Building and Environment 113 (2017) 22e38

Contents lists available at ScienceDirect

Building and Environment

journal homepage: www.elsevier.com/locate/buildenv

A novel circadian daylight metric for building design and evaluation

Kyle Konis
University of Southern California, United States

a r t i c l e i n f o a b s t r a c t

Article history: This paper extends the applicability of emerging frameworks for evaluating the non-visual effects of light
Received 9 June 2016 through the development of a novel area-based daylighting metric addressing goals of human circadian
Received in revised form stimulus and entrainment in buildings. Procedures using annual, climate-based daylight modeling of
8 November 2016
eye-level light exposures are developed to analyze and map indoor environments in regard to spatial and
Accepted 12 November 2016
seasonal changes in the availability of a circadian-effective daylight stimulus. Because the biological
Available online 14 November 2016
effects of light exposure are not instantaneous, a novel approach is developed to assess the duration of an
effective stimulus on a daily basis, as well as the frequency an effective stimulus is present over the
Keywords:
Daylighting
course of a year. Results can be used to identify and visually examine building zones where long-term
Circadian system occupancy may lead to disruption of the circadian system in the absence of supplemental electrical
Climate based daylight modeling lighting capable of effective circadian stimulus. The metric and visualization techniques are implemented
Daylighting metrics in a parametric, simulation-based workflow utilizing publicly available software tools. The workflow can
Health be used to assess and differentiate the performance of various daylighting strategies during the design
phases of a project, or to examine existing spaces. The applicability of the workflow is demonstrated
using two example models: a portable school classroom, and a generic open-plan commercial office floor
plate.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction average [7], lighting is often provided by electrical sources that are
adequate for performance of visual tasks (i.e. stimulation of the
Standards and practices for lighting design were developed to visual system), but can lack the appropriate spectral composition
serve human visual needs prior to scientific understanding of the and intensity required to stimulate the circadian system. All zones
important role light plays in maintaining healthy human biological within a building that do not regularly achieve the lighting condi-
functions. The discovery of a third class of photoreceptors in the tions necessary for effective circadian stimulus can be labeled as
human retina [1e4], referred to as Intrinsically Photoreceptive biologically dark, and considered as zones where sustained occu-
Retinal Ganglion Cells (ipRGCs), has led to a growing interest in the pancy over extended time periods (e.g. regular workday schedules)
non-visual effects of light on human health and well-being. In may present a risk for disruption of the circadian system in the
contrast to rod and cone photoreceptors, which serve as inputs for absence of supplemental electrical lighting capable of effective
low-light and color vision, the ipRGCs serve no visual (image- circadian stimulus.
forming) function. Instead, ipRGCs play a critical role in synchro- As evidence of the health impacts of light exposure grows, it is
nizing human circadian rhythms to the 24-h light/dark cycle of the import for designers to have metrics and guidance to evaluate
local environment. Notably, the action spectrum of light for the project performance in regard to the non-visual effects of light
circadian system is shifted towards shorter wavelength (~490 nm) alongside more commonly used lighting metrics related to visual
“blue” light relative to the visual system, which is maximally sen- task performance (e.g. horizontal workplane illuminance and illu-
sitive to (~555 nm) “green” light [5,6]. As a result, humans are not minance uniformity), visual discomfort (e.g. probability of glare),
well equipped to self-report the presence or intensity of circadian- lighting energy savings (e.g. electrical lighting energy reduction
effective light based on visual perception. from photocontrols), and Indoor Environmental Quality (IEQ) (e.g.
Inside buildings, where adults spend 87% of their lives on compliance with the U.S. Green Building Council's LEED Daylighting
Environmental Quality (EQ) credit) [8].
There are currently no minimum requirements for daylight ac-
E-mail address: [email protected].
cess in buildings to support circadian entrainment. However, the

http://dx.doi.org/10.1016/j.buildenv.2016.11.025
0360-1323/© 2016 Elsevier Ltd. All rights reserved.
 K. Konis / Building and Environment 113 (2017) 22e38 23

International WELL Building Institute has recently developed a building design. These include the timing, intensity, duration,
building certification system with the stated objective of, wavelength and past history of light exposures [11]. The following
“measuring, certifying and monitoring the performance of building sections describe the rationale and assumptions made in regard to
features that impact health and well-being” [9]. One of the pre- each of these parameters to develop a procedure to analyze and
conditions for certification, (entitled “Circadian Lighting Design”), map a space in terms of the frequency of a circadian-effective
is the provision of sufficient melanopic light intensity for work daylight stimulus.
areas. The term “melanopic” refers to a new photometric measure
of light intensity weighed by the sensitivity of the melanopsin- 2.1. Spectrum and intensity of light exposure
containing ipRGCs, and is discussed further in Section 2.1. While
the precondition does not require the use of daylight to meet the To study the potential circadian effects of various light sources it
requirement, the contribution of daylight can be included in is first necessary to quantify light exposure in biologically mean-
simulation-based predictions (the specific compliance criteria and ingful units. Fig. 1 shows the spectral efficiency function of the
their current ambiguities are discussed in greater detail in the melanopsin-containing ipRGCs (black curve) developed by Enezi
following sections). Despite the fact that compliance can be ach- et al. and Lucas et al. [12,13], referred to as the melanopic spectral
ieved exclusively through the use of electrical lighting, it is antici- efficiency function (annotated here as C-lambda). The melanopic
pated that designers will seek to meet such requirements to the spectral efficiency function can be used to calculate melanopic
extent possible through the use of daylight, and supplement illuminance (reported in units of Equivalent Melanopic Lux (EML))
insufficiently-daylit zones with appropriate electrical lighting. for various light source Spectral Power Distributions (SPD) [14].
The emergence of requirements for circadian lighting design Fig. 1 also shows the spectral efficacy function of the visual (phot-
signals a growing interest in the challenge of translating scientific opic) system (V-lambda) along with the SPDs of three Commission
knowledge into actionable information that can be applied to Internationale de l'Eclairage [15 CIE] daylight illuminants, (D55)
improve the well-being of building occupants. It should be sunlight, (D65) overcast sky, and (D75) north sky daylight. Fig. 1
emphasized that the development of a circadian daylight metric shows that the maximum efficacy of the circadian system (C-
relies on a combination of available scientific information and lambda) is more closely aligned with the maximum power of the
expert judgments related to the timing, intensity, duration, wave- three daylight SPDs compared with the photopic function (V-
length and past history of light exposures. A rationale for how each lambda). In contrast, Fig. 2 compares the spectral response of the
of these factors is addressed is included in this paper along with visual system (V-lambda) and the circadian system (C-lambda) to
discussion of additional factors that are not directly related to the spectral power distribution of a “standard” fluorescent lamp,
building design, such as age and work schedule requirements. The (CIE illuminant F11), which represents a narrow tri-band fluores-
reader should expect the judgments made in this paper to be cent of 4000
Kelvin color temperature. Fig. 2 shows that the peak
revisited as scientific understanding of the human non-visual power of two of the three most prominent wavelength bands fall
response to light evolves. Nevertheless, it is important for de- largely outside sensitivity of the circadian system (C-lambda). The
signers to have access to design support tools and performance introduction of EML as a unit enables designers to differentiate the
criteria developed on available knowledge to specifically address relative “circadian efficacy” of various light sources (such as
non-visual effects of light during design. Such tools can help de- daylight vs. fluorescent) that may produce the same visual effect.
signers to better assess, understand and improve the circadian Several researchers have proposed models of the spectral
effectiveness of various daylighting strategies. sensitivity of the circadian system that can be used to relate the SPD
This paper extends the applicability of emerging frameworks for from various light sources to objective and subjective stimulus ef-
evaluating the non-visual effects of light through the development fects. The model developed by Rea et al. [16] is based on published
of a novel area-based daylighting metric addressing goals of human studies of nocturnal melatonin suppression using lights of various
circadian stimulus and entrainment in buildings. Procedures using SPDs. The model relates a given SPD to a Circadian Stimulus (CS)
annual, climate-based daylight modeling of eye-level light expo- effect from 0% (no effect) to 70% (maximum suppression level
sures are developed to analyze and map a space in terms of the achievable after 1-h) characterizing the relative effectiveness of the
frequency of a circadian-effective daylight stimulus. Because the source as a stimulus. The model can be applied to convert various
biological effects of light exposure are not instantaneous, a novel light sources to units of Circadian Lux (CLA) for relative comparison
approach is developed to assess the duration of an effective stim- using a publically available circadian stimulus calculator [17]. The
ulus on a daily basis, as well as the frequency an effective stimulus model developed by Andersen et al. [18] is based on both nighttime
is present over the course of a year. Results can be used to identify, [19] and daytime [20] studies and “sets a tentative lower and upper
quantify and visually examine building zones where long-term bound for the likelihood that a given light exposure will have an
occupancy may lead to disruption of the circadian system in the effect on alertness,” with a liner ramp-function applied to interpret
absence of supplemental electrical lighting capable of effective intermediate values. The upper and lower bounds of the model can
circadian stimulus. The metric and visualization techniques are be converted into the standard photometric unit of illuminance
implemented in a parametric, simulation-based workflow utilizing (lux) using the approach described in Pechacek et al. [21] for any
publicly available software tools. The workflow can be used to SPD of interest by applying a conversion factor. For example, for
assess and differentiate the performance of various daylighting D65, the lower bound is 190 lux, and the upper bound is 870 lux.
strategies during the design phases of a project, or to examine Finally, Amundadottir et al. [22] have developed a framework to
existing spaces. The workflow files are available for download here describe the circadian effectiveness of light that can be explored
[10]. The applicability of the workflow is demonstrated using two using an online calculation and visualization tool [23]. The frame-
example models: a portable school classroom, and a generic open- work incorporates dose-response models of melatonin suppres-
plan commercial office floor plate. sion, melatonin phase shift, and perceived alerting effect, enabling
users to predict and compare the biological effect for various light
2. Previous work source SPDs. The framework incorporates a lens transmittance
model [24] and requires the user to specify the age of the observer
There are a number of parameters known to control the circa- to account for the relative loss in retinal exposure due to age. In
dian system's response to light that are directly impacted by specifying any threshold level, the age of the occupants is an
24 K. Konis / Building and Environment 113 (2017) 22e38

