Professional Documents
Culture Documents
To cite this article: Amir Nezamdoost & Kevin G. Van Den Wymelenberg (2017) Revisiting the
Daylit Area: Examining Daylighting Performance Using Subjective Human Evaluations and
Simulated Compliance with the LEED Version 4 Daylight Credit, LEUKOS, 13:2, 107-123, DOI:
10.1080/15502724.2016.1250011
KEYWORDS annual daylight metrics, annual sunlight exposure, LM-83, spatial daylight
autonomy
1. INTRODUCTION
The importance of occupant satisfaction in spaces with daylight is undeniable [Borisuit
and others 2014; Konis 2013; Van Den Wymelenberg and others 2010; Veitch and
others 2007]. Benefits related to occupant comfort, satisfaction, and well-being in daylit
Received 26 March 2016; revised 8
October 2016; accepted 14 October 2016 spaces such as offices and classrooms can be substantial [Heschong Mahone Group
Address correspondence to Amir 2003; Leslie 2003], and several codes and standard organizations have begun to promote
Nezamdoost, Integrated Design Lab, annual daylight simulation to support improved design (IESNA Daylight Metrics
University of Idaho, 306 S. 6th St., Boise, ID
83702, USA. E-mail: amirn@uidaho.edu Committee 2012; U.S. Green Building Council (USGBC) LEED V4 2013).
Color versions of one or more of the However, the interpretation of metrics from annual daylight simulation results is in its
figures in the article can be found infancy. Daylight simulation results need to be correlated to how occupants perceive the
online at www.tandfonline.com/ulks.
107
world [Reinhart and Weissman 2012]. A few studies have manual blind operation setting based upon LM-83 and point-
been conducted to that end. Referencing an extensive field in-time simulations were conducted with both blinds open
study of 61 real spaces with daylight and corresponding and closed. Sensitivity studies for point-in-time simulations
annual simulations and human factors results from experts were conducted to see which illuminance indicator thresholds
and occupants, the IES Daylighting Metrics Committee pub- ranging from 50 to 450 lx produced the smallest difference
lished LM-83. Following this, serving as a preliminary valida- when comparing students’ assessments of fully daylit, partially
tion for aspects of LM-83, Reinhart and others [2014] daylit, and non-daylit areas within the spaces.
published a field study of 13 real spaces where a team of
invited researchers from multiple universities compared grad- 2.1. Field Study
uate architecture students’ evaluations of a space’s “daylit The field study included three common study space types
area” to annual simulation results. However, this paper did following Heschong and Van Den Wymelenberg [2012]. A
not address patterns of blind use as required by LM-83. There detailed classification of space types (office n = 11, classroom
is a need to understand the relationship between point-in- n = 5, and other n = 6) is documented previously [Heschong
time and annual simulation results relative to human evalua- Mahone Group 2012]. Several characteristics for each study
tions. Though this article does not definitively address all of space is documented elsewhere [Nezamdoost 2015].
these gaps, it makes substantial contributions to several areas
of inquiry to improve the science and usefulness of both 2.2. Experimental Setup
point-in-time and annual daylight simulation.
From January 2012 to May 2015 field studies were
conducted by architectural graduate students in a seminar
2. METHODS course titled Daylighting Design and Simulation at the
The research plan was designed to address one primary University of Idaho in Boise [Van Den Wymelenberg 2012–
objective and several secondary objectives. The article’s pri- 2015]. Following Reinhart and others [2014] and Reinhart
mary objective is to compare simulated daylight results with and Weissman [2012], at the beginning of each field study,
daylit area boundary drawings conducted by students of students spent about 30–60 minutes in the room experiencing
architecture to determine what simulated parameters most it from multiple points of view and understanding the func-
closely correspond with student qualitative assessments of tional purpose of the space before being given the following
non-daylit, partially daylit, and fully daylit areas within instruction:
study spaces. This follows up and builds upon two recent In this exercise you are asked to follow your own intuition and
papers in this area [Reinhart and others 2014; Reinhart and divide the designated study area into a daylit and a non-daylit area.
