Help me calculate cumulative irradiation from a sun that hasn't risen (in the Arctic)

JonathanASP · May 3, 2022, 2:38pm

So I’ve recently gotten into Ladybug Tools after taking a course in daylighting at my university last year, and I’ve been experiencing an issue with accurately simulating winter daylight in the Arctic (I specialize in architectural engineering aimed at constructions in GreenlandI).

My issue is that all cumulative/annual simulations do not seem to account properly for lighting from a sun that hasn’t risen, which is a big issue as a lot of the hours of the year for which this is the case are actually pretty bright. Basically I get outputs like the ones below where my HDR images and point in time (when using CIE Standard Sky only, all other models also give invalid results) are fine, but all of my cumulative results are inaccurate (providing me 0kWh for all hours before sunrise).

Does anyone have any ideas of how to fix this inside the current Ladybug tools?

chris · May 4, 2022, 12:15am

Hey @JonathanASP ,

This is a really good question and I would be interested in hearing the thoughts of @mostapha , @mikkel and @sarith on it. In the meantime, I can explain what currently happens in the “Annual Daylight” and “Annual Irradiance” recipes.

In order to get the simulation to run as fast as possible, the annual daylight and annual irradiance recipes currently only run the calculation for hours when the sun is above the horizon. We figure out these hours by means of the honeybee_radiance.sunpath module, which is essentially using the ladybug.sunpath module under the hood.

The ladybug Sunpath is certainly capable of computing how far below the horizon the sun is. So it would not be very difficult for us to just change these sun-up hours to include hours where the sun is 5 degrees below the horizon, for example. However, I don’t know enough about Radiance to say for sure that the gendaymtx command that we use under the hood to account for the diffuse sky irradiance is still going to produce non-zero values even when the sun is below the horizon. Nor do I have a sense about whether these types of dawn/twilight conditions produce anything close to a meaningful level of light, though I feel intuitively that they do as I sit in civil twilight as I write this.

Perhaps with some input from the people above, we can decide whether a change the the annual recipes is merited for this case.

In the meantime, the Point-in-Time workflows are probably the right way to go since, as you say, the CIE model is robust enough to model dawn/twilight conditions.

JonathanASP · May 4, 2022, 8:54pm

Thank you so much for the reply @chris,

I can’t speak very much to the technical implementation of a model that takes dawn/dusk into account as I’m still very much in the process of learning the basics of how our daylighting models work, but having lived multiple years north of the Arctic circle I probably can provide some useful input as to whether or not the light present in these hours represents meaningful daylight for a building design process.

Dawn/dusk in the Arctic and Subarctic can be multiple hours long (SunCalc - sunrise, sunset, shadow length, solar eclipse, sun position, sun phase, sun height, sun calculator, sun movement, map, sunlight phases, elevation, Photovoltaic system, Photovoltaic , Suncalc reference for Sisimiut, where I’ve lived multiple winters), and these winter hours, in my experience, provide daylight autonomy for about 4-6 hours a day for regions close to the Arctic circle. Of course this decreases further north, but having talked to a classmate raised in a city located at 73deg northern latitude, I feel fairly comfortable claiming that most Arctic settlements experience some daylight autonomy at midday throughout the winter.

I’ve provided a December picture from Sisimiut below, so everyone can see what I’m talking about.

Unfortunately it seems the original timestamp got lost in the process of getting my pictures onto the PC, so I can’t be sure of exact date/time, but the picture represents a slightly underexposed version of my subjective experience of Sisimiut in winter during the interval of 9am-10am/2pm-3pm and it seems to me that this light I experience living there doesn’t translate into the current models.

How best to change that… Well I don’t know but if I can be helpful, I’d really like to be.

wim · May 5, 2022, 9:04am

Just a sideline comment - because of refraction in the atmosphere, the sun can visually appear to be above the horizon, while, in fact, it is below the horizon. I’ve spent enough summers about 100 km south of the arctic circle to know that the midnight sun doesn’t stop at that arctic circle.
-wim

chris · May 5, 2022, 5:00pm

Thank you both. The Ladybug Sunpath that we use to calculate the sun positions right now already accounts for the refraction of the sun over the horizon as you can see here in the source code:

github.com

ladybug-tools/ladybug/blob/master/ladybug/sunpath.py#L275-L289

    
      
