For building owners, facilities managers, and commercial roofing contractors

Insulation, air, and vapor control on a low-slope roof

Low-slope · commercial, industrial, and institutional

This is the part of a commercial roof that an owner is least equipped to interrogate, and the part a contractor can change most cheaply without anyone noticing. It is also where the roof either dries out or rots from the inside.

30-second answer

A contractor wants to change the insulation, drop the cover board, or move the vapor retarder. How do I tell whether that is value engineering or damage?

Above a commercial deck sit four control layers — water, thermal, air, vapor — plus a cover board that protects them. The R-value on the submittal is a laboratory number measured at one temperature; the finished roof delivers less, because polyiso loses R-value in the cold and because fasteners and board joints bleed heat around the foam. Where the vapor retarder belongs is decided by climate, not by habit.

Safety first

Going up to look is occupational work, not an errand

OSHA’s general-industry rule for work on a low-slope roof, 1910.28(b)(13), requires a guardrail system, safety net, travel restraint, or personal fall arrest system for employees working less than 6 feet from a roof edge. From 6 feet to less than 15 feet it requires the same range of systems, and permits a designated area only where the work is both infrequent and temporary. At 15 feet or more it permits one of those systems or a designated area, and — where the work is both infrequent and temporary and the employer implements and enforces a work rule keeping employees back from the edge — it does not require fall protection at all. Read the paragraph itself before relying on any of that; the obligations follow the employer, which on your own building can mean you.

Nothing on this page requires you to be on the roof. Insulation type, cover board, attachment method, and vapor retarder placement are all questions you settle in a specification and a submittal review, at a desk.

Send a qualified roofing contractor or consultant, and ask what fall protection they will use before they go. If a moisture survey has found saturated insulation over a steel deck, treat the deck’s condition as an open structural question and keep traffic off that area until someone qualified has looked at it.

Learning paths and saved lessons

At a glance

The short versionSection link

Every number below carries the test condition that produced it. Where a figure has no sourced test condition, this page does not print it.

The three rigid foams used above commercial decks, and one sourced number that distinguishes each. These are not equivalent kinds of number — read the fourth column before comparing them.
Insulation	The number that distinguishes it	The test condition behind that number	What the number cannot tell you
Polyisocyanurate (polyiso)	Labelled by LTTR, which is reported against the product's own thickness and rises with it — typically 5.6 per inch for 1-inch product, 5.7 for 2-inch, 5.8 for 3-inch and 5.9 for 4-inch. NRCA instead recommends designers calculate using an in-service R-value of 5 per inch.	LTTR is a laboratory projection under ASTM C1289, drawing on CAN/ULC-S770, of a time-weighted average over service life. NRCA's 5 per inch is a design figure that, in NRCA's words, accounts for known losses over time and in changing temperature conditions.	What this roof delivers. Both numbers describe a board, not an assembly with fasteners and joints in it.
Expanded polystyrene (EPS)	ASTM C578 permits maximum water absorption of 2%, 3% or 4% by volume depending on type. NRCA recommends roof-board EPS have a minimum nominal density of 1.25 pcf — Type VIII.	A short-term laboratory immersion test written into the product standard.	How much water a board holds after years in a wet assembly. IIBEC's authors note that samples retrieved from below-grade work after several years are often far above the short-term maximum.
Extruded polystyrene (XPS)	ASTM C578 caps water absorption at 0.3% by volume for the roofing types. NRCA recommends roof-board XPS have a minimum compressive strength of 15 psi — Type X.	The same short-term immersion test; compressive resistance is also an ASTM C578 specification minimum, not a measured field value.	That it is immune. The same authors record a predictable 1% to 2% R-value decline over a product life spanning decades as blowing agent leaves the cells.

Read this table one item at a time

Polyisocyanurate (polyiso)

The number that distinguishes it: Labelled by LTTR, which is reported against the product's own thickness and rises with it — typically 5.6 per inch for 1-inch product, 5.7 for 2-inch, 5.8 for 3-inch and 5.9 for 4-inch. NRCA instead recommends designers calculate using an in-service R-value of 5 per inch.
The test condition behind that number: LTTR is a laboratory projection under ASTM C1289, drawing on CAN/ULC-S770, of a time-weighted average over service life. NRCA's 5 per inch is a design figure that, in NRCA's words, accounts for known losses over time and in changing temperature conditions.
What the number cannot tell you: What this roof delivers. Both numbers describe a board, not an assembly with fasteners and joints in it.

Expanded polystyrene (EPS)

The number that distinguishes it: ASTM C578 permits maximum water absorption of 2%, 3% or 4% by volume depending on type. NRCA recommends roof-board EPS have a minimum nominal density of 1.25 pcf — Type VIII.
The test condition behind that number: A short-term laboratory immersion test written into the product standard.
What the number cannot tell you: How much water a board holds after years in a wet assembly. IIBEC's authors note that samples retrieved from below-grade work after several years are often far above the short-term maximum.

Extruded polystyrene (XPS)

The number that distinguishes it: ASTM C578 caps water absorption at 0.3% by volume for the roofing types. NRCA recommends roof-board XPS have a minimum compressive strength of 15 psi — Type X.
The test condition behind that number: The same short-term immersion test; compressive resistance is also an ASTM C578 specification minimum, not a measured field value.
What the number cannot tell you: That it is immune. The same authors record a predictable 1% to 2% R-value decline over a product life spanning decades as blowing agent leaves the cells.

Compressive resistance across all ASTM C578 types runs 5.0 to 60.0 psi for the EPS types and 15.0 to 100.0 psi for the XPS types; for a given density the XPS types are stronger. Depending on type, EPS can absorb seven to ten times as much water as XPS and still meet the standard — which is why EPS is not used in protected-membrane assemblies, where the insulation sits above the waterproofing and lives wet. No compressive-resistance or water-absorption figure for polyiso is printed here, because this page did not verify one against its own product standard.

Tradeoffs

When this page's advice is right, and when it is wrongSection link

The general case on a commercial roof is polyiso in two layers with offset joints under a cover board. There are real buildings where each part of that is the wrong answer.

Best when

You are reroofing a heated building in a cold or mixed climate and someone is proposing to change the insulation build after the price was set.
The submittal states a total R-value and you want to know what the assembly will actually deliver, rather than what the board is labelled.
A contractor has proposed deleting the cover board, or substituting a different one, as a cost saving.
The building runs a high interior humidity — a pool, a food plant, a museum, a laundry, a printing operation — and nobody has said the words vapor retarder out loud yet.
The roof is going back on over a concrete deck, which holds construction moisture and vents it upward for a long time.

Think twice if

The building is unheated and unconditioned. A vapor retarder discussion for an open equipment shed is a discussion about nothing, and thermal performance may not be the point of the roof at all.
The assembly is a protected-membrane roof, where the insulation sits above the membrane and stays wet by design. Almost every rule on this page inverts there — that is the one place XPS is chosen over polyiso for the reason polyiso would normally win.
The design is a compact assembly a qualified building enclosure consultant has already modelled hygrothermally for this building. A model built on this building's real interior conditions outranks any general rule, including the ones here.
The roof carries a large photovoltaic array. Mount penetrations are a thermal defect on a different scale from fasteners, and the array changes both the surface temperature and who is walking on the roof.
Cost of downtime dominates. On some occupied buildings the right answer is the assembly that goes on fastest in a short weather window, and thermal optimisation is a secondary conversation.

What changes the answer

The adopted energy code in your jurisdiction, its edition, its local amendments, and your climate zone. Those set the prescriptive R-value, and no national page can.
Whether the interior is humidified in winter, and to what relative humidity. This single fact moves the vapor retarder decision more than the climate zone does in borderline cases.
Deck type. Steel, concrete, wood, and gypsum decks each change attachment, vapor behaviour, and what a fastener is doing thermally.
Attachment method. A fully adhered build removes most fastener thermal bridging and changes hail behaviour; a mechanically attached build is cheaper and faster.
Hail frequency and the presence of rooftop equipment that brings service traffic. Both are cover-board arguments, and both are site facts rather than material facts.
Whether anybody has run a moisture survey. Adding insulation over wet insulation buries the problem under a longer warranty.