Weissman 2012]. The secondary objectives are to determine Within the daylit area indoor illuminances levels due to natural
light should be adequate, useful and balanced for most of the year.
whether point-in-time or annual simulations more closely Please conduct your assessment individually without consulting
matched students’ perceptions of daylight sufficiency and with other students.
which illuminance thresholds provide the greatest predictive
ability for fully and partially daylit spaces. Additionally, Then, students conducted illuminance measurements and
an examination of whether each study space met the require- completed a three-page questionnaire. Whenever possible,
ments of LEED V4 Indoor Environmental Quality (EQ) evaluations were conducted with all electric lights off and
Daylight Credit was conducted. These secondary objectives any movable shading systems fully retracted (opened). In the
were not explicitly explored in previous publications. spaces that had automated shading systems (#20 and #21),
A substantial human factors field study was organized in the blinds remained in the automatically controlled position.
22 real spaces where graduate architecture students evaluated The research protocol (project 15-724) was reviewed by the
the spaces and documented the daylit area and non-daylit area university’s Institutional Review Board and was certified as
using plan drawings (Fig. 1). Studies were conducted over a exempt.
period of 3 years from 2012 to 2015 in Boise, Idaho, and
2.3. Daylit Drawings Evaluation
Seattle, Washington, in three space types (classroom, office,
other) including 260 individual daylit area occupant evalua- All paper-based daylit area evaluations completed during
tions. Daylight simulations were conducted using both annual the study period were gathered, digitally scanned, and
and closely equivalent point-in-time simulation settings for translated to vector line drawings in AutoCAD (Fig. 2,
comparative purposes. Annual simulations were run with a left As expected, the boundary areas drawn by each
Fig. 2 Area-based averaging function, classroom space #008. Left: Overlaid daylight boundary drawings of all students. Right: Resulting
fully daylit areas (white), partially daylit areas (grey), and non-daylit areas (dark grey).
Ambient bounces Ambient division Ambient sampling Ambient accuracy Ambient resolution Direct threshold
for comparison, such as continuous on-site irradiance data occupied hours (8 AM–6 PM) and partially daylit areas
collection and historical weather data retrieval, were had between 150 and 299 lx for at least 50% of the time.
considered. We crosschecked the student sky condition Figure 3 demonstrates the simulation-based analysis in one
descriptions with historical data and found reasonable example study space showing the fully, partially, and non-
agreement. However, we acknowledge that any of these daylit areas.
methods, including ours, have various limits to absolute
accuracy. 3. RESULTS
Similar to the students’ paper-based evaluations, we
determined a need to define a partially daylit area based
3.1. Comparing Daylight Availability Plots
upon the annual simulation results. Therefore, two levels Table 2 illustrates and compares the fully daylit (white),
of annual daylight sufficiency performance criteria based partially daylit (gray), and non-daylit (dark gray) areas in all
upon the metric spatial daylight autonomy (sDA) [IESNA 22 study spaces based on (1) students’ daylit evaluations
Daylight Metrics Committee 2012] were used. Similar to during field studies, (2) simulated daylight availability at the
the point-in-time analyses, fully daylit areas were desig- exact time of student’s evaluation, (3) simulated annual
nated as those having 300 lx for at least 50% of the daylight based on LM-83, (4) simulated annual daylight if
Fig. 3 Simulation-based analysis. Left: Standard daylight autonomy plots with 150 and 300 lx at 50% of time basis with windows always
retracted. Right: Resulting fully daylit areas (white), partially daylit area (grey) and non-daylit areas (dark grey).
Type of blind Louver blind Louver blind Louver blind + Roller shade
Students’ evaluation
Point-in-time
simulation
IES LM-83
Always open
Always closed
blinds were always open, and (5) simulated annual daylight drawings and the point-in-time simulation results for 12
if blinds were always closed. Table 2 is organized with out of 18 study spaces with students’ evaluations.
spaces having blinds of some type appearing first and spaces It is interesting to note some features of the six spaces
without blinds appearing last. A few spaces did not get where the students’ drawings noticeably do not align with
student daylit area evaluations but are included for compar- the point-in-time simulation results. Spaces #010 and
ison of annual and point-in-time analyses. #012 are located in the core of a building with a couple
Table 2 routinely reveals noticeable differences in the of interior relite windows and no direct daylight available.
