Back to Case Studies

Part 1 ] [ Part 2 ] Part 3 ] Conclusion ]

Case Study

Roof Ice Dam Remediation for Northeast Ski-Area Condominiums

Part II

by Henri C. Fennell
President, H. C. Fennell, Inc. & FOAM-TECH
North Thetford, Vermont, USA.


Phase I Diagnostics

Four of the oldest units with full-time residents (non-rental units) were selected to maximize project access. The four dwellings comprised all of the units in one of the typical four-family clusters. The four-unit cluster chosen provided two "end units" and two "middle units" to evaluate and serve as baselines for comparison with other unmodified units. (The only other type of cluster layout was comprised of three similar units - two end units and one middle unit).

Once the test units were established, an experienced local contractor was selected and the process was implemented. The video and photographic evidence was presented by the author to the owners, the contractor, the architect, and the property manager - hereinafter referred to as the team. From this review, the most severe areas of interior heat loss or roof surface warming were located. Architectural drawings were examined and plans were made for the contractor to develop access to inaccessible areas. The contractor then made inspection openings in soffits, chases, attics, and dropped ceiling areas to allow visual identification of the actual construction details at the areas of significant heat loss. These problem areas were organized into the following four categories:

  1. General energy conservation problems not related to ice dam formation such as missing insulation, unsheathed wall areas, or major air-leakage sites.

  2. Pipe freeze-up causes such as un-insulated rim joists, unsealed penetrations, or second floor overhanging floor areas.

  3. Ice dam formation causes – cathedral roof slope warming.

  4. Ice dam formation causes – attic roof slope warming.

While remedying many of the energy conservation and freeze-up problems (Items 1 and 2 above) was included in the work and resulted in fuel savings and reduced repairs, only the ice dam formation causes (Items 3 and 4 above) will be addressed in this paper.

Cathedral Roof Slope Warming – Problems Identified in Phase I

In this project, "cathedral" roofs are areas where both the inside and outside of a sloped roof are finished surfaces (no attic space). These areas were typically at the lowest roof elevation and always had attics above them.

The cathedral roof construction, from the inside to the outside, consisted of gypsum board or horizontal tongue and groove wood paneling, 4 mil polyethylene vapor retarder, 2 x 12 wood rafters at 24" O.C., 9" un-faced batt insulation (R=30), continuous molded polystyrene vent chutes, plywood sheathing, and asphalt shingles on roof felt. The soffits were narrow in most locations and had continuous ventilation strips in the bottom trim. The ventilation design included a ventilated ridge cap.

Most of the ice dams were located along the lower vented cathedral ceiling eaves and behind the zero-clearance fireplace boxes. Ice was concentrated at the valleys where the runoff from both roofs combines. In spite of the ice build-up and water infiltration, these lower roof areas were performing the best thermally. It was observed that the outside surfaces of the cathedral roof areas were generally cooler than the attic roofs above. This was indicated on the outside by snow and frost patterns. Typically, after light snow falls (6" or less) or on frosty spring or fall mornings, the upper attic roofs were melted bare. Under the same conditions the lower cathedral slopes were covered with snow or frost, showing only "spot" melt areas related to specific interior conditions. These patterns were constant from bottom to top on cathedral spans as long as twenty feet, stopping abruptly at a line where the vented attics began (see Detail / Photo Essay # 4).

The snow "melt" water flowing down from the warmer attic roofs froze when it reached the cooler cathedral roof creating ice dam formations and water infiltration (leakage) damage. In similar unmodified units, the property managers were forced to remove snow from the entire roof area to stop leakage as the snow depths increased. This revealed ice build-up on the roof surface extending all the way up the roof from the eaves to the attic floor height, indicating that the entire cathedral roof surfaces, not just the eaves, were cool enough to re-freeze the melt water from the upper (attic) roof surfaces Interior infrared analysis of cathedral roof slope areas during the Phase I testing revealed the following:

Snow Melt Pattern on end units showing melt on attic roofs above cathedral slopes - foreground and background units same pattern.

Snow melt pattern on middle units showing melt on attic roofs above cathedral slopes - melt spot at party wall, behind chimney, and at bath vent.

Roof with snow removed showing ice on entire cathedral area and ice dams along eave edge. Cathedral roof area showing melt above skylight rough opening leakage and above fireplace vents.