How it works

Four control layers, one cover board, and the defects nobody drawsSection link

A low-slope roof is a stack of control layers: water out, heat in or out, air stopped, vapor managed. Most of the failures on this page are not failures of a layer. They are failures of the joints, fasteners, and interfaces between layers.

Section through a mechanically attached low-slope assembly. The bands are the control layers; the numbered defects are the four things that take R-value and durability back out of them. Schematic and not to scale — this is a way to see the argument, not a detail to build from.Original diagram, Understanding Roofing. Every callout is written out in full below the figure.

What the numbered callouts are

The structural deck. Steel, concrete, wood, or gypsum. It carries the load and it is what everything above is anchored to. Deck type also decides vapor behaviour, and concrete is the clearest case: NRCA’s 2026 manual recommends a vapor retarder rated at a maximum of 0.01 perms over all concrete roof decks in climate zones 2 through 8 — far broader than its general criteria, and worth raising by name if your deck is concrete.
The air and vapor control layer, applied at the deck. On a compact roof these two jobs are usually done by one sheet in one place — on top of the deck, under the insulation. An air barrier stops bulk air movement; a vapor retarder slows diffusion through the material itself. They are different jobs, and one product can do both. Which one matters more is not close: Joseph Lstiburek’s summary of six decades of the research is that “air leakage was and is more important than vapor diffusion.”
Two insulation layers with offset joints, not one thick layer. Every board has edges, and edges are where heat escapes the foam. Splitting the same thickness into two layers and offsetting the joints in both directions means the slot from one layer is capped by solid board in the other. The U.S. Department of Energy’s Building America guidance is explicit about the practice: install multiple sheets with joints offset horizontally and vertically, and seal or tape the joints.
A through joint. The failure the offset is supposed to prevent. Where a joint in the top layer lands over a joint in the bottom layer, the slot runs continuously from the cover board down to the deck. National Research Council Canada testing found that through gaps produce nearly triple the R-value loss of a staggered gap, and that a single-layer assembly with a quarter-inch through gap lost about 9% of its effective R-value — 13.5% if the gap widened to half an inch.
A fastener through the bottom layer only. Its plate is buried under the top layer of foam, so the metal shaft ends inside the insulation rather than a fraction of an inch under the membrane. This is a design decision, and it is the one on this drawing that most reliably buys back performance for money.
A fastener driven from the top through everything. Cover board, top layer, bottom layer, deck, in one run. Cheaper and faster. It is also a metal path from the deck to just under the membrane, and — as the hail section below sets out — it is the single worst place on the roof for a hailstone to land.
The cover board. A thin, dense board between the foam and the membrane, from an eighth of an inch to an inch thick depending on material. It is the layer that gets deleted first when a price needs to come down, and it does at least five separate jobs.
The membrane. The water control layer, and the surface every other decision on this page is hidden under. Which membrane — TPO, EPDM, or PVC — is a separate decision from everything under it, and the two decisions are often made by different people who do not talk.
The winter condensing plane. Somewhere in the assembly, on a cold night, the temperature crosses the dew point of the air inside the building. If that plane lands on a surface where moist indoor air can reach it, water forms there. The whole point of putting insulation above the deck is to push that plane up into the foam, where there is no moist air to condense — and the whole point of the air and vapor control layer at item 2 is to stop indoor air getting past it. This is the same vapor-drive mechanism that runs hot-humid climate moisture failures, pointed the other way round.

Why the order of the layers is not negotiable

Almost every rule below follows from one fact: water vapor moves from warm and humid toward cool and dry, and it condenses on the first cold surface it can reach. A roof assembly succeeds when the cold surfaces are unreachable and the reachable surfaces are warm.

Insulation above the deck is what makes that arrangement possible, because it warms everything below it. That is also why swapping a proposed above-deck insulation build for a below-deck one is not a like-for-like substitution, and why NRCA’s current guidance for assemblies insulated below the deck is that they should still carry continuous above-deck insulation of R-5 to R-10 for condensation protection.

Worked example

What the submittal says R-30, and what the roof deliversSection link

This is the calculation nobody puts in front of an owner, done with published numbers and with every step of the arithmetic shown so you can disagree with it.

Start with a build the polyiso industry itself publishes. PIMA — the trade association of polyisocyanurate manufacturers, and therefore not a neutral party on polyisocyanurate — states that in climate zones 4, 5 and 6, two layers of 2.6-inch polyiso deliver a total R-value of 30. That is 5.2 inches of foam and an implied rate of about 5.77 per inch. LTTR is reported against a product’s own thickness and rises with it: NRCA gives typical minimum LTTR values of 5.6 per inch for 1-inch product, 5.7 for 2-inch, 5.8 for 3-inch and 5.9 for 4-inch. An implied 5.77 for a 2.6-inch board sits inside that pattern, so nothing about the figure is irregular. What matters here is only that it is a labelled laboratory value. Call it the submittal number.

Step one — the material, in service rather than on a label

LTTR is a laboratory projection of a time-weighted average, produced under ASTM C1289 using accelerated ageing methods that stand in for fifteen years of service. NRCA recommends designers not use it as the design figure — it recommends polyisocyanurate be specified by thickness rather than by LTTR or R-value at all. Since January 2016 it has recommended designers calculate at an in-service R-value of 5 per inch, because that figure — in NRCA’s own words — accounts for known losses in R-value over time and in changing temperature conditions.

5.2 in × 5.0 per in = R-26.0

The submittal said R-30. Designing at NRCA’s figure gives R-26 before a single board is laid. That is a 13% difference, and it is entirely a difference of opinion about which laboratory number describes a roof.

Step two — the joints between boards

Every board has four edges and they do not close perfectly. National Research Council Canada’s ERCR consortium, which NRCA supports — named Energy Resistance of Commercial Roofs in its 2020 reporting and Energy Rating of Commercial Roofs in its 2024 reporting — measured what that costs. NRCA’s 2024 summary of the Phase II results reports average losses in effective R-value of 2% to 7% for quarter-inch gaps and 4% to 9% for half-inch gaps.

low-loss chain: R-26.0 × (1 − 0.02) = R-25.5

high-loss chain: R-26.0 × (1 − 0.07) = R-24.2

Step three — the fasteners

The same research programme measured through-fastener thermal bridging at four, six and ten fasteners per four-by-four-foot board, across R-26, R-31 and R-36 insulation thicknesses. It found thermal losses ranging from about 5% up to 15% depending on fastener density and insulation R-value.

low-loss chain: R-25.5 × (1 − 0.05) = R-24.2

high-loss chain: R-24.2 × (1 − 0.15) = R-20.6

The two chains land on the same rounded number at different stages: R-24.2 is where the optimistic chain finishes and where the pessimistic one still has a step to go. That is a coincidence of rounding, not a finding.

The result, and what is wrong with it

A submittal that says R-30 describes an assembly that plausibly delivers somewhere between about R-20.5 and R-24 — roughly a fifth to a third below the number on the paperwork.

Now the honest part. No study read for this page measured all three of those losses together on one assembly. Multiplying percentages taken from separate pieces of research is an estimate of an order of magnitude, not a measurement of a roof, and the true figure for any particular building is measured rather than calculated from a website. Two specific weaknesses are worth naming. NRCA’s 5 per inch already folds temperature effects into a design figure, so the cold-weather numbers in the table below are a different view of the same material and must not be stacked on top of it. And the ERCR gap and fastener percentages were measured on assemblies at R-26 to R-36, which is the range this example sits in — applied to a much thicker or thinner build they would be extrapolation.

