Repeating Honeybee Simulation Results in Different Shortwave MRT Delta Everytime

Hello everyone!

I have a very strange problem with Honeybee UTCI component Simulations.

Everytime I repeat the simulation, without doing any changes, I get different Shortwave MRT delta, why does this happen?

Sometimes there are even random negative values in the middle.
However, the Longwave MRT does not change.

I will attach screenshots of the results of 3 repeated simulations without doing any changes.

I appreciate your help.

Thanks a lot.
Regwan

1
2
3

Hello @chris,

I would appreciate your help to understand what is happening here. As I cannot seem to validate the outcome against field measurements.

Thanks a lot.

Hi @regwan,

The difference in SWMRT between runs is due to that aspect being derived from the radiance simulation. Radiance simulations are stochastic, ie they have some inherent randomness over each simulation - you can up the radiance settings to reduce variation between runs, but it’ll take much longer to simulate. The variation is typically seen as OK.

If you search for negative lux values you’ll see some other conversations on why Radiance can give negative values.

Personally I don’t think the shortwave MRT deltas should ever be negative. My preference would be for the comfort workflows to set any negative values to equal 0.

For Chris’s reference, I don’t mind seeing negative values for daylight simulations as it’s a good indicator of how appropriate the radiance parameters are, and negative values don’t contribute to CBDM metrics (I believe)

Edit:
Energy plus gives the long wave results and is a deterministic process, ie an identical simulation will always give the same results.

Hi @charlie.brooker ,

It’s good to hear from you again. Thank you for your prompt response to my post.

The problem is, the differences are not small as you can see in the 3 trials. One had a maximum shortwave MRT delta of 27 and another had max 20. This is kind of a large difference.

Another thing is, those values are for the hours 7am to 7pm of one day. I expect the MRT to be increasing gradually up to a certain point then decreasing. But the First trial shown above, that had a negative value, does not follow this sense. You find it increasing then decreasing the next hour, then increasing again the hour after. This is strange.

I uploaded a screenshot of the radiance parameters I used. I followed the recommended parameters for a higher accuracy.

Thanks a lot Charlie.

Hey @regwan, from my fairly limited radiance knowledge, optimal radiance parameters will vary based on the model size of the model you’re simulating and the size of the grid.

The default parameters are supposed to be roughly good for a wide range of scenarios. To test how well they’re working for your scenario I’d recommend doing a HB annual irradiance simulation and looking at the individual hourly results from that (or you should be able to load the radiance results that have been produced from comfort analysis is another option). If your results are patchy / generally look inaccurate you should be changing your parameters to make them accurate enough for your use case. Particularly the parameters related to scene size (I’d have to look it up but I’m thinking ambient resolution and ambient division).

Bear in mind that shortwave mean radiant temperature delta is the impact on comfort from solar radiation. Bodies I don’t think emit shortwave radiation so the delta should never be negative.

Radiance can be quite sensitive to direct radiation contribution (hence the negative value errors). Something to be particularly aware of. The most significant drive of shortwave delta will be direct radiation, indirect bounces (unless you have very reflective surfaces nearby) will be fairly minor. I think this is why Chris and the team have set the default number of ambient bounces for the comfort workflows to 2.

Hey @charlie.brooker,
Thanks for the recommendations regarding radiance parameters, my knowledge about the details are not quite much at all. I thought I’d follow the recommended parameters for accuracy here: SETTING RENDERING OPTIONS

Here is a screenshot of the model:

So basically, I am doing simulations at 1 point in the middle of the canyon, the ground and canyon buildings are modelled as HB rooms, you can see the ground surface is cut into 1*1m grid, the building surfaces is a bit larger.

For time issues, I modelled all rest of surrounding ground and other context buildings as ‘Context’.

I will try to look into everything you said here, and see how it goes.

Thanks a lot.

Best,
Regwan

1 Like

Thanks for sharing that link - looks very useful!

1 Like

Oh I am glad, thanks. It is included in the ‘HB Radiance Parameter’ component description. But it was sent to me by another researcher as well.

Hello Everyone,

I am re-opening this discussion again, as I am still facing quite a few problems.

First, the Direct solar exposure hours at the point of study (in the middle of an urban canyon), is different between Radiance and Ladybug. I will attach screenshots.

For radiance, it shows exposure from 10:00 to 13:00, while for LB it shows exposure from 10:00 to 14:00.

Another thing is, I still get extremely varying results for the HB annual irradiation and the shortwave MRT delta, when repeating simulations without any changes.

Sometimes the results make sense, and are gradually increasing then decreasing throughout the day, but sometimes they are not following any sense and I even get random negative results.

I noticed that the odd results are usually on hours (9:00) and (14:00), which are supposed to be 1 hour before and after direct exposure according to Radiance.

Any ideas?
I would appreciate your help.

Thanks
Regwan


Hello again,

I just made a discovery! Radiance does not read the ‘North Vector’ input for the direct solar exposure.

So when I removed the North Vector from the LB human to sky relation component, to get the direct exposure without customizing the North direction. The direct exposure was the same as Radiance (from 10:00 to 13:00).

Now, when I input the new North Direction in both HB direct sun hours and LB human to sky relation, the LB’s new direct solar exposure hours change, but the HB (radiance) does not!

@chris, would you mind please checking that?

Thank you
Regwan

Hey @charlie.brooker,

Seems that Chris is away for a while.

Did you get to check the latest posts here? Seems like there is a bug regarding the north direction input in HB / radiance.

update: I rotated the whole model, to avoid having to input a North vector, and HB still reads direct exposure from 10:00 to 13:00, instead of 10:00 to 14:00 (like ladybug)

I also have a question regarding ‘Radiance modifiers’. Whenever I used them to modify the surface reflectivity of materials, there is absolutely no change in the shortwave MRT outcome, is this normal? were they made for other purposes? (daylight studies?)