size and shape of the fully, partially, and non-daylit areas The simulations suggest that the spaces are non-daylit, but
between the students’ daylit area drawings relative to the students found substantial fully daylit areas across the
simulated annual daylight results. Table 2 also shows whole length of space with some partial and non-daylit
reasonable similarity between the students’ daylit area areas adjacent to the relight windows. In spaces #011 and
Type of
Roller fabric shade Roller blind Curtain accent
blind
Students’
evaluation
Point-in-time
simulation
IES LM-83
Always open
Always
closed
#013, the students’ evaluations suggest that the spaces are responses as a function of wavelength when it is adapted
fully daylit with some partially daylit areas adjacent to the to light conditions (photopic vision) and dark conditions
window openings, whereas the simulations predict that the (scotopic vision) [Bass 1995].
first two thirds of the space (next to the windows) is daylit The two remaining unique spaces, #004 and #005,
and the last third (far from window) is partially daylit. The have clerestory openings facing to the south and north,
results for spaces #011 and #013 suggest a compelling respectively, that bring more natural light into the space.
example of how the nature of human visual perception South-facing clerestory windows provide top lighting to
for daylight is complicated. Students found substantial the space. Light is directed onto an interior wall facing the
daylit areas in a relatively dark space, which could be clerestory openings. The bright wall may indicate to the
due to visual adaptation. The human eye has different students that the space is “brighter” than it is when
Students’ evaluation
Point-in-time
simulation
IES LM-83
Always open
Always closed
considering only horizontal illuminance and consequently spaces #002, #004, #005, and #006. Accordingly, the
caused students to evaluate the entire area as daylit. These annual sunlight exposure (ASE) metric [IESNA Daylight
two spaces, in conjunction with #010–#014, begin to Metrics Committee 2012] was calculated for each space
support a hypothesis about the perception of vertical sur- separately and shown in Table 2. ASE reported zero for a
face brightness as a moderating variable, which is elabo- few spaces, #003, #010, and #012, which is not surpris-
rated in the Discussion section. ing given that they are located in the core zone of the
The spaces without installed blinds did not examine buildings without direct sunlight. Furthermore, spaces
blind use in simulation, per LM-83, and therefore no #002, #011, and #013 face north and also have low–
calculation of sDA by manual operated blinds (LM-83) visual light transmission windows, resulting in zero
was conducted in eight spaces: #002, #003, #004, #005, annual sunlight exposure during the occupied period, 8
#006, #007, #017, and #022. In several study spaces AM–6 PM. For spaces #019 and #022, all of the open-
without blinds, other strategies were used to control the ings are translucent and direct sunlight did not hit the
amount of sunlight, such as exterior shading devices in simulated sensors.
Students’ evaluation
Point-in-time
simulation
IES LM-83
Always open
Always closed
Table 3 indicates that 9 of 18 (50%) spaces are cor- [USGBC LEED V4 2013]. According to LEED V4, accep-
rectly predicted as either fully (preferred) or partially table spaces are those with at least 55% sDA300/50% and no
(nominal) daylit by the LM-83 blind control pattern more than 10% ASE1000,250. It is noteworthy that LM-83
relative to students’ evaluations. The other nine spaces does not propose formal absolute performance criteria for
were consistently underpredicted by LM-83, suggesting ASE1000,250 as it does for sDA300/50%; rather, it suggests
that students found them to be either partially or fully that ASE should be reported, compared in a relative man-
daylit but the simulation predicted the spaces to be non- ner, and higher ASE values are more worrisome. A similar
daylit 50% of the time. point has been made in other recent articles [Nezamdoost
Table 3 also shows sDA values for always open, always and Van Den Wymelenberg 2015; Reinhart 2015]. The
closed, LM-83, and ASE in all study spaces. In this table, absolute criteria of a 10% limit on ASE1000,250 was a LEED
study spaces were also evaluated based on LEED V4 V4 committee determination, and at the time of this
# 012 # 013
ASE = 0 ASE = 0
Students’ evaluation
Point-in-time simulation
IES LM-83
Always open
Always closed
publication, it is under committee consideration for amend- plotted, comparing the percentage difference of fully daylit
ment to moderately relax the limit. areas for point-in-time simulation and annual simulation
Figure 4 shows the sDA values of blinds operated (using 300 lx based on IES LM-83 documentation) versus
[IESNA Daylight Metrics Committee 2012] plotted next students’ evaluations.