What the arithmetic is for is the decision it changes. Fasten the bottom layer only and adhere the top layer and cover board, and step three largely disappears. Offset the joints in both directions and step two shrinks toward its low end. Those are choices made at a desk, before the roof is priced, and they are worth more than an extra half inch of foam.

How R-value moves with the temperature at which it is measured. The two columns are different kinds of number and are not a product comparison — read the footnote.
Mean temperature at test	Polyiso — NRCA test of 2-inch boards, per inch	XPS — ASTM C578 specification minimum, per inch	What it means on a roof
25 °F	4.05 (average of seven specimens)	5.60	The coldest nights are when polyiso is furthest below its label and polystyrene is furthest above its 75 °F figure. This is the temperature at which the heating bill is decided.
40 °F	4.91 (average)	5.4	A typical winter mean temperature through a roof assembly in much of the country.
75 °F — the label condition	Below the products' published LTTR values; NRCA published no average figure	5.0	The one temperature at which the number on the submittal is measured, and a temperature a roof assembly rarely averages.
110 °F	Tested, but no average published in the source	4.65	Membrane surface temperatures on a dark roof run far above air temperature, so the hot end is not hypothetical either.

Read this table one item at a time

25 °F

Polyiso — NRCA test of 2-inch boards, per inch: 4.05 (average of seven specimens)
XPS — ASTM C578 specification minimum, per inch: 5.60
What it means on a roof: The coldest nights are when polyiso is furthest below its label and polystyrene is furthest above its 75 °F figure. This is the temperature at which the heating bill is decided.

40 °F

Polyiso — NRCA test of 2-inch boards, per inch: 4.91 (average)
XPS — ASTM C578 specification minimum, per inch: 5.4
What it means on a roof: A typical winter mean temperature through a roof assembly in much of the country.

75 °F — the label condition

Polyiso — NRCA test of 2-inch boards, per inch: Below the products' published LTTR values; NRCA published no average figure
XPS — ASTM C578 specification minimum, per inch: 5.0
What it means on a roof: The one temperature at which the number on the submittal is measured, and a temperature a roof assembly rarely averages.

110 °F

Polyiso — NRCA test of 2-inch boards, per inch: Tested, but no average published in the source
XPS — ASTM C578 specification minimum, per inch: 4.65
What it means on a roof: Membrane surface temperatures on a dark roof run far above air temperature, so the hot end is not hypothetical either.

Units are R-value per inch of thickness. The polyiso column reports measured averages from NRCA's testing of seven specimens of newly manufactured 2-inch boards from six manufacturers, aged three to nineteen months, tested per ASTM C518. The XPS column reports ASTM C578 minimum specified values for Types X, IV, VI, VII and V — a specification floor, not a measurement of any product. Comparing a measured average against a specification minimum does not tell you which foam performs better; what the table shows is the direction each moves, and that the two directions are opposite. No EPS row appears because this page could not source the equivalent figures. No polyiso figure appears at 75 °F or 110 °F because the source does not publish one.

The layer that gets deleted

What a cover board is actually forSection link

It is a thin, dense board between the foam and the membrane — an eighth of an inch to an inch thick depending on material — and it does at least five separable jobs. Deleting it is sometimes defensible. Deleting it without naming which job is now unmet is not.

The five jobs a cover board does, what actually decides each one, and what a cover board cannot fix.
The job	What actually decides it	What a cover board cannot fix
Impact and puncture resistance	Board material and thickness, and — as the hail research shows — whether a fastener plate sits directly under the point of impact. A membrane over a plate has nothing to give.	Attachment geometry. If the membrane and cover board are fastened through from the top, the roof still has a grid of hard spots under it.
Contribution to the assembly's fire classification	The listing. A cover board may be the component that makes a specific assembly achieve its Class A, B or C classification under ASTM E108 or UL 790.	A substitution. Fire classifications are limited to the materials, products and configurations described in the listing, and swapping a board takes the roof outside it.
A stable substrate for adhering the membrane	Facer type and surface. Adhesives are formulated to bond to particular facers, and an adhered system is only as good as what it is stuck to.	A wet or unsound substrate underneath. Adhesion to a cover board over saturated foam buys nothing.
Protecting the insulation from traffic	Where people actually walk. Rooftop equipment service routes are the real load case, and walkway pads on those routes matter as much as the board.	The absence of walkway pads. Impact work found standard polyiso crushing about half an inch deep and gypsum cover board cores shattering under sufficient impact.
Enabling two-layer insulation with offset joints	The build. A cover board is what makes it practical to run the insulation in two layers with staggered board joints and still present a flat, uniform surface to the membrane.	Joints that line up anyway. The offset has to be specified and inspected; the board does not enforce it.

Read this table one item at a time

Impact and puncture resistance

What actually decides it: Board material and thickness, and — as the hail research shows — whether a fastener plate sits directly under the point of impact. A membrane over a plate has nothing to give.
What a cover board cannot fix: Attachment geometry. If the membrane and cover board are fastened through from the top, the roof still has a grid of hard spots under it.

Contribution to the assembly's fire classification

What actually decides it: The listing. A cover board may be the component that makes a specific assembly achieve its Class A, B or C classification under ASTM E108 or UL 790.
What a cover board cannot fix: A substitution. Fire classifications are limited to the materials, products and configurations described in the listing, and swapping a board takes the roof outside it.

A stable substrate for adhering the membrane

What actually decides it: Facer type and surface. Adhesives are formulated to bond to particular facers, and an adhered system is only as good as what it is stuck to.
What a cover board cannot fix: A wet or unsound substrate underneath. Adhesion to a cover board over saturated foam buys nothing.

Protecting the insulation from traffic

What actually decides it: Where people actually walk. Rooftop equipment service routes are the real load case, and walkway pads on those routes matter as much as the board.
What a cover board cannot fix: The absence of walkway pads. Impact work found standard polyiso crushing about half an inch deep and gypsum cover board cores shattering under sufficient impact.

Enabling two-layer insulation with offset joints

What actually decides it: The build. A cover board is what makes it practical to run the insulation in two layers with staggered board joints and still present a flat, uniform surface to the membrane.
What a cover board cannot fix: Joints that line up anyway. The offset has to be specified and inspected; the board does not enforce it.

Cover board materials divide first on combustibility. Wood fibreboard and high-density polyisocyanurate are combustible; perlite, glass-mat-faced gypsum, fibre-reinforced gypsum, mineral fibre and cement boards are noncombustible; asphalt-core boards are described as moisture-resistant. Which of those belongs on a given roof is a question for the listing that assembly is being built to, not a general ranking.

Where the cover board argument is genuinely weak

On a fully adhered assembly in a low-hail region, with no rooftop equipment, a fire classification already satisfied without it, and a membrane manufacturer who will warrant direct application to the insulation facer, a cover board is a real cost for a modest return. Some roofs are that roof. The way to find out is to ask which of the five jobs above is unmet and get the answer in writing, rather than to treat the board as either mandatory or as padding.

The opposite is also worth saying plainly. In a hail region the more consequential decision is usually not which cover board but whether the assembly is fastened through from the top at all. The GAF impact work found TPO of any thickness failing when struck directly above a fastener and plate, and passing in the field of the sheet where no fastener sat beneath — and the authors’ recommendation follows from that: consider fully adhered methods without through fastening in high-frequency hail areas. That is a manufacturer’s laboratory finding rather than a code determination or a field-loss study, and it points at attachment, not at a product. Read it alongside what hail ratings do and do not establish.

Control layers

Where the vapor retarder belongs, and how getting it wrong condenses water inside the roofSection link

This is the decision on the page with the widest gap between how confidently it is made and how well it is understood. There is no universal answer, and the wrong universal answer is expensive.

