment, supplies and camera
settings used for the fieldwork. The fieldwork involves
taking two hemispherical HDR photographs
simultaneously.
Figure 1: Diagram showing six positions where
photographs facing the tree were taken.
Figure 2: Camera positions facing the tree and
facing the sky for photographs taken from the sides.
Figure 3: Camera positions facing the tree and
facing the sky for photographs taken from under the
crown.
Taking HDR photographs: Each of the multiple
exposure photographs were taken using the Canon
EOS 5D mounted on a tripod, with settings described
in Table 1. The different exposures were achieved
using a fixed aperture size (f/5.6) and varying only the
shutter speed in manual exposure mode as described
by Inanici (2006).
Positioning the camera: Cameras are positioned as
shown in Figure 2 and 3. The position A faces the tree
and position A’ is placed in the same line, opposite
position A but facing the sky. Both the cameras face
the same portion of the sky just that one includes the
tree and the other does not. The camera at position A
captures the light passing through the tree and camera
A’ captures the sky luminance and hence both the
captures need to be synchronised. Figure 3 shows the
camera positions for shots from under the crown. The
position U faces the under crown of the tree and U’
faces the open sky without the tree.
Table 1: Equipment, supplies and camera settings
Equipment/
Settings
Information #
Camera Canon EOS 5D (full
frame sensor)
2
Hemispherical
Lens
Canon Fisheye
Lens f 1/4L EF
8-15mm
2
Fisheye
Projection
Equisolid projection –
Tripod – 2
Camera settings:
- Aperture size fixed at f/5.6
- Shutter speed varies from 4 to
1/8000 seconds
- Sensitivity ISO 100
- White balance Daylight
Though HDR photography combines multiple
exposures to capture wide luminance variations within
a scene, clear sky photography with the sun can lead
to loss of information in the image and thresholding
difficulties while image processing (Jakubiec, 2014;
Stumpfel et al., 2004). It is best to conduct the
fieldwork with an overcast sky or when the sun is low
in the horizon (i.e. early morning or evening before the
sunset)
Image Processing – Thresholding and Masking
The multiple exposure sequence of photographs for
every set is processed using a software called
Photosphere (Ward, 2015). Photosphere uses a
computationally derived camera response function
(Inanici, 2006) to fuse the sequences of image into a
single HDR image. The HDR format is used to
determine the transmittance percentage whereas it is
also converted into a commonly used image format
with a balanced exposure for thresholding, masking
and determining the gap percentage.
Thresholding is the process of converting an image
into a binary image of black and white pixels. This is
the first and a significant step in image processing to
separate the crown from the background sky. The
precision in separation while thresholding depends on
conditions while the photograph is captured and the
method of thresholding used.
Two different methods to threshold the image were
tried: manual and automatic. Manual thresholding
requires each photograph to be processed individually,
removing the sky colour pixels in an image editing
software such as Photoshop. Once the sky pixels are
removed, the image can be converted using the
maximum threshold value. Automatic thresholding is
based on Nobis and Hunziker’s (2005) method which
Figure 4: (a) Scanning method of the computational script showing the minimum and maximum position
values of the black pixels. (b) The output image of the stored position values.
Figure 5: Steps of image processing. (a)Photograph taken onsite, with a blue cloth band around the trunk.
(b)Photograph converted into a thresholded image. (c)Cropped area of interest of the crown. (d)Tight outline
of the crown. (e) Area bounded by the outline within which the gaps will be calculated. (f) White pixels as
gaps within the outline.
searches for the threshold values that gives the highest
local contrast at the edges between the crown and the
sky. Manual thresholding allows individual control of
picking sky pixels, especially in images where the
contrast between sky and foliage are not high, but it is
not objectively reproducible and is time consuming.
On the other hand automatic thresholding is run as an
objective and procedural process, hence is consistent
and quicker with the thresholding results. Therefore,
in the method described here, the authors propose the
use of automatic thresholding. Once the images are
thresholded, the crown must be isolated from the rest
of the image. This process is called masking and is the
second step in image processing.
Masking requires two actions, one determining the
area of interest and the second drawing the outline of
the crown. Determining the area of interest is
achieved by cropping the crown from the rest of the
image for processing. This is a subjective step because
a) there is no definite boundary that separates foliage
of the crown from the trunk and branches, b) there can
be open gaps between the foliage and sky, which can
cause significant differences in gap percentages if
included or excluded and c) in cases where a part of
the crown’s background or foreground scene have
other objects like trees, buildings, lamp posts, etc.,
then cropping around these areas can differ.
