Bruce Kinzey:

Welcome, everyone. This is Bruce Kinzey with the Pacific Northwest National Laboratory. Welcome to today's webinar, A Technical Discussion of DOE's Sky Glow Study, Modeling Methods, and Key Variables, brought to you by the U.S. Department of Energy's Solid-State Lighting program. This is the second in a two-part series discussing our recently released investigation of LED streetlighting's effect on sky glow. Last week's webinar provided a high-level overview of the effort and its findings. And if you missed it, the slides are actually posted on the Solid-State Lighting Program website today. The video is not there yet, but it should be posted by the end of next week

Today's webinar is going to provide a deep dive into the modeling effort, and will compare the influence of individual variables. Our presenter today is Tess Perrin. Tess is a lighting scientist with the Advanced Lighting Team at the Pacific Northwest National Lab. She's currently focused on the evaluation of emerging technologies for a variety of lighting applications, ranging from sky glow to flicker. She's an emerging professional in the Illuminating Engineering Society of North America and a member of the International Dark-Sky Association Technical Committee. Tess?

Tess Perrin:

Good morning, and thank-you for tuning in to part 2 of the Solid-State Lighting Program Sky Glow Webinar. So today's presentation is going to offer a technical discussion of our sky glow study, the various modelling methods used, and key variables. So as Bruce started to share, this webinar is by no means comprehensive. But it is an attempt to dive a little deeper into the weeds. And I'll just say upfront, for more information, the full report, and other papers that I'll cite throughout this presentation will provide more of a wealth of knowledge after the presentation.

So as Bruce showed in his presentation last week, blue light is receiving significant attention of late, primarily due to its perceived role in a variety things, which include visual acuity, light scattering, circadian rhythm entrainment or disruption, as well as environmental influence. So alongside these impacts, blue-rich LED sources are increasingly more prevalent due to their significantly reduced energy use, along with their improved color rendition properties, their high precision of optical control, and potential use for control strategies.

So we essentially have impacts of blue-rich light to weigh against an increase in adoption. So most specific to this particular sky glow study, we've asked ourselves the question, how does the blue content of these lighting sources impact night sky visibility? Many current impacts take an over simplified approach by I recall an example of substituting an SPD of an LED source to that of a high-pressure sodium source, while holding all other factors constant. But this doesn't really reflect the reality of actual installations that we've seen in the US.

So our study has attempted to quantify the actual contributions to sky glow by representing real-world scenarios and then isolating variables to look at the relative impact of specific design characteristics. And we also realize that this study has focused solely on streetlighting, which is only one of the many applications that contribute to light at night.

So just to start, it's why we recognize in fields like weather forecasting, that there will be inaccuracies given the atmosphere represents essentially infinite combinations of rather complex elements. So significant simplifications are required in any really good modelling effort. And even then, models that are used for these purposes are necessarily complex from often a mathematical standpoint, and they are usually very data-intensive. But with that being said, models offer us the opportunity to represent relative A to Z comparisons, where many of the complexities are held constant. So in an ideal world, this allows for a better grasp of potential impacts of change.

So Miroslav Kocifaj developed the Model Sky Glow Simulator back in 2007, so a decade ago. An example of the best graphical user interface is shown on the right, as is the geometrical setup of the model on the left. So this model is an improvement over point source approaches because it's designed for realistically shaped emission surfaces. And when we talk about emission surfaces, we're really talking about cities or other defined areas.

So at first stepping through the model, the shape and the size of a city is defined through its geographic coordinates. So Sky Glow Simulator models the light emitted by ground sources — in our case, it's streetlights — based on the angular distribution of light. So we're going to step through a schematic here, hopefully to boil down the model in a more simplistic way. So to characterize the emitted light, the number of individual light sources is specified, along with their corresponding spectral and distribution characteristics. The model then calculates the total ground emissions based on luminaire wattage or light output and sometimes even the city population, and this is based on lumens per person. And this has historically been a common way to estimate the total emissions from an area.