Air first, vapor second

The two jobs are constantly conflated. An air barrier stops air moving through the assembly; a vapor retarder slows vapor diffusing through a material. Vapor retarder classes are set by permeance measured to ASTM E96 Test Method A: Class I at 0.1 perm or less, Class II above 0.1 and up to 1.0 perm, Class III above 1.0 and up to 10 perms.

The order of importance is not close, and it has been settled for decades. Lstiburek’s summary of the Canadian research that established it is that air leakage was and is more important than vapor diffusion, because the leakage of air from inside a building through constructions — rather than diffusion alone — is often the principal means by which water vapor reaches cold surfaces. On a commercial roof that translates into something concrete: sealing the deck plane, the flutes of a steel deck, the wall-to-roof junction, and every penetration matters more than the perm rating of the sheet you seal them with.

When NRCA says to consider a vapor retarder

NRCA’s criteria for considering a vapor retarder layer in a low-slope assembly are these, and the verb matters:

Climate zones 6A, 7 and 8.
An outside average temperature during the coldest month below 40 °F, combined with an expected interior relative humidity during winter of 45% or more.
A building with a high interior relative humidity — NRCA names swimming pools, museums, and specific manufacturing facilities.
Consistency with the wall: where a vapor retarder is deemed necessary in the exterior wall assembly, NRCA says the designer should consider one in the roof.

Separately, the 2026 NRCA Roofing Manual recommends a vapor retarder rated at a maximum of 0.01 perms over all concrete roof decks in climate zones 2 through 8 — a much broader recommendation than the criteria above, and one that reaches most of the country. If your deck is concrete, ask about it by name.

The failure that follows from copying a cold-climate detail south

A low-perm layer is not a one-way valve. It slows vapor in both directions, which means it also stops the assembly drying through it. Place one at the deck in a building that is air-conditioned most of the year, and the physics reverse: warm humid outdoor air drives inward, and the underside of that retarder — cold, because the conditioned space is beneath it — becomes the first cold surface the moisture can reach. The layer installed to protect the assembly is the surface the water forms on. Lstiburek’s standing objection to interior polyethylene is the same objection: in an air-conditioned building it is a vapor barrier on the wrong side.

The corresponding cold-climate failure is the mirror image. A humidified building with no effective air control at the deck pushes interior air up into the insulation, where it meets a surface below the dew point and condenses inside the foam. Neither failure announces itself. Both show up years later as a moisture survey full of wet areas and a deck that has been corroding quietly.

The general design principle behind both cases is the one the DOE guidance states for flat roofs: keep the assembly able to dry in at least one direction. Its residential low-slope guide accepts a Class II or Class III vapor retarder on the interior side and prohibits a Class I polyethylene sheet there, precisely because the polyethylene would eliminate drying. The same logic governs a commercial compact roof, and it is why a second low-perm layer is a decision rather than a default: an insulation layer with a low-perm sheet above it and another below it has no drying path left at all.

What a defensible answer looks like

Three answers are acceptable when you ask why a vapor retarder is or is not in the design: a named climate-zone criterion, a named interior humidity condition, or a hygrothermal analysis run for this building with its real interior conditions. In borderline cases the third is the only one that settles it, and it is a qualified design professional’s work rather than a contractor’s or this page’s. What is never acceptable is a habit — including a habit this page might be read as endorsing. Vented and correctly designed unvented assemblies are both legitimate; so are roofs with and without a vapor retarder. The variables decide, and the variables are in your building, not in a table.

Before the submittal is approved

The half-hour that is worth more than the extra inch of foamSection link

None of this requires you to be on the roof, and none of it requires a technical background. It requires asking for documents and reading the answers.

Ask for the moisture survey before anything else. If there is not one, the insulation conversation is premature. A condition assessment with a marked plan of wet areas and core-cut photographs is the deliverable, and it is what makes the recover-versus-tear-off decision real rather than rhetorical.
Ask what R-value went into the calculation and on what basis. Labelled LTTR or an in-service figure. Both are defensible; only one is usually disclosed.
Ask how many insulation layers, and whether joints are offset in both directions. This is a line on a drawing and a line in a specification, and it costs almost nothing to require.
Ask which layers are fastened and which are adhered. Fastening the bottom layer and adhering the rest is the single change on this page that improves thermal performance and hail behaviour at the same time.
Ask for the energy code edition, section, and climate zone in writing. Then ask who confirmed it with the authority having jurisdiction, and when.
Ask what the building’s winter interior relative humidity is, and whether it was used. If the contractor does not know the number, nobody used it.
Require every substitution as a written change order. Product removed, product substituted, manufacturer’s written acceptance, and confirmation that the assembly remains inside its fire and wind listings.

If you are comparing more than one proposal, the mechanics of making them say the same thing are covered in normalising scope before price. The insulation section is where two bids most often differ without appearing to.

Considerations

What changes this on a real buildingSection link

The axes below are the ones that actually change this assembly on a real building. Each carries the guardrail that keeps it honest.

Code and jurisdiction: The minimum R-value for a commercial reroof is set by the energy code your jurisdiction has adopted, in the edition it has adopted, as amended, for your climate zone. The International Energy Conservation Code is a model code: it has no force anywhere until a jurisdiction adopts it, on that jurisdiction’s own effective date and usually with amendments, and different states are currently on different editions. In the 2021 IECC, the prescriptive R-value path runs through Table C402.1.3, “Opaque Thermal Envelope Insulation Component Minimum Requirements, R-value Method,” which gives minimum R-values for building components including roof assemblies. Later editions exist and the section numbering in the code your jurisdiction enforces may not match. That edition also offers an alternative compliance option — complying with ANSI/ASHRAE/IESNA 90.1-2019 instead — so “which code” and “which path within it” are two separate questions to put to whoever approves the submittal.
There is no nationwide energy code for site-built construction. Which edition applies at your address, whether it has been amended, when it took effect, and whether your reroof scope triggers it at all are determined by your authority having jurisdiction. This page prints no R-value-by-climate-zone table, because the only correct table is the one your AHJ enforces. Ask for the code edition and section by number, in writing, before the submittal is approved.
Moisture and ventilation: NRCA’s stated criteria for considering a vapor retarder in a low-slope assembly are: climate zones 6A, 7 and 8; or an outside average temperature during the coldest month below 40 °F combined with an expected winter interior relative humidity of 45% or more; or a building with a high interior humidity such as a swimming pool, a museum, or particular manufacturing facilities. NRCA also notes that where a vapor retarder is judged necessary in the wall assembly, the designer should consider one in the roof. Separately, its 2026 manual recommends a maximum 0.01-perm vapor retarder over all concrete roof decks in climate zones 2 through 8.
These are trade recommendations, not code, and they are criteria for considering a vapor retarder rather than a rule for installing one. A low-perm layer placed on the wrong side of the insulation in a cooling-dominated building is not a neutral addition — it is a cold surface that indoor air cannot dry. Where the answer is not obvious, the honest route is a hygrothermal analysis for this building by a qualified design professional, not a rule of thumb.
Fire: Roof assemblies are classified for exterior fire exposure by ASTM E108 or UL 790 as Class A, B or C — severe, moderate and light test exposure respectively. A cover board is frequently the component that makes a particular assembly achieve its classification, and cover board materials differ on this axis: perlite, glass-mat-faced gypsum, fibre-reinforced gypsum, mineral fibre and cement boards are noncombustible, while wood fibreboard and high-density polyiso are combustible.
The classification belongs to the tested assembly, never to a board on its own. In NRCA's words, fire classifications are limited to the specific materials, products and configurations described in a listing, and substitution of materials or products not identified in that listing is not permitted. A cover board substitution approved on price is a substitution out of the listing unless somebody has checked the listing.
Hail and impact: Impact testing of TPO over commercial insulation, published by researchers at GAF, found that membranes of any thickness failed when the impact landed directly above a fastener and plate, while membranes passed in the field of the sheet provided no insulation fastener sat under the point of impact. Beneath the membrane, the same work reported gypsum cover board facers cracking and the board core shattering, and standard polyiso crushing to a depth of about half an inch with extensive facer breakage.
That study was authored by researchers at a membrane and insulation manufacturer and reads as product-specific; its finding that high-density polyiso appeared more resilient than glass-mat-faced gypsum is a manufacturer's comparison of a product category it sells. What survives that caveat is the mechanism: in hail country the attachment method, not the cover board alone, decides where the roof is weakest. Impact classifications belong to tested assemblies, and no rating makes a roof hail proof.
Structural weight: Insulation adds dead load, and a recover adds it on top of load already there. Compressive resistance also matters here, but as a serviceability question rather than a structural one — foot traffic to rooftop equipment is a point load on a board, repeated for decades. ASTM C578 compressive resistance runs 5 to 60 psi across the EPS types and 15 to 100 psi across the XPS types, and for a given density the XPS types are stronger.
Nothing here is a structural determination. Whether your deck and frame can carry an added assembly, and whether the existing roof can stay in place under one, are questions for a licensed structural engineer looking at this building's drawings and this building's condition.
Maintenance: The insulation is the layer you cannot inspect. Everything you will ever learn about it comes from a moisture survey, a core cut, or a leak. That is an argument for surveying before a recover rather than after, and for walkway pads on the routes service technicians actually take rather than the routes the drawings assume.