In-order to reduce the subjectivity involved in this
step, the following measures are taken. a) A coloured
band is tied around the trunk of tree as close to the
crown as possible during on site photography such that
the band becomes the boundary that separates the
crown from the tree trunk. In all sets of photographs
this band becomes the cropping line. b) Where the
crown meets the sky; the cropping line is loosely
drawn. In the next step of masking, a script is
implemented to tightly fit the cropping line around the
crown. This automated script determines whether the
open gaps are included as a part of the crown or not
and c) In areas where the crown’s
background/foreground scene have other objects, a
tight cropping line is drawn manually to exclude the
objects from the area of interest of the crown.
After the area of interest is determined and cropped,
the second step is drawing a tight outline of this such
that it becomes a boundary within which the variables
of gap and transmittance percentages are measured.
The computational script implemented to draw the
tight outline is explained in Figure 4(a). The script
scans every row and column pixels of the thresholded
and cropped image of the crown. It then stores the
minimum and maximum position values of black
pixels in every row and column giving a canopy
outline depicted in Figure 4(b). Figure 5 summarises
all of the steps in image processing.
Image Analysis – Determining the Measured
Variables
Gap percentage of the crown is the percentage of the
gap sizes within the crown. The higher this percentage,
the higher the transmittance of light through the crown
will be. As shown in Figure 5(f) the white pixels
represent the gaps and black pixels the foliage. Hence
the gap percentage is:
As described earlier, the multiple exposure
photographs captured are converted to HDR format.
Hence the photometric properties of the entire scene
can be read pixel by pixel in all the six sets of
photographs ’facing the tree’ and ’facing the sky’.
Transmittance percentage is defined as the
luminance values of the gap pixels (i.e. white pixels in
the thresholding phase) facing the tree divided by the
luminance values of the sky pixels in the photographs
facing the sky. Figure 6 shows the gap pixels and the
corresponding sky pixels from the pilot study.
.
.
This method can be compared to the one proposed and
illustrated in Wilkinson et al. (1991) and McKennan
(1988) where discrete luminance measurements were
taken in rows across the crown and of the sky
immediately on either sides. The significant difference
in this method is that sky luminance values are not
measured from the sides of the crown rather from the
portion of the sky behind the crown.
Figure 6: (a) Luminance of gap pixels from
photograph facing tree (b) Luminance of sky pixels
from photograph facing sky
Table 2 illustrates the gap and transmittance
percentage for the tree crown profile measured as a
pilot study. The position number is shown in Figure 1.
Table 2: Measured Variables of pilot study - Gap
Percentage and Transmittance Percentage.
Position
Number
Gap
Percentage
Transmittance
Percentage
1 NE 6.9 5.8
2 SW 9.0 4.1
3 S 7.2 3.9
4 N 3.6 2.5
5 under
canopy 12.6 8.6
6 under
canopy
15.3 9.7
Predictive Lighting Calculations Using Measured
Data
Once taking measurements of trees, to make the
results useful it is imperative to translate the
measurements into a format for use within predictive
lighting simulation tools. The authors use the
Radiance (Ward 1994) reverse raytracing software to
perform such calculations through (1) generating
simplified canopy geometry and (2) applying an
appropriate reflectance value to said canopy.
In order to generate a simplified canopy model, a
triangulated mesh hemisphere is generated using
Marsaglia’s point choosing method to create 20,000
random, uniformly distributed vertices. Then the mesh
triangles are filled randomly by adding either single
triangles or clusters of triangles in order to maintain a
visual similarity to the actual tree canopy. Triangles
are removed until the area removed is equal to
Vertical gap percentage0.5 × Total hemispherical
Canopy Area. The vertical gap percentage used in this
process is the average of measurements 1–4 in Table
2, 6.7%. This approximates the situation where light
passes through two surfaces which each attenuate the
transmission by a set amount to arrive at the final gap
percentage. The resulting perforated hemisphere can
be scaled in the Z-axis in order to approximate an
appropriate canopy shape. Figure 7 presents a visual
comparison between the under canopy measurement 5
(from Table 2) and the resulting simulation canopy
geometry generated using vertical gap percentage
viewed from below. The tree trunk and major branches
were overlaid on top of the generated geometry in
black as they occlude the view but may not be included
in the vertical measurements.
