The Non-Image-Forming Effects of Daylight: An Analysis for Design Practice Purposes

Abstract

Daylight plays a crucial role in human health, and as research into its effects expands, it is essential for designers to estimate the non-image-forming impacts of various daylighting and lighting strategies. This allows them to create indoor environments that are both pleasant and comfortable. To support this, daylight measurements were taken in five Chinese cities, focusing on spectral power distribution, correlated color temperature, and illuminance. The study calculated the non-image-forming effects of daylight exposure using metrics like melanopic Equivalent Daylight Illuminance and Circadian Light. A key finding was the development of the action factor S_AI, which estimates the potential non-image-forming effects of light in built environments. This factor serves as a proxy for understanding how non-image-forming effects relate to correlated color temperatures. Additionally, the research suggests the possibility of creating a time-variational daylighting and lighting strategy with four distinct periods of non-image-forming effects throughout a 24 h day. These insights could be valuable for architects and designers in optimizing indoor lighting systems.

1. Introduction

Practices for luminous environment design inside buildings have been developed to satisfy human visual needs, following a greater depth of understanding of intricate photoreceptive systems that influence underlying human photobiological effects. Daylight, through its non-image-forming (NIF) effects, plays a critical role in human health [1]. With the discovery of a third class of photoreceptors in human’s retina [2,3], known as Intrinsically Photoreceptive Retinal Ganglion Cells (ipRGCs), much more attention has been paid to the NIF effects of light on human health and well-being [4,5]. Results from several fields of science [6,7,8] demonstrate that ipRGCs are central to circadian system phototransduction, and in this phototransduction process, S-cones, L-cones, M-cones, and rods provide excitatory/inhibitory signals to ipRGCs or each other via ON/OFF cone or rod bipolar cells [9,10]. Our Earth rotating on its axis, with its alternating 24 h light–dark cycles, has a significant impact on the physiology and behavior of organisms from other mammals to human beings [11]. In the presence of light, action mechanisms are generated through the preceding synergistic–antagonistic pathways, and therefore daylight participates in regulating and maintaining multiple circadian rhythms, such as that of cognitive performance, hormones secretion/suppression, and the sleep/wake cycle [12,13,14]. Daylight is known to synchronize the periodic timing to the local, 24 h solar day [15,16]. Of particular interest, the melatonin level in humans’ bodies during a day is a widely used indicator for variable luminaires’ impact on the circadian system [17]. The sleep/wake cycle is closely related to the 24 h melatonin cycle. Habitual bedtime occurs approximately 2 h after the onset of melatonin secretion, while habitual awakening usually occurs 10 h after [18].

What is more, it is now established that the melatonin secretion response to light exposure is dependent on the intensity of light and on the time of day [19,20]. The higher the light intensity, the stronger the suppression of melatonin secretion, and the shorter the required time. Melatonin suppression is correlated with both light intensity and the cumulative time of light stimulation [21]. The circadian rhythm of the human body is influenced by light during the morning and evening periods. Morning light advances the secretion of melatonin, whereas nocturnal light exposure delays it [22,23]. The studies on polychromatic light [24,25] elucidate greater melatonin suppression for high correlated color temperature (CCT) light than for low CCT light. Due to the different spectral sensitivities of each type of photoreceptor cell, light sources with different spectral power distributions (SPD) have significant effects on melatonin suppression in varying degrees as well [26,27], and it is generally agreed that light around the blue–green spectrum has a significant impact on circadian rhythm [28] and short wavelengths have a stronger impact on melatonin suppression than long wavelengths [29,30]. According to the spectral sensitivity of known retinal photoreceptors, melanopic Equivalent Daylight Illuminance (m-EDI) [31] and Circadian Light (CL_A) [32] are independently proposed to support the quantification and evaluation of current spatial lighting environments, providing a scientific basis for healthy daylighting/lighting design. These models are available at different levels for evaluating the circadian impact of a light stimulus, and appear to be more accurate and more succinct to apply, due to continuous optimization [33,34].

Although the current knowledge about the circadian system is incomplete, it is clear that the NIF effects of daylight are important to human health and wellness [35,36,37,38]. During pandemic control for COVID-19, Figueiro et al.’s [39] research revealed that the higher the exposure to light during the day (indoors or outdoors), the better the self-reported sleep of the individuals. Diurnal light exposure enhances the secretion of melatonin at night, stabilizes the biological clock, and reduces sensitivity to nocturnal light [40]. The widespread use of artificial light and luminescent devices has led to prolonged exposure to inappropriate light [20], which can readily interfere with the normal secretion of melatonin [41], leading to issues like circadian rhythm disturbances and sleep disorders [42]. Appropriate or adequate daylight exposure is beneficial in eliminating fatigue, reducing stress, and promoting working efficiency [43]. In recent years, the importance of daylight utilization in the built environment has been spotlighted by a significant amount of research. One important research focus, grounded in the assessment of visual sufficiency and comfort [44], is to deeply explore a dynamic lighting evaluation system [45,46] and discourse upon different approaches to replacing the daylight factor in order to define the quality and quantity of daylight indoors [47]. Another significant area of research concerns energy efficiency and sustainability [48,49]. Identifying the optimal artificial lighting configurations based on natural lighting conditions, as well as harnessing daylight effectively, holds substantial potential for energy savings [50]. Additionally, in non-image-forming effects research, given the dynamic shifts of daylight influenced by factors like time, seasons, and weather conditions, computer simulations are predominantly employed to gather spatial lighting data. By establishing virtual or laboratory settings and modifying various parameters within these settings [51,52], the exploration of the targeted research is successfully carried out.

