DISCOMFORT GLARE METRICS

Transcription

1 DISCOMFORT GLARE METRICS Investigating their accuracy and consistency in daylight glare evaluation by using human subject study data Jae Yong Suk The University of Texas at San Antonio Marc Schiler University of Southern California Karen Kensek University of Southern California ABSTRACT Daylighting has been widely adopted in buildings to reduce building energy consumptions and to promote occupant productivity and comfort. Many daylighting and discomfort glare metrics and tools have been developed to provide the desired amount of natural light while avoiding excessive visual discomfort. Considerable research has been done to examine how to evaluate glare and how to mitigate the problem for building users from a design standpoint. Confounding the problem is the fact that glare is a subjective phenomenon, and people do not always agree on what constitutes glare. For more accurate glare evaluation, it is critical to make a clear understanding of the existing glare metrics and tools. This study performs validation studies on five glare metrics including Daylight Glare Probability (DGP) and Daylight Glare Index (DGI) that have been developed specifically for daylight glare issues. A parallel human subject study has been performed to collect subjective discomfort glare evaluations. In addition, high dynamic range (HDR) imaging was used to capture and analyze the glare scenes that were experienced by those human subjects. More than 450 daylight glare scenes and subjective surveys were collected in a closed office setting. The collected data has been processed in hdrscope and Evalglare to obtain glare scores, and the results were compared for statistical analysis of subjective evaluations. The results show only one or none of the five metrics correctly matches to the subject s evaluation for each glare scene. This evaluation comparison study shows that the five glare metrics have significant inconsistency and inaccuracy issues. KEYWORDS Daylighting, discomfort glare, shading, occupant comfort, architectural glass, human subject study, Evalglare INTRODUCTION Daylighting provides significant benefits in buildings such as energy saving, occupant comfort, and occupant productivity. Different daylighting analysis methodologies and tools have been developed to ensure desired amount of natural light in buildings. Energy simulation tools calculates estimated amount of energy savings from different daylighting strategies. Daylighting and lighting simulation tools allow checking illuminance levels inside a building. Discomfort glare analysis tools and metrics evaluate different levels of visual discomfort in indoor spaces (Konis, 2013; Suk et al., 2013; Hirning et al., 2013 and 2014). Although many people cannot accurately define the causes and types of glare, most people experience it daily inside and outside of buildings. Researchers agree that glare occurs when the eyes have adjusted to a certain general level of brightness, and some annoying, distracting, or blinding light appears within the visual field. It is also well acknowledged that excessive levels of glare source luminance cause discomfort glare regardless of contrast levels. Veiling 2016 WORLD CONGRESS 1

2 reflection is also classified as a type of discomfort glare. Glare can be described in one of three main ways: according to the process that created the glare, according to an individual s perceived degree of glare intensity, and according to the results of the glare. Many existing glare metrics including DGP (Daylight Glare Probability), DGI (Daylight Glare Index), UGR (Unified Glare Rating), VCP (Visual Comfort Probability), and CGI (CIE Glare Index) focus on evaluating perceived degree of glare intensity. BACKGROUND DGP is the most recent glare metric that was developed to evaluate daylight glare. For determining glare, its formula combines the vertical eye illumination as a glare measure with the central term of existing glare metrics. It also considers the influence of the glare source. Compared with the other existing glare metrics, DGP shows a very strong correlation with the user s response regarding glare perception (Wienold and Christoffersen, 2005). DGI was also developed for daylight glare as it evaluates a large glare source such as high luminance levels of windows (Bellia et al., 2008). UGR, VCP, and CGI were developed to address glare issues caused by electrical light sources. Even though they are not designed for daylight glare, several studies claim potential use of these metrics for daylit spaces (Isoardi et al., 2012). Therefore, all five existing glare metrics were examined in this study. For automatic glare evaluation of the luminance images captured