Essentially, the model then divides ground-based light sources into many horizontal pixels. So if you look at the drawing, the red cube is an example of one of these pixels. And then you'll see that gradiated red area is just the ground source as an example. So these horizontal pixels are combined then with vertically stacked pixels, illustrated by the grey cubes. And these make up the atmosphere. So the atmosphere is then characterized by specific properties, like altitude dependent scattering and absorption properties. The green cube now is an example of just one atmospheric pixel, the specific scattering properties. And an example of these scattering properties is delineated by the green arrows in the upper left illustration.

So each pixel is characterized by the position with respect to the observer. So the observer is just a blue dot here. And this is as a function of the zenith angle, shown by that orange arrow, and that azimuth angle shown by the purple arrow. And we'll talk about these more in the next slide. But essentially the zenith arrow angle represents an observer looking directly above and then tilting their head down to the horizon. Whereas the azimuth angle represents an observer standing in place and then spinning in a circle. So those are the two angles that we're looking at.

So this is then completed for all zenith and azimuth combinations. And it eventually creates a polar plot. And the exercise we just went through is just one combination just to illustrate the calculation. Essentially, it's how the scattered light from that one red pixel adds to what the observer sees. And this is all part of the sum of all such pixels.

So the resulting angular distribution of the sky glow is either unweighted or scotopically weighted, depending on what specific impact you are investigating. And then, as I just shared, it's represented by a polar plot. The total intensity is calculated for zenith angle. And you'll see here that's the arrow that is measured from the center to the margin of the polar plot and then the azimuth angle, which is the angle along the circle.

So some of the outputs of the model are diffuse irradiance or scotopic illuminance on the horizontal surface. Another output is the ratio of zenith radiance, or luminance to diffuse radiance, or illuminance.

So just to break the down a little further, the diffuse horizontal irradiance, or illuminance value, is essentially associated with the overall flux directed to the observer, and it's really independent of the direction at which the light beam is propagated, because you're looking at the whole horizontal surface. So this value doesn't really continue to reflect information on dominant light change.

And then the ratio of the zenith luminance relative to the horizontal irradiance or illuminance — this value can reveal how much the sky changes with distance or angle. But this study, for simplicity's sake, really exclusively analyzes the diffuse irradiance and illuminance values on the horizontal surface.

So this model has been verified multiple times and in a variety of different manners. The model was compared against a more complex one, specifically running all these model illumina, which itself has undergone its own verification.

Sensitivity studies were also conducted through a variety of model runs to test whether or not the model predicted long-run behaviors. And then the calculated values were also compared against actual measurements. So this process in particular looked at the impacts, like we did, of changing streetlighting on sky glow for different observer distances. And this happened at two observatories in Slovakia.

These measurements were taken back in August of 2009, and they really were the last step in verifying and validating the model performance. So you'll see the measurements on the far left graph. So that's the polar plot of the actual measurements there.

And then the reconstructed data is in the middle. And then on the right, this is where all of the measurements were taken. So just for reference, the luminance values were all relative to zenith, which were assigned a value of 1. So blue indicates a value that's three times zenith. Green was 12 times, and red was 40 times. The main point of that was just to see that the left and the middle graph are very similar.

So now, focusing right on our experiments, the design of our experiment was really based on predominant contribution factors to sky glow that were included in the model specifically. So at the fixture level, these include percent uplight, the SPD of the source, and then the lumen output per fixture.

So also included are the city specifications, which were mainly the dimensions and the geographic location. We also looked at the location of the observer and then a variety of atmospheric conditions. And then really bridging the gap between the fixture level and then looking at either city variables or the greater environment at large is the emission function, which we'll go into shortly.

So all together, these factors amounted to over 215,000 runs. So just to bring this into perspective, each full-spectrum run took about 15 to 20 minutes on a Windows desktop computer. So just those 4,000 trials alone would have taken about 40 days.