Fig. 1. Comparison of the relative spectral power distributions of three CIE daylight illuminants: (D55) sunlight, (D65) overcast sky, and (D75) north sky daylight along with
normalized photopic (V-Lambda) and circadian (C-Lambda) spectral efficiency curves. Note: Both response curves are scaled to have equal area under the curves.

100

90

80
C (λ) F11
Relative Power (or Efficacy)

70

60
V (λ)
50

40

30

20

10

390 410 430 450 470 490 510 530 550 570 590 610 630 650 670 690

Wavelength [nm]

Fig. 2. Comparison of spectral response of the visual system (V-lambda) and the circadian system (C-lambda) to the spectral power distribution of the CIE illuminant F11. Note: Both
response curves are scaled to have equal area under the curves.

important consideration, as the relative level of light reaching the (melatonin suppression level) is provided in Table 1, which are
retina decreases due to age. Given that buildings should be ex- derived using a 65-year-old observer model. The illuminants “A”,
pected to effectively accommodate a range of ages, the author “F11” and “D65” are standard CIE illuminants. Outcomes for a 9500

compared threshold criteria discussed in the following section Kelvin color temperature LED are provided in the final column. For
against a 65-year-old observer model. reference, the 65-year-old observer model requires a 6% greater
A comparison of the EML for various light sources and photopic stimulus intensity relative to a 32-year-old to achieve an equivalent
illuminance levels, along with the resulting biological effect effect, or a 11% greater intensity relative to a 10-year-old observer.

Table 1
Biological impact of various light sources and photopic illuminances.

Melatonin suppression (%) EML A (Lux) F 11 (Lux) D 65 (Lux) LED 95 (Lux)

0.5 17 29 27 16 14
5.0 34 56 52 31 27
25.0 56 95 87 52 45
50.0 77 129 118 71 62
75.0 105 176 161 97 84
95.0 176 296 272 162 142
99.5 341 575 526 315 275
 K. Konis / Building and Environment 113 (2017) 22e38 25

At present, there is no consensus for the appropriate minimum rest/activity patterns [28]. Humans possess an internal biological
light exposure threshold to ensure effective circadian stimulus in clock that regulates daily patterns of activity following the natural
buildings, or for the duration at which the effects of light exposure 24-h light/dark cycle. The suprachiasmatic nuclei (SCN) hosts the
saturate. The WELL Building Standard's Circadian Lighting Design circadian clock (or circadian system) responsible for orchestrating
precondition (option 1) implements a minimum threshold of 250 the daily timing of physical, mental and behavioral changes, for
EML (equivalent to 226 lux from D65), which must be available for example, sleep/wake, alertness level, mood, hormone suppression/
at least 4 h each day and can be provided at any point during the secretion, and core body temperature. The internal period of the
day. As noted previously, this requirement can be met with human circadian rhythm can range between 23.5 and 24.7 h, with
daylight, electrical light (exclusively), or a combination of both an average of 24.2 h among healthy adults [29] and relies on a
sources. The 250 EML threshold and 4-h exposure requirement resetting response driven by light received at the retina to maintain
currently implemented in the WELL Building Standard are based on entrainment with the local 24-h light/dark cycle. Lack of an effec-
best judgments derived from recent studies [25,26] and should be tive light stimulus at the appropriate time during the day can
expected to be refined as the relationships between spectral dis- disrupt the circadian system. For example, most individuals who
tribution, duration, timing, and intensity of light exposure for exhibit total blindness and consequently lack access to a daily
optimal circadian health are further clarified by the research resetting response from light suffer from “non 24-h sleep wake
community. For comparison, Figueiro et al. recommend exposure to disorder,” in which the internal circadian rhythm becomes out of
a CS of 0.3 or greater at the eye for at least 1 h in the early part of the phase with the local 24-h light/dark cycle [30,31]. Disruption of the
day (equivalent to 180 lux, D65) [27]. circadian clock can lead to poor sleep, reduced alertness, and
increased risk of a range of health maladies including diabetes,
2.1.1. Definition of circadian effect thresholds obesity, cardiovascular disease and cancer [32]. As noted by Zelinski
While it is arguably too early to propose a precise minimum et al. [32] the most severe health risks, such as cancer, diabetes, and
threshold for adoption in an international standard, it is possible to coronary problems are identified from studies of populations
establish a threshold derived from available research and emerging engaged in shift work or employment on a rotating schedule,
health-based standards (e.g. WELL) that can serve as an indicator of where an individual is required to deviate his/her sleep/wake
the presence (or lack) of an effective circadian stimulus at a pattern from the local 24-h light/dark cycle. For these populations,
particular measurement location over a specified time period. The health risks cannot be addressed by improved access to an effective
approach taken by the author is to use the same 250 EML threshold circadian stimulus during the regular (i.e. 6:00e10:00 a.m.) circa-
implemented by the WELL Building Standard, but segment the dian resetting period. However, these findings illustrate the risk of
analysis to the circadian resetting period of the day (6:00e10:00 circadian disruption in humans. Therefore, although the long-term
a.m.). As shown in Fig. 3, the effect of light on melatonin suppres- health risks of insufficient or inappropriately-timed light exposure
sion follows a nonlinear dose-response curve. Using the model on healthy building occupants working on regular schedules is less
provided by Amundadottir et al. [22] (see Fig. 3), which is based on established, a precautionary approach warrants the provision of a
the findings of Cajochen et al. [19] and Zeitzer et al. [25], The 250 circadian-effective and appropriately timed light stimulus on a
EML threshold is shown to predict nearly full saturation (98.5%) in daily basis throughout the year to avoid the risk of circadian
melatonin suppression for a 65-year old observer. For comparison, disruption.
the light stimuli (in EML) required to achieve 5% and 50% effects are For a typical well-rested and regularly-sleeping individual, a
shown as vertical lines in Fig. 3 and listed in Table 1. light stimulus in the early morning will advance the circadian clock,
causing earlier wake-up time and earlier sleep onset. Light received
in the evening will delay the circadian clock, causing later wake-up
2.2. Timing and duration time and later sleep-onset. Light received in the middle of the
biological day will have limited effect on circadian advancement or
The timing of light exposure during the day plays an important delay, but has been shown to cause reduced levels of sleepiness and
role in synchronizing the rhythm of the circadian system with daily

5% 50% 98.5% Suppression

100

90
34 77 250
Melatonin Suppression Level (%)

80

70

60

50

40

30

20

10

0 50 100 150 200 250 300

Light Dose (EML)