by High Dynamic Range Imaging (HDRI), the software program Evalglare was developed (Wienold and Christoffersen, 2006). Evalglare identifies potential glare sources based on a threshold value, which can be 1) specified by the user manually as a fixed luminance value; 2) computationally determined based on average luminance in the field of view; or 3) computationally determined based on a user specified task location (Inanici, 2004 and 2005). The threshold value in Evalglare is by default a multiplier 5 times of the mean total image luminance or the mean task area luminance. Evalglare can be plugged into several daylighting analysis software programs making glare analysis easier, but its use has not been widely adopted yet. Evalglare calculates five glare metrics from an image in either HDR or PIC format in either a 180 degree angular fisheye or a normal perspective image. By typing Evalglare commands in a DOS window or using hdrscope software, it calculates luminance value and location of each pixel of the image, then uses the information to calculate the crucial values such as background mean luminance, glare source luminance, glare source position, solid angle of glare sources, vertical illuminance, and direct vertical illuminance, etc. Based on the information taken from an image, Evalglare provides glare scores and visual representation of potential glare source locations and sizes. The following table shows different glare score ranges of the five glare metrics to categorize different levels of perceived glare from imperceptible glare to intolerable glare. Higher score represent worse discomfort glare issues in DGP, DGI, UGR, and CGI. Unlike the other glare metrics, the higher score represents better visual comfort in VCP, since it is the probability of comfort. Degree of Perceived Glare DGP DGI UGR VCP CGI Imperceptible < 0.35 < 18 < < 13 Perceptible Disturbing Intolerable > 0.45 > 31 > 28 < 40 > 28 Table 1. Degree of Glare in Different Glare Metrics (Jakubiec, 2010) Based on simulation-only study, Jakubiec and Reinhart claim that DGP shows the most robust results for most daylight situations among the five metrics (Jakubiec and Reinhart, 2010). Their study also found that VCP is not intended to be used for daylight glare calculations, and CGI tends to show much higher glare levels than the other metrics. DGI and UGR can be used for daylight glare evaluation, but they work only when the direct sunlight does not enter (Jakubiec, 2010). Based on human subject study performed in a large open office space, Hirning et al. claim that DGP and DGI were unable to provide accurate evaluations of discomfort glare experienced by the participants (Hirning et al., 2013 and 2014). These previous findings and claims on the application of the existing glare metrics were expected to be verified in this study. Evalglare is one of the most practical tools for calculating daylight glare, but previous research has shown that the existing glare metrics provide inconsistent glare evaluations for a same glare scene, which makes users suspicious of their evaluation accuracies (Suk and Schiler, 2012). The previous study was done entirely without subjective inputs; recent research utilizing extensive human subject study data was performed to find out levels of accuracy and consistency of the five metrics. METHOD After reviewing a number of precedents (Luckiesh and Guth, 1949; Hopkinson, 1957; Ngai and Boyce, 2000; Velds, 2002; Osterhaus, 2005; Linney, 2008; Wymelenberg, 2012; Hirning et al, 2013; Konis, 2013), a human subject study methodology was developed to collect 2 FACADE TECTONICS INSTITUTE

3 accurate and consistent subjective evaluations of discomfort glare caused by daylight. Electrical lighting sources were completely excluded so that the visual discomfort experienced by participants is known to be the sole result of daylight glare, rather than glare contributed from artificial light sources. HDRI was used to capture the visual information that was experienced by human subjects. HUMAN SUBJECT STUDY AND HIGH DYNAMIC RAN GE PHOTOGRAPHY Three male and three female subjects were recruited for the human subject study. No participants with vision-related illness or color blindness were included in the study. Similar to Hopkinson s (1957) research, a small number of subjects was chosen instead of a large number of random subjects because of the unique characteristics of daylight glare research. Hopkinson s study compared the findings from an experienced group of six people to the findings from fifty random subjects and found that the experienced group provided more consistent glare evaluation data than the random subjects. As the sun and sky conditions constantly change, it is