So fortunately, the PNNL institutional computing program has a Windows HP cluster with about 400 cores. So the full set of jobs took about 5,000 core hours, which was less than 24 hours in actual calendar time. So we selected three cities and assigned for city area, the radius, the population, and the number of fixtures.

And we modeled these cities essentially based on real cities in the US, and they intentionally scanned a small city all the way up to a large metropolis. And they had a large range in lighting density. Now, one of the key decisions for that — instead of modeling the exact shape of the city, which you can do in this model, we actually just need the circle. So that variation in city shape was not a factor in this study, so that study was based on the radius, and then that circle was based on the radius of the city.

We picked two observer locations for each city. So the first was located at the perimeter of the city. And then the second was always 40 kilometers from the center of the city, about 25 miles. And for all cities, this was outside the city. And this, for reference, among the modeling community is believed that most sky glow models are accurate up to around 50 kilometers for that light-emitting area. So that's largely why 40 kilometers was selected.

So various atmospheric conditions can be analyzed in the model ranging from cloudless, to overcast, to cloudy skies. The probability that a photon will be scattered into the angle of observation is really dependent on the optical properties of the atmosphere, which is characterized by a scattering function.

Aerosol particles and the air molecules are the most important elements of the atmosphere responsible for the scattering and attenuation of that visible radiation. So the overall transmission function reflects the optical properties be it absorption, extinction, and scattering of that molecular aerosol atmosphere.

So in this model, the molecular scattering is simulated in accordance with Rayleigh theory. And then the aerosol scattering is approximated by a phase function.

So just for reference, aerosols are really one of the most unstable atmospheric constituents, because their optical properties depend on the critical type be it the reflected index or the chemical composition, and then they also depend on the size.

So just as an example, black soot in the air absorbs much of any incident light and thus it creates a darker sky glow. But then when you think about salt aerosols that we find in coastal regions, these tend to be diffuse reflectors and make the sky brighter, and then thus reduce the contrast of the stars.

So the characteristics of aerosol particles really collectively influence the turbidity of the atmosphere which then impacts the behavior of the light that's traveling through that atmosphere. So in order to kind of boil this down, we needed a set of characteristics in order to allow for accurate representation of atmospheric conditions in the US. So these included the asymmetry parameter, single-scattering albedo, and optical thickness.

So we have selected five atmospheric conditions, four of which are cloudless with varying degrees of turbidity, and this is really stipulated by different values for aerosol, optical thickness, as well as the angstrom exponent. And we picked one cloudy condition.

So the range in aerosol content was based on ground-based, remote sensing data. And this was provided by the Aerosol Robotics Network, which is the program that was established by NASA and PHOTONS. And then, again, these conditions are only loosely based on US cities, because we really wanted to represent a realistic range of temperature.

So returning to the inner workings of the model, this angles of emission function is really a key property for evaluating a specific city's sky glow as different installations, urban designs, and other geographies dictate the need for multiple functions. The light-emitting function is just simply how to write properties into the atmosphere.

So back in the '80s, Roy Garstang found an approximate formula for the angular behavior of radiation produced by surface sources. And this has really become the city emission function that most sky glow models use, including this one. So this function separates the distribution as a percent downlight and uplight. And uplight here is just defined as any light emitted above 90 degrees.

So just breaking this down further, the downlight portion, the yellow curve on the left, is reflected light, and that's based on a Lambertian distribution. And it's defined as percent reflected. So for all of our runs, we chose 15%. The percent uplight, shown by the red curve, is proportional to the angle measured from the zenith to the fourth power.

And then the composite distribution is represented by the orange curve. So we selected three values for percent uplight. The 0% represents roughly what used to be known as full cut off fixtures. 2% and 5% represent a dropped lens cobrahead giving around 2% to 4% uplight.

And then 10% is about as close to a good quality acorn top light post. And we also selected 10% because it allows for class model comparisons since most of the other sky glow models in use today with core values at 10% uplight.