Detail of beam at top of cathedral slopes. Frost melt pattern showing melt line along beam at top of cathedral slope.
  1. There was insulation in the slopes, but there was wind washing in the lower eave area, i.e., the cold air from the eave soffit vents is coming in under and through the lower slope batt insulation. As the outside air travels up the roof on the warm inner side of the insulation and through the insulation, it picks up heat and exits into vent space and eventually into the attics above. Interior infrared images showed a gradual temperature gradient from bottom to top in wind-washing cathedral roof slope areas. (This pattern was not visible after the rafter tails had been blocked and sealed except along the un-insulated gable walls).

  2. The scans also showed cold air infiltration along the lower wall to ceiling connection of the gable (end units) or party walls (middle units), skylight rough openings, and at the zero-clearance fireplace box walls . This indicated that the air barriers in these areas were incomplete or unsealed (3). As with the wind washing in the main cathedral areas, this was coldest at the bottom of the slope, indicating the ex-filtration of warm interior air into the roof cavities at the higher ceiling (stack-effect) elevations. These theories were verified by the melt patterns above.

Plan of end unit fireplace showing unsheathed and unsealed wall insulation. Section at fireplace box showing warm air paths to roof vent space.

In the two end units there are bathrooms with knee walls and dropped ceilings. Between the knee walls and the dropped ceilings are short cathedral slopes. Melt patterns showed that this short cathedral area was performing better than the attic roofs around it, except in the bay with a recessed fan/light fixture.

Melt pattern showing location of bathroom fan/light and heat flow in vent space above.

Infrared scan of fan/light prior to retrofit.

Recessed fan/light after being relocated to a surface-mount position.

Interior infrared analysis and physical inspections of these areas during the Phase I testing showed the following:

  1. The fan / lights in the Guest Bathroom (including a shower) slopes were a thermal disaster for the following reasons:

  2. The depth of the enclosure occupied the insulation space.

  3. The enclosure box and its connection to the gypsum ceiling (the air barrier in this case) were unsealed, leaking warm, moist bathroom air directly up into the vent space and insulation above. Cool air from the vent space and wind washing below the light leaked into the bathroom around the fixture trim.

  4. The 4" un-insulated flexible duct from the fan disrupted the insulation all the way up to the attic, and the warm room air flowed unrestricted up this duct against the roof sheathing. The long duct runs that vent to an exit point in the soffit on the other eave side of the building were exposed, thus dissipating any remaining heat into the attic space. The un-dampered termination cap was installed facing down in the eave soffit on the windward side of the building, encouraging the exhaust air to flow back up into the adjacent soffit vent strip and into the attic. This flow was visible during sub-zero weather.

    The light fixture had a high-wattage incandescent bulb that created a hot spot where there was little or no roof insulation. The infrared tests clearly showed the infiltration of cold air into the fixture, the outside melt pattern showed the fixture location and the warm air travel path in the roof’s vent space above the fan.

  5. The attics behind the bathroom knee walls and the boxed-down soffit walls were unsheathed on the attic (cold) side.

Cold air bypass below roof insulation at strapping above unsheathed knee wall bay. Frost pattern above entry at bathroom knee wall attic.

Original knee wall insulation without sheathing - none in floor thickness.

Knee wall insulation after air barrier sheathing retrofit.

Section showing unsheathed knee wall attic over entry - fan/light usually in skylight location.

Insulation above knee wall open to wind washing through and under batts.
Section at top of party wall in cricket attic. Typical existing soffit bays including un-insulated narrow skylight double rafter bays - note inaccessible bays at top of fireplace b

Unsheathed wall inside soffit.

Typical soffit bay after blocking and sealant

Open soffit and rafters - before. Soffit walls and blocked rafters - after.
Un-insulated rake bay over gable end wall before retrofit of insulation and air sealing.

These attic-type roofs melted sooner than the surrounding cathedral slopes. Some were un-insulated or the insulation had fallen out of the stud bays. There was no air barrier in the plane of the knee walls between the floor and ceiling below. There were un-insulated recessed fixtures in the floor of this attic. While none of these surfaces were in the roof slopes themselves, they were assumed to be the causes of the warmed roof surfaces in these attics and in the cathedral slopes directly above them. Warmed ventilation air from the eave soffits below flowed through these attic spaces into the cathedral slope vent chutes starting at the tops of the knee walls. These interior "rafter tail" type conditions were not blocked below the vent chutes, allowing wind washing in the roof bays.