However, there are few studies that offer non-image-forming insights into the analysis of daylight characteristics in typical climates. Compared to computer simulation, through field measurements, daylighting characteristics in a space can be collected more realistically, enabling a more scientific and objective evaluation of various indicators. Through hundreds of millions of years of evolution, human eyes have adapted to function in an environment with daylight. As evidence of the health impact of daylight grows, and also considering that a large part of the population in industrial countries spends most of their time inside buildings [53], it is increasingly crucial for designers to have access to estimates regarding the NIF effects of diverse daylighting/lighting strategies and to plan reasonable daylighting/lighting systems to make luminous environments indoors pleasant and comfortable. In the present paper, after a series of daylight measurements were carried out in five cities in China and the NIF effects of retinal daylight exposure were calculated by using m-EDI and CL_A, a global analysis of daylight in a visible spectral range (380~780 nm) regarding both visual and non-image-forming issues was performed, and from there, the characteristics of findings were discussed, an action factor concerning the NIF effects was proposed, and a time-varied daylighting/lighting strategy with four distinct ‘non-image-forming effect’ periods over a day was developed, with the hope of providing potent implications for architects and designers.

2. Materials and Methods

2.1. Daylight Measurements

Daylight measurements were carried out in five cities with different solar surface conditions (SSC). Measurement for each city was conducted over a day. Five cities are selected, including Kunming, Xining, Beijing, Nanchang, and Chongqing, and they are all the capitals of their own provinces and located in different daylight-climate regions, from I to V, respectively [54].

For comparison purposes, it was indispensable to have each measured timing arranged in terms of the same solar altitude, since the five cities with a broad range of geographical distribution in China are obviously at odds with the solar altitude angles (SAA) at the same standard time (China Standard Time, UT+8:00). Firstly, the maximum solar altitudes on each measurements’ days had to be calculated; next, depending on the maximum solar altitudes acquired, the local times corresponding to SAAs at an interval of 10° could be calculated, together with the additional local times corresponding to 5°, −5°, and the maximum solar altitude; finally, the local times were transformed to the standard time (China Standard Time, UT+8:00), and then each measured time was obtained. The SAAs during measurement periods were all defined as positive before noon (in the east), while negative in the afternoon (in the west), with the maximum solar altitude as a boundary. Every measured time from 5° to −5° could be determined through the following method, using Equation (1) [55]:

$$\sin h_s=\sin \left(\phi \right)·\sin \left(\delta \right)+\cos (\phi )·\cos \left(\delta \right)·\cos \left(\Omega \right)$$

(1)

where

• $h_s$ is the solar altitude (°);
• $\phi$ is the observer’s latitude;
• $\delta$ is the solar declination angle relevant to date;
• $\Omega$ is the hour angle relevant to the observer’s longitude;

Daylight-climate region, latitude, longitude, date, the maximum solar altitude h_smax, and the sunrise, sunset, and noon time (China Standard Time UT+8:00) are listed in

Table 1. Summary of information regarding daylight measurements in five cities.

City	Daylight-Climate Region 1	Latitude	Longitude	Sunrise (h)	Sunset (h)	Noon (h)	hsmax (°)
Kunming	I	25.05 N	102.73 E	6:29	19:49	13:09	85.15
Xining	II	36.57 N	101.75 E	6:23	20:03	13:13	69.75
Beijing	III	39.92 N	116.42 E	4:51	19:37	12:14	72.93
Nanchang	IV	28.40 N	115.55 E	5:25	19:11	12:18	84.75
Chongqing	V	29.35 N	106.33 E	6:33	19:15	12:54	69.83

¹ In terms of the meteorological monitoring data, there are five daylight-climate regions in China [54], numbered from I to V, where the lower number is indicative of more natural light resources.

.

Moreover, it was necessary to select an open space unobstructed by surrounding buildings, structures, or vegetation during each measurement period, and then a Konica Minolta T10 luxmeter (made in Japan) and a PR-650 spectrascan spectroradiometer (made in USA), as shown in

Table 2. Parameters of PR-650 spectral photometer.

Performance Parameter		Photograph
Spectral range	380–780 nm
Spectral accuracy	±1 nm
Spectral bandwidth	8 nm
Illuminance accuracy	±2% (2856 K,23 °C)
Chrominance	±0.0015 CIE 1931 x, ±0.001 CIE 1931 y
Measurement and observation angle	1° (Measurement)/7° (observation)
Overall dimension	355 mm (length) × 201 mm (width) × 81 mm (height)

, were used to acquire the following data at each measured time: daylight SPDs, daylight CCTs, and horizontal and vertical illuminances (facing sunlight direction) at the eye level of a person standing on the ground. At the same time, weather conditions corresponding to each measured time such as the SSC, cloud cover, cloud shape, and atmospheric visibility were also estimated and registered by using the approaches proposed in related publications [56].

It should be noted that the PR-650 spectrascan was placed under the 2° visual field, with 0/45° standard illumination/observation condition as recommended by the International Commission on Illumination (CIE), and its attachment, the RS-2 reflectance standard whiteboard, was positioned facing direct sunlight so that the intersection angle between them had a value of 90°.