hard to control daylight glare conditions to be identical for a large number of subjects. Therefore, it is ideal to have a small number of subjects and have them all experience many different daylit environments. Testing the same subject multiple times helps to ensure evaluation accuracy and consistency. Further study with an expanded number of subjects should be pursued, expanding on the findings of this study. The human subject study to assess discomfort glare issues was performed inside a closed office space at the University of Southern California. There is no exterior visual obstruction that is closely located to the office. The room is a corner office with two clear glass windows facing southwest and southeast (Figure 1). Each window has two adjustable blinds: a venetian blind and a roller blind. The corner office condition was selected to allow more natural light inside the space (to be more exposed to the outside) and to avoid the severe contrast issues that can occur in a closed room with small aperture windows. This research setting represents many office buildings with large amount of fenestrations to allow direct view to outside and natural light inside buildings. The allowance of adequate daylight also helped to avoid the necessity of electrical lighting usage. Another advantage of using the corner office condition is that it provides more opportunity to experience potential glare sources. Since the office has front and side windows, it can have potential glare sources from different directions. The research setting was used for the subjects and HDR photography (Figure 1). The room is 11'-3'' high by 9'-6'' wide by 11'-4'' long. The height of the windows goes from task height (2-6 A.F.F.) to ceiling. A desk was located adjacent to the windows facing southwest and southeast; a desktop monitor was placed on top of the desk, in front of a southwest-facing window. A total of four illuminance sensors and data loggers were installed: one behind the monitor facing the window, one above the camera facing the window, and sensors on either edge of the desk facing up. Data loggers were provided with thirty-second recording intervals in which to record data from the sensors. The equipment was carefully calibrated and normalized prior to the study. A glare scene was captured using various exposures by a Nikon Coolpix 4500 camera and angular fisheye lens. More than +- 2 full stop exposures from normal exposure were taken to capture the dynamic ranges of human eyes by adjusting shutter speeds only. For the scenes with a direct view of the sun, much wider exposure ranges were taken to capture the extreme ranges of the luminous environment: the sun s disk and interior surfaces. The luminance values on the captured HDR images were also calibrated with field measured luminance values by a Cooke luminance meter. Illuminance sensors (Li-Cor LI-210R) and data loggers recorded the vertical illuminance values arriving into the subject s eyes and horizontal illuminance at the task height. The captured HDR images were then processed in hdrscope and Evalglare to compare DGP, DGI, UGR, VCP, and CGI index scores WORLD CONGRESS 3

4 N With HDR Photography With Human Subject Figure 1. Interior glare study research setting: 1) Camera; 2, 3, 4, 5) Li-Cor sensors for vertical illuminance; 4, 5) Li-Cor sensors for horizontal illuminance; 2, 3, 4, 5) HOBO sensors; 6) Tripod; 7) Cooke luminance meter Each subject was tested under three different lighting conditions. For each, the subjects were asked to perform three different tasks: no task, typing task, and writing task. The room had both venetian blinds and roller blinds on the windows. There were three lighting conditions: Fully open: both roller and venetian blinds were fully opened on both front and side windows and could not be adjusted (Figure 2, top). Roller blinds only: the subjects were able to separately adjust front and side roller blinds as they preferred. The venetian blinds were fully open and could not be adjusted (Figure 2, middle). Venetian blinds only: the subjects were able to separately adjust front and side venetian blinds as they preferred. The roller blinds were fully open and could not be adjusted. Unlike with the roller blinds, venetian blinds can also be set to different angles. This allowed the subjects to introduce more natural light into the room if they wanted higher light levels and to block incoming natural light if they wanted lower light levels (Figure 2, bottom). After performing each task, the subjects provided subjective evaluations on each glare scene. Visual comfort and visual satisfaction levels were asked in a seven point Likert scale. Also, perceived glare categories such as imperceptible, perceptible, disturbing, and intolerable glare were asked to the subjects. The collected subject s responses to different questions were compared each other to verify consistency of their responses on a same glare scene. 4 FACADE TECTONICS INSTITUTE