Thanks a lot,
Regwan

Hey @regwan, I was doing some HB radiation studies earlier and didn’t notice anything obviously wrong, but I haven’t explicitly tested the north angle yet - will try to remember tomorrow.

If you could share an example demonstrating the problem to save me putting one together that would be great and I’ll happily test

Hi @charlie.brooker,

Thanks for the prompt response. I was un-installing Radiance and re-installing. Just making sure that it is not a software issue.

I tried inputting a different direction of North, and the results matched between LB and HB. So it seems like an unfortunate coincidence that this specific hour has the issue in this specific orientation.

So no need to re-test.

However, I noticed that the values that are usually off greatly are for the 1 hour right before and right after the direct exposure. Those are the ones that sometimes become negative too.

Also, any idea on radiance modifiers if they should actually change the shortwave MRT?

Thanks a lot,
Appreciate your help

Best,
Regwan

Hi @regwan,

The negative values occurring around these hours make sense - have you looked at this thread?

HB uses the enhanced 2-phase method for daylight/radiation simulation, which from my understanding is the root cause of negative lux / irradiance values. My rough understanding is the sun patch and sun beam simulations aren’t guaranteed to fully align, especially with coarse radiance parameters. Have a look at other posts from @justinfmccarty and @mikkel for more information.

The full tutorial by Sarith here is a big help in understanding the calculation methodology.
https://www.radiance-online.org/learning/tutorials/matrix-based-methods

My only potential cause for concern, without actually testing, is that the sky dome oddities reported by @lionpeloux for certain weather files could be resulting in misaligned sky patch intensity and sun beam vectors. I’m likely wrong with that concern and I’m sure someone like @chris or Mikkel could easily correct me - my knowledge of how sky patches and sun beam vectors are produced for radiance is close to none.

On the Radiance modifiers, I would expect them to have an impact, but that impact could be relatively minor. Are you explicitly modelling a ground surface? I’d assume the ground reflectance will be one of the most significant variables you could change to see an effect. I would set up a simple model to test.

Hey @charlie.brooker,

I haven’t seen those threads before, thanks for sharing.
I thought with a highly accurate Radiance Parameters, it should give out better results, or atleast a consistency in the simulation repetition. But I have tried a lot of varities and still same issue always comes up.

I will take a closer look at the threads you sent.

As for Radiance modifiers, I did try on a simple ground and a point above it at a height of 1.1 m.

No matter how much I change the reflectivity of surface, the shortwave MRT is more or less the same, unlike when I change it through material properties in construction sets.

I will give it another go, and maybe upload a clean file here to see where the problem is.

Thanks a lot Charlie.

Regwan

Hi @regwan ,

These are good questions and @charlie.brooker 's response is generally correct. The most important thing to note about these negative values is that they result from Radiance’s use of stochastic methods to generate rays. So the negative values will go away if you use better Radiance parameters.

You can kinda think of these negative values as an indication of how large the error is given your choice of Radiance parameters. So, if they seem large to you, then you should use some better ones until they are acceptable low.

One other thing that I wanted to clarify is that, even though the recipe uses Radiance under the hood and accepts Radiance parameters as input, the Radiance modifiers assigned to the Honeybee geometry do NOT affect the thermal comfort mapping simulation. Because the thermal mapping recipes can use Energy plus to compute surface temperatures and indoor temperatures. This requires full layered definitions of constructions and so it’s the Energy plus constructions assigned to the Honeybee geometry that determine the output of the thermal mapping recipes. There’s actually a step that runs as part of the thermal mapping recipes, which assigns all of the radiance modifiers using the solar reflectance and transmittance of the Energy plus constructions before starting the radiance simulation.

Hope that helps.

Thanks @chris, really useful to know that the radiance modifiers are changed to match the E+ constructions as part of the recipe.

Hi @chris, thanks so much for the useful info!

Regarding adjusting Radiance Parameters, I have tried dozens of variations and still the repeated simulations result in negative or low values sometimes (for the hour right before and after the direct solar exposure). I checked the online slides on radiance parameters as well as the recommended values for accuracy. Nothing is working for me, any advice on how to optimize radiance parameters for my model?

Also, radiance is showing direct solar exposure from 10:00 to 13:00, unlike Ladybug that’s showing it from 10:00 to 14:00, hence the hour 14:00 has low shortwave MRT even though it is supposed to be exposed to direct radiation. Is there a way I can fix the direct solar exposure hours?

Thanks a lot
Appreciate your help.

Regwan

Hey @regwan ,

If you are adjusting the correct radiance parameters (using the “rfluxmtx | annual” option from the HB Radiance Parameter component), then you should see the negative values get smaller and smaller the higher quality that you use. At “high” quality the simulation will take a while but the negative values should be small enough to be barely noticeable. However, they may not go away completely since they result from the fact that the randomly generated rays between the “total” sky and the “direct” sky steps of the recipe are not perfectly equal. The 2.5 phase method that we use subtracts the direct results from the total before adding in a more accurate version of the direct sun results. So that subtraction can result in some negative values around where this direct subtraction occurred.

As for this:

The comfort maps (and all annual radiance recipes in the LBT plugin) evaluate conditions on the half-hour mark (eg. 10:30). NOT the hour mark (eg. 10:00). The reason for this is probably more complex than this post merits but it results from the fact that the radiation values in the EPW file are essentially for the half-hour. Long story short, just make sure you are running your direct sun study using sun positions on the half-hour instead of the hour (eg. 10.5 instead of 10) and the direct results should match exactly between your comfort maps and the direct sun study.

2 Likes

Thanks a lot @chris :))

1 Like