Fig. 3. Dose response curve showing relationship between light exposure and melatonin suppression level.
26 K. Konis / Building and Environment 113 (2017) 22e38

higher levels of subjective alertness [20,33]. Table 2 provides a towards climate-and-context-representative Climate Based
summary of the three daily time periods proposed by Ref. [18] to Daylight Modeling (CBDM) [37,38]. While one of the stated objec-
categorize time-varied light exposures according to their expected tives of the current LEED Daylight EQ credit compliance approach is
non-visual effect. Notably, the WELL Building Standard does not to “reinforce circadian rhythms,” the reliance of the compliance
specify the time during the day when the light stimulus must be process on the metric Spatial Daylight Autonomy (sDA) [39] is
present. Therefore, a space that achieves the compliance criteria problematic as a proxy indicator for non-visual effects of daylight
(250 EML for 4 h each day) from 12:00e4:00 p.m. could meet the for several reasons. First, the procedures to calculate sDA are not
requirement while potentially failing to provide adequate stimulus sensitive to timing of light during the day. For example, morning
for circadian resetting during the morning. Finally, past history of light exposure will have an advancing effect on the circadian clock,
light exposure has an effect on sensitivity of the circadian system to while early evening exposure will have a delaying effect. sDA is
light [34]. Higher levels of light exposure during the day cause the calculated by taking a count of all analysis grid point locations that
sensitivity of the circadian system to decrease over time, and lower achieve the specified Daylight Autonomy (DA) threshold of 300 lux
exposure levels causes sensitivity to increase. A detailed review of horizontal illuminance for at least 50% of occupied hours during the
the parameters that control the response of the circadian system to year (eg. 8:00e18:00, Monday-Friday), and dividing the count by
light can be found in Amundadottir et al. [35]. the total number of analysis grid points to determine the spatial
Finally, the regularity of exposure to an effective light stimulus is average. Thus, a workspace could meet the DA criterion without the
an important consideration. It is known that sleep and other daily presence of any daylight during the morning circadian resetting
rhythms in physiology and behavior evolved in the natural light/ period (6:00e10:00 a.m.) [18]. Second, the sDA metric relies on
dark cycle. And, it is known that circadian rhythms reflect the horizontal illuminance measurements which are recommended to
programming of biological activities to the periodic nature of the be acquired on a workplane located 0.72 m (30in) from the floor.
natural environment [36]. Because the natural light/dark cycle However, the circadian system responds to vertical light exposure
follows a recurring (24-h) pattern, it can be argued that the design at eye level. Third, the current illuminance threshold of 300 lux
objective should be to provide an effective circadian stimulus on a exceeds the level of daylight illuminance needed for circadian
daily basis throughout the year. However, circadian rhythms are stimulus and thus discounts hours that may be effective for circa-
endogenous in nature, meaning that they are self-sustaining and dian stimulus. Fourth, the procedures to calculate sDA allow for
can persist for a significant period of time in the absence of an substantial temporal periods where daylight illuminances are
external timing cue. And, circadian rhythms take time to adjust allowed to fall below the specified thresholds. For example,
when the period of the external timing cue is changed. The assuming a project is occupied 2000 h per year, a grid point location
observation of an air-traveler's circadian rhythms remaining syn- could fail to meet a given threshold illuminance for 999 of the 2000
chronized with the traveler's original time zone is an example. total occupied hours and still be considered Daylight Autonomous.
Therefore, it may be possible to maintain circadian entrainment in Such an outcome is problematic for the maintenance of the human
buildings even when a stimulus is not present for an individual day, circadian system due to the need for regular (i.e. daily) exposure to
or perhaps multiple days, so long as the stimulus is periodically sufficient light. Finally, the LEED compliance procedures reward
available within a window of days (e.g. on a weekly or monthly projects where only a portion of the regularly occupied space must
basis). While humans exhibit the ability to adapt and adjust to meet the compliance criteria (e.g. 55% or 75% of floor area). Due to
changes in environmental conditions, it can be argued that a metric the significant health risks of circadian disruption, regularly occu-
addressing the effectiveness of daylight for circadian stimulus pied areas of a project that do not meet minimum performance
should be capable of assessing the regularity of light exposure over requirements for circadian stimulus should be clearly delineated to
a period of time, and capable of differentiating measurement lo- ensure that alternate means of effective circadian stimulus are
cations where a stimulus is available frequently (e.g. 5 days per provided to occupants.
week) from infrequently (e.g. 2 days per week). Addressing this task Preliminary steps have been taken to develop calculation
requires the development of methods to assess stimulus frequency methods and workflows utilizing lighting simulation software to
for particular view vectors. While the minimum frequency needed investigate the non-visual effects of daylight in buildings. Working
to maintain healthy circadian stimulus is not known (e.g. is a in collaboration, Andersen et al. [18] and Mardaljevic et al. [40]
stimulus on only Monday, Wednesday, and Friday sufficient?), it have proposed a preliminary framework for predicting the non-
can be argued that measurement locations that have more frequent visual effects of daylight, informed by a review of outcomes of
availability of an effective stimulus should be valued over those photobiology research. The framework incorporates a
where availability is less frequent. A preliminary approach to photobiology-based lighting model to predict the magnitude and
measuring stimulus frequency is presented in Section 3.5 and in- direction of a circadian effect based on parameters of intensity of
tegrated into the circadian daylight metric proposed by the author. vertical illuminance at the eye, light source spectrum, and timing
(over a 24-h period). Mardaljevic et al. implemented the model in
2.3. Previous work relating circadian entrainment to architecture an “in house” climate-based daylight modeling software workflow.
The workflow is sensitive to the spectral character of the estimated
Daylighting guidance has only recently begun to shift from the sky condition to the degree that it differentiates between three CIE
near universal application of the Daylight Factor (DF) approach illuminant types (D55, D65 and D75) to account for relative

Table 2
Subdivision of Day Based on Non-visual Effect (After Andersen et al.).

Daily time Non-visual effect Description

period

6:00e10:00 Circadian resetting Sufficient daylight illuminance can serve to phase advance the clock in the majority of people.
10:00e18:00 Alerting effects of High levels of daylight illuminance may lead to increased levels of subjective alertness without exerting substantial phase shifting
daylight effects on the clock
18:00e6:00 Bright light avoidance Daylight exposure that might trigger the non-visual effect is to be avoided so as not to disrupt the natural wake-sleep cycle
 K. Konis / Building and Environment 113 (2017) 22e38 27

differences in their contribution to circadian stimulus. The output occupant behavioral adjustments (e.g. moving to better daylit zone
of the framework is a graphic, color-coded visualization which for a period of time during the morning). Addressing these limi-
presents a cumulative, annual summary of the occurrence of non- tations requires the additional assumption of a minimum accept-
visual effects (the “circadian potential”) for a specified set of loca- ability threshold for daylight exposure and consideration of how
tions and for four viewing directions per location, each split by time-varying annual daylight exposures from multiple views for
periods of the day and reported on a scale ranging from 0% (low each location should be summarized. Finally, the “in house”
likelihood of alerting effect) to 100% (high likelihood). workflow developed by Mardaljevic is not publically available and
More recently, Inanici et al. [41] developed a simulation pro- thus cannot be incorporated into the workflows used by designers
cedure to more accurately compute the spectral content of light for to inform decision-making.
the purpose of analysis using circadian lighting indicators such as
EML. The procedure is referred to as multi-spectral lighting simu-
lation and adopts the technique developed by Ruppertsberg and 3. Method
Bloj [42,43] who where focused on improving the color simulation
accuracy of the lighting simulation engine Radiance [44]. The The following sections describe the procedures developed for
procedure is implemented in a free software tool (Grasshopper quantifying and mapping circadian effectiveness in terms of the
plugin) entitled “Lark Spectral Lighting” which can be used by de- frequency of a circadian-effective daylight stimulus. Area-based
signers to analyze luminance renderings and irradiance data to summaries of each spatial category can then be used as indicators
obtain point-in-time calculations of EML [45]. The primary limita- to assess and differentiate the performance of various daylighting
tion of this tool is that it currently cannot be applied to annual, strategies during the design phases of a project, or to quantify the
hourly analyses. It is anticipated that promising tools such Lark will circadian effectiveness of existing spaces.
become integrated into CBDM workflows, enabling more accurate
grid-based time-series lighting data. The procedures presented in 3.1. View-point orientation and spatial considerations
this paper are essentially a post-process on such lighting data and
serve as a metric for annual evaluation that will become more The first step in the procedure is to obtain hourly illuminance
precise as simulation methodologies continue to evolve and measurements for each view direction. A horizontal measurement
improve. grid is defined and positioned at a distance from the floor specified
Even with the capability to generate spectrally accurate, spa- by the user to represent either seated or standing eye-height. Fig. 4
tialized (e.g. grid-based) time-series data over an annual period, presents an example showing the measurement grid (0.5 m
there is still the task of appropriately interpreting, summarizing spacing) generated for analysis of a portable school classroom
and visualizing performance outcomes to inform the design pro- where eye-height is assumed to be 1.2 m. At each grid point,
cess. The approach developed by Andersen and Mardaljevic [18,40] multiple vectors are arrayed at even increments to represent all
has several limitations in this regard that should be considered. The possible view directions. In Fig. 4, eight (8) vectors are specified,
procedure reports a cumulative measure of “circadian potential” for however the user can increase this number for greater precision.
each view direction (separated by time of day) that is calculated by Each vector is then used to define the orientation of a global vertical
taking the arithmetic mean of all illuminance values recorded illuminance sensor in Radiance.
annually. Therefore, an outcome of 30% could indicate 100% circa- The room geometry and project orientation are modeled in the
dian potential for 30% of the year, or 30% circadian potential for 3-D modeling software Rhinoceros [46] and imported into Grass-
100% of the year. While the latter outcome would indicate an hopper [47], where Radiance materials are assigned to all surfaces
acceptable location for regular and effective circadian entrainment, using the open-source plugin Honeybee [48]. Relevant information
the former case would be problematic because the level of light related to the model is presented in Table 3 and Radiance param-
exposure is insufficient for 70% of the year. When exterior daylight eters in Table 4. An annual climate-based daylighting simulation is
is available during the year, designers should work with the goal of then performed with Radiance and Daysim [49] to obtain hourly
providing an effective level of daylight stimulus to regularly occu- illuminance measurements for each view direction.
pied areas on a daily basis. Therefore, designers need feedback
indicating the periods of the year when a given design option re-
sults in sufficient or insufficient daylighting conditions for circadian
stimulus. This information is needed to delineate the presence or
quantify the relative size of zones (e.g. floor area) that should be
considered biologically dark, which designers would seek to
minimize during the design process. Second, the procedure
developed by Andersen and Mardaljevic to summarize the annual
performance of a given project reports the arithmetic mean of all
sensor locations and all view directions. Taking the arithmetic
mean of grid-based illuminance results (i.e. all points and all views
over all analysis hours) is ambiguous in regard to the presence and
relative size of zones of biological darkness and can easily mask the
presence of biologically dark zones in comparisons between
various design alternatives. For example, strategy 1 may produce a
space with a zone of high circadian potential (e.g. side-lit perimeter
zone) and a zone of biological darkness (e.g. core zone), but could
result in the same overall mean score (e.g. 50%) as a space having no
biologically dark zones, where in fact these lighting environments
are significantly different. Importantly, the former could pose sig-
nificant potential health risks to occupants in the absence of sup- Fig. 4. Perspective view of analysis grid (0.5 m grid-spacing) with eight view vectors
plemental electrical lighting capable of circadian stimulus, or per grid point, positioned at a 1.2 m distance from the floor.
28 K. Konis / Building and Environment 113 (2017) 22e38