5 Fully open No Task Typing Task Writing Task Roller binds No Task Typing Task Writing Task Venetian blinds No Task Typing Task Writing Task Figure 2. Nine different glare scenes under three different task and lighting types EVALGALRE SIMULATION S Human subject study was performed inside the office from February 18, 2013, to June 17, Each subject experienced at least 75 different daylit conditions (25 different conditions per task). Most of the tests were performed under clear sky conditions, but each subject performed at least one test under an overcast sky condition. With the various sky conditions, subjects experienced various levels of discomfort glare, even when the blinds were fully opened. The collected human subject study data was analyzed using the five existing glare metrics (DGP, DGI, VCP, UGR, and CGI). The captured glare scenes in HDR image format were properly edited in hdrscope. Then, they were processed in Evalglare to obtain glare scores calculated by each metric. Comparison of the calculated glare scores to subjective evaluations was expected to show the accuracy of each glare index s analysis of various daylit conditions. As the existing metrics have an inconsistent evaluation issue on one set of data (Suk and Schiler, 2012), the expanded set was evaluated first to confirm the findings in the previous study and also to see whether or not they would have an inaccuracy issue. DATA More than 450 captured glare scenes were analyzed in Evalglare to calculate glare evaluations using the five glare metrics. The glare scores from Evalglare were transferred to the perceived glare degree categories based on the glare score ranges that were developed for each glare index. Then, the glare evaluation results from each glare index were compared to the collected human subject evaluation data to see whether they matched each other or not. Subjective glare evaluation data was compared to the glare scores calculated by the existing glare metrics to check what glare metrics match best to the subjective evaluation for the no-task glare scene for fully open blinds, roller blinds, and venetian 2016 WORLD CONGRESS 5

6 blinds (Figure 3). Subjective evaluation and the glare metrics matching subjective glare evaluations are in bold. The fully open blind scene was evaluated as disturbing glare by a human subject but no glare metric matches it. DGI and UGR underestimate the glare scene while the other metrics apparently overestimate the scene as intolerable glare. This example supports the previous finding that DGI and UGR cannot be used for a glare scene with direct sunlight. DGI s evaluation on this scene is surprising as it reported imperceptible glare even though the sun is visible in the field of view. For the roller blind scene, UGR matches to subjective evaluation, as it evaluates perceptible glare. DGP and DGI underestimate the glare scene while VCP and CGI overestimate it. None of the metrics evaluates the venetian blind scene as imperceptible glare, and all of them overestimate the glare to be perceptible, disturbing, or intolerable compared with the subjective evaluation. Subjective evaluation: Disturbing glare DGP = (Intolerable glare ) Fully open DGI = (Imperceptible glare) UGR = (Perceptible glare) VCP = (Intolerable glare ) CGI = (Intolerable glare) Subjective evaluation: Perceptible glare Roller blinds DGP = 0.26 (Imperceptible glare) DGI = (Imperceptible glare) UGR = (Perceptible glare) VCP = (Disturbing glare) CGI = (Disturbing glare) Subjective evaluation: Imperceptible glare Venetian blinds DGP = (Perceptible glare) DGI = (Perceptible glare ) UGR = (Disturbing glare ) VCP = (Intolerable glare) CGI = (Intolerable glare ) Figure 3. No-task condition scenes processed in Evalglare, compared accurately to participants subjective evaluation Glare evaluations between human subjects and the existing glare metrics were compared for typing task glare scenes (Figure 4). The fully open blind scene shows that DGP, VCP, and CGI evaluate the scene as intolerable glare, which matches what the subject experienced. As shown in this example, the glare metrics show better accuracy and consistency for an extreme glare condition. DGI and UGR still underestimate the scene even though it includes direct sunlight in the field of view. This confirms that DGI and UGR are not suitable to analyze indoor daylit conditions with high potential of direct sunlight (or direct view of the sun). Again, none of the metrics matches the subjective evaluations for the roller blind scene. All of the metrics clearly underestimate the glare source through the roller blind. Also, only UGR matches to the subjective evaluation for the venetian blind scene in its evaluation of the scene to have disturbing glare. DGP reports imperceptible glare, DGI reports perceptible glare, and VCP and CGI report intolerable glare. This example clearly shows inconsistency issue as five glare metrics evaluate from imperceptible to intolerable glare on a same scene. 6 FACADE TECTONICS INSTITUTE