So now stepping through that function in the bottom right a little bit more, if you can just turn your attention there, if there is uplight, in other words lowercase q is greater than 0, then the right side of our famous equation, the one that's underlined in red, is employed. And the emission function is the sum of both the reflected and the direct uplight quantities.

So in this case, large amounts of optical radiation are emitted from those smaller elevation angles, which are essentially within about 10 degrees above horizontal. And this is often where the greatest scattering occurs. But now, if lower case q is equal to 0 — in other words, there is no uplight — then the reflected light is emitted. It is emitted only according to the cosine function. And this is represented by the red line in the graph.

So cosine light fixtures emit most of the radiation in an upward direction. So a nearby observer only sees a relatively small amount of backscattered light. And just getting into the weeds a little bit, as the distance between the observer and the source grows, then the amount of the light seen by the observer grows due to the increase in side scatter.

So at the time of this research, the only available city emission function in the model was Garstang. But it's worth putting out that recent studies have actually questioned the validity in using Garstang's original emission function, and they have started to propose other modified functions.

And really the shortcomings that are being claimed with Garstang's function are essentially centered around if the function overestimates emissions at low-elevation angles. So then these trajectories are often blocked be that by trees or buildings.

We selected 11 SPDs. And these included incumbent technologies like high-pressure sodium and metal halides along with LEDs of various CCTs. We also included an equal energy SPD, by the dot there. And we included this in order to look at the resulting sky glow spectrum when emissions was equivalent at all wavelengths.

So just for reference, the scotopic to photopic ratios relative to high-pressure sodium for the LEDs ranged from 0.7 to 3.1. And then just listed here, because this metric is often referenced as a means to evaluate light sources for sky glow potential in some approaches to streetlighting design. And generally, light sources with more emission around 507 nanometers, which was the peak of the human scotopic lumen efficiency function, will have higher SP ratios.

So just going through some of the last variables that we considered — the SPDs were either left unweighted. So the resulting sky glow was reported in irradiance. Or they were scotopically weighted. So the results were recorded in scotopic illuminance. So as presented earlier, other measures were also output in the model, including the radius or luminance at specific observation angles. This is how the polar plots are created.

The model currently does not export the SPD of the sky glow. So in order to get around that, we actually conduct our trials using the full SPD from 380 to 780 nanometers. So then we also broke them down into 5 nanometer bands So these increments are run for each condition. We were then able to piece them together to form an SPD of the resulting sky glow.

And if you look at that, the diagram above, you'll see that the bottom of the diagram is not as tall as the top part simply because the incremented runs were only completed for that unweighted output, as they could then be adjusted further to reflect the topic impact.

So with those two values for lumen output, and these were 1,000 lumens per fixture and 500 lumens per fixture. And given we're only looking at relative comparisons in this study, the actual values really aren't as important. Their ratios are. And our intention was just simply to allow for one scenario to have half the lumen output of another.

So now, we're going to spend the rest of the presentation going even deeper into these weeds as we looked at the individual impacts of these variables. And so we're going to step through this figure, just to give a general outline of the figure. The figure separates the relative impact of variables independent of the lighting atmosphere, or the lighting system — mainly the atmospheric conditions from those that can be altered through the luminaire selection. And these are really SPD, percent uplight, and light output.

So in this figure, atmospheric conditions are all relative to the clearest condition that we modeled. SPD is relative to high-pressure sodium. The percent uplight values represent a change from 2% to 0% uplight. And the light output values essentially represent a range from 100% to 50% output.

So comparing the relative sizes of these corresponding columns can provide a useful indication of the different levels of influence between variables, if you figure that there is a difference in the y-axis scales where the ones on the left goes from 0% to 18%. And the ones on the right go from 0% to 4%.

And just breaking this down a little further, the solid bars represent unweighted results, while the hatched bars are scotopically weighted. And then you'll see that specific locations of individual LED SPDs are displayed. And this is simply to facilitate comparisons between them. And then other SPDs, like low-pressure sodium and metal halides, are included to delineate what SPDs had the minimum and maximum. performance.