Table 3 following the schema developed by Andersen et al. [18] (Table 2) to

Model properties. assess circadian potential. These are 6:00e10:00 a.m. (circadian
Property Value resetting), 10:00e18:00 (alerting effects of daylight), and 18:00e6:00
Window size 0.9 m by 1.5 m
(bright light avoidance, dim light only). Access to bright, circadian
Floor plate length (E-W) 10 m effective light in the morning is most critical for circadian resetting.
Floor plate depth (N-S) 7m However, it is important to note that exposure to bright light during
Ceiling height 3m the 10:00e18:00 period may be desirable (and preferred) by oc-
Surface reflectance (interior floor) 0.3
cupants for its potential to increase alertness. In the present
Surface reflectance (interior wall) 0.5
Surface reflectance (interior ceiling) 0.8 example, the analysis focuses on the circadian-resetting period
Glazing visible light transmittance (VLT) 0.65 (6:00e10:00 a.m.) annually, and the hour 6:00e7:00 a.m. is
Climate San Francisco, CA removed from analysis on the basis that the space would be un-
Analysis grid spacing 0.5 m
occupied during this hour.
Number of view vectors per grid point 8

3.4. Daily spatial assessment of circadian stimulus

Table 4
Model radiance parameters. Hourly EML results within the time period 7:00e10:00 a.m. are
evaluated to determine all vectors that achieve the minimum
Parameters
stimulus requirement (250 EML) for all three analysis hours. Such
av ¼ 0 ps ¼ 4 pj ¼ 0.9 vectors are then defined as Circadian-Effective (CE) for that day.
ab ¼ 6 ar ¼ 64 dp ¼ 256
dc ¼ 0.5 as ¼ 2048 dt ¼ 0.25
Fig. 6 shows the result for March 19 (predominantly overcast skies)
aa ¼ 0.2 ds ¼ 0.25 lr ¼ 6 as an example, where red is used to identify each CE view vector.
ad ¼ 2048 pt ¼ 0.1 dj ¼ 0.5 Fig. 7 shows the result during the same time period for the
st ¼ 0.5 dr ¼ 1 lw ¼ 0.01 following day (March 20, clear skies). Comparison between Figs. 6
and 7 shows how daily variations in sky conditions can lead to
significant changes in the location, orientation, and quantity of CE
3.2. Calculation of EML based on the relative direct and diffuse views. The visualization of daily CE view results can provide valu-
magnitudes able feedback to designers on the dynamic nature of circadian
effective area in the space. However, it would likely become
Illuminance measurements (reported in lux) are converted to burdensome to visually examine 365 individual images. The chal-
EML using a conversion factor (ranging from 1.0 to 1.10) to predict lenge is to understand and quantify daily variations in the avail-
the actual source efficacy for each hour based on the relative direct ability of an effective stimulus for each measurement location in a
and diffuse illuminance magnitudes reported in the climate data space to begin to examine the “quality” of circadian entrainment.
file. Diffuse illuminance is assumed to correspond to D65, which This challenge is addressed in the following section.
has a “circadian efficacy” of 1.10 (i.e. the conversion factor from
photopic lux to EML is 1.10). Direct illuminance is assumed to 3.5. Calculation of stimulus frequency
correspond to D55, which has a circadian efficacy of 1.0. Thus, a
direct illuminance of 60,000 lux concurrent with a diffuse illumi- While the availability of a circadian-effective stimulus can be
nance of 15,000 lux leads to a conversion factor of easily determined on a daily basis, additional assumptions are
(0.8  1.0 þ 0.2  1.10 ¼ 1.02). Fig. 5 shows the hourly variation in required to assess the effectiveness of the stimulus in maintaining
the modified conversion factor using climate data from a location in entrainment over period of time, such as a week, month, or year.
San Francisco, CA. It is important to note that the current approach Examination of data from a single grid-point location from the
to determine eye-level EML exposures from illuminance results classroom model illustrates this challenge. Fig. 8 uses a polar plot to
does not account for the modification of the relative spectrum of display each day of the year (365 analysis days) where the daily
light by glass (e.g. spectrally selective glass) or non-neutral internal circadian stimulus criteria was met (grey) by at least one view from
surfaces (e.g. colorful walls, floors or ceilings). grid point location # 159 (see asterisk in Fig. 7). In this example,
compliance for each grid-point is based on a best vector approach,
3.3. Time periods of analysis which considers a stimulus present for a given day if at least one of
the 8 available view vectors meets the daily stimulus criteria (i.e.
In the present workflow, three times of day were defined 250 EML from 7:00e10:00 a.m.). The best vector approach is

Fig. 5. Annual variation in spectral weighting coefficient derived from San Francisco, CA direct and diffuse illuminance data.
 K. Konis / Building and Environment 113 (2017) 22e38 29

Fig. 6. Plan view of classroom showing views that meet the minimum stimulus criteria (250 EML from 7:00 to 10:00 a.m.) for March 19 under predominantly overcast sky
conditions.

Fig. 7. Plan view of classroom showing views that meet the minimum stimulus criteria (250 EML from 7:00 to 10:00 a.m.) for March 20 under predominantly clear sky conditions.
The yellow asterisk indicates the viewpoint location examined in Section 3.5. (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)

developed to identify the potential provided by a given daylighting current day (day 7) and trailing 6 days (days 1e6). The 7-day
strategy for maintaining healthy circadian entrainment in early window is a judgment made by the author based on an assump-
stage design, prior to definition of specific (e.g. fixed) views that tion of the “length” of light history that is necessary to examine in
may be imposed by furnishings or other elements. Even with the order to make a reasonable assessment of the health impact on the
best vector approach, an effective stimulus is only available for 58% occupant. For example, lack of an effective stimulus over 7 days can
of days during the year (210 of 365 days) for grid-point #159. be considered problematic for circadian disruption. Similarly, an
Because the stimulus is not available during December and January, effective stimulus for only one or two days within the analysis
the location can be easily identified as problematic for that seasonal window may also be problematic, but is less of a risk relative to zero
period. The remaining 10 months are more challenging. During days. As the calculated frequency increases, the level of risk de-
these months, February through November, a CE stimulus is pre- creases. Using a moving analysis window of 7 days, monthly and
sent, but its daily availability is not continuous. For example, there seasonal stim.freq trends can be examined to assess variations in
is a regular daily CE stimulus during the first 1.5 weeks of June, but the frequency of an effective stimulus. Times of the year when a
then it becomes more fragmented. daily stimulus is more frequent can then be graded higher than
To quantify variations in the availability of a CE stimulus over a times when stimulus is less frequent, or not present. It should be
period of time (e.g. one week), the author developed the indicator noted that in the following examples, the analysis is applied to all
stimulus frequency (stim.freq). A stim.freq score is calculated for each days of the year for the purpose of simplicity and to account for the
view on a daily basis. The score is calculated as a percentage of the fact that commercial buildings may be occupied on weekends.
current and trailing X-days, and serves as an indicator of how Where the scheduled occupancy of a project is known to not
frequently a daily stimulus was present over the specified window include weekends, these days could be excluded from analysis.
of time (e.g. 1 week). Fig. 9 presents an example of daily exposure Fig. 9 shows a provisional grouping of stim.freq levels into five
scenarios leading to varying stim.freq outcomes, based on the letter grades (A,B,C,D,F) that represent variations in the entrainment
30 K. Konis / Building and Environment 113 (2017) 22e38

