We've been working on a 1960s ranch for what seems like forever, and are finally tackling the last main section.

It has a cold deck flat roof--so continuous soffit vents on all sides, 2x10 joists with a plywood deck and built-up (asphalt/gravel) roof. The joist cavities have some 2-3" foil-backed (mineral wool or fiberglass?) insulation in them, and no blocking at the ends of the bays to keep air from moving through the insulation.

The roof needs replacing, and I am trying to decide whether it makes more sense to completely remove the roof deck and insulate from the top, or sacrifice the plaster ceilings and remove a lot of trim from inside and insulate from the bottom.

What I know:
-In order to get a respectable R value, I will probably have to convert to a warm roof deck, ie make the roof sandwich non-vented. (assuming fiberglass batts, the best I could get is an R25 in the cavity while retaining the gap for airflow under the deck);

-Although the roofers can add ISO board on top of the sandwich, I am resisting this because it will change the roof edge profile (currently it looks like it could have been done by Frank Lloyd Wright)

I have been reading this article on Building Science that shows the two acceptable ways of building a non-vented roof, and include the two pictures.

The second version I like better--spray foam from below, but can dense pack cellulose be put into a ceiling cavity as shown?

I would also have to adjust either model to accommodate the large overhangs that I have--any suggestions on how to do that?

I also have quite a few pot lights in this section--is it worth getting rid of them?

Any thoughts or suggestions, or anyone retrofitted a house like this?
 

FWIW, here's a sketch of a cold roof deck from the UK that gives an idea of the current construction.

Dense-packing cellulose-only without foam at the roof deck is problematic in cold climates, but can work fine in warmer climates (US zone 3 or lower). You may be able to do a low-rise foam pour from the top, but at some risk of cracking the plaster with the pressure. (It's expensive too.)

The reason you need the closed cell foam between the cellulose and the roof deck is that without venting the roof deck will otherwise accumulate moisture whenever it's below the dew point of the conditioned space air, since the cellulose is both water-vapor and air permeable. Closed cell foam is air-impermeable and somewhat vapor permeable, but forms a condensing surface that will not wick moisture up to the susceptible wood, dramatically reducing the rate of moisture accumulation, yet still allowing the wood to dry toward the interior during warmer weather. In very cold climates it may be necessary to pay attention to the ratio or foam/fiber R values to fully protect, but when the condensing surface is closed cell foam it takes orders of magnitude more seasonal hours below the dew point of the interior air for the moisture to end up in the roof deck compared to the un-foamed roof deck configuration. The cellulose wicks and redistirbutes the moisture too, and can buffer quite a bit of water before losing function. The roof deck has no chance of drying toward the exterior through a membrane or torch-down roof, so going too thick with the foam can form a moisture trap, if the vapor permeance of the foam layer goes too low. The minimum amount of foam required to be protective varies with climate, but even 1-2" does a lot. See: http://www.buildingscience.com/documents/reports/resolveuid/b892656b9ff291aa5d87e055bbc6e5c1

Pot lights on a thin roof under a membrane roof in a snowy climate creates localized freeze/thaw areas in the roof, stressing the membrane. Getting rid of them reduces the heat loss and better protects the roof. If you must, replacing R30 & R40 fixtures with air-tight low profile R20-sized LED fixtures maximizes R value and minimizes the localized heat source, limiting the problem.

Putting rigid foam above the roof deck increases performance by insulating over the thermally-bridging rafter/joist elements, and even a small amount does a world of good from a thermal point of view. But to protect the roof deck from wintertime moisture accumulation it needs to have sufficient R value relative to the fiber layer that the amount of time the roof deck dwells below the dew point of the interior air is very limited. The amount of exterior foam that it takes to achieve that varies by climate.

Thanks so much for your thoughtful reply--the reason for the foam is now much clearer in my mind. We are Zone 5a, so we get real winter and humid air at least some of the time.

But a couple of further questions:

1) If I work from below, I imagine I can box in the ends of the joist bays, and spray the foam on those ends, and across the underside of the roof deck. Seems straightforward. But how would you add the cellulose? I've seen them use netting on a sloped/cathedral ceiling or walls, but can they use that on the underside of a flat ceiling?

Or do the installers have some trick where you would leave a strip out of the drywall and fill from the center out to each side?

Or?

2) Is there much information out there about how hard it would be to remove the plywood decking once you have foamed it in place if you had to replace it down the road? I'm thinking it would be a hard job... (This may influence whether or not I replace the deck now as we do the roof).

Cheers,
Dean

Closed cell foam is like glue- at 3" of thickness it's even structural. Once it's stuck to the roof deck, it's bonded (unless the contractor really screws it up, which happens.)

Cellulose is easily dense packed in netting from below if you're ripping out the plaster ceiling to spray-foam the deck. Be sure to specify that the contractor roll it flat to the joists rather than leaving it pillowed out or it'll be tough to make a flat ceiling. Use 5/8" or 3/4" gypsum to limit bowing over time.

For climate zone 5 you could instead use a denser open cell foam such as 0.7lb Demilec to achieve similar vapor retardency to the 1" of closed-cell + cellulose approach, and it may even be cheaper & easier than the two-step. It's about 5 perms @ 5", but about half that in a full cavity-fill for 2x10s, which puts it at ~ 2x the vapor permeance of 1" closed cell, but since it's air-impermeable through it's entire depth the condensing surface is the back side of the gypsum, guaranteed to be above the dew point of the room air in winter. In the BSC unvented roof document they did the simulations with half-pound foam, which more vapor permeable than 0.7lb or 1lb foam.