Warranty and repair

What is covered, and what can actually be repairedSection link

Insulation and cover board sit inside the warranty in a way owners routinely discover only at claim time.

The system warranty covers what was installed, as listed: A manufacturer’s system warranty is written around a specific assembly. Fire and wind classifications work the same way: a listing describes particular materials, products and configurations, and substituting something not named in it takes the roof outside the listing. A cover board changed in the field, an insulation brand swapped for availability, or a fastening pattern altered to save a day are all substitutions. Whether they are permitted substitutions is a question with a written answer, and the time to get it is before the work, not after the storm.
Value engineering is a warranty event: The most common route to an uncovered roof is not a defect. It is a sequence in which a price is agreed, a substitution is proposed to hold that price, the substitution is accepted verbally, and nobody revises the submittal. Require every substitution as a written change order naming the product being removed, the product replacing it, and the manufacturer’s written acceptance.
Wet insulation and the maintenance condition: Warranties carry maintenance and inspection obligations, and they exclude conditions the owner knew about. A moisture survey that finds saturated insulation creates knowledge. Acting on it protects the roof; filing it does the opposite. See how roofing warranties are actually structured before relying on one.

Repairability

Insulation is repairable only by opening the roof. A localised wet area can be cut out and replaced in kind, which is routine work when it is found early and caught within one or two board widths. What is not repairable is a field-wide saturation discovered at year twelve: at that point the choice is a tear-off, and the money spent on a recover that buried the problem is gone. This is the argument for a condition assessment with an actual moisture survey before a recover-versus-tear-off decision, not after it.

A warranty is a contract between a reader and whoever wrote it. What it covers, what voids it, whether it transfers, and how it is enforced are set by that document and by the law where the reader lives. Read the actual warranty for the product and the installer in front of you — not a summary of one, including this one.

Ask before you sign

Questions to ask an installerSection link

These are submittal-review questions. Every one of them has a documentary answer, and a contractor who cannot produce it is telling you something.

What R-value did you use in the calculation — the labelled LTTR, or an in-service value?
A good answer names the number and the basis, and knows that NRCA recommends designing at 5 per inch for polyiso rather than at the labelled LTTR. A weak answer is the total from the submittal with no idea where it came from. This is also the question that reveals whether anybody has thought about the assembly at all.
Is the insulation going on in two layers with offset joints, or one thick layer?
One layer is faster and cheaper. It also puts every board joint on a single continuous plane from cover board to deck, which is the condition the NRC Canada testing found costs roughly triple what a staggered joint costs. If the answer is one layer, ask what is being bought with the saving.
Which layers are mechanically fastened and which are adhered, and how many fasteners per board?
Fastener density and location are a thermal decision and a hail decision at the same time. NRCA-reported research put through-fastener thermal losses at roughly 5% to 15% depending on fastener density and insulation R-value, tested at four, six and ten fasteners per four-by-four-foot board.
What cover board is specified, in what thickness, and is it in the fire and wind listings for this assembly?
“A cover board” is not a specification. Material, thickness, combustibility, and its presence in the specific listing are four separate facts, and only the listing settles whether the assembly still classifies.
Is a vapor retarder proposed, where does it sit, and what is the basis for that decision?
Three answers are acceptable: a stated climate-zone criterion, a stated interior humidity condition, or a hygrothermal analysis. One answer is not: “we always do it that way.” A low-perm layer in the wrong place is worse than none.
What is our humidified winter interior relative humidity, and did you use it?
This is a question about your building that your contractor cannot answer without asking you. If nobody asked, nobody modelled. On a pool, a plant, or a printing operation, this single number can reverse the design.
Has a moisture survey been done, and can I see the plan showing where the wet areas are?
Adding insulation over wet insulation is the most expensive mistake available on this page, because it is invisible for years and terminal when found. A survey plan with marked areas, core-cut locations and photographs is the deliverable; a verbal “it looked dry” is not.
Which energy code edition and section are we complying with, and who confirmed it with the AHJ?
A named edition and section number is a checkable answer. “It meets code” is not. The reroof may also trigger requirements the original roof never had.

Require these in writing

Insulation type by product standard and, where applicable, type and grade — not just a brand name
Thickness of each layer, and the number of layers, stated separately
Whether joints are offset, and in both directions
The design R-value used in the calculation, and whether it is a labelled LTTR or an in-service value
Cover board material, thickness, and the listing it appears in
Attachment method for each layer, with fastener type and density per board
Vapor retarder product and perm rating, its position in the assembly, and the stated basis for including or omitting it
The energy code edition, section, and climate zone the design is complying with
The moisture survey report, with the plan of wet areas and the core-cut photographs
A written statement that every product listed appears in the fire and wind listings claimed for the assembly

What goes wrong

Misconceptions and failure modesSection link

These are the beliefs that survive because the layer they are about is buried the day it goes in.

Common misconceptions

Common belief
The submittal says R-30, so the roof is R-30.
What is actually true
R-30 is what the boards were labelled. It is not what the assembly delivers, and the gap is not small — the worked example above puts it at roughly a fifth to a third on a plausible mechanically attached build. Nothing about that is a defect or a scandal. It is the ordinary difference between a laboratory value for a material and the performance of a roof with joints and fasteners in it, and the point of knowing it is to make the right choices about layers and attachment while they are still choices.
Common belief
Polyiso has the highest R-value per inch, so it is always the best choice.
What is actually true
Polyiso does carry the highest labelled R per inch of the three foams here — a typical minimum LTTR of 5.6 per inch for 1-inch product, rising to 5.9 per inch for 4-inch product, against an ASTM C578 minimum of 5.0 per inch for the XPS roofing types at the same 75 °F condition — and on most commercial roofs it is the right choice. But its R-value falls as the mean temperature drops below the 75 °F condition its label is measured at, which means it is furthest from its label on the coldest nights of the year. Polystyrene moves the other way: ASTM C578 specifies higher minimum R-values per inch for the XPS types at 25 °F and 40 °F than at 75 °F. That does not make XPS the answer — it makes the honest comparison one you have to do at the temperature that matters for the building, not at the label temperature.
Common belief
A vapor barrier is always a good idea. More is safer.
What is actually true
A vapor retarder is a control layer, and control layers are directional. A low-perm layer stops moisture crossing it in both directions, which means it also stops an assembly drying through it. Lstiburek’s standing warning is about exactly this: an interior polyethylene sheet becomes a vapor barrier on the wrong side of the assembly in an air-conditioned building. The building-science instruction is not “add a vapor barrier” but “avoid trapping the assembly between two of them and leave it able to dry in at least one direction.”
Common belief
The cover board is an upsell.
What is actually true
It is sometimes optional and it is never decorative. It does at least five separable jobs — impact and puncture resistance, contribution to the assembly’s fire classification, a stable substrate for adhering the membrane, protection of the foam from traffic, and the practical enabling of two-layer insulation with offset joints. Deleting it may be defensible on a specific roof. Deleting it as a line-item saving without naming which of those five jobs is now unmet is not.
Common belief
Wet insulation will dry out.
What is actually true
Above a deck, sealed under a membrane, with a cover board over it and often a vapor retarder under it, there is no drying path worth the name. This is the deliberate consequence of a well-built compact roof: it is designed not to let vapor through. That is a virtue while the water stays out and a trap once it is in, which is the whole reason a moisture survey precedes a recover rather than following it.
Common belief
Both roofs got the same insulation, so they will perform the same.
What is actually true
Same boards, different assemblies. One fastened through from the top at ten fasteners per board with joints that line up in places; one with the bottom layer fastened, the top layer adhered, and joints offset in both directions. Those are the same materials and measurably different roofs, and the difference is set by decisions made in a submittal review that took an hour.