7 Subjective evaluation: Intolerable glare DGP = (Intolerable glare ) Fully open DGI = (Perceptible glare) UGR = (Disturbing glare) VCP = (Intolerable glare ) CGI = (Intolerable glare) Subjective evaluation: Disturbing glare Roller blinds DGP = (Imperceptible glare) DGI = (Imperceptible glare) UGR = (Perceptible glare) VCP = (Perceptible glare) CGI = (Perceptible glare) Subjective evaluation: Disturbing glare Venetian blinds DGP = (Imperceptible glare) DGI = (Perceptible glare) UGR = (Disturbing glare) VCP = (Intolerable glare) CGI = (Intolerable glare) Figure 4. Typing task scenes processed in Evalglare Three examples were chosen for the writing task condition glare scenes (Figure 5). With direct sunlight in the field of view, the fully open blind scene has no matching glare metric evaluation to intolerable glare. Again, the glare metrics provide a wide range of evaluations from imperceptible to disturbing glare, which is still underestimated compared to human subject s evaluation. Again, DGI makes extremely underestimates this glare scene as imperceptible glare. The roller blind scene was correctly evaluated by UGR only. Similar inconsistent evaluation issue is shown in this example. The venetian blind scene was correctly evaluated by DGP as it reports perceptible glare. The other metrics overestimated the scene as disturbing or intolerable glare. Subjective evaluation: Intolerable glare Fully open DGP = (Perceptible glare) DGI = (Imperceptible glare) UGR = (Perceptible glare) VCP = (Disturbing glare) CGI = (Disturbing glare) Subjective evaluation: Perceptible glare Roller blinds DGP = (Imperceptible glare) DGI = (Imperceptible glare) UGR = (Perceptible glare) VCP = (Intolerable glare) CGI = (Disturbing glare) 2016 WORLD CONGRESS 7

8 Subjective evaluation: Perceptible glare Venetian blinds DGP = (Perceptible glare) DGI = (Disturbing glare) UGR = (Intolerable glare) VCP = 4.35 (Intolerable glare) CGI = (Intolerable glare) Figure 5. Writing task scenes processed in Evalglare EXPLANATION The rest of 450 captured glare scenes show a similar pattern as these examples. At most one of the five metrics correctly matches to the subject s evaluation for each scene. This evaluation comparison study supports the findings that the five glare metrics have vast inconsistency issues. Furthermore, it indicates that the existing glare metrics have significant inaccuracies in their evaluations of glare scenes. Higher accuracy and consistency were observed in extreme glare conditions such as intolerable or imperceptible glare scenes. Based on the analysis of 450 glare scenes, DGP shows the highest accuracy rate at 54.0% and DGI shows the second highest rate at 42.2% of matching evaluation. UGR, VCP, and CGI shows lower evaluation accuracy as 37.8%, 35.8%, and 27.8%. Similar to the findings from the study performed by Jakubiec and Reinhart, the result shows that DGP is the most reliable daylight glare evaluation metric among the currently available metrics even though its accuracy rate is not satisfactory. It is also found that VCP and CGI are inappropriate to analyze discomfort glare caused by natural light as they mostly overestimate glare levels. This is not surprising as they were originally developed for glare issues caused by electrical lighting. As previously noted, DGI and UGR do not work with a glare scene with direct sunlight. This significantly limits their capability and reliability as successful daylighting design cannot be achieved without accurately evaluating excessive sunlight penetrations. Further analysis is required to verify the findings. CONCLUSION AND FUTURE WORK More than 450 daylight glare scenes were analyzed by the five existing glare metrics and were compared to subjective surveys in statistical analysis software. The analysis results indicate that DGP shows the best evaluation accuracy among the five metrics when the subjective evaluations are used as the baseline for determining accuracy. Unfortunately, the evaluation accuracy of all the existing glare metrics is too low to be trusted as the highest accuracy level is slightly over 50%. The accuracy levels of VCP and CGI shows that these glare metrics are not appropriate