The under-crown horizontal gap percentage generated
based on vertical measurements gives a value of
14.2%, a result between actual measured values at
positions 5 and 6.Material reflectance values are
applied to the triangular surfaces based on measured
data from a portable spectrophotometer. In the case of
the tree presented in this manuscript, the average of
the front and back leaf reflectance is 14.05%.
Image captured from position 5 with crown outline overlaid
Gap percentage: 12.6%
Geometric canopy model with branches and trunk overlaid
Gap percentage: 14.2%
Figure 7: Visual and gap percentage comparison
between reality and geometric model
The method outlined in this subsection can be
compared to the one proposed by Tregenza and
Wilson (2011), which uses a spherical surface with a
Radiance trans material applied (Ward 1994)
approximating tree canopy properties. Following our
method above, a trans material was created in
Radiance such that the diffuse reflectance is 14.05%
and the specular transmittance is 6.70.5, or 25.9%. The
comparison is made using a standard south-facing
reference room with an 87% window-to-wall ratio
located in Singapore (Reinhart, Jakubiec & Ibarra,
2013). A version of the model without trees is included
as a reference value. The results are displayed for 9am
on December 21st under a CIE clear sky in Figure 8.
Through an area-based luminance averaging
technique, the total transmitted amount of light is
found to be similar; however, the distribution of light
is very different. This can have significant impacts on
aesthetics as well as visual comfort evaluations. For
example, two standard visual comfort indices,
Daylight Glare Probability (DGP) (Wienold and
Christoffersen 2006) and the Unified Glare Rating
(UGR) (CIE 1995) were calculated for the images.
The speckled lighting from the detailed geometric tree
results in a significantly greater UGR glare value of
just acceptable compared to the trans material method.
DISCUSSION
As shown in Table 2, the gap and transmittance
percentage of position 5 and 6, which are horizontal
under canopy photographs, are almost twice the
percentage of that analysed from other positions which
are vertical crown profile photographs. This shows the
significance of measuring light through tree crown
profiles for evaluating daylight availability in
buildings rather than the commonly used under
canopy measurements.
The advantage of hemispherical HDR photography
over hemispherical photography is that we are able to
perform both photographic and photometric
measurements at the same time. Though the challenge
in photometric measurements is aligning the camera
‘facing the sky’ to cover the exact portion of the sky
covered by the canopy. Alignment is difficult mainly
due to topographical differences and other on-site
obstructions. However, this difficulty is not an issue
with under canopy photographs, where the camera
only needs to point upwards, and is a low-cost and fast
method to analyse LAI and gap fraction.
Hemispherical HDR photography would be
categorised as a non-contact indirect method to
measure LAI in the plant sciences and combines the
measurements—quantity of visible light and gap
percentage—currently taken using PAR
(Photosynthetically Active Radiation) sensors and
hemispherical photography. However, PAR sensors
measure quantity of visible light with a spectral
response required for plant photosynthesis and HDR
photography measures it using a spectral response
required for human vision.
Another benefit of utilizing HDR photography is that
it can achieve a better exposure than regular low
dynamic range photographs for the purposes of image
thresholding. Unfortunately, under windy conditions
the leaf edges will be blurred in the final outcome. This
is a limitation in the method which results in
necessitating the use of a single exposure image for
calculating gap percentage measured in windy
situations.
For design purposes, the proposed method can be used
to create a database of measurements of several kinds
of trees with varying canopy shape, gap and
transmittance percentage as well as gap patterns.
Measurements can also be taken under different
seasons and recorded. This information will help
designers such as architects and landscape planners to
predict and choose the right kind of tree for different
orientations of buildings, to evaluate the lighting and
to design for daylight availability while avoiding glare
simultaneously.
Perceptual qualities of spaces using dappled shadow
effects of vegetation or trees such as in vegetated
trellis walkways, walls, gardens etc. can also be
analysed using this method. From Figure 8, it is
abundantly clear that trees should be included
somehow in lighting simulation models when present
as they reduce the visible light by up to 75% in the
depicted situation. The benefits of utilizing a detailed
geometry based on gap fraction measurements are
aesthetic improvements and increased accuracy and
utility in the analysis of visual discomfort.
ACKNOWLEDGEMENT
The authors would like to thank Thommen George
Karimpanal for his help in image processing code in
Matlab. The authors further acknowledge support for
equipment from the SUTD-MIT International Design
Centre (IDC). Any opinions expressed herein are those
of the authors and do not reflect the views of the IDC.