How it actually fails

Condensation on the underside of a vapor retarder in a cooling-dominated building: A low-perm layer placed at the deck in a building that is air-conditioned for most of the year. Warm humid outdoor air drives inward, reaches the underside of the retarder — which is cold, because the conditioned space is below it — and condenses there. The layer that was installed to protect the assembly is the surface the water forms on, and it also prevents the assembly drying downward.What you can see: Staining or drips at the deck in summer rather than winter. Rust streaks on a steel deck with no leak above. A pattern that follows cooling season rather than storms.
Winter condensation in the insulation of a humidified building with no air control: Interior air at 45% or higher relative humidity moving into the assembly through unsealed penetrations, deck flutes, or wall-to-roof junctions, and reaching a surface below the dew point. Air leakage carries far more moisture than diffusion does, which is why sealing the ceiling and deck plane outranks arguing about perm ratings.What you can see: Ceiling stains that appear in cold weather and stop in spring. Frost on the underside of a metal deck. Core cuts that come up wet at the bottom of the insulation rather than the top.
Thermal short circuit through fasteners and aligned board joints: Metal fasteners with plates conducting from deck to just under the membrane, combined with joints in a single layer or through joints in two. NRCA-reported research put through-fastener losses at about 5% to 15% and board-joint losses at 2% to 7% for quarter-inch gaps and 4% to 9% for half-inch gaps.What you can see: Regular grid patterns of melted snow or frost on the roof surface, or a matching grid on an infrared survey. Energy bills that never quite match the modelled improvement after a reroof.
Hail puncture directly over a fastener plate: A membrane over a fastener plate has no give: the plate is a hard, unyielding backer, so impact energy that would deform the foam elsewhere goes into shearing the membrane instead. Manufacturer researchers found TPO of any thickness failed at impacts above a fastener and plate, while passing in the field of the sheet where no fastener sat beneath.What you can see: Punctures on a regular grid rather than randomly scattered. Leaks after hail that show a spacing matching the fastener pattern.
Crushed foam and a shattered cover board under service traffic: Repeated point loading from technicians servicing rooftop equipment, on routes that were never given walkway pads. Impact work found standard polyiso crushing to a depth of about half an inch and glass-mat-faced gypsum cover board cores shattering under sufficient impact.What you can see: A visible path of depressions from a roof hatch to a unit. A crunching underfoot that was not there last year. Membrane splits that follow a walking route.
A recover installed over saturated insulation: A moisture survey skipped, or run and disregarded, before adding a layer. The water has nowhere to go, the wet foam has lost the R-value being paid for, and on a steel deck the corrosion continues out of sight under a new warranty.What you can see: Nothing, for years. Then a fastener that spins in the deck, a core cut that comes up dark and heavy, and a capital plan that has to be rewritten.

Sources and further readingSection link

Understanding Roofing / Published August 26, 2026

Scope and limitations

It cannot tell you the minimum R-value required at your address.
That is set by the energy code your jurisdiction has adopted, in the edition it adopted, as amended, for your climate zone, and by whether your scope of work triggers it.
In the 2021 IECC — a model code, not law anywhere until adopted — the prescriptive table for roof assemblies is Table C402.1.3; later editions have been published and the numbering in your adopted code may differ.
This page could not open the adopted text for any jurisdiction, or the model text itself (codes.iccsafe.org refused automated access), and therefore publishes no values from that table and no description of its individual rows.
It does not publish a cost figure.
Above-deck insulation is priced by thickness, board type, attachment, deck condition, tapered layout, staging, and local labour, and no defensible national dataset separates that from the rest of a low-slope reroof.
A per-square-foot number here would be decoration.
It does not publish a compressive-resistance or water-absorption figure for polyisocyanurate.
Those exist in ASTM C1289, but this page did not open C1289 or a source that quotes it, and a number without a verified test condition is exactly what this page argues against.
It does not publish minimum R-value-per-inch figures for EPS by type and temperature.
The IIBEC paper cited below gives that data for the XPS types and this page could not obtain the EPS equivalents from a source it read.
The EPS row of the temperature table is therefore absent rather than estimated.
It cannot tell you whether your building needs a vapor retarder.
That depends on your climate, your interior humidity in winter, your deck, and your assembly, and in the cases that matter it is settled by a hygrothermal analysis for that building rather than by a criterion list.
It cannot tell you whether your deck and structure can carry an added assembly.
That is a licensed structural engineer's determination for this building.
It does not evaluate spray polyurethane foam, protected-membrane assemblies, vegetative or ballasted roofs, or tapered insulation design, all of which change the rules on this page in ways that need their own treatment.