Attic Roof Warming – Problems Identified in Phase I

In this project, attic warming was the largest source of melt water; and, as a result, heat loss to the attics was the most significant cause of ice dams. Warm air rising into the attic melts the snow and creates the bare roof areas as shown below.

Multiple units with common pattern of melted attics above snow covered cathedral slopes.
Original melt pattern showing attic melt above cathedral slopes. Attic melt pattern reduced to “spots” after the Phase I repairs.
Section showing cathedral and attic areas of a middle unit.
Infrared image showing heat in attic end wall.
Background: Unmodified units with attic melt. Foreground: Modified units with no melt.

Attic melt pattern with snow remaining on short cathedral slopes.

Assuming a constant snow depth, the volume of water generated for ice dam formation is directly related to the surface area of snow-covered roof that is melted by heat loss. However, the physical size of the heat loss source (i.e. the air-leakage hole size) is not necessarily directly proportional to the roof area it can melt. In the cathedral slopes, the area influenced by specific air leakage sites was limited by the area of the rafter bay that the leak was in, and the volume of air that could flow out of the leakage area through the ventilation chute or space. In the case of the attic, a relatively small number of square feet of air-leakage area (hole or penetration size) could allow the unrestricted flow of conditioned air to warm the entire roof surface of the attic (for example, an open attic hatch). Heat from the attic floor’s conductive losses was also distributed over the entire roof area. (Note that the use of strapping behind the gypsum board in both the cathedral and flat attic ceilings serves to create a cross-connection between communicating wall and ceiling framing bays and increases the air leakage from perimeter sources.)

The attic construction, from the inside (below the ceiling) to the outside, was as follows: gypsum board, 4-mil polyethylene vapor retarder in the bedroom ceiling areas (not in boxed chases or dropped ceilings), 1 x 3 strapping, 2 x 8 wood ceiling joists - 24" O.C., a single 9" layer (R=30) or two layers, one of 6" and one of 3.5" unfaced batt insulation (total R=30). (Note: The second layer of batt insulation was turned across the framing and stopped at the intersection with the rafters, several feet short of the first layer (R=19) which went to the edge of the ceiling bays where they met the slope insulation). The ridge had a continuous vent.

There was no ice formation observed on the upper attic roofs. Soon after a snow or frost event had occurred, a few hot spots would melt above specific areas such as the bathroom vents and the party walls then, the entire roof would melt clear over the attics of all four units. The combined losses from the sources listed below resulted in attic temperatures as high as 55 degrees (Fahrenheit) when outside temperatures were below 10 degrees F. Infrared analysis inside the attics during the Phase I testing showed warm air leakage into the attics from numerous sources as well as reduced R-values in the loose-fill fiberglass and batt insulation (4):

  1. Dropped ceiling areas over bathrooms, closets, hallways, and recessed window tops had not been sheathed (no air barrier) above or below the un-faced batts. Only strapping or friction supported the batts between the ceiling joists in these areas. In some cases the batts had fallen into the recess below.

  2. Partition tops were unsealed at framing connections and at plumbing and wiring penetrations.

  3. Party wall tops at the ceiling plane were unsealed and un-insulated. At the intersection of adjoining middle unit and end unit, the party wall was under an isolated "cricket" attic (a section of roof built over a lower roof slope to prevent water from running down the lower roof against a vertical wall where it cannot drain properly). One side of the party wall was above the rafter line of the adjoining unit.

Section at top of party wall in cricket attic. Melt pattern at end unit cricket area above valley.

Looking down along the middle unit end wall into the open top of the party wall.

Cricket party wall condition after air sealing with rigid foam.
Section showing isolated cricket attic and location of detail 4/A11 shown above. Interior infrared image of cricket attic area.

The end unit side of the double wall was unsheathed (no outside air barrier) and not blocked or sealed at the lower top plate. The party walls were spaced apart by a 1" space, creating a continuous passage from the interior party walls below - up into the cricket attic. The conductive loss through the top of the party wall was evident on the interior infrared tests as a cool band at this location. The scan from the attic showed the flow of heat up along the party wall and roof slopes. When viewed from the outside, the party walls and cricket roof areas had the highest rate of melting.

  1. Attic hatches were unsealed and uninsulated.

  2. Batt insulation was installed up to, but not over the boxed I-