          # approx atmospheric refraction used to correct the altitude
          if altitude > 85:
              atmos_refraction = 0
          elif altitude > 5:
              atmos_refraction = 58.1 / math.tan(math.radians(altitude)) - \
                  0.07 / (math.tan(math.radians(altitude))) ** 3 + \
                  0.000086 / (math.tan(math.radians(altitude))) ** 5
          elif altitude > -0.575:
              atmos_refraction = 1735 + altitude * \
                  (-518.2 + altitude * (103.4 + altitude * (-12.79 + altitude * 0.711)))
          else:
              atmos_refraction = -20.772 / math.tan(math.radians(altitude))
          
          
atmos_refraction /= 3600
          altitude += atmos_refraction

So we still count an hour as “sun-up” when the sun isn’t physically above the horizon of the earth but you can still see the disc of the sun. The first-hand accounts of life close to the arctic circle are also helpful but I can think of two things that might be a bit more rigorous if you would like to help us decide whether a change in the recipe is merited here:

Using a Point-in-Time CIE simulation where the sun is completely below the horizon, are you able to get a case where the illuminance is more than 100-200 lux? I’d say that this is probably the lower limit for what we would consider “meaningful”.
Within these subarctic weather files, can you find hours of the year where the Diffuse Horizontal Radiation is greater than zero, even though it is for an hour where the Ladybug Sunpath shows the sun is completely below the horizon for the whole hour?

If both of these end up being true, then I think I’ll look deeper into this when I get the chance.

josephyang · May 7, 2022, 8:51am

This is a really interesting photo considering the location. One thing that also seems to be happening here is the reflection of sunlight from the bottom of the cloud. Wondering if in your experience there is a noticeable difference in brightness when the sky is completely clear vs. somewhat cloudy like what’s shown in the photo (e.g. enough to reflect some light but not enough to completely obscure the sun)?

JonathanASP · May 13, 2022, 10:26am

@chris I’ve been looking over the sunpath and gendayz files for the last couple of days and I’ll start running some tests including the ones you suggested using the DOE dataset for Resolute, Canada, as I’m confident in the accuracy of that specific dataset. I’ll upload the gh file when I’m done

@josephyang Thinking about it, all of the brightest seeming days from my decembers in Sisimiut involved underlit clouds so I think you’re onto something. I think it also helps that Sisimiut has a very specific commonly occuring cloud pattern that involves cloud banks forming just offshore while the rest of the sky remains clear.

JonathanASP · May 13, 2022, 8:03pm

@chris I completed the tests you suggested first as they seemed like a good place to start.

I unfortunately can’t provide the gh file as I intended, as my forum account isn’t old enough, so I’ll have to provide screenshots instead.

Picture above is from solar noon on the 3rd of February (hoy 805) which is the brightest hour on the first day of the year in the Resolute dataset that has measured daylight. As seen interior surface illuminance exceeds 100 lux in certain areas, but is mostly below that figure. View is facing south at an elevation of 1.6m and geometry is a 10x10x3m box with 0/0.2/0.4/0.2 apertures by ratio and 0.1m extruded border shades resting on a 100x100m face with 70% reflectance and specularity and roughness 0.1 (my regular stand-in for snow-covered ground, which is my assumption, also if anyone has a good source on optical parameters for snow-covered cityscapes I’d love to read it, I can only find literature pertaining to remote sensing applications), and ambient bounces are set to -ab 8. All other radiance parameters are default.

The ladybug sunpath module gives the first hour of sunlight as solar noon on the 5th of February for the location which excludes 6 hours of measured daylight in the dataset on the 3rd and the 4th of february, as well as two hours on the 5th of February. I haven’t combed through the entire dataset, but I’d guess we’re losing 2 hours of twilight each day of February and November when comparing with the dataset.

Above is a grid-based point-in-time for the same model using a 0.5m grid at the standard 0.8m height above floor, and we find that the entire room falls within the UDI space.

The above picture represents the same geometry but with the view set due north, and we get a much better illumination of the interior when viewing in this direction, with a fairly even 70 lux on the back wall.

And finally an exposure adjusted HDR image of the north facing view.

So this seems like useful daylight to me, which is what I expected to find, although I actually find the results brighter than I expected. It seems like the CIE sky provides results that are brighter than I’d expect from reading through the EPW file, which is a discrepancy I’ll have to wait on investigating till I’m back in Greenland and can collect some data.

I also have a question for @chris as to how the HB cumulative simulations use the EPW data. I can tell that according to the sunpath module the sun will be above the horizon for a lot of hours where the Onebuilding datasets available for Greenland have no daylight (these datasets are all inaccurate, at least the Greenlandic ones, the DOE datasets consistently have measured light much closer to the winter solstice than the Onebuilding datasets for similar latitudes, and the DOE sets much better match both CIE standard sky and suncalc.org dawn/dusk data), so what happens in the event that the sunpath module finds that the sun is above the horizon, but the EPW data says there is no light?