Oct
Sep Nov

Aug Dec

Daily Stimulus >= 250 EML

Jul Jan
(7:00−10:00 AM)

Jun Feb

May Mar
Apr

Fig. 8. Daily variation in presence of CE stimulus for a single grid point location (grid point #159) using the best vector approach to determine if location meets daily stimulus
requirement.

Fig. 9. Hypothetical daily exposure scenarios leading to varying stimulus frequency outcomes. Shown on the left are the five grades (A, B, C, D, F) developed to differentiate levels of
entrainment quality.

quality achieved for a given location. A regularly occurring “A” least 5 days within any 7-day period (equivalent to maintaining an
grade for a grid point location over a period of the year (e.g. the “A” or “B” entrainment quality grade, Fig. 9). This assumption is
month of June) indicates a location in space that can be easily based on the rationale that an effective stimulus should be available
labeled as circadian effective. Alternatively, a grid point that regu- on a daily basis, but that one or two days per week where an
larly achieves an “F” indicates a location that can easily be labeled effective stimulus is not present would present a minimal risk for
as biologically dark. The intermediate cases represent variations in circadian disruption. Presently, the relationship between exposure
performance that are more challenging to label. For example, Fig. 10 to various stim.freq levels and risk for circadian disruption is not
presents the resulting stim.freq outcomes for grid point #159 based known. However, it can be argued that a greater stim.freq level (e.g.
on the frequency of daily exposure shown in Fig. 8 and using a 7- 86%, 6 of 7 days) will always be valued over a lower level (e.g. 29%, 2
day analysis window. Not surprisingly, the stim.freq outcome is of 7 days). Therefore, stim.freq can be applied as a provisional in-
often variable. To assess when during the year the stim.freq is dicator of variations in the entrainment quality of a daylit space. The
acceptable and unacceptable for maintaining healthy circadian following sections present an approach for visualizing and quan-
entrainment, an additional assumption is required for the mini- tifying spatial and temporal variations in stim.freq.
mum acceptable stim.freq level. In Fig. 10, a threshold of 71% is
applied, indicating the requirement of an effective stimulus for at
 K. Konis / Building and Environment 113 (2017) 22e38 31

Oct
Sep 100%
Nov

80%

60%
Aug Dec

40%

20%
Stimulus Frequency
71%
Jul Jan % of 7−day Window
Daily Stimulus >= 250 EML
(7:00−10:00 AM)

Acceptability Threshold
(71% = 5 of 7 Days)
Jun Feb

May Mar
Apr

Fig. 10. Stimulus Frequency calculated for grid point #159. Calculation is based on assessment of circadian stimulus level on a daily basis within moving 7-day window (current day
and trailing 6 days).

3.6. Calculation of circadian effective area (CEA) that in both climates, the daily CEA is less variable as the threshold
is reduced (e.g. from 7d/wk to 4d/wk) and that neither location is
The Circadian Effective Area (CEA) of a space is defined as the capable of achieving 100% CEA at any point during the year,
percentage of analysis area that meets or exceeds the minimum regardless of the minimum stim.freq level specified. Polar-plots of
acceptable stim.freq threshold on a daily basis. Fig. 11 presents an daily CEA, such as Figs. 11 and 12 can be used to visually assess daily
annual visualization of daily CEA (0e100%) for the classroom model and seasonal variations in performance and are applicable for
based on varying stim.freq threshold requirements ranging from 7- comparisons between projects in different climates. The annual
days/week to 1-day/week for a San Francisco climate during the mean CEA (Table 5) for a given threshold (e.g. 5d/wk) serves as a
circadian resetting period (7:00e10:00 a.m.) and including all (365) summary indicator that can be used to assess the relative perfor-
days of the year. Fig. 12 presents results for the same model situated mance of various daylighting strategies for a single climate. Table 6
in Helsinki, Finland. The annual mean CEA (0e100%) for both cli- presents annual mean CEA achieved within each entrainment
mates are presented in Table 5 for a range of various stim.freq quality grade (A,B,C,D,F). When reported by grade (Table 6), the
threshold possibilities. Comparison between Figs. 11 and 12 shows percentages of analysis area that fall within each grade sum to

Stim.freq = 7d/wk Stim.freq >= 6d/wk Stim.freq >= 5d/wk Stim.freq >= 4d/wk
Oct Oct Oct Oct
Sep 100% Nov Sep 100% Nov Sep 100% Nov Sep 100% Nov

Aug Dec Aug Dec Aug Dec Aug Dec

50% 50% 50% 50%

Jul Jan Jul Jan Jul Jan Jul Jan

Jun Feb Jun Feb Jun Feb Jun Feb

May Mar May Mar May Mar May Mar

Apr Apr Apr Apr

Stim.freq >= 3d/wk Stim.freq >= 2d/wk Stim.freq >= 1d/wk

Oct Oct Oct
Sep 100% Nov Sep 100% Nov Sep 100% Nov

Aug Dec Aug Dec Aug Dec

50% 50% 50%

Jul Jan Jul Jan Jul Jan

Jun Feb Jun Feb Jun Feb

May Mar May Mar May Mar

Apr Apr Apr

Fig. 11. Annual visualization of daily Circadian Effective Area (0e100%) for classroom model based on varying stimulus frequency threshold requirements ranging from 7-days/week
to 1-day/week (San Francisco climate data).
32 K. Konis / Building and Environment 113 (2017) 22e38

Stim.freq = 7d/wk Stim.freq >= 6d/wk Stim.freq >= 5d/wk Stim.freq >= 4d/wk
Oct Oct Oct Oct
Sep 100% Nov Sep 100% Nov Sep 100% Nov Sep 100% Nov

Aug Dec Aug Dec Aug Dec Aug Dec

50% 50% 50% 50%

Jul Jan Jul Jan Jul Jan Jul Jan

Jun Feb Jun Feb Jun Feb Jun Feb

May Mar May Mar May Mar May Mar

Apr Apr Apr Apr

Stim.freq >= 3d/wk Stim.freq >= 2d/wk Stim.freq >= 1d/wk

Oct Oct Oct
Sep 100% Nov Sep 100% Nov Sep 100% Nov

Aug Dec Aug Dec Aug Dec

50% 50% 50%

Jul Jan Jul Jan Jul Jan

Jun Feb Jun Feb Jun Feb

May Mar May Mar May Mar

Apr Apr Apr

Fig. 12. Annual visualization of daily Circadian Effective Area (0e100%) for classroom model using Helsinki climate data.