to the always open and always closed blind conditions for
study spaces with blinds installed. 3.2.1. Point-in-time illuminance threshold
Given the increased accuracy of point-in-time simulations
3.2. Point-in-Time vs. Annual Simulation
with regard to student evaluations outlined in Section 3.1,
As noted above and detailed in Table 2, the general shapes a varied range of illuminance thresholds was chosen and
and sizes of the fully, partially, and non-daylit areas of examined in order to discover the most accurate threshold
students’ daylight evaluations and the simulated point-in- based on students’ evaluations for the delineation of fully,
time daylight illuminances show notable similarities. In partially, and non-daylit areas. The sensitivity analysis
contrast, greater discrepancy can be seen between the went through illuminance levels from 50 to 450 lx (50,
annual illuminance plots relative to students’ evaluations. 75, 100, 125, 150, 175, 200, 250, 300, 350, 400, 450)
In order to test this visually identified trend, Fig. 5 was for fully, partially, and non-daylit conditions in all study
Students’
evaluation
Point-in-time
simulation
Always open
Students’
evaluation
Point-in-time
simulation
Always open
# 017 # 022
ASE = 3.76 ASE = 0
Students’ evaluation
Point-in-time
simulation
Always open
spaces separately and then compared with students’ per- error in simulated horizontal illuminance plots relative to
ceptions (daylit boundary drawings) and the percentage students’ evaluations. This error generally occurs when
total difference, percentage correct prediction, false nega- vertical glazing or clerestory openings provide high levels
tive errors, and false positive errors were analyzed. of daylight washing vertical surfaces and/or the ceiling of
Figure 6 shows an example of this procedure in the non- the study space. Consequently, this surface brightness can
daylit category in space #001. In each plot, the black color impact students’ perceptions of daylit/non-daylit areas in a
shows the similarity between students’ evaluations and point- room and the horizontal task illuminance (simulation grid)
in-time simulation with specific thresholds. The dark grey does not necessarily exceed the required threshold to desig-
color illustrates false negative error, meaning that simulations nate that area as daylit. To test this concept, two categories
disagree that the area is fully daylit, partially daylit, and non- of study spaces were compared. The first group includes all
daylit (the result is negative). Nevertheless, it is present in study spaces and the second group consists of only spaces
students’ daylit evaluations. This scenario is reversed for the that do not meet the characteristics described above as
light grey zone, meaning that simulations determined a fully contributing to the human brightness perception phenom-
daylit, partially daylit, and non-daylit area in point-in-time enon (thus excluding #004, #011, and #013). Figure 8
simulations (the result is positive) though it is not present in illustrates this comparison using the average difference of
students’ drawings (the result is false). each group for specific illuminance thresholds. It is inter-
Figure 7 graphically displays the sensitivity study to esting to see that at higher illuminance levels (more than
determine which illuminance indicator thresholds (ranging 200 lx), the spaces without conditions likely to produce the
from 50 to 450 lx) produced the smallest difference with occupant brightness perception phenomena (group 2) pro-
students’ assessments of fully daylit, partially daylit, and duce lower discrepancy. The groups are statistically differ-
non-daylit areas within the study spaces. Additional ana- ent above 200 lx (P < 0.00075).
lysis is revealed for sky condition and fenestration type.