No trees Trans material trees Detailed geometric trees
Ev (lx) 5,719 1,343 1,470
DGP 0.498 (intolerable) 0.238 (imperceptible) 0.269 (imperceptible)
UGR 18.1 (perceptible) 1.4 (imperceptible) 21.7 (just acceptable)
Figure 8: Visual comparisons of 9am on December 21st in Singapore under a CIE Clear Sky lighting condition
REFERENCES
Al-Sallal, K. A., & Al-Rais, L. (2013). A novel method
to model trees for building daylighting
simulation using hemispherical photography.
Journal of Building Performance Simulation,
6(1), 38–52.
Commission Internationale de l'Eclairage, CIE
Technical Committee 3-13. 1995. Discomfort
Glare in Interior Lighting. CIE Publication
117, Vienna.
Federal, U., & Catarina, D. S. (2012). The
Characterization of Masses of Vegetation as
Elements in Lighting Design - Test Method
Light through vegetation. In Plea2012 -
opportunities, limits & needs towards an
environmentally responsible architecture.
Lima.
Inanici, M. N. (2006). Evaluation of high dynamic
range photography as a luminance data
acquisition system. Lighting Research &
Technology, 38(2), 123. doi:
10.1191/1365782806li164oa
Jakubiec, J. A. (2014). The use of visual comfort
metrics in the design of daylit spaces (Doctoral
dissertation, Massachusetts Institute of
Technology). Retrieved from
http://hdl.handle.net/1721.1/91295
Jonckheere, I., Fleck, S., Nackaerts, K., Muys, B.,
Coppin, P., Weiss, M., & Baret, F. (2004).
Review of methods for in situ leaf area index
determination: Part i. theories, sensors and
hemispherical photography. Agricultural and
forest meteorology, 121(1), 19–35.
McKennan, G. (1988). Light Attenuation by Trees
(Tech. Rep. No. April). Department of
Architecture and Landscape Manchester
Polytechnic.
McKennan, G. (1995). The use of digital images for
data collection and analysis of light attenuation
by tree canopies. Arboricultural Journal, 19(3),
267–274.
Nobis, M. (2005). Sidelook 1.1-imaging software for
the analysis of vegetation structure with
truecolour photographs. Dostupné z http://www.
appleco. ch.
Nobis, M., & Hunziker, U. (2005, February).
Automatic thresholding for hemispherical
canopyphotographs based on edge detection.
Agricultural and Forest Meteorology, 128(3-4),
243– 250. doi: 10.1016/j.agrformet.2004.10.002
Reinhart, C. F., Jakubiec, J. A., & Ibarra, D. (2013).
Definition of a reference office for standardized
evaluations of dynamic façade and lighting
technologies. In Proceedings of BS2013: 13th
Conference of Intl. Building Performance
Simulation Assoc. (pp. 3645-3652).
Rich, P. M. (1990). Characterizing plant canopies with
hemispherical photographs. Remote Sensing
Reviews, 5(1), 13–29. doi:
10.1080/02757259009532119
Sadeghi, R., & Mistrick, R. (2015, aug). Modelling the
exterior surround - how much detail is
necessary? (14th International Radiance
Workshop ed.) [Computer software manual].
Stumpfel, J., Tchou, C., Jones, A., Hawkins, T.,
Wenger, A., & Debevec, P. (2004). Direct hdr
capture of the sun and sky. In Proceedings of the
3rd international conference on computer
graphics, virtual reality, visualisation and
interaction in africa (pp. 145–149).
Tregenza, P., & Wilson, M. (2011).
Daylighting:architecture and lighting design
(first ed.). Routledge.
Ward, G. (2015, feb). Photosphere, computer software
2005. Retrieved from http://www.anyhere.com/
Ward, G. J. (1994). The RADIANCE lighting
simulation and rendering system. In
Proceedings of the 21st annual conference on
Computer graphics and interactive techniques
(pp. 459-472)
Wienold, J., Christoffersen, J. 2006. Evaluation
methods and development of a new glare
prediction model for daylight environments with
the use of CCD cameras. Energy and Buildings
38: 743-757.
Wilkinson, D. M. (1991). Can photographic methods
be used for measuring the light attenuation
characteristics of trees in leaf? Landscape and
urban planning, 20(4), 347–349.
Wilkinson, D. M., Yates, D., & McKennan, G. (1991).
Light attenuation characteristics of seven
common british trees. Arboricultural Journal,
15(1), 37–44.
