This otherwise quite sensible article says:

Designers tend to settle for the simple metric of the U-value of the window, the heat transfer coefficient of the whole system including conduction, convection and radiation. It represents the heat flow in watts per hour through each square metre of the window for a 10°C temperature difference between the indoor and outdoor air temperature.

What?

I really expect and hope Prof. Roaf did not write that. I bet she's more than a tad fed up with whoever mangled her article in this way.

Ooh dear! Yes, didn't notice this! And I usually do notice the proof-reading errors!

Good point Tom. Same question could be asked for U-values of more opaque parts of the building fabric. We know that it's better to put more insulation in roofs because they can be exposed to radiation losses to clear skies but do we take that into account with HDD calculations, etc?

My guess is that radiation loss, and therefore this kind of uncertainty about it, is far less significant for opaque fabric (walls, roof etc) than for transparent. And particularly important because in low-energy buildings, esp ones that take account of 'passive solar' (e.g. PH), the whole concept of solar gain thro south windows, falling on heavy internal 'storage' elements (floor, walls etc) depends on such assumptions about re-radiation loss, as soon as the sun goes in.

For opaque fabric, radiation loss (if any) is from internal room/objects to the internal wall/roof surface, then from external wall surface to scenery/sky etc. The temp gradient from inside room/object to scenery/sky is divided into three, in which the middle segment (thro the wall) greatly dominates, leaving v little gradient to drive radiation in the first and last segments.

Whereas for transparent fabric, the full temp gradient from internal room/object to scenery/sky etc is in full force, driving the radiation loss. It's true that the glass, and its coatings, do obstruct that radiation loss, by being relatively transparent to hot (high frequency) incoming solar radiation, and relatively opaque (a lot or a little) to tepid (low frequency) outgoing radiation from room/objects.
However that doesn't make it same as opaque fabric - the full temp gradient applies, albeit the resultant radiant heat flow encounters resistance at the glass line.

What's more, generally, those internal room/objects (incl the room's heat-source surfaces) that can 'see' the outdoors, thro the glass, are warmer than room air temp,
while those internal room/objects (incl the room's heat-loser surfaces such as external walls) that can't 'see' the outdoors, thro the glass, are cooler than room air temp.
So the room/objects that emit the radiation-loss are generally, by definition, warmer than would be expected by simply assuming that radiation loss is proportional to room air temp (same as conductive/convective loss is). Thus radiation loss is generally under estimated.

Similarly, when the weather is critical, the scenery/sky is generally colder than external air temp.
So the scenery/sky that 'pulls' the radiation-loss is generally, by definition, colder than would be expected by simply assuming that radiation loss is proportional to external air temp (same as conductive/convective loss is). Thus radiation loss is again generally under estimated.

I think so.
Presumably, in hot box type tests, the internal surfaces of the box on the hot side will equilibriate at about equal to hot-side air temp, likewise on the cold side.
Then they measure the actual heat transfer from hot side to cold side, at given maintained delta-t, without distinguishing between conductive, convective or radiant modes of transfer.
That must give an over-optimistic heat transfer coefficient U, because as I've suggested above, in real life
the delta-t between object temps, as affecting radiation will be different from
the delta-t between air temps, as affecting conduction/convection.

In warming weather, the radiant delta-t between object temps may actually be smaller than
the conductive/convective delta-t between air temps,
but when it matters, in weather that's getting colder, and/or when the sun goes in,
the radiant delta-t may be v much greater than the conductive/convective one, on which hot box testing is based.

So when it matters, radiation loss thro windows may be much higher than anyone has suspected.
That has great bearing on the typical assumption in passive solar design, that heat stored into heavy masses inside south windows, actually stays there - maybe it's re-radiated away much quicker than expected.

Posted By: Ed DaviesOn the outside the interface resistance is usually taken to be small, because at least some wind is assumed

That's another, comparable approximation that we accept trustingly. How long ago were these assumptions and approximations established, and do they still hold good in the new regime of ultra-low U values? Are all our so-precise calcs (esp in e.g. PHPP) in fact way out?