The SSCs, which refer to the relationship between the formation of sunlight and shadows on the ground, reveal the basic weather conditions, denote the spectral radiant intensity of daylight to a certain extent, and also correlate closely with cloud cover and cloud shape in the sky, can be represented by the four following specific symbols with different meanings: Л, which denotes that the sun’s silhouette can hardly be observed through the thick clouds and there is no shadow on the ground; Θ, which denotes that the sun’s silhouette can be observed faintly through thin clouds and there is no shadow on the ground; Θ⁰, which denotes that the sun’s silhouette can just be observed and there is a soft shadow on the ground; Θ²,which denotes that the sun’s silhouette can be clearly observed and there is a hard shadow on the ground.

2.2. Data Analysis

On the basis of obtained data, the procedures developed independently by CIE S 026 [31] and Rea et al. [32] can be used to calculate the possible NIF effects of daylight reaching the retina.

According to CIE S 026, m-EDI values, the melanopic Equivalent Daylight Illuminance, can be obtained starting from smel(λ) based on the actual knowledge about nocturnal melatonin suppression in humans [19,30], using the equation reported below [31]:

$$\mathrm{m}-\mathrm{E}\mathrm{D}\mathrm{I}={\displaystyle \frac{\int {\mathrm{E}}_{e,\mathsf{\lambda }}\left(\mathsf{\lambda }\right){\mathrm{s}}_{mel}\left(\mathsf{\lambda }\right)\mathrm{d}\mathsf{\lambda }}{K_{mel,V}^{D65}}}$$

(2)

where

• $K_{mel,V}^{D65}$ = 1.3262 mW/lm;
• E_e,λ(λ) is the light source spectral irradiance;
• s_mel(λ) is the melanopsin spectral efficiency function;

Rea et al. [32] argue that the functional impact of the circadian system is expressed in circadian stimulus (CS) percentages that are depended upon CL_A, and CL_A values can be calculated by the equations that follow:

$${\mathrm{C}\mathrm{L}}_{\mathrm{A}}=1548\left\{\begin{array}{c}\left[\int {\mathrm{M}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }-{\mathrm{a}}_{\mathrm{r}\mathrm{o}\mathrm{d}1}\left({\displaystyle \frac{\int {\mathrm{V}}_{\mathsf{\lambda }}^{\prime }{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }}{\int {\mathrm{V}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }+{\mathrm{g}}_1\int {\mathrm{S}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }}}\right)\left(1-e^{{\displaystyle \frac{-\int {\mathrm{V}}_{\mathsf{\lambda }}^{\prime }{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }}{\mathrm{R}\mathrm{o}\mathrm{d}\mathrm{S}\mathrm{a}\mathrm{t}}}}\right)\right]+\\ \left[\begin{array}{c}{\mathrm{a}}_{\mathrm{b}-\mathrm{y}}\left(\int {\mathrm{S}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }-\mathrm{k}\int {\mathrm{V}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }\right)\\ -{\mathrm{a}}_{\mathrm{r}\mathrm{o}\mathrm{d}2}\left({\displaystyle \frac{\int {\mathrm{V}}_{\mathsf{\lambda }}^{\prime }{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }}{\int {\mathrm{V}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }+{\mathrm{g}}_2\int {\mathrm{S}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }}}\right)\left(1-e^{{\displaystyle \frac{-\int {\mathrm{V}}_{\mathsf{\lambda }}^{\prime }{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }}{\mathrm{R}\mathrm{o}\mathrm{d}\mathrm{S}\mathrm{a}\mathrm{t}}}}\right)\end{array}\right]b-y>0\\ \left[\begin{array}{c}\int {\mathrm{M}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}d\lambda -\\ {\mathrm{a}}_{\mathrm{r}\mathrm{o}\mathrm{d}1}\left({\displaystyle \frac{\int {\mathrm{V}}_{\mathsf{\lambda }}^{\prime }{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }}{\int {\mathrm{V}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }+{\mathrm{g}}_1\int {\mathrm{S}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }}}\right)\left(1-e^{{\displaystyle \frac{-\int {\mathrm{V}}_{\mathsf{\lambda }}^{\prime }{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }}{\mathrm{R}\mathrm{o}\mathrm{d}\mathrm{S}\mathrm{a}\mathrm{t}}}}\right)\end{array}\right]b-y\le 0\end{array}\right.$$

(3)

where

• $\mathrm{b}-\mathrm{y}=\int {\mathrm{S}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }-\mathrm{k}\int {\mathrm{V}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }$;
• ${\mathrm{V}}_{\mathrm{c}\mathsf{\lambda }}={\displaystyle \frac{\left(\frac{{\mathrm{V}}_{\mathsf{\lambda }}}{{mp}_{\mathsf{\lambda }}}\right)}{max\left(\frac{{\mathrm{V}}_{\mathsf{\lambda }}}{{mp}_{\mathsf{\lambda }}}\right)}}$; ${\mathrm{S}}_{\mathrm{c}\mathsf{\lambda }}={\displaystyle \frac{\left(\frac{{\mathrm{S}}_{\mathsf{\lambda }}}{{mp}_{\mathsf{\lambda }}}\right)}{max\left(\frac{{\mathrm{S}}_{\mathsf{\lambda }}}{{mp}_{\mathsf{\lambda }}}\right)}}$;
• ${\mathrm{E}}_{\mathsf{\lambda }}$ is the light source spectral irradiance;
• ${\mathrm{M}}_{\mathrm{c}\mathsf{\lambda }}$ is the melanopsin sensitivity (corrected for crystalline lens transmittance);
• ${\mathrm{S}}_{\mathsf{\lambda }}$ is the S-cone fundamental;
• ${\mathrm{V}}_{\mathsf{\lambda }}$ is the photopic spectral efficiency function;
• ${\mathrm{m}\mathrm{p}}_{\mathsf{\lambda }}$ the is macular pigment;
• ${\mathrm{V}\prime }_{\mathsf{\lambda }}$ the is scotopic luminous efficiency function;
• K = 0.2616; a_b−y = 0.7; a_rod1 = 3.3; a_rod2 = 1.60; g₁ = 1.00; g₂ = 0.16; RodSat = 6.5 W/m²;

To carry out these calculations, the spectral radiances obtained from daylight measurements need to be first transformed into spectral irradiances, considering that both m-EDI and CL_A are based on spectrally melanopsin-weighted irradiance. This transformation could be conducted using the procedure demonstrated in detailed in related publications [32].