So let's start with atmospheric conditions. Atmospheric conditions really play a significant role in both the level and the distribution of sky glow that was talked about earlier. And they have the potential to significantly increase the impact on sky glow from your observers.

And just as a reminder, the near observer was located at the perimeter of the city in the study. So in this graph for the near observer only, the relative impact of select atmospheric conditions are compared to this clearest condition. And this is indicated by the red line.

And these are all based on your percent uplight value be it 0%, 2%, or 10%. And then we use the data from each city and for five select SPDs. And you can see how they are all labeled on the x-axis. And then just in this graph, only the scotopically weighted results are shown just to narrow the datasets.

So climate conditions really have a large potential impact on sky glow within the defined city area. Clouds essentially reflect, transmit, or absorb all wavelengths of light fairly evenly. And the resulting diffusion of the reflected light then contributes to localized sky glow.

So for the near observer, the brightness of the night environment noticeably increases under cloudy conditions. The atmosphere also affects the impacts of uplight from the luminaire. So for a near observer under cloudless conditions, decreasing uplight slightly decreases sky glow because of the associated increase in scatter. So, then, under cloudy conditions, decreased uplight from the luminaire actually increases the visible sky glow.

So this might seem like a contradiction at first, but this occurs for two primary reasons. Under cloudy conditions, the downward portion of the light from the luminaire essentially bounces multiple times between the ground and the overhead clouds. So it's effectively trapped. And then also, most of the increased uplight is emitted at low elevation angle again with that yellow curve, and the diagram, and thus it does not reach the near observer.

So now, let's look at the distant observer. And just as a reminder, this was always 40 kilometers from the center of city. And just note the different scale on this y-axis where this goes up to about 1.2. Whereas the previous graph for the near observer went all the way up to 18.

So as we discussed in the previous slide, since clouds act as immediate reflectors, they really work to confine light locally and thus effectively reduce the light observed at some distance from the city. Under cloudy conditions, distant observers only receive light that is scattered either through the cloud layer or directly emitted at angles only slightly above the horizontal because it needs to skirt the cloud layer.

A decrease in uplight reduces sky glow for the distant observer. And really, overall, this figure illustrates the significant impacts of sky glow from more turbulent atmospheres as well as those with cloud cover as seen in the bottom half of the graph. It's worth noting that clearer skies — if you look at the orange points clustered around the red baseline at the top of the graph, these clearer skies display almost none of the effects that I just discussed. With the further observation distances, it's really difficult to project what the difference in aerosol optical thickness will do as the optical properties depend on not only the size of the particles, but the type.

So this can just start to explain the fluctuating behavior of the clearer conditions around one another. So we won't dwell on these graphs. They're just included here to show all of the variables instead of the subset that we discussed in the previous two slides.

But one more thing that's worth noting when we're zoning in on the atmosphere is just that the larger the city, the more difficult it is to forget the scattering behavior in the atmosphere. The spread of light beams becomes more complicated. And there is a greater chance that beams from opposite sides of the city relative to the observer will travel at lower elevation angles and along the optical path.

So these graphs show the relative impacts for SPDs — percent uplight and lumen output. So just sticking with lumen output for a moment, while nothing below a 50% reduction with models, as Bruce noted last week, sky glow directly scales up or down with light output for all observer positions.

So this suggests that, for example, during certain hours when lower light levels might be acceptable, depending on local policy or other criteria, a lighting system can really be dimmed to any level, including off, for further decreased sky glow contribution.

So these graphs now show the impact of reducing percent uplight under cloudless and cloudy conditions. From left to right, each bar shows the minimum — the first quartile, the median, the third quartile, and then the maximum values for all combinations of modeled variables.

And the colored area within each of the bars represents the data between the first and third quartiles, the 25th and 75th percentiles. These top two graphs and the blue bars are for the near observer, while the bottom two graphs and the yellow-orange bars represent the distant observer.