Table 5 3.8. Spatial mapping of annual performance

Annual mean circadian effective area (0e100%) exceeding various stimulus fre-
quency thresholds (7d/wk to 1d/wk).
The visualization format shown in Fig. 13 is more complex than
Climate Days/week when EML stimulus was achieved the standard practice of generating a single annualized perfor-
7d 6d 5d 4d 3d 2d 1d mance summary. Fig. 14 presents an annual visualization of the
circadian effectiveness for the classroom model (San Francisco
San Francisco 20.5 29.9 38.1 45.5 50.4 53.7 56.2
Helsinki 18.9 23.3 26.6 29.1 31.2 33.4 35.7 climate) by reporting the percentage of the year (0e100%) where a
minimum stim.freq of 71% (5 of 7 days) is achieved (or exceeded)
during the circadian-resetting period (7:00e10:00 a.m.). In Fig. 14,
Table 6
the best vector approach is used to determine the presence of an
Annual mean circadian effective area (0e100%) achieved for each entrainment effective stimulus at each grid-point location. Annual mapping is
quality grade. useful for understanding the percent of the year (0e100%) when
Climate A B C D F
the specified minimum stim.freq is achieved (or exceeded) for a
particular location within the space. The same technique could also
7d/wk 5e6d/wk 3e4d/wk 1e2d/wk 0d/wk
be applied to examine other periods of the day, such as the
San Francisco 20.5 17.6 12.3 5.8 43.8 10:00e18:00 period, where an alerting effect is of interest. The
Helsinki 18.9 7.7 4.6 4.5 64.3
annual summary allows for variations in performance throughout
the space to be identified, where locations that fail to achieve the
minimum acceptable stim.freq threshold for significant portions of
100%. Therefore, the design objective is to increase the amount of the year can be assessed as higher risk for circadian disruption (in
analysis area falling within the higher grades (e.g. A and B), which the absence of circadian effective supplemental electrical lighting)
will then reduce the amount falling within the lower grades (e.g. C relative to locations that regularly meet or exceed the threshold.
and below). Designers can also use annual mapping to identify the view ori-
entations that present the greatest potential for circadian entrain-
3.7. Spatial mapping of seasonal variations in circadian effective ment to inform the location and orientation of workstation views.
area (CEA) However, the annualized format does not preserve information
indicating seasonal variations in performance, or the percentage of
Fig. 13 presents a proposed visualization format that includes a the year various grid-point locations achieve other (e.g. more strict
“stacked” version of the stim.freq data in Fig. 11, but grouped within or more lenient) stim.freq thresholds, and should be relied on after
each of the five entrainment quality grades defined in Fig. 9. The the design team has examined spatial and seasonal variations using
twelve floor plans arrayed around Fig. 13 present a spatial mapping the format shown in Fig. 13.
of the monthly mean CEA for each of the five grades. This format
allows designers to visually understand spatial variations in the 3.9. Treatment of spaces where occupant view orientations are
presence and quality of the daylight stimulus over an annual fixed
period, with reasonable representation of seasonal effects. It should
be noted that this image is somewhat compressed so that it can fit While using the best vector approach is suitable for identifying
within a standard journal paper format, and would likely be the circadian “potential” of a given grid point location, the approach
expanded to a larger size when viewed by the design team. is unlikely to be appropriate for spaces where occupants have a
 K. Konis / Building and Environment 113 (2017) 22e38 33

Fig. 13. Combined performance visualization showing daily variation in Circadian Effective Area (CEA) based on varying entrainment quality grades (A, B, C, D or F) surrounded by
spatial mapping of monthly mean CEA within each of the five grades.

Fig. 14. Annual visualization of the circadian effectiveness of the classroom model (San Francisco climate) showing the percentage of the year (0e100%) where a minimum stim.freq
threshold of 5d/wk is achieved during the circadian-resetting period (7:00e10:00 a.m.).

relatively fixed viewpoint, such as the present classroom example, appropriate to constrain analysis to a single view direction, or
where the pre-determined furniture layout indicates a west-facing subset of view directions. Fig. 15 presents an alternate mapping of
orientation for those seated at desks. For examination of spaces the classroom using the same vector data from Fig. 14, but
pre-programmed with largely fixed viewpoints, it may be more restricting analysis to the results of only the west-facing view
34 K. Konis / Building and Environment 113 (2017) 22e38

Fig. 15. Alternate spatial mapping of circadian effectiveness based on fixed view direction.

vectors. Comparison between results from the two different ap- Table 7
proaches (Fig. 14 vs. Fig. 15) illustrates the substantial sensitivity of Model properties.

analysis outcomes to the assumed view orientation(s). Property Value

Window to wall ratio (WWR) 0.30 (or 0.50)

Floor plate length (E-W) 64 m
4. Application of spatial classifications to inform design
Floor plate depth (N-S) 40 m
Ceiling height 3m
The following example demonstrates the applicability of the Surface reflectance (interior floor) 0.3
proposed metric for making relative comparisons between design Surface reflectance (interior wall) 0.5
options in early-stage design. For this example, one floor plate from Surface reflectance (interior ceiling) 0.8
Glazing visible light transmittance (VLT) 0.65
a hypothetical medium-size commercial office building located in
Climate San Francisco, CA
San Francisco, CA (Fig. 16) is analyzed to compare the outcomes for Analysis grid spacing 2m
two different Window-to-Wall Ratio (WWR) options, 1) a WWR of Number of view vectors per grid point 8
0.30 (shown in Fig. 16) and, 2) a WWR of 0.50. The WWR is
consistent for all four facades, the Visible Light Transmittance (VLT)
of all windows is (0.65), and the models are analyzed without the morning circadian-resetting period (7:00e10:00 a.m.) using the
presence of interior shading devices. Structural columns and an best vector approach and a minimum stim.freq threshold require-
opaque core zone volume are included in the models to represent ment of 5d/wk (Fig. 9). As a result, the spatial mapping indicates the
the level of detail for an early stage of schematic design. The core percentage of the year where each grid-point location achieves at
zone (e.g. elevators, stairs, bathrooms and storage), along with a least a “B” entrainment quality grade (see Fig. 9). Fig. 18 shows the
buffer zone of circulation are excluded from the analysis grid. annual result for the 0.50 WWR option. Comparison between
Relevant model properties are shown in Table 7 (Radiance param- Figs. 17 and 18 allows designers to visualize the spatial implications
eters are identical to the previous example, Table 4). A relatively of a change in WWR from 0.30 to 0.50, for example, the elimination
large analysis grid spacing (2 m) is applied for the purposes of of the small region of analysis area (shown in black in Fig. 17) that
visualization. never achieves a stim.freq 5d/wk. Notably, both options show that
Fig. 17 shows the annual result for the 0.30 WWR option for the a significant percentage of the analysis area is circadian effective
(achieves a stim.freq 5d/wk) for large portions of the year.
Fig. 19 presents a side-by-side comparison between design op-
tions (0.30 and 0.50 WWR) where performance can be examined
more precisely in terms of daily variations in CEA (0e100%) for each
entrainment quality grade (A,B,C,D, (F is not plotted)). Fig. 19 shows
that the 0.30 WWR option achieves a level of entrainment quality
above an F grade over nearly 100% of the analysis area for
approximately 8 months of the year (March through November).
However, the entrainment quality is often poor (e.g. C, or D). By
comparison, the 0.50 WWR option achieves significant improve-
ments in entrainment quality over the 0.30 WWR option, as well as a
greater percentage of CEA. For example, Fig. 19 shows that the 0.50
WWR option leads to an “A” grade of entrainment quality for nearly
100% of the analysis area over the seasonal period from May
through September, and relatively better grades for the remaining
period of the year. Notably, both options fail substantially during
the month of January, and as more options (e.g. 0.95 WWR) are
Fig. 16. Perspective view of one floor of medium commercial office building model examined the designer will be able to establish the limits in
(0.30 WWR) showing analysis grid and view vectors.
 K. Konis / Building and Environment 113 (2017) 22e38 35

Fig. 17. Plan view of one floor of medium commercial office building model (0.30 WWR) showing annual availability of circadian stimulus (stim.freq 5d/wk) based on best vector
approach.

Fig. 18. Plan view of one floor of medium commercial office building model (0.50 WWR) showing annual availability of circadian stimulus (stim.freq 5d/wk) based on best vector
approach.

Fig. 19. Comparison of daily variations in Circadian Effective Area (0e100%) between design options (0.30 and 0.50 WWR) for each stimulus frequency grade (A,B,C,D, (F is not
plotted)).

performance imposed by latitude and climate. Finally, summarized to achieve a solution where 100% of the CEA is achieved with an “A”
outcomes reported in terms of mean CEA achieved within each grade. Given that this optimum may be impossible to reach due to
entrainment quality grade (Table 8) can be applied in iterative climatic and programmatic constraints, the design team would
performance-based design workflows (and in objective functions then favor the option that achieves the greatest percentages of CEA
used by optimization tools) to differentiate the performance of for grades A and B (e.g. the 0.50 WWR option achieves a total CEA
competing design options. For example, a design team would seek for A and B grades of 62.3% þ 21.5% ¼ 83.8% (Table 8), which is
36 K. Konis / Building and Environment 113 (2017) 22e38