So long as CL_A has been acquired, the CS, which is the relative effectiveness of CL_A for nocturnal melatonin suppression, can be calculated by the following equation [32]:

$$\mathrm{C}\mathrm{S}=0.7\ast \left[1-{\displaystyle \frac{1}{1+{\left(\frac{{\mathrm{C}\mathrm{L}}_{\mathrm{A}}}{355.7}\right)}^{1.1026}}}\right]$$

(4)

Meanwhile, given that m-EDI and CL_A values are all expressed in terms of nocturnal melatonin suppression, which lay the foundation for relative comparison, and also considering the meaningful differences from circadian responses between the two values [31,32], the percentage deviation calculated by Equation (5) could be used to estimate the complex synergistic–antagonistic interactions (S_AI) between photoreceptors in an intact retina, through the synergistic–antagonistic pathways, triggered by daylight stimulus at the cornea.

$${\mathrm{S}}_{\mathrm{A}\mathrm{I}}=\left({\displaystyle \frac{{\mathrm{C}\mathrm{L}}_{\mathrm{A}}-\mathrm{m}-\mathrm{E}\mathrm{D}\mathrm{I}}{\mathrm{m}-\mathrm{E}\mathrm{D}\mathrm{I}}}\right)\times 100\%$$

(5)

3. Results

During measurement periods, all the cities were sunny, with good weather conditions. There existed direct sunlight opposite to the reflectance standard whiteboard at most times, except at a few measured times caused by cloud interference. After measurements, the results are shown as follows.

3.1. Daylight SPD

The human retina has two distinct types of visual photoreceptors: cones and rods, each possessing unique spectral characteristics that can be depicted by spectral efficiency curves. The spectral sensitivity of rods is captured by V’(λ), while cones encompass L-cones, M-cones, and S-cones, whose spectral responses are, respectively, represented by L(λ), M(λ), and S(λ). Additionally, the circadian efficiency curve C(λ), which reflects the influence on melatonin suppression, characterizes the spectral sensitivity of ipRGCs. Given that different light sources exhibit vastly different SPDs, exposure to light can result in sensitivity curves where the relative contributions of photoreceptors carry varying weights. As shown in Figure 1, Figure 2, Figure 3, Figure 4 and Figure 5, it was generally observed that the trends in Daylight SPDs were basically the same with CIE standard illuminance [31]: from 380 nm to 475 nm, they rose quite steeply; from 475 nm to 625 nm, they declined smoothly; from 625 nm to 780, they fluctuated frequently. The maximum peak appeared at about 475 nm, gathering in short wavelengths, and seemed to be much closer to that of C(λ) curve than those of other spectral efficiency curves, such as S(λ), M(λ), L(λ) and V’(λ), which shifted relatively to the left or the right. It can also be seen that, on the whole, variations in daylight spectral radiant intensity accorded with the SAAs’ changes, and daylight SPD curves in Xining, Beijing, and Chongqing showed a peak at or near the maximum solar altitude, whereas in Kunming and Nanchang the curves corresponding to the maximum solar altitudes were lower than those at other solar altitudes. This is probably a consequence of the variations in the SSCs: while the sun surface is in the condition of Θ², with less cloudiness in the sky, there is little impact on the spectral radiant intensity of daylight. In Xining, Beijing, and Chongqing, daylight SPD curves at the maximum solar altitudes with the SCCs being Θ² were therefore at or near the top. As cloud cover increases, the SSCs are gradually transformed into Θ⁰ and Θ, leading to enhance the impact on the spectral radiant intensity of daylight. Consequently, in Nanchang, the daylight SPD curve at the maximum solar altitude with the SCC being Θ was in a middle position rather than at the top. While the sun surface is in the condition of Л, the sun is completely obscured by thick clouds and the impact on the spectral radiant intensity of daylight reaches the maximum. Thus, in Kunming, the daylight SPD curve at the maximum solar altitude with the SCC being Л was much lower than those at other solar altitudes with the SCCs being Θ².

3.2. Daylight Illuminance

In Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10, horizontal and vertical illuminance values for daylight in five cities are reported. Illumination is one of the key physical characteristics of composite white light. A lower illuminance indicates weaker light, and vice versa. In general, resembling the variations of daylight SPDs described, these values at eye level also increased or decreased as the SAAs increased or decreased in a day; in particular, in Kunming they fluctuated frequently given the variable SSCs. Moreover, it was found that in every city the trends among the three values were, in most cases, similar, but sometimes quite different; for example, at approximately 50°and 70° in Beijing, in the range comprised between approximately 50° and −50° in Nanchang, and at approximately −30° in Chongqing, trends in vertical illuminance values were at odds with those of horizontal illuminance values, probably resulting from the influence of the SSCs’ variations at these measured hours.