Now digging down right even further, the solid bars show the impact of changing from 2% to 0% uplight. The vertically hatched bars show the impact of changing from 5% to 0% uplight. And then the diagonally hatched bars show the impact from 10% to 0% uplight.

So for example, a value of 0.5 indicates that reducing uplight to 0% from the corresponding baseline, which again is shown by that red line, would result in half the sky glow. So these charts show that reducing uplight decreases sky glow under all conditions except for the near observer under cloudy skies.

This reduction in sky glow results from eliminating uplight. This is really particularly pronounced for the distant observer. So as you can see at the far left of the bottom part, sky glow for the distant observer has been reduced by at least 95%, simply by switching from 2% to 0% uplight. So really, the take-home message here is that eliminating uplight has by far the single largest influence of any variable to reduce sky glow for a distant observer.

So now turning to SPD, these graphs on the left show the SPDs on the right versus that were included in the study. So the top chart displays the SPDs at equal lumen output. Then the middle chart zooms in better just to reveal the variation among really all of them except the low-pressure sodium. And then the bottom chart just shows the LEDs. So you can see that significant variation among spectral content is really evident in all of these charts.

Now, these charts on the right display the data in the same manner as the percent uplight graph. So the graphs show the range of impact for SPD relative to high-pressure sodium indicated by the vertical red lines again at the baseline. And this is for all cities and all atmospheric conditions. So essentially, you'll see that all SPDs increase sky glow compared to high-pressure sodium except for the low-pressure sodium and the PC amber.

The greater the amount of short wavelength content, the greater the range in impact, as you'll see by those specific choices that the arrows are pointing to. And then scotopic lighting, as you see in the far right graph, this really further emphasizes these impacts due to the human eyes increased sensitivity to wavelengths near the peak of this scotopic sensitivity curve. So you'll see that those bars are moved more to the right in the graph.

So now, this figure on the left shows the resulting sky glow spectra for the equal energy source. And just as a reminder that the SPD with equal energy [INAUDIBLE] at all wavelengths. And these spectra really represent the relative power of each wavelength in the original source that would be present in the sky glow for the near observer.

And these graphs were all made from those incremented runs of equal energy source.

So you can notice that except for the cloudy conditions which is in that purplish color, very short wavelengths dominate in the sky glow spectrum. The atmosphere really more readily scatters short wavelengths. So the color of the sky glow under cloudless conditions is likely to be bluer than the color of light emitted from the streetlighting. And this is just natural sunlight where the sky, if you're blue, compares to the light coming from the sun.

Under cloudy conditions, however, wavelengths are scattered more evenly. So there's less spectral variation. And the sky glow spectrum thus will appear closer to that of the original source. You can see that those curves are more flatlined.

So now, looking at the figure on the right, we see the sky glow spectra for the distant observer. So here, short wavelengths affect sky glow less for the distant observer, because really shorter wavelengths are more likely to be scattered or reduced simply because they are traveling longer distances.

And then this effect becomes even more pronounced when there is greater turbidity and clouds in the atmosphere. So now moving from individual impacts of variables back to the overall picture, as Bruce showed last week, these charts really provide an overall summary of our results.

So the baseline again is represented by the dashed red line at a normalized value of 1. And the last set of three charts is for the near observer, while the right is for the distant observer. And then within each set of charts, there are really two graphs.

The left graph shows the unweighted results. While the right graph shows the scotopically weighted results. And then taking that further, you have the clear conditions, which are in white, and then you have the cloudy condition, which is gradiated in grey.

So the top charts really show the isolated effects of replacing the baseline high-pressure sodium SPD with the various other LEDs that were modeled. And this is with no other modification, just the SPD was changed. The middle charts then add the effects of reducing luminaire lumen output by half compared to the baseline high-pressure sodium. And this is a typical result for conversions here in the US.