Tech Today — polyisocyanurate R-value testing at four mean temperatures
Mark S. Graham, Professional Roofing (National Roofing Contractors Association) / 1 March 2015
That NRCA had seven specimens of newly manufactured 2-inch-thick polyisocyanurate from six U.S. manufacturers, aged three to nineteen months, tested by R&D Services Inc. of Cookeville, Tennessee using ASTM C518, at mean temperatures of 25 F, 40 F, 75 F and 110 F; that the average of the 25 F R-values was 4.049 against 4.744 in 2009, and the average at 40 F was 4.905 against 5.39 in 2009; that review of the 75 F data reveals the average of the results are less than the products' published LTTR values; and NRCA's then-current recommendation of an in-service R-value of 5.0 per inch thickness in heating conditions and 5.6 per inch in cooling conditions.
Trade technical guidance, not adopted code, and a snapshot of product then on the market rather than a standing property of every polyiso board. The article states the averages numerically without repeating the unit in the same sentence; this page reads them as R-value per inch because the 2009 comparison values and the surrounding NRCA recommendations in the same article are per inch of thickness. No average was published for the 110 F results, and this page therefore prints none.
Specifying specifics
Mark S. Graham, Professional Roofing (National Roofing Contractors Association) / 1 March 2020
That minimum LTTR values are typically 5.6 per inch thickness for 1-inch-thick polyisocyanurate products, 5.7 per inch for 2-inch-thick, 5.8 per inch for 3-inch-thick and 5.9 per inch for 4-inch-thick products; that NRCA recommends designers determine insulation R-value using an in-service R-value of 5 per inch thickness; that this in-service approach accounts for known losses in polyisocyanurate insulation's R-value over time and in changing temperature conditions; and that NRCA recommends polyisocyanurate be specified by thickness rather than by LTTR or R-value.
A trade association's design recommendation, not a code requirement and not a manufacturer's warranty position. This page did not open the 2026 edition of the NRCA Roofing Manual and cannot confirm the recommendation's current wording there.
Tech Today — NRCA revises its polyisocyanurate R-value recommendation
Mark S. Graham, Professional Roofing (National Roofing Contractors Association) / 1 June 2016
That in January 2016 NRCA revised its recommendation to an in-service R-value of 5.0 per inch for polyisocyanurate, published as an interim update to the NRCA Roofing Manual, and that the recommendation rests on R-value testing taking into account real-world conditioning during shipment and storage including exposure to changing temperature and humidity.
Records the change and its date. It does not itself set out the temperature data behind it.
Info-502: Temperature Dependence of R-values in Polyisocyanurate Roof Insulation
Building Science Corporation / 12 April 2013
That the thermal performance of polyisocyanurate decreases as mean temperatures deviate from 75 F, the mean temperature used for label R-value tests; that testing at mean temperatures of 25 F, 40 F, 75 F and 110 F showed all samples decreasing in R-value as outside temperatures go below freezing; and the mechanism — condensation of the gases trapped in the cells during manufacture, where cell walls coated in a highly conductive condensate increase heat transfer and lower the R-value.
A building-science information sheet describing a set of tested samples. It contains no comparison with XPS or EPS, and this page draws none from it.
New LTTR Values and What They Mean for Roofing Industry Professionals
Michael Cusick, Lance Wang and Christopher Griffin (Johns Manville), Roofing Contractor / 7 February 2014
That LTTR is a fifteen-year time-weighted average of the foam's R-value; that ASTM C1289 identifies two methods for calculating it, ASTM C1303 and CAN/ULC-S770, both using slicing and scaling to accelerate ageing and predict R-value after five years as a proxy for fifteen-year performance; that the design R-value for permeable-faced polyiso moved from 6.0 to 5.7 per inch with the change from S770-03 to S770-09; and that PIMA's QualityMark certification began reporting the new values on 1 January 2014.
Written by researchers at an insulation and membrane manufacturer and published in trade press. It is used here only for the description of the test standards and the dated change in reported values, not for any performance claim.
Considerations for Specifying Rigid, Cellular Polystyrene Insulations in Various Applications
Rob Brooks, Tiffany Coppock, Matt Dillon, Mike Fischer, Meng Guo and Valentina Woodcraft, IIBEC Interface / February 2024
Which ASTM C578 types are typically EPS and which are typically XPS; that compressive resistance ranges from 5.0 to 60.0 psi for the EPS types and 15.0 to 100.0 psi for the XPS types, and that for a given density XPS types are stronger; that maximum water absorption values of 2%, 3% or 4% by volume are seen for EPS types against a maximum of 0.3% by volume for the XPS roofing types; that depending on type, EPS can absorb seven to ten times as much water as XPS and still meet the standard; that its water absorption rate is one of the main reasons EPS is unsuitable for protected-membrane roofing assemblies; that in general the R-value of rigid cellular polystyrene decreases as temperature increases, with R-values consistently higher at 25 F and 40 F than at 75 F and 110 F; that five XPS types are specified in ASTM C578 to have the same minimum R-values per inch, plateauing at 5.60, 5.4, 5.0 and 4.65 at 25 F, 40 F, 75 F and 110 F; that XPS R-value typically decreases by a predictable 1% to 2% over a product life often spanning several decades; and that moisture absorption measured on samples retrieved from below-grade applications after several years is often much higher than the short-term maximum values specified in ASTM C578.
A technical article in a building-enclosure institute's journal, several of whose authors work for insulation manufacturers. The R-value figures quoted are ASTM C578 specification minimums for the named types, not measurements of any particular product, and this page labels them as such. The paper does not give the equivalent minimum R-values for the EPS types, so this page publishes none.
Tech Today — polystyrene insulation types for roofing
Mark S. Graham, Professional Roofing (National Roofing Contractors Association) / 1 May 2015
That ASTM C578 identifies EPS products as Types I, II, VIII, IX, XI, XIV and XV and XPS products as Types IV, V, VI, VII, X, XII and XIII; that EPS Type XI and XPS Types XII and XIII are generally not intended for roofing applications; that NRCA recommends EPS used as rigid board roof insulation have a minimum nominal density of 1.25 pounds per cubic foot, such as ASTM C578 Type VIII; and that NRCA recommends XPS used as rigid board roof insulation have a minimum compressive strength of 15 psi, which complies with ASTM C578 Type X.
Trade recommendations rather than code, and they do not address R-value, moisture behaviour, or temperature limits, none of which this page draws from this source.
Mind the gap
Sudhakar Molleti, Ph.D., National Research Council Canada, in Professional Roofing / 1 September 2020
That NRC Canada runs an industry consortium, Energy Resistance of Commercial Roofs, whose partners include NRCA; that in more than 75 experiments the presence of gaps reduced effective thermal resistance of R-31 assemblies by 3% to 6% depending on gap features; that quarter-inch gaps across insulation heights of 2 to 9.5 inches lowered overall thermal performance by 2.4% to 6.9% and half-inch gaps produced a thermal bypass of 3.5% to 9.3%; that a through gap in single-layer 3⅓-inch insulation cost about 9% of the effective R-value, rising to 13.5% at a half-inch gap width; that through gaps result in nearly triple the R-value loss of a staggered gap; and that assemblies with a 6-inch gap offset averaged 3% lower effective R-value than assemblies with a 24-inch offset.
Laboratory results for specific tested configurations under a defined research programme, not a prediction for any particular roof. The percentages describe the effect of gaps alone and do not include fastener effects.
New roofing research
Mark S. Graham, Professional Roofing (National Roofing Contractors Association) / 1 March 2024
Reporting Phase II of NRC Canada's Energy Rating of Commercial Roofs research: that average losses in effective R-value from insulation board gaps ranged from 2% to 7% for quarter-inch-wide gaps and 4% to 9% for half-inch-wide gaps; that through-fastener thermal bridging produced losses from about 5% up to 15% depending on fastener density and insulation R-value, tested at densities of four, six and ten fasteners per four-by-four-foot board across R-26, R-31 and R-36 thicknesses; that photovoltaic mount penetrations produced thermal losses up to 50% depending on mount type and penetration frequency; and that the overall effect of several combined thermal losses on a roof assembly's thermal performance can be significant.
A summary of conference research reported by a trade association. The percentages are results for tested configurations, and no source read for this page measured board-joint loss, fastener loss and material in-service loss together on one assembly — which is why the worked example on this page is presented as an estimate with its arithmetic shown, not as a measurement.
Covering your options
Jason Wilen, AIA, NCARB, CDT, RRO, Professional Roofing / 1 May 2014
That cover boards are flat or tapered stock materials ranging from 2 by 4 feet to 4 by 12 feet with thicknesses from one-eighth of an inch to one inch depending on material composition; that benefits include increased roof membrane impact and puncture resistance and that a cover board may be required to achieve a fire-resistance classification for a roof assembly; that it allows installation of insulation board layers with staggered board joints and can improve a roof assembly's overall thermal performance; and the eight material families and their combustibility — wood fibreboard (combustible), perlite, glass-mat-faced gypsum, fibre-reinforced gypsum, mineral fibre and cement (noncombustible), high-density polyisocyanurate (combustible), and asphalt core (moisture-resistant).