#001 50.63 0.25 41.6 10.33 57.16 11.10 32.29 Incorrect Fail Fail
(Underpredict)
#002 15.05 0.49 NB 0 53.82 23.11 23.15 Incorrect Fail Pass
#003 1.20 1.20 NB 0 0.00 41.15 58.85 Correct Fail Pass
#004 33.20 0.78 NB 1.18 92.18 5.67 2.21 Incorrect Fail Pass
(Underpredict)
#005 18.08 0.38 NB 12.02 47.37 47.69 4.94 Incorrect Fail Fail
#006 97.62 87.3 NB 5.2 90.10 10.08 0.00 Correct Pass Pass
#007 97.22 2.38 NB 15.0 83.91 16.35 0.00 Correct Pass Fail
#008 98.97 93.85 97.95 2.59 86.50 11.67 1.83 Correct Pass Pass
#009 85.50 51.48 83.73 2.08 88.53 8.00 3.47 Correct Pass Pass
#010 0.37 0.37 0.37 0 95.45 3.93 0.63 Incorrect Fail Pass
(Underpredict)
#011 62.5 0.74 62.5 0 79.33 17.05 3.62 Near correct Pass Pass
(Underpredict)
#012 0.64 0.64 0.64 0 65.89 24.36 9.75 Incorrect Fail Pass
#013 86.32 0.85 86.32 0 94.80 3.99 1.21 Correct Pass Pass
#014 73.33 10.88 64.91 10.6 No evaluations No evaluations No evaluations — Pass Fail
#015 98.19 13.32 87.58 32.48 92.41 5.88 1.71 Correct Pass Fail
#016 67.05 3.88 57.53 18.89 No evaluations No evaluations No evaluations — Pass Fail
#017 6.51 0.47 NB 3.76 25.94 15.80 58.26 Incorrect Fail Pass
#018 93.55 93.55 93.55 36.89 No evaluations No evaluations No evaluations — Pass Fail
#019 93.24 90.54 93.24 0 No evaluations No evaluations No evaluations — Pass Pass
#020 81.28 1.02 79.87 6.28 60.93 39.07 0.00 Incorrect Pass Pass
#021 93.16 85.02 92.88 12.96 100.00 0.00 0.00 Correct Pass Fail
#022 98.22 54.38 NB 0 85.55 14.73 0.00 Correct Pass Pass
NB: Spaces with no blinds installed (No Blinds). Underlying values show two sides of comparisons between simulation results and students’ evaluations to find students’ prediction
status. Bolded “Pass” words indicate to those spaces which pass LEED V4 daylight credit (Both sDA and ASE are passed).
90
80
60
50
40
30
20
10
0
#001
#020
#016
#011
#014
#009
#013
#015
#021
#019
#018
#008
Study space
Always Open Always Closed LM83
Fig. 4 sDA plot by blinds always open, closed, and operated [IESNA Daylight Metrics Committee 2012] sorted by LM-83 values.
# 022
# 021
# 020
# 017
# 015
# 013
Study space
# 011
# 009
# 008
# 007
# 006
# 005
# 004
# 002
# 001
0 10 20 30 40 50 60 70
Percent of difference between annual and point-in-time simulation results
versus students’ evaluations (%)
Annual Point in time
Fig. 5 Comparing percentage difference of fully daylit areas in point-in-time and annual simulation results versus students’ evaluations.
In addition, manually operated blinds [IESNA Daylight operated condition required by LM-83 is. A few other
Metrics Committee 2012] plus always open and always findings include the following:
closed blind conditions were compared and analyzed
according to sDA (in Fig. 4). At first glance it is interesting High values of sDA for the always closed blind condition in
those spaces with skylights show the importance of top lighting
to see how close to the always open condition the manual
strategies to provide robust daylight performance even while all of
Fig. 6 Non-daylit area discrepancy plots in range of 25–200 lx illuminance level thresholds in open office study spaces. Black, correct; dark
grey, false negative error; light grey, false positive error).
the windows are covered with closed blinds. At the same time, predictive ability and less discrepancy was recorded regard-
spaces with only vertical fenestration with blinds always closed ing students’ drawings. Using these indicator values the
show very low sDA values, indicating how critical blind use is in
annual daylight performance evaluation in these spaces.
difference decreased by 18.39%, from 21.6% to 17.6%
Blind operation is most forgiving (or least sensitive) to spaces for point-in-time, and by 21.47%, from 27% to 21.2%,
with north-facing fenestration and top lighting. Spaces that used for annual simulation.
both top lighting and side lighting strategies are robust against blind It is interesting that space #015 shows lower discre-
closure (spaces #008, #009, #018, #019, and #021). Spaces facing
north (spaces #011 and #013) show no considerable difference for pancy in annual simulation than for point-in-time simula-
blind operation, whereas sidelit spaces with south, east, and west tion results, and this is the students’ studio space where
orientations show greater difference by blind pattern. they spent a great deal of time on an annual basis. The
reduced error for point-in-time results, with this excep-
In all study spaces (except space #015), point-in-time tion, suggests that students cannot accurately integrate
simulation results had less difference between student day- evaluations across imagined annual sky conditions. The
lit area evaluations than did annual simulation results. foundational research behind LM-83 relied on feedback
Considering all study spaces, the average percentage dif- from both daylighting experts and naïve, but long-term,
ference in point-in-time simulation was 21.6%, whereas building inhabitants. It is difficult to assess how well
annual simulation produced a 27% difference (with blinds experts and building inhabitants can integrate evaluations
always open). Using modified indicator values of 100 and across annual sky conditions, but this finding suggests the
250 lx (partially and fully daylit, respectively) and consid- importance of permanent space occupants in human fac-
ering both point-in-time and annual simulation increased tors research regarding annual daylight performance.