And then the bottom charts further add the impact of eliminating uplight from the luminaire. So this is seen as the the typical high-pressure sodium baseline value of 2%. So these bottom charts really represent typical conversions, an incumbent high-pressure sodium cobrahead product with 2% uplight. That is then replaced by each of the LED products listed at half the light output and no uplight.

So kind of putting all of the discussion on those individual variables together, you can see that for the near observer, short wavelength content does contribute toward increased sky glow. But CCT is not always a reliable predictor of impacts.

And this is especially important given much of the current public discussion reflects this comparison of SPD in isolation of other factors. On an unrated basis, typical conversions to LED that include a 50% reduction in lumen output and eliminate uplight will significantly reduce sky glow compared to a high-pressure sodium incumbent delivering 2% uplight.

But scotopically weighting the results does reduce the number of products that can make this claim. For now for the distant observer, there is greater variability compared to the near observer simply due to the impact of different atmospheric effects and more length to travel.

The transition to no uplight essentially removes the contribution of streetlighting to sky glow for the distant observer by at least 95%. And this is for all products and atmospheric conditions. Similarly, cities can use this knowledge of the combined effect to balance streetlighting conversions and new streetlighting installations to either maintain current levels of sky glow or to reduce the total sky glow over time.

So as you're looking for more, you're welcome to see our full report. The link is on the left. And then the various papers that really document Miro’s model. And all of those can be found on the Unified Sky Luminance Model website. And now, we'll open up the lines for questions.

Bruce Kinzey:

Thanks, Tess. That was excellent, really. So we have only received a very few number of questions. I'm going to take this first one to give Tess a water break here — really excellent job — and maybe a few more will come in while we're discussing this.

"So have you evaluated the effects of sky glow using other measures such as its impact on human health, bird migration, or plant growth?" No is the short answer. We stayed away from these possible downstream impacts, because they involve a host of new additional assumptions, some of which do not yet have even their own scientific consensus. So first there are things like what's the age of the individual and their general health level? And how much exposure to sunlight did they receive over the previous 24 hours and so forth? Of course, these are the gross assumptions we would have to make.

And they are also more like individually based variation. Like what particular species are we talking about? Often these things don't have the same impact among different species. So all of this is stepping into an entirely different realm of areas that we don't really have expertise in, so we have drawn our boundary at this analysis at this quantitative assessment of the change to the amount and characteristics of light in the sky. Our numbers can then be used for looking at these other downstream kinds of impacts.

But because it's fraught with hazards, we didn't try to go there ourselves. OK, so the second question is — I'll just start this one off, and then Tess, you can take this after you can add to this if necessary. So this reflects a great deal of technical effort. But I'm wondering if you have suggestions for how we might use the results in our day-to-day work. How do we practically apply this?

The first thing I want to say is, we're engineers and scientists. And you're asking us for something practical? Well, we'll do our best here. Actually, after doing all these runs, we decided there should be some way to take advantage of all the data we had developed. So we are in the process of developing a tool that's probably going to be in a spreadsheet format that will accept any SPD and enable the user to compare the contribution of sky glow of one product versus another. The purpose of this will be to enable someone to add sky glow to their selection criteria when they're designing an outdoor lighting system.

And the accuracy is subject to some limitations relating to the assumptions that we've used to produce these results. And here's where Tess I think can really step in. But these should still give us a good ability to conduct some pairwise comparisons. So Tess, is there anything you want to add to that?

Tess Perrin:

Sure, I'd just add that in the slide that I showed toward the end of the presentation — the different sky glow SPDs that are based on the equal energy spectrum — what we've essentially done is we've gone through all of those SPDs for all the different runs.

And then what we've done is — they essentially work as scattering functions, which can then be applied to different SPDs. So we're still limited within kind of the universe of the runs that we used in this effort.

So a user would only be able to select from, for example, like 0%, 2%, 5%, and 10% uplight. But they'll be able to choose all of the different variables that we ran — all those different conditions — and then actually put in their own SPDs to then compare the sky glow spectrum for different scenarios that they're interested in modeling.