3.3. Daylight m-EDI, CL_A and CS

Figure 11, Figure 12, Figure 13, Figure 14 and Figure 15 serve to illustrate m-EDI, CL_A, and CS values calculated from spectral radiances detected at the eye level of a person standing on the ground. Atmospheric conditions may lead to asymmetry in the spectral characteristics on either side of solar noon. Compare to m-EDI following smooth bell-shaped curves on an ideal clear day [13], different weather conditions introduce volatility and variability.

It can be observed that m-EDI, CL_A, and CS values during the day varied closely with the SAA changes, and their variations were generally consistent with those of daylight SPDs and illuminances as previously stated. Furthermore, the three values appeared to share a common trend during a day. Specifically, for the SAAs from approximately 5° to 30° at the beginning, m-EDI, CL_A, and CS values in five cities overall showed a fast rise. Then, from approximately 30° to −30°, it was found that the three values in Kunming fluctuated continuously due to the variable SSCs and most of CS values varied around nearly 60%, except at 40°, 60°, and 70°, where the corresponding SSCs were all Л and the CS values dropped considerably. In contrast, m-EDI, CL_A and CS values in the rest four cities generally changed with SAAs, and peaked at or near the maximum solar altitude, but, particularly in Nanchang, dropped to varying degrees near the maximum solar altitude. This different trend can also be seen in Beijing’s morning and Chongqing’s afternoon: in Beijing, the three values slid slightly at approximately 50° and 70°, and in Chongqing the three values increased at approximately −30°, while the SAAs dropped to the contrary. The reason for the different trends between the three values and the SAAs may be attributable to the obvious variations of the corresponding SSCs. Meanwhile, it was also found that compared to m-EDI and CL_A values, CS values varied more slightly within this range, which were almost around the baseline level shown as horizontal lines and even the same at different solar altitudes. This perhaps means that during this period, the impact of daylight on melatonin suppression comes close to full saturation. Eventually, for the SAAs from approximately −30° to −5°, downward trends were obvious as the SAAs decreased. Interestingly, in Kunming, the CS value rose again at approximately −10°, where the corresponding SSCs varied largely; in Beijing, there was a sudden climb after the angle of 20°, probably due to the fact that the spectral radiant intensity of daylight increased considerably from approximately −20° to −5°.

It can also be seen the trends in illuminance values in the five cities were basically in accordance with those of m-EDI, CL_A, and CS values. Moreover, compared to horizontal illuminances, the three values appeared to show trends much closer to those of vertical illuminances, which were obviously found at approximately 50° and 70° in Beijing, at approximately −30° in Chongqing, and notably in the range comprised between approximately 50° and −50° in Nanchang.

3.4. Daylight CCT and S_AI

As can be seen from Figure 16, Figure 17, Figure 18, Figure 19 and Figure 20, as the SAAs changed, with ranges from approximately 3000 K to 7000 K over the day, daylight CCTs fluctuated to a greater degree at dawn and dusk, and varied to a comparatively milder degree in most of the daytime. Particularly, daylight CCTs in Kunming tended to vary very much due to the variable SSCs. The CCT values were highly dependent on solar elevation. When the sun was low in the sky, a high horizon may obscure the brightest part of the sky or the sun. In this way, buildings and the landscape can give rise to deviations from the smooth curves that would otherwise be observed. Moreover, it can be observed that daylight CCT trends were, in most cases, consistent with S_AI, except at approximately 5°, 10°, 60°, 85.15°, and −80° in Kunming, −5° in Xining and Beijing, and 5° in Nanchang, where their trends were reversed.

To investigate the hypothetical link between daylight CCTs and S_AI, a graph illustrating CCTs and S_AI values in the five cities is shown in Figure 21, scaling all the vertical illuminances to 1000 lux at the cornea, which is the minimum threshold to guarantee the demand for non-image-forming effects in humans. It can be seen that the relation between daylight CCTs and S_AI did not adhere to a simple linear dose–response curve and there was an obvious disconnection partitioned roughly at 4000 K. Depending on the data sets obtained, it was found that no matter whether daylight CCT values were in the range of 3000~4000 K or 4000~7000 K, the S_AI values overall increased as daylight CCT values increased, and vice versa, following independent curves. Furthermore, to different daylight CCT values, the same S_AI values corresponded by occasion. For example, while CCT values were 3444 K (at Point P) and 5947 K (at Point Q), the corresponding S_AI values were both 41%. Fitting the data by the Origin 2024, the relation can be expressed independently on a set of a piece-wise function as follows:

$${\mathrm{S}}_{\mathrm{A}\mathrm{I}}=\left\{\begin{array}{c}3.8\times {10}^{-4}{\times T}_C-0.91\mathrm{If}3000<T_C\le 4000\mathrm{K}\hfill \\ 0.39\times \ln \left(T_C-3380.5\right)+0.04\mathrm{If}4000<T_C\le 7000\mathrm{K}\hfill \end{array}\right.$$

(6)

where $T_C$ denotes daylight CCT value; the coefficient of determination (R²) is greater than 0.95 if CCTs are from approximately 3000 K to 4000 K, and 0.97 if CCTs are from approximately 4000 K to 7000 K, receiving a good correlation.

