I am writing to seek your assistance regarding a simulation issue we encountered while using Daysim. We have been using Daysim to simulate solar irradiation on building surfaces for a small case study in Singapore based on a weather file collected from the local climate station. However, we found that the simulation results for facades are confusing.
Specifically, we have generated figures for each month, at a fixed time, to show how the facades’ irradiation changes based on global horizontal irradiation (GHI). For a given month and time of day, there is little difference in the angle between sun and buildings, so we would expect the larger the GHI, the larger solar irradiation. However, we have found that for facades, the curve initially increases and then decreases, particularly at times of day when there is no direct irradiation on facades, such as the East facades in the afternoon and West facades in the morning. This occurs for all facades and buildings we have simulated in Singapore.
We are writing to inquire about the cause of this phenomenon and how it works in Daysim. We would greatly appreciate it if you could provide us with any insights or explanations regarding this issue.
Welcome to the Radiance forum. I’m just taking a wild guess here, but how are you accounting for weather? I assume that when GHI is lower, the sky is overcast, so it is more uniform, and when GHI is higher, the sky is clear, so the lighting is more directional. This would explain why you see this happening more on the facades facing away from the sun. You could separate direct normal and diffuse horizontal irradiance in your analysis, instead of using GHI, to account for this.
If it turns out that isn’t the problem, then check your ambient bounce settings. Having too few ambient bounces, or too low an ambient accuracy, could throw off your results. Also, consider using the two-phase method instead of Daysim to eliminate some of the issues that irradiance caching could introduce.
Dear Nathaniel Jones:
Thank you very much for your suggestions. We rerun the simulation using Daysim, increasing ambient bounces=6, simulating GHI, diffuse horizontal radiation, direct normal radiation. The down-trend is still there (see GHI, direct normal radiation). For diffuse horizontal radiation, the curve is even strange.
May I know your idea about this? I’m very appreciated for it. Thank you very much for your help.
I was not referring to the number of ambient bounces you use, but rather to the other ambient settings, particularly ambient accuracy. Again, to eliminate this as a source of concern, you are probably better off using the two-phase method instead of Daysim.
I’m unable to read the graphs you uploaded, as the resolution is too low for text legibility and I can’t read any of the axis labels. However, my previous point still stands: I don’t think you are accounting for the weather. The distribution of radiance from the sky dome changes depending on how clear or overcast the sky is, and on how close the sun is to the zenith. I think you are expecting your graphs to follow a linear shape, but that would only happen if there were no variation in sky condition.
Perhaps you could render the sky and measure the brightness of the portion of the sky faced by each facade.
Dear Nathaniel Jones:
Thank you for your reply. Here are higher resolution versions of the plots for GHI, direct normal radiation and diffuse horizontal irradiation. As you can see, the higher the GHI and the higher the direct normal irradiation, there is a decrease in the irradiation on one building surface (similar behavior observed for other surfaces and other buildings). This behavior is very consistent for east facing facades in the evening (i.e., when you wouldn’t really expect direct normal radiation other than due to reflections). The question is, if this behavior is solely due to weather conditions, wouldn’t that imply that for all evening hours where GHI/direct normal radiation was high it was always cloudy?
Regarding the values used for the ambient settings, here’s the values used originally (the only change for the new version was changing ab to 6):
ab = 4
ad = 512
as = 32
ar = 20
aa = 0.15
Thanks for your feedback on the two-phase method. We’ll look into it as well.
Dear Nathaniel Jones:
The graph resolution is low due to the website issue. So I upload them on Github with the hyperlink below. It’s high-resolution and clear. Thank you.
DHI DNI
(NB: Dots with different colors represent different months)
Looking at your parameters, the ambient resolution (ar) is very low, which could lead to inappropriate blending of cached irradiance values. This is only an issue if you have shading geometry or a ground surface, and I’m not sure if you do. If that is the case, a smaller ambient accuracy (aa) value would also be helpful. I also think you need a larger ambient division (ad) setting. With a value as low as 512, there may be some chance that your sampling misses the sky patch containing the sun, although I don’t think that’s your issue.
No, it implies the opposite. When the sky is clear, there will be less irradiance on the east side of the building late in the day. When the sky is cloudy, the irradiance on the east side will be higher, more similar to the other facades. Again, you could produce a rendering of the sky to verify this.
Were you expecting the sky to be darker when it is overcast than when it is clear? Take a look at this set of photos of the sky, for example. On a clear day, the sky is not very bright.