for daylight glare analysis. DGI and UGR show slightly higher accuracy rates than VCP and CGI but they are not capable of analyzing glare scenes with direct sunlight. This makes all four glare metrics except DGP inappropriate for daylight glare analysis. Besides the accuracy issue, the inconsistent evaluation issue found from the previous study arose again throughout this analysis of human subject data. Furthermore, each glare index shows higher or lower evaluation accuracy depending on the glare levels and blind conditions of a scene, indicating inter-index inconsistency. The findings might be applicable only to a closed office with large glazing in a dominantly sunny sky condition. Further study is required to verify the findings in an open office setting or a closed office with small glazing. Further distance from windows to human subject or a field of view parallel to windows should be also tested to verify the findings. Based on the precedents and current study, it is recommended to use DGP for daylight glare analysis. It is also important to thoroughly check luminance distributions and levels in a glare scene so that inaccurate evaluations can be avoided. Overall, these findings can help users of the existing glare metrics to find a better understanding of what they can expect from each of the metrics. And, it will help developing a better daylighting design process which would create visually comfortable daylit environments in buildings. REFERENCES Konis, Kyle. Evaluating daylighting effectiveness and occupant visual comfort in a side-lit open-plan office buildings in San Francisco, California. Building and Environment 59 (2013): Hirning, Michael, Isoardi, Gillian, Coyne S, Cowling, Ian, and Garcia-Hansen, Veronica. Post occupancy evaluation relating to discomfort glare: A study of green buildings in Brisbane. Building and Environment 59 (2013): Hirning, Michael, Isoardi, Gillian, and Cowling, Ian. Discomfort glare in open plan green buildings. Energy and Buildings 70 (2014): Wienold, Jan and Christoffersen, Jens. Towards a New Daylight Glare Rating. Lux Europa. Berlin, 2005 Bellia, Laura, Arcangelo Cesarano, Giuseppe F. Iuliano, and Gennaro Spada. Daylight glare: a review of discomfort indexes., visual quality and energy efficiency in indoor lighting: Today for tomorrow, International Workshop, Rome, Italy, 31 March 31, Isoardi, Gillian, Veronica Garcia-Hansen, and Michael Hirning. Evaluation of the Luminous Environment in Open-plan Offices with Skylights. The World Renewable Energy Forum, SOLAR 2012, Denver, Colorado, May 13-17, FACADE TECTONICS INSTITUTE

9 Wienold, Jan and Christoffersen, Jens. Evaluation methods and development of a new glare prediction model for daylight environments with the use of CCD cameras. Energy and Buildings (2006); Volume 38, Issue 7 Inanici, Mehlika and Galvin, Jim. Evaluation of High Dynamic Range Photography as a Luminance Mapping Technique. Lawrence Berkeley National Laboratory LBNL (accessed April 3, 2010) Inanici, Mehlika. Per-Pixel Lighting Data Analysis. Lawrence Berkeley National Laboratory NBNL (accessed April 3, 2010). Suk, Jae Yong and Schiler, Marc, Investigation of Evalglare Software, Daylight Glare Probability and High Dynamic Range Imaging for Daylight Glare Analysis. Lighting Research and Technology (2012) Jakubiec, Alstan and Reinhart, Christoph. The Use of Glare Metrics in the Design of Daylit Spaces: Recommendations for Practice. 9th International Radiance Workshop, (accessed April 1, 2011) Luckiesh, Matthew and Guth, Sylvester K. Brightness in Visual Field at Borderline between Comfort and Discomfort (BCD). Journal of Illuminating Engineering Society November (1949): Hopkinson, Ralph G. Evaluation of glare. Illuminating Engineering, (1957): Ngai, Peter and Boyce, Peter. The Effect of Overhead Glare on Visual Discomfort. Journal of Illuminating Engineering Society November (2000): Velds, Martine. User Acceptance Studies to Evaluate Discomfort glare in Daylit Room. Solar Energy 73, No. 2 (2002): Osterhaus, Werner. Discomfort glare assessment and prevention for daylight applications in office environments. Solar Energy 79-2 (2005): Linney, Andrew Scott. Maximum Luminances and Luminance Ratios and their Impact on Users Discomfort Glare Perception and Productivity in Daylit Offices. Master thesis, Victoria University of Wellington Wymelenberg, Kevin Van Den. Evaluating Human Visual Preference and Performance in an Office Environment Using Luminance-based Metrics. PhD diss., University of Washington, Suk, Jae Yong, Schiler, Marc, and Kensek, Karen. Development of new daylight glare analysis methodology using absolute glare factor and relative glare factor. Energy and Buildings 64 (2013): WORLD CONGRESS 9