The set of the piece-wise function gave the likelihood of assessing the synergistic–antagonistic interactions between photoreceptors under different circumstances of daylight CCTs: if daylight CCT values are approximated to around 4278.4 K (at Point O), the combined effects of cones and rods through the synergistic–antagonistic pathways gradually served no function (i.e., S_AI = 0) to the response of the ipRGCs; if daylight CCT values were higher than 4278.4 K or lower than 4000 K, the combination of cones and rods responses through the synergistic–antagonistic pathways generally served synergistic functions (i.e., S_AI > 0) to the response of the ipRGCs; if daylight CCT values were lower than 4278.4 K but at the same time higher than 4000 K, the sum of cone-response and rod-response through the synergistic–antagonistic pathways mainly served an antagonistic function (i.e., S_AI < 0) to the response of the ipRGCs.

4. Discussion

4.1. Daylight SPD Concerning NIF Effects

Over the course of extensive evolution, the intricate human photoreceptive system has evolved to better adapt to the composition of daylight spectra [57]. The spectral power distribution of the light has effects on human circadian rhythm and consequently on health. The photosensitivity of all photoreceptor cells in the human eye, including ipRGCs, is affected by wavelength. In terms of the photosensitivity mechanism, the spectral composition of light entering the eyes inevitably influences non-image-forming effects. The blue light spectrum was believed to notably suppress melatonin secretion [29,58], particularly during nighttime exposure [59,60], which can significantly affect sleep. Comparing 3 h of blue light exposure to orange light exposure at night, it was found that blue light effectively suppresses melatonin secretion and alters circadian rhythms [61]. Another study has also confirmed that white light illumination enriched with the blue spectrum plays a role in reducing memory errors and reaction times in work settings [62]. Berman et al. [63] noted that the photoreceptors’ peak spectral sensitivity lies within the blue light range, while the maximum sensitivity of circadian retinal photoreceptors lies in the range of the visible light spectrum with a peak wavelength of 480 ± 5 nm, while Brainard et al.’s [64] experiment exhibited peak sensitivity at a wavelength of 460 nm. In this study, it showed that the non-image-forming effects of light are mediated primarily via the ipRGCs in an intact retina with a peak sensitivity that is blue-shifted (λ_max$\cong 480\mathrm{n}\mathrm{m}$) relative to photopic visual system (λ_max$=555\mathrm{n}\mathrm{m}$), which is consistent with the results of Berman et al. [63]. Based on the blue-shifted sensitivity of the ipRGCs, the result that the maximum peak range of daylight SPDs nearly overlapped with circadian spectral efficiency curves C(λ) makes the proof clear that daylight rich in short wavelength (blue) radiation has significant influences on the NIF effects. Therefore, posing attention to the potential health benefits of daylight, and at the same time avoiding the retinal damage caused by the blue part of the spectrum [65], would be a judicious way to address the comfort and health of occupants in the built environment. Castilla N et al. [35] demonstrated the significance of maintaining a balance between daylight and electric light to create an ideal learning environment that can significantly impact students’ academic performance. Although it is currently impossible to accurately quantify the physiological and psychological effects of daylight on the human body, research based on specific health conditions can establish the correlation between daylight and human physical and mental health, which can then be applied to the design of interior spaces.

4.2. Daylight Intensity Concerning NIF Effects

In addition to the solar altitude, factors such as the SSCs, cloud cover, cloud shape, and atmospheric contents are also important considerations while studying daylight [66]. The results reported in the previous section showed that the SSCs variations among Θ², Θ⁰, Θ, and Л probably made the spectral radiant intensity of daylight quite different and consequently triggered diverse the NIF effects in humans. This is problematic for current worldwide standards depending mainly on idealized, static sky conditions, such as the CIE standard overcast sky, without considering that the SSCs vary widely in different daylight-climate regions. Therefore, the SSCs could also be considered as a basis for the next generation of building guidelines to realize comprehensive evaluation of building lighting environments [67], thus addressing a more realistic and effective potential of daylight NIF effects. However, the SSCs are always undergoing stochastic and dynamic changes [68]; it is not possible to access the only data measured for the five cities and then forge a systematic link between the data obtained and the results reported in this work until more comprehensive collections of meteorological data founded on a wide range of observations over the years at a given site are acquired, and the same happens for other sky conditions, such as cloud cover, cloud shape, and the atmospheric contents mentioned in this paper.

It is now established that the intensity of light sources is critical to achieving circadian effects [24,69]. As lighting intensity increases, the suppression of melatonin secretion in the human body becomes more pronounced. Chen et al. [70] augmented daylight-deficient indoor environments with various artificial lighting methods, and the findings indicate that vertical lighting enhances the uniformity of light intensity within the space, thereby elevating the rhythmic stimulus to a critical threshold. The results from the present paper also revealed that vertical illuminance rather than horizontal illuminance shows trends close to those of m-EDI, CL_A, and CS. This means that while lighting designers predict a light source impact on the human circadian system, it is important for them to evaluate vertical illuminances at the cornea to better assess the NIF effects of any luminous environment; the horizontal working plane measurements can only guarantee vision. Moreover, from the analysis of CS, CL_A, and m-EDI trends, it can be noted that CS and CL_A values tended basically to differ from m-EDI values. In fact, the m-EDI model is developed without respect to any underlying physiology and mainly characterizes the spectral sensitivity at which ipRGCs [32], excluding other photoreceptors, respond to a light stimulus. Instead, CS and CL_A values incorporate circadian responses of different photoreceptors, spectral opponency, and shunting inhibition implied by the known neuroanatomy and electrophysiology in the human retina [71,72], and these results are in accordance with a non-redundant role for cones, rods, and ipRGCs in mediating human non-image-forming photoreception [9,27]. Furthermore, owing to the fact that the relative spectral efficiency function is normalized with respect to the photopic luminous efficiency function, m-EDI has a great inhibition on the yellow–green band of the spectrum in which M-cones and L-cones have peak spectral sensitivity [31], whereas CS and CL_A generally tend to avoid such situations, given that CS and CL_A values are determined by the value of the results of this equation: $\int {\mathrm{S}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }-\mathrm{k}\int {\mathrm{V}}_{\mathrm{c}\mathsf{\lambda }}{\mathrm{E}}_{\mathsf{\lambda }}\mathrm{d}\mathsf{\lambda }$, choosing different relations in Equation (3) for calculation in terms of the proportion of photopic regions in a spectrum [32,73]. Therefore, the approach of using a single parameter to address the measurement of NIF effects cannot be straightforwardly generalized to the application of composite light sources in buildings. Overall, the m-EDI model offers greater convenience in evaluating the daylight NIF effects, while the CL_A model offers more accuracy.