Table 8 5. Discussion and conclusions

Mean circadian effective area (0e100%) achieved annually for each entrainment
quality grade.
As knowledge of the relationship between lighting parameters
WWR A B C D F and health outcomes increases, new metrics, design frameworks,
7d/wk 5e6d/wk 3e4d/wk 1e2d/wk 0d/wk and field-based evaluation techniques are needed to ensure that all
design interventions lead to indoor environments that effectively
30% 46.0 25.3 12.4 4.6 11.7
50% 62.3 21.5 7.6 2.9 5.7 support the health and well-being of occupants. The area-based
circadian daylight metric described in this paper is applicable for
informing decision-making in multiple contexts including all
valued over the 0.30 WWR option of 46% þ 25.3% ¼ 71.3%). phases of the design process, as well as for assessing existing
spaces. In early stage design, the goal is to maximize the CEA that
achieves the highest entrainment quality grades (e.g. A and B)
through manipulation of architectural building parameters such as
4.1. Consideration of fixed viewpoints building form, massing, aperture size (e.g. WWR), ceiling height,
and other parameters that are set early in the design process. The
The results shown in Section 4 are based on the best vector same goal can be addressed in later stages of design by informing
approach, which is developed to identify the potential provided by the selection of building components (e.g. glazing and facade
a given daylighting strategy for maintaining healthy circadian shading systems) and controls for lighting and automated shading
entrainment. In many real projects, programmatic constraints and systems. For the evaluation of existing buildings, the metric can be
other design considerations (e.g. glare) may limit the ideal orien- used to assess and differentiate the performance of multiple spaces
tation of the occupant view. Figs. 20 and 21 illustrate the impact of that a prospective tenant may choose to occupy, or for a building
various fixed view orientations on performance outcomes for the owner to identify and inform the retrofit of poorly performing
0.30 and 0.50 WWR options respectively. Each figure presents zones within a building.
annual outcomes for four of the eight specified view orientations The metric and procedures developed in this paper rely on a
and reveals significantly different outcomes relative to the ideal number of assumptions for the timing, spectrum, intensity, dura-
case (e.g. Figs. 17 and 18). For example, the decision to orient tion and past history of light exposure needed to maintain effective
workstation views towards the core of the floor-plate would result circadian entrainment. These assumptions are developed from
in “F” grades for a significant number of cases, while views oriented theoretical knowledge, findings from photobiology, and judgments
parallel to the facade would achieve the stim.freq threshold (5d/ made by the author. Notably, the calculation of stimulus frequency
wk) for a significantly smaller percentage of days during the year (stim.freq) and letter grades proposed for entrainment quality
relative to views oriented towards the facade. While it is not sur- require assumptions for both the appropriate length of past light
prising that the orientation of the view significantly impacts eye- history to examine (e.g. current and trailing 6 days) and for the
level daylight exposure, it should be noted that current horizontal minimum acceptable number of days within that period when an
illuminance measurement practices, such as those applied in LEED effective stimulus must be present for the exposure scenario to be
Daylight Credit compliance procedures and in various climate- considered supportive of health circadian entrainment. In this
based daylighting metrics do not account for this potential vari- context, the approach developed to assess the duration of an
ability. Early in design, analysis of view orientation impacts on effective stimulus on a daily basis as well as the frequency an
circadian daylighting performance can help to inform baseline as- effective stimulus is present over the course of a year can be
sumptions for the appropriate orientation of workstation views.

Fig. 20. Plan views of four possible view orientations for the medium commercial office building model (0.30 WWR) showing impact of view orientation (black arrow) on annual
availability of circadian stimulus (stim.freq 5d/wk).
 K. Konis / Building and Environment 113 (2017) 22e38 37

Fig. 21. Plan views of four possible view orientations for the medium commercial office building model (0.50 WWR) showing impact of view orientation (black arrow) on annual
availability of circadian stimulus (stim.freq 5d/wk).

considered as a preliminary step in addressing the expanded set of illuminance measurements to new metrics based on vertical, view-
assumptions needed to establish health-based lighting metrics. based assessments requires new approaches to account for occu-
Such assumptions are needed as lighting design shifts from metrics pant views that may range from dynamic or selectable to fixed. The
that are based on an assumption of an instantaneous effect (i.e. examples presented in this paper present two basic approaches to
visual task performance or glare discomfort) to metrics that seek to identify an appropriate view orientation for analysis. However,
predict a biological effect that manifests over a length of time and more sophisticated approaches can be explored. For example,
particular pattern of light exposures. The assumptions made in this analysis could be performed on a subset of view vectors for each
paper are expected to be refined in response to additional research grid point location, based on an assumption that an occupant has
describing the complex relationship between human biological the ability to adjust his/her view within a specified viewing arc (e.g.
lighting needs to maintain (or enhance) health and well-being in 45
to the left or right of the primary task view).
buildings. The user can adjust the light stimulus thresholds and The metric and visualization techniques presented in this paper
temporal criteria in the parametric workflow shared with this pa- enable designers to specifically address non-visual effects of light
per [10] to address specific user populations that may warrant during design by making relative comparisons between design al-
alternate assumptions. For example, it may be determined that a ternatives. The metric can also be incorporated into simulation-
higher minimum daylight illuminance threshold is appropriate for based workflows incorporating multi-objective optimization,
adults older than 65, due to age-related health effects which reduce which seek to balance daylighting objectives with whole-building
eye sensitivity to light (e.g. cataracts, glaucoma, macular degener- energy use and other performance goals [54]. As a result, the po-
ation). Alternatively, a user can calculate performance results for tential health benefits of daylight can begin to be added to the
the alerting period of the day (10:00e18:00 alerting effects of multiple performance objectives used to design and evaluate
daylight), in addition to the circadian resetting period, or for any buildings.
arbitrary daily time period.
While theoretical knowledge and scientific findings are suffi-
cient to begin to propose metrics and procedures to classify indoor References
daylit spaces in terms of anticipated circadian effectiveness, the [1] I. Provencio, I.R. Rodriguez, G. Jiang, W.P. Hayes, E.F. Moreira, M.D. Rollag,
task of validating the numerous assumptions that underlie such A novel human opsin in the inner retina, J. Neurosci. 20 (2) (2000 Jan 15)
metrics will require a substantial level of field-based validation 600e605.
[2] J.J. Gooley, J. Lu, T.C. Chou, T.E. Scammell, C.B. Saper, Melanopsin in cells of
involving mechanisms capable of collecting feedback from large
origin of the retinohypothalamic tract, Nat. Neurosci. 4 (12) (2001 Dec) 1165.
subject populations in buildings in use [50] in order to compare [3] J. Hannibal 1, P. Hindersson, S.M. Knudsen, B. Georg, J. Fahrenkrug, The pho-
physical measures of light stimulus with individual health out- topigment melanopsin is exclusively present in pituitary adenylate cyclase-
comes. It should also be acknowledged that window views can activating polypeptide-containing retinal ganglion cells of the reti-
nohypothalamic tract, J. Neurosci. 22 (1) (2002 Jan 1). RC191.
often become a source of glare, which can lead to the deployment of [4] Hattar S1, H.W. Liao, M. Takao, D.M. Berson, K.W. Yau, Melanopsin-containing
shades, informal occupant modifications to workstations, and in retinal ganglion cells: architecture, projections, and intrinsic photosensitivity,
some cases, permanent retrofits to the building facade which Science 295 (5557) (2002 Feb 8) 1065e1070.
[5] G.C. Brainard, J.P. Hanifin, J.M. Greeson, B. Byrne, G. Glickman, E. Gerner,
reduce daylight transmission [51]. Therefore, efforts to improve the M.D. Rollag, Action spectrum for melatonin regulation in humans: evidence
circadian effectiveness of a space should incorporate the use of for a novel circadian photoreceptor, J. Neurosci. 21 (2001) 6405e6412.
glare indicators [52] and assessment procedures [53] to understand [6] K. Thapan, J. Arendt, D.J. Skene, An action spectrum for melatonin suppres-
sion: evidence for a novel non-rod, non-cone photoreceptor system in
the relative trade-offs between circadian daylighting and glare. humans, J. Physiol. 535 (2001) 261e267.
The transition from metrics based on horizontal workplane [7] N.E. Klepeis, W.C. Nelson, W.R. Ott, J. Robinson, A.M. Tsang, P. Switzer,
J.V. Behar, S. Hern, W. Engelmann, The national human activity pattern survey
38 K. Konis / Building and Environment 113 (2017) 22e38