35
30
25
20
15
10
100
150
200
250
300
350
400
450
25
50
75
Fig. 7 Sensitivity study on illuminance indicator thresholds (ranging from 50 to 450 lx) for point-in-time data.
Simulation sensitivity studies were conducted to see with the geographic areas studied. As the sample of build-
which annual illuminance indicator thresholds (ranging ings published in the literature increases in number and
from 50 to 450 lx) produced the smallest difference when diversity (geographic location, building use type, sky con-
comparing students’ assessments of a daylit area. The fol- dition, aperture type, et cetera), it is likely that these
lowing horizontal illuminance thresholds emerged from the indicator values will become accordingly more nuanced.
sensitivity analysis as representing consistently low discre- Further space-by-space results interpretation is available
pancy between student evaluations and point-in-time simu- elsewhere [Nezamdoost 2015].
lation results; the threshold between non-daylit and partially Finally, only 8 of 22 spaces studied meet both the sDA
daylit produced 100 lx and the threshold between partially and ASE requirements of the LEED V4 criteria for the
daylit and fully daylit produced 250 lx. These are slightly EQ daylight credit [LEED V4 2013]. However, 14 of the
lower than previously published indicator values (150 and 22 spaces met the sDA criteria portion of the LEED V4
300 lx). This is possibly due to the relatively young sample EQ daylight credit. Therefore, 6 of the 14 spaces that
of human participants in this study or may be connected failed the LEED daylight credit is due to the limit of the
400
300
250
200
0 5 10 15 20 25 30 35
Percent of difference of fully daylit areas for point-in-time simulation results
and students’ evaluations (%)
All spaces Spaces without conditions likely to produce the occupant brightness perception phenomena
Fig. 8 Point-in-time simulation comparison of spaces designated with and without characteristics contributing to the occupant brightness
perception issue.
ASE 10% required by LEED V4. Interestingly, of the 18 Our research suggests that student evaluations correlate
spaces with student daylit area drawings (a subset of the 22 better with point-in-time data and that they have a more
total spaces), students designated that 11 spaces had more difficult time integrating their evaluations across year-
than 75% of the floor area as “fully daylit,” suggesting that long daylighting phenomena. The inverse was true,
they felt that these spaces had sufficient daylight available. however, in the space that served as their design studio.
Nine of these 11 spaces also met the sDA component of Furthermore, for point-in-time analyses, student evalua-
LEED V4. The two that students said were daylit but did tions conducted under clear skies and in study spaces
not achieve a high sDA are possibly explained by the with only vertical fenestration showed greater discre-
vertical surface brightness phenomenon. It is our opinion pancy with simulated results for the non-daylit and
that this ASE threshold, as it is currently defined by LEED partially daylit designations than those conducted
V4 and calculated via simulation, is too restrictive and under overcast skies and those in study spaces with top
warrants further research before these criteria are imple- lighting. On the other hand, student evaluations con-
mented in additional reach standards or codes. However, ducted under overcast skies and in study spaces with
the recommendations within LM-83 for its establishment only vertical fenestration showed greater discrepancy
as a metric and that it can be used at present as a relative with simulated results for the fully daylit designation.
performance metric (as opposed to an absolute metric) in This reinforces the difficulty that students had integrat-
lighting design applications. ing several factors in their assessments of a daylit area,
and these findings suggest the value of the opinions
gathered from long-term occupants in this type of field
5. CONCLUSION research.
In recent daylighting field research [Heschong Mahone This article identifies and explains inconsistencies in
Group 2012; Reinhart and others 2014; Reinhart and manual blind operation with regard to the determination
Weissman 2012] it has become an increasingly common of spaces as “nominally daylit” or having “preferred day-
method to collect subjective point-in-time comfort data light” according to LM-83 terminology. Additional
and correlate these data with annual simulation results. human factors blinds operation behavior research is