4.3. Daylight CCT Concerning NIF Effects

In general, the light with a high CCT has a stronger impact on melatonin suppression than the light with a low CCT [26,74]; individuals operate more efficiently under a high CCT light environment, which enhances alertness [60,75], boosts positive emotions [76], and reduces pressure [77]. In the low CCT lighting environment, individuals experience greater relaxation [78,79], reduced mental workload [80], and a significant decrease in subjective negative emotions [81]. Conversely, Li et al. [82] found that higher color temperatures were associated with fewer pleasant emotions and more sad emotions among participants. Additionally, Kulve et al. [83] showed that variations in correlated color temperature had no significant impact on subjective alertness or body surface temperature, and Bellia et al. [84] indicated that there was no connection between eye-level CCTs and melatonin suppression. In this study, it is worth noticing the relationship between daylight CCT values and S_AI values: the synergistic–antagonistic interactions between S-cones, L-cones, M-cones, rods, and ipRGCs tended to be relevant for different CCT values of natural dynamic daylight in a non-linear way, expressed by a set of a piece-wise function. This may be due to the fact that in the long course of evolution, human photoreceptor cells with different spectral sensitivities have been somewhat suited to daylight, with full spectral content of light sources in the visible spectrum [57], with its CCTs varying regularly during the day, maintaining a reasonable spectral match. Despite many uncertainties about circadian phototransduction, it appears that the spectrally opposed blue-versus-yellow color mechanism that defines the irregular discontinuity of spectral sensitivity in nocturnal melatonin suppression is physiologically responsible for the segment, consistent with independent assessments of human photopic color vision, which shows an analogous behavior [9,31,32]. The mathematical expressions in Equation (6) are characterized by the merit of being reasonable in regard to the synergism–antagonism of human photoreceptors, consistent with retinal neurophysiology and with considerations to the functional characteristics of human circadian phototransduction. The proposed expressions bring with them new perspectives on human circadian photostimulation, and promote the discussion of architectural lighting praxis and of the relationship between light and health. The present expressions simply endow designers with more accurate characterization of circadian responses to a light stimulus concerning the CCT parameter.

From the application perspective, the CCT values of polychromatic light sources used for general illumination or of daylight entering buildings are frequently encountered in the range comprising 3000 K and 7000 K. It is therefore possible to consider S_AI as a circadian action factor in the design context, enabling designers to explicitly address and distinguish the NIF effects of light exposure, and to effectively evade the negative range of CCTs, where the corresponding S_AI values are lower than zero, prospectively, so as not to create a disadvantageous lighting environment for occupants. The S_AI factor allows designers to make relative comparisons between CCT alternatives (such as 6000 K and 3500 K, of which the corresponding S_AI values show good agreement) to broaden more strategies during design. Clearly, too, the S_AI factor makes it possible to predict the circadian effectiveness of a given luminous environment in terms of CCT values. So far as the circadian effect of luminous environments indoors is concerned, this parameter can serve as a reasonable proxy to represent the inter-relationship between the NIF effects and CCTs.

However, due to the unpredictability stemming from the fact that the opponent blue-versus-yellow mechanism is represented by the difference among cones’ responses and that rods control the spectral sensitivity of the cones [26,85], and also taking into account the complexity between CCT values and Circadian Light values, as graphically stated by Bellia L et al. [86], there is a lack of sufficient data to show the trends of S_AI at around 4000 K that seem to be far more complicated and irregular. Moreover, the factor S_AI, derived from an analysis of a full spectrum of daylight in this work, also cannot be taken as evidence of the circadian response to the combination of monochromatic light exposure since human circadian photostimulation does not follow Abney’s law like the linear behavior of luminance in orthodox photometry [73,87]. Along with collecting more empirical data, further refinements may be proposed and better evaluations may be obtained.

4.4. A Four-Period Schedule Concerning NIF Effects

In modern society, individuals spend a significant portion of their lives indoors [53,88]. The predominant lighting in most living spaces is constant static illumination throughout the day, leading to a growing disparity between natural light cycles and individual physiological rhythms. Dynamic lighting patterns, compared to static lighting, elicit superior non-image-forming effects [89]. Exposure to dynamic light throughout the day reduces an individual’s alertness before bedtime compared to static lighting [90,91]. Rahman et al. [92] found that dynamic lighting patterns facilitate the adjustment of circadian rhythms to the shift in sleep–wake phases. However, other studies have not observed significant effects of dynamic light on nocturnal sleep, with some even indicating the opposite [93]. The results reported in Section 3.1 and Section 3.3 showed that the SAAs and SSCs probably made the spectral radiant intensity of daylight quite different and consequently triggered diverse NIF effects in humans. While the SSCs were relatively stable, variations of daylight NIF effects during the day were regular and dynamic with the SAAs. Intrinsic melatonin suppression responded accordingly to daylight as discussed in the results section; it is thus possible to differentiate three distinct ‘non-image-forming effect’ periods in daytime as follows (as shown in Figure 22):

−

At sunrise to early morning (at about 8:00 h according to local civil time), for the SAAs from approximately 5° to 30°, in vivo melatonin suppression shows a fast rise, and sufficient daylight can ensure the rousing of one’s awareness.