Posted By: Ed DaviesI've no idea how much energy is really transferred by radiation at typical temperatures

We ought to know, shouldn't we.

Yeah, that's another thing again.

How do dynamic results compare with U value based calcs using degree-days i.e. averaged over a 'typical' month (assuming the 'typical' month that's described in the degree-days figures is derived from the same date-range as the 'typical' month that's described in the weather data fed into a Tas calc).

As most windows are vertical, and most houses are surrounded by a other houses, isn't the temperature difference just the difference between the inside of the house and the outside of the house it is looking at rather than the black cold space?
So you are probably only talking about 10 to 20°C for a few days rather than a couple of hundred °C.

If the window can see the sky....

These are strictly radiation losses?

Posted By: tony200K for a window with a view of the horizon.

I am not so sure it is that high when looking at the horizon, there is a lot of warm air and particles before you reach cold space.
I think that if it was a serious effect then we would be getting windows icing up in air conditioned buildings in the West Indies.

Posted By: fostertomNot so sure - AFAIK it's just a %age opaque.

That's correct, about 2% of long wave IR will pass through glass.

Posted By: SteamyTeaAt a 20 K temp difference the losses are 0.0083 W/m^2
40 K it is 0.13 W/m^2

I think, so pretty minor

That's for a black body, in reality the emissivity will mean it's even lower. For example a cotton sofa at ΔT of 20K will emit about 0.007Wm^-2. Low-e windows can have an emissivity of something like 0.05 IIRC, so you'd be looking at 0.0005Wm^-2. There's a reason we don't really bother going out of our way to separately model radiative heat loss in the standard models for buildings. It's just not a big deal at the temperatures involved.

As for radiation to the sky (space being what, 3K or somthing?), as ST points out the atmosphere isn't particularly transparent, especially at low angles. If it wasn't so we wouldn't have seasons, and in fact life on Earth wouldn't be possible. The insulating effect of the atmosphere stops the Earth radiating away all the heat it gets from the sun and raises the average temperature from what it's supposed to be (about -15º) to a nice comfy 15º. Which kind of makes the idea of a "habitable zone" for exoplanets a bit moot, as technically not even Earth is in the sun's habitable zone. But I digress...

Posted By: SeretThat's for a black body

I used 0.92 as that is what is claimed for glass. But that is close to 1.
Would a cotton sofa at 20K shatter when sat on

Was a bit about 'habitable zones' in New Scientist and they have extended the zones now because water vapour plays a crucial role in climate, also explains, to some extent, Earth early climate.

Posted By: SteamyTeaWould a cotton sofa at 20K shatter when sat on

He actually said

Posted By: Seretcotton sofa at ΔT of 20K will emit about 0.007Wm^-2

I'm suggesting that at times when these considerations really matter - i.e. when outside air temp is falling, or it's a cold clear-sky night, or just when the sun goes in, the delta-t as affecting radiation will be a lot more than 20K delta-t. It will be between the sofa at 20C, and

Posted By: Ed Daviesthe effective radiant temperature of the sky then is around -40 °C or so

- so 60K delta-t, or heading that way.

Posted By: fostertom

Posted By: Seretabout 2% of long wave IR will pass through glass

Sounds a bit too general - if that's true of ordinary glass, why bother with coatings?

You're confusing emissivity with transmittance. Low-e coatings are to stop the window itself radiating, not to stop IR radiated from other sources. Ordinary float glass is perfectly good at blocking long wave IR, which is why it works for greenhouses, but it's emissivity is above 0.9, so it's an excellent radiator. Hence low-e coatings, they cut radiative heat loss from the windows themselves enormously.

I've just realised I did my maths wrong anyway, it's T_hot⁴-T_cold⁴ not ΔT⁴.

So the sofa at 20°C, if it were taken outdoors into freezing temps would radiate at about 75Wm^-2, but through the window of your living room it would only manage 1.5Wm^-2. Reflection and the low-e coating on the window would mean that the window would itself radiate bugger all of the other 73.5W.

If the sofa were radiating to a -40°C sky through the window it would manage about 3.8Wm^-2.