Bruce Kinzey:

OK, thanks, Tess. And I'll read this one, and you can take it. Or I'm happy to take this one too. "Is there a correlation between the sky glow information and the AMA report on the impact of streetlights on humans?"

So I can respond to that. So of course, these two issues — the health issues have been raised by AMA and sky glow — these are both related to light at night. There's a lot of the same kind of underlying issues that are brought up — the blue light content of these and what effects it might have. I kind of referred to this in the first question above. These are — these health impacts and so on are things that we did not try to get into in this particular assessment, because we're not experts in that area is the main thing.

But just one other thing I would say here is if you think about the — we're talking about totally different intensities. If you're standing underneath a streetlight, or you've got a streetlight coming in your window, the amount of light coming in that window compared to the amount coming in the window from a change in sky glow are different orders of magnitude.

And so some of the same kinds of issues would apply. But In general, the different intensities really — really really change things. Is there anything else you want to add to that, Tess?

Tess Perrin:

Yeah, Bruce, I would just add that perhaps one of the greatest considerations that came out of this study was just opportunity for all of us to realize that streetlighting is only one part of the puzzle. So even when you're thinking about health impacts, there are so many different sources of light. So it's really looking at the whole picture and not just — zoom in on one variable that seems to be a really important next step for all of us.

Bruce Kinzey:

OK, thanks. On the last slide of charts, it was shown that some LED products increased sky glow. And others — and actually it says increase — and others decreased it. Could you comment on this a bit more?

Tess Perrin:

Sure. So this last set of charts — Bruce actually stepped through this really extensively in his presentation last week. I just attempted to bring everything back to center by going back to how all of these variables coalesced together. So if you'd like more information, I'd actually suggest that you look at his presentation or even the summary of our reports.

But essentially, whether or not an LED product increases sky glow or decreases sky glow, you can see that that changed depending on if you're looking at just changes in SPDs, if you're then adding a change in lumen output or if you're then taking percent uplight into account.

So what we really found, when you're looking at what we qualified as the fulcrum area, and that is a change in uplight output and the SPD, we saw that when the spectrum is unweighted, then all LED products decrease sky glow.

When you then start to scotopically weight it, there is a breaking point where some LED products increase it and others decrease it. And to find out the specifics on the CCTs where that happens — that's all detailed within the report.

Bruce Kinzey:

Yeah, and I'll add that we've tried to de-emphasize particular SPDs. This something else I also talked about last week is that there is far from being a monolithic product you actually have all kinds of variation in spectral content of LEDs, even at the same color temperature. That color temperature is basically just sort of a weighted average that comes as a result of adding up all the different spectra at a particular source and multiplied by their actual power being emitted at each of those wavelengths.

And so there is any number of combinations of those that can produce a given CCT. And so just to say all 3,000 CCT products have these characteristics is inaccurate, because actually, CCT only describes the appearance of a color source, and the issues that are being raised are not caused by colors or by the appearance, but they're caused by the actual spectral content. And it's not a very good measure of that actual spectral content.

Tess Perrin:

Can I add one more thing?

Bruce Kinzey:

Yes, please, go.

Tess Perrin:

So if you refer back to the slide that was just looking at SPDs, where I showed all of the SPDs branched on the left and then basically the histograms for the performance relative to high-pressure sodium. You'll see that the bars for each of the SPD — those are actually done in terms of increasing CCT for the LED sources.

And you'll see that the range between the minimum and the maximum for most of those bars — they all overlap. So it's just another illustration of how — depending on other atmospheric variables or uplight, CCT can essentially be manipulated to be more or less, depending on the other variables.

Bruce Kinzey:

OK, so that completes all of the questions that we have received. People are either spellbound, or they're confused, or you've answered all their questions. Thanks again, Tess, that was an excellent presentation. And I want to thank everybody again for participating in today's webinar. That's brought to you by U.S. Department of Energy Solid-State Lighting program. Thanks again. You may all now disconnect.