(NHAPS): a resource for assessing exposure to environmental pollutants, related to circadian phase in the blind, Sleep 22 (5) (1999) 616e623.
J. Expos. Anal. Environ. Epidem. 11 (3) (2001) 231e252. [32] Erin L. Zelinski, Scott H. Deibel, Robert J. McDonald, The trouble with circadian
[8] U.S. Green Building Council, LEED v4 for Building Design Construction, April 5, clock dysfunction: multiple deleterious effects on the brain and body, Neu-
2016, p. 161. http://www.usgbc.org/sites/default/files/LEED%20v4%20BDC_04. rosci. Biobehav. Rev. 40 (March 2014) 80e101. http://www.ncbi.nlm.nih.gov/
05.16_current.pdf. pubmed/24468109.
[9] International Well Building Institute. WELL Building Standard (v1). May 2016. [33] M. Ruuger, M.C.M. Gordijn, D.G.M. Beersma, B. de Vries, S. Daan, Time-of-day-
[10] Zip archive of simulation workflow used in this paper. https://dl. dependent effects of bright light exposure on human psychophysiology:
dropboxusercontent.com/u/4262841/workflow.zip Accessed 9 August 2016. comparison of daytime and nighttime exposure, Am. J. Physio. Reg. I 290 (5)
[11] S.W. Lockley, G.C. Brainard, C.A. Czeisler, High sensitivity of the human (2006).
circadian melatonin rhythm to resetting by short wavelength light, J. Clin. [34] A.M. Chang, F.A. Scheer, C.A. Czeisler, The human circadian system adapts to
Endocrinol. Metab. 88 (9) (2003 Sep) 4502e4505. prior photic history, J. Physiol. 589 (5) (2011) 1095e1102.
[12] Enezi, et al., A “Melanopic” spectral efficiency function predicts the sensitivity [35] Amundadottir et al. Modeling non-visual responses to light: unifying spectral
of melanopsin photoreceptors to polychromatic lights, J. Biol. Rhythms 26 sensitivity and temporal characteristics in a single model structure. In Proc. of
(2011) 314. CIE Centenary Conference ”Towards a New Century of Light”, Apr 15-16, 2013,
[13] R.J. Lucas, S.N. Peirson, D.M. Berson, T.M. Brown, H.M. Cooper, et al., Paris, France, p. 101e110.
Measuring and using light in the melanopsin age, Trends Neurosci. 31 (1) [36] C.S. Pittendrigh, Temporal organization: reflections of a Darwinian clock-
(2014) 1e9. watcher, Annu. Rev. Physiol. 55 (1993) 16e54.
[14] Lucas et al. Excel-based melanopic illuminance calculator. http://lucasgroup. [37] Mardaljevic J. CIBSE National Conference 2006: Engineering the Future. 21e22
lab.ls.manchester.ac.uk/research/measuringmelanopicilluminance/Accessed March 2006, Oval Cricket Ground, London, UK.
25 July 2016. [38] J. Mardaljevic, L. Heschong, E.S. Lee, Daylight metrics and energy savings,
[15] Commission Internationale de IEclairage (CIE). http://cie.co.at/Accessed 19 Light. Res. Technol. 0 (2009) 1e23.
October 2016. [39] Illuminating Engineering Society (IES), Approved Method: IES Spatial Daylight
[16] Rea, et al., Modeling the spectral sensitivity of the human circadian system, Autonomy (SDA) and Annual Sunlight Exposure (ASE), 2013, p. 14. ISBN: 978-
Light. Res. Technol. 44 (4) (December 2012) 386e396. 0-87995-272-3.
[17] Lighting Research Center. Circadian stimulus calculator. http://www.lrc.rpi. [40] J. Mardaljevic, M. Andersen, N. Roy, J. Christoffersen, A framework for pre-
edu/programs/lightHealth/Accessed 8 June 2016. dicting the non-visual effects of daylight - Part II: the simulation model, Ligh.
[18] M. Andersen, J. Mardaljevic, S.W. Lockley, A framework for predicting the Res. Technol. 2013 (0) (2013) 1e19.
non-visual effects of daylight - Part I: photobiology-based model, Light. Res. [41] Inanici M, Brennan M, and Clark E. “Multi-spectral Lighting Simulations:
Technol. 2012 (44) (2012) 37. Computing Circadian Light,” International Building Performance Simulation
[19] C. Cajochen, J. Zeitzer, C. Czeisler, D. Dijk, Dose-response relationship for light Association (IBPSA) 2015 Conference, Hyderabad, India, December 7-9, 2015.
intensity and ocular and electroencephalographic correlates of human alert- [42] A.I. Ruppertsberg, M. Bloj, “Rendering complex scenes for psychophysics using
ness, Behav. Brain Res. 115 (2000) 75e83. radiance: how accurate can you get? J. Opt. Soc. A. 23 (4) (2006) 759e768.
[20] J. Phipps-Nelson, J. Redman, D. Dijk, S. Rajaratnam, Daytime exposure to [43] A.I. Ruppertsberg, M. Bloj, Creating physically accurate visual stimuli for free:
bright light, as compared to dim light, decreases sleepiness and improves spectral rendering with radiance, Behav. Res. Methods 40 (1) (2008) 304e308.
psychomotor vigilance performance, Sleep 26 (6) (2003) 695e700. [44] G. Ward, The radiance lighting simulation and rendering system, in: Pro-
[21] C.S. Pechacek, M. Andersen, S.W. Lockley, Preliminary method for prospective ceedings of SIGGRAPH 94. Computer Graphics Proceedings, Annual Confer-
analysis of the circadian efficacy of (day) light with applications to healthcare ence Series, 1994, pp. 459e572.
architecture, Leukos 5 (2008) 1e26. [45] Lark Spectral Lighting. http://faculty.washington.edu/inanici/Lark/Lark_
[22] M.L. Amundadottir, S.W. Lockley, M. Andersen, Unified framework to evaluate home_page.html, Accessed 25 July 2016.
non-visual spectral effectiveness of light for human health, Light. Res. Tech- [46] Rhinoceros 3-D Modeling Program. Robert McNeel & Associates. http://www.
nol. 0 (2016) 1e24. en.na.mcneel.com/contact.htm, Accessed 19 October 2016.
[23] 
Ecole Polytechnique Fe derale de Lausanne, Interdisciplinary Laboratory of [47] Grasshopper Algorithmic Modeling Program for Rhino. http://www.
Performance-Integrated Design. SpeKtro, Interactive dashboard for exploring grasshopper3d.com/, Accessed 19 October 2016.
non-visual spectrum lighting. http://spektro.epfl.ch/Accessed 24 July 2016. [48] Sadeghipour Roudsari Mostapha, Michelle Pak, Ladybug: a parametric envi-
[24] J. van de Kraats, D. van Norren, “Optical density of the aging human ocular ronmental plugin for grasshopper to help designers create an
media in the visible and the UV, J. Opt. Soc. Am. A 24 (7) (2007) 1842. environmentally-conscious design, in: Proceedings of the 13th International
[25] J.M. Zeitzer, D.J. Dijk, R.E. Kronauer, E.N. Brown, C.A. Czeisler, “Sensitivity of IBPSA Conference Held in Lyon, France Aug 25e30th, 2013.
the human circadian pacemaker to nocturnal light: melatonin phase resetting [49] Daysim, Advanced Daylight Simulation Software, Accessed 19 October 2016,
and suppression, J. Physiol. 526 (3) (2000) 695e702, http://dx.doi.org/ http://daysim.ning.com/.
10.1111/j.1469-7793.2000.00695.x. [50] Konis, K. Leveraging ubiquitous computing as a platform for collecting real-
[26] Chang, et al., Human responses to bright light of different durations, J Physiol. time occupant feedback in buildings. Intell. Build. Int. 5(3), 150e161.
590 (Pt 13) (2012) 3103e3112, 2012 Jul 1. [51] Konis, K. Evaluating daylighting effectiveness and occupant visual comfort in a
[27] Figueiro, et al., Designing with Circadian Stimulus. Lighting Design and Ap- side-lit open-plan office building in San Francisco, California. Build. Environ.
plications (LDþA), 2016. The magazine of the Illuminating Engineering Society 59, 662e677.
of North America (IESNA). Published October 2016. [52] J. Wienold, J. Christoffersen, Evaluation methods and development of a new
[28] S.B.S. Khalsa, M.E. Jewett, C. Cajochen, C.A. Czeisler, A phase response curve to glare prediction model for daylight environments with the use of CCD cam-
single bright light pulses in human subjects, J. Physiol. 15 (2003) 945e952. eras, Energy Build. 38 (7) (2006) 743e757.
[29] C.A. Czeisler, J.F. Duffy, T.L. Shanahan, E.N. Brown, J.F. Mitchell, D.W. Rimmer, [53] J.A. Jakubiec, C.F. Reinhart, The ‘adaptive zone’ e a concept for assessing
et al., Stability, precision, and near 24-Hour period of the human circadian discomfort glare throughout daylit spaces, Light. Res. Technol 44 (2012)
pacemaker, Science 284 (5423) (1999) 2177e2181. 149e170.
[30] S.W. Lockley, D.J. Skene, H. Tabandeh, A.C. Bird, R. Defrance, Relationship [54] K. Konis, A. Gamas, K. Kensek, Passive performance and building form: an
between melatonin rhythms and visual loss in the blind, J. Clin. Endocrinol. optimization framework for early-stage design support, Solar Energy 125
Metab. 82 (11) (1997) 3763e3770. (February 2016) 161e179.
[31] S.W. Lockley, D.J. Skene, L.I. Butler, J. Arendt, Sleep and activity rhythms are