A survey article describing the product category. It gives no wind-uplift numbers and no hail-classification values, and this page draws none from it.
A lasting impression
Thomas J. Taylor, Ph.D., Sarang Bhawalkar, Ph.D. and Tammy Yang, Ph.D. (GAF), Professional Roofing / 1 June 2016
That when impacted above a metal fastener with an appropriate plate, TPO membranes of any thickness always failed with a rating of 3 or worse; that all membranes passed in the sheet's field provided there was no insulation fastener underneath the point of impact; that beneath the membrane, standard polyisocyanurate suffered extensive facer breakage and crushing of the foam to a depth of about half an inch; that gypsum cover board facers suffer large cracks and the board core shatters; and the authors' recommendation that fully adhered methods of adhesion without through fastening should be considered when installing single-ply roof systems in high-frequency hail areas.
Manufacturer research, published by authors at a membrane and insulation manufacturer, and product-specific. Its finding that high-density polyisocyanurate appeared more resilient than glass-mat-faced gypsum board is a manufacturer's comparison of a product category it sells and is not repeated on this page as a recommendation. It is a laboratory impact study, not a hail classification and not a field-loss dataset.
Tech Today — roof assembly fire classifications
Mark S. Graham, Professional Roofing (National Roofing Contractors Association) / 1 April 2015
That roof assembly resistance to exterior fire exposure is tested and classified using ASTM E108 and UL 790; that Class A designates resistance to relatively severe fire-test exposure, Class B to relatively moderate and Class C to relatively light; and that fire classifications are limited to the specific materials, products and configurations described in a listing, with substitution of materials or products not specifically identified in a listing not permitted.
Explains the test methods and the listing rule. Which classification a building requires is set by the building code the jurisdiction has adopted, and this page makes no determination about any building.
Do you need a vapor retarder?
Mark S. Graham, Professional Roofing (National Roofing Contractors Association) / 1 October 2021
NRCA's stated criteria for considering a vapor retarder layer in a low-slope roof assembly: climate zones 6A, 7 and 8; an outside average temperature during the coldest month below 40 F combined with an expected interior relative humidity during winter of 45% or more; and buildings with high interior relative humidity such as swimming pools, museums and specific manufacturing facilities. Also that if a vapor retarder layer is deemed necessary for a building's exterior wall assembly, the designer should consider one as a roof system component.
Criteria for considering a vapor retarder, not a requirement to install one, and not code anywhere. The article directs readers to Chapter 2 of the NRCA Roofing Manual for placement detail, which this page did not open.
Changing guidance
Maciek Rupar, Professional Roofing (National Roofing Contractors Association) / 1 February 2026
Changes in the 2026 edition of the NRCA Roofing Manual: that NRCA recommends the use of a vapor retarder rated at maximum 0.01 perms over all concrete roof decks in climate zones 2 through 8; that low-slope assemblies with below-deck insulation should include continuous above-deck insulation of R-5 to R-10 for condensation protection; that design drawings should specify exact locations of joints, interconnections, penetrations and termination point heights; that liquid-applied vapor retarders exhibit a wide range of vapor permeability ratings depending on product and tend to require a high level of application skill; and that the manual incorporates an updated U.S. climate zone map sourced from ASHRAE 169.
A summary of a trade association's own manual revisions. It is guidance, not adopted code, and this page did not open the manual itself.
BSD-106: Understanding Vapor Barriers
Joseph Lstiburek, Building Science Corporation / 15 April 2011
The vapor retarder classes measured by ASTM E96 Test Method A — Class I at 0.1 perm or less, Class II greater than 0.1 and up to 1.0 perm, Class III greater than 1.0 and up to 10 perms; that vapor barriers installed on the interior of assemblies prevent assemblies drying inward, which creates problems in air-conditioned spaces; and the instruction to avoid double vapor barriers so that an assembly retains drying potential in at least one direction.
A building-science digest written largely around walls and residential assemblies. Its class definitions are general; its zone-by-zone tables are not a substitute for the adopted code or for an analysis of a specific commercial building.
BSI-073: Macbeth Does Vapor Barriers
Joseph Lstiburek, Building Science Corporation / 2013
That air leakage was and is more important than vapor diffusion; that research established the leakage of air from inside a building through constructions, rather than vapor diffusion alone, was often the principal means by which water vapor moved to cold surfaces; and that an interior polyethylene sheet carries a large liability in air-conditioned buildings because it results in a vapor barrier on the wrong side of the assembly.
An opinion and history article by a named building scientist. It argues a hierarchy — air control before vapor control — rather than giving a quantified comparison, and this page presents it that way.
Water Managed Roof — Re-roofing and Adding Insulation over a Flat Roof
U.S. Department of Energy, Building America Solution Center (PNNL) / Last updated 28 December 2015
That when installing multiple layers of rigid foam, the seams are offset in two directions with a continuous bead of sealant at the roof perimeter between each layer and the seams of each layer taped; that an insulation cover board, typically fibreglass-faced gypsum board, is installed over the rigid foam; and that on the interior side a Class II or Class III vapor retarder is acceptable while a Class I polyethylene vapor barrier is prohibited because it would eliminate drying of the assembly.
Best-practice guidance written for residential and light-commercial flat roofs, not adopted law, and its code references are to specific IRC editions that are model provisions rather than the law anywhere. Its detailing recommendations transfer to commercial work as principles; its assembly is not a commercial specification.
Low-Slope (Flat) Roofs
U.S. Department of Energy, Building America Solution Center (PNNL) / Last updated 18 September 2020
That rigid insulation above the deck should be limited to a maximum thickness of 10 inches and installed using multiple 2-inch-thick sheets with joints offset horizontally and vertically; and that tapes for sealing the joints in rigid insulation should be acrylic-based. Only the offset-and-seal practice is drawn from this source on this page.
Guidance for low-slope roofs at residential and light-commercial scale. It is not a commercial specification, not code, and it does not address vapor retarder perm ratings.
29 CFR 1910.28 — Duty to have fall protection and falling object protection, paragraph (b)(13), work on low-slope roofs
U.S. Occupational Safety and Health Administration
That where work is performed less than 6 feet from a roof edge, the employer must ensure each employee is protected from falling by a guardrail system, safety net system, travel restraint system or personal fall arrest system; that the same range of systems applies at least 6 feet but less than 15 feet from the edge, where a designated area may be used for work that is both infrequent and temporary; and that at 15 feet or more the employer must protect each employee by one of those systems or a designated area, or is not required to provide fall protection where the work is both infrequent and temporary and the employer implements and enforces a work rule prohibiting employees from going within 15 feet of the edge without fall protection.
A federal general-industry occupational safety standard binding on employers, including building owners with their own employees on a roof. It is not homeowner or occupant guidance, and construction work is governed by a separate set of standards in 29 CFR 1926. It says nothing about roof assemblies.
Polyiso possibilities
Marcin Pazera, Ph.D., senior technical director, Polyisocyanurate Insulation Manufacturers Association, in Professional Roofing / 1 June 2024
The build used as the starting point of the worked example on this page: PIMA's statement that in climate zones 4, 5 and 6, two layers of 2.6-inch-thick polyisocyanurate insulation deliver a total R-value of 30, alongside its comparable statements of two 2.3-inch layers for R-25 in zones 2 and 3 and two 3.1-inch layers for R-36 in zones 7 and 8.
Written by the technical director of the trade association of polyisocyanurate manufacturers, which is not a neutral source on polyisocyanurate. It is used here only as the source of a stated example build, never as evidence of what any jurisdiction requires. Whether that build satisfies the energy code at a given address is determined by the adopted code and the authority having jurisdiction.
A compendium of changes: Part two
Mark S. Graham, Professional Roofing (National Roofing Contractors Association) / 1 November 2021
That in the 2021 IECC, Table C402.1.3, 'Opaque Thermal Envelope Insulation Component Minimum Requirements, R-value Method,' provides minimum R-values for building components including roof assemblies under the prescriptive compliance option; that an alternative compliance option is also available, namely complying with ANSI/ASHRAE/IESNA 90.1-19; and that requirements in that edition increased in named categories — R-49 in climate zones 4 and 5 and R-60 in zones 7 and 8 for the attic and other category.
Establishes that the prescriptive roof R-value table exists and where it sits in the 2021 IECC. It is a trade summary, not the code text, and this page publishes no values from Table C402.1.3 for the insulation-entirely-above-deck category because it did not open the adopted text of any jurisdiction. The IECC is a model code with no force until a jurisdiction adopts it, on that jurisdiction's own effective date and usually with amendments.

Suggest a correction