−

Early morning (at about 8:00 h according to local civil time) to mid-afternoon (at about 16:00 h according to local civil time), for the SAAs from approximately 30° to −30°, in vivo melatonin suppression tends to be nearly full saturation, and high levels of daylight have potential to add alertness and vigilance.

−

Mid-afternoon (at about 16:00 h according to local civil time) to sunset, for the SAAs from approximately −30° to −5°, in vivo melatonin suppression decreases significantly, and appropriate light exposure may eliminate fatigue and stress.

Additionally, in terms of the CS metric, it could be possible to predict that in vivo melatonin suppression does not respond to daylight during night time, from sunset until sunrise, without the presence of any daylight, where the CS values is zero [87]. Thus, light exposure that might induce detrimental NIF effects should be obviated, in order to maintain the natural sleep–wake pattern.

Daylight is always undergoing stochastic and dynamic changes from dawn until sundown and human beings have adapted to it. As noted, the aforementioned schedule over the 24 h day described four changing tendencies of melatonin suppression and seemed to better accord with human modes of production and way of life under the influence of daylight and the light–dark cycle that the sunrise makes, as well as the sunset. It could be determined that during every period of the day, human demands for Circadian Light exposure are distinct. This means that the regular timing and durations of light exposure could be applicable in correctly entraining the circadian effect in workplaces or domestic environments. In practice, a designer could develop a time-varied lighting strategy for a day-long period in healthy buildings, in terms of each time-scheduling category, for a quintessential well-rested and regularly sleeping individual. Initially, exposure to morning light shifts the phase of an individual’s circadian rhythms earlier [94], during the circadian enhancement period in the early morning, alertness and vigilance may be aroused by intense light exposure [95]. Then, during the circadian saturation period, light stimulus could be gradually transformed into a lower level, and at high noon drop to the minimum illuminance threshold of 500 lux so that it would be propitious to physical and mental relaxation as well as for visual tasks. After lunch, light stimulus can be strengthened again, and then gradually lowered. Next, during the circadian attenuation period close to end-of-day, a mild rise in light stimulus could be seasonable to ameliorate weariness and fatigue. Daylight exposure duration influences the quantity of light an individual is exposed to [19], with a notable cumulative effect over time [96]. During these two periods, it is essential to enhance light exposure during the day to ensure sufficient light intake, thereby mitigating the adverse effects of nighttime lighting. Eventually, exposure to high levels of light during the night can delay the phase of an individual’s circadian rhythm [16] and should be averted so as not to disturb the natural sleep–wake cycle, especially at pre-bedtime.

However, it should be noted that pros and cons of the categorization will vary very much depending on different individual circumstances. Due to occupational differences, significant variations exist in the daytime light exposure experienced by shift workers [19,97] and those working in underground spaces (e.g., subways, mines) [98]. During the circadian enhancement period, contrary to the previous statement, they require minimized morning exposure to facilitate sleep in daytime, and instead bright-light stimulus to increase alertness during the night. Future research can also focus on the unique requirements of various populations and workplaces, integrating the four-period schedule concerning NIF effects to create a healthy and comfortable lighting environment, sustain circadian rhythms, and enhance the quality of physical and mental health.

5. Conclusions

As outcomes of the relationship between human physiology and lighting parameters grow, what seems sure is that new approaches and design schemes are required for luminous environments indoors that can effectively maintain both user’s health and well-being. The research presented in this paper can provide new insights into these approaches and schemes. To analyze daylight characteristics for design practice purposes, focusing particularly on their non-image-forming effects in humans, daylight measurements were conducted in five cities in China to acquire the data of spectral power distribution, correlated color temperature, and illuminance. One of the main contributions of this research was that an action factor S_AI was developed to estimate the potential for the non-image-forming effects of light stimuli in the built environment. This factor could be considered as a reasonable proxy to represent the inter-relationship between the non-image-forming effects and the correlated color temperatures. Another important finding indicated that it was likely to develop a time-varied daylighting and lighting strategy with four distinct periods of non-image-forming effects over a day.

To acquire more universal findings would necessitate conducting the same research in other geographic locations, under other weather conditions, to study if the results are somehow similar, and to therefore further verify the validity of such hypotheses. At the same time, characteristics of daylight entering the room are impacted by the architectural features of the buildings, such as form, massing, orientation, and window-to-floor ratio, and the optical behaviors of the environment such as the transmission of glasses, absorption, and reflection of indoor surfaces. This would require further efforts to investigate the inter-relationship between daylight and environment and its influences on circadian responses, and also to validate the results of this work.

The study here has laid the foundation for further research and the ultimate goal is to develop easily feasible approaches and strategies that provide useful recommendations and considerations for design practices. It is hoped that the findings of these studies will foster further progress in architectural applications where daylight/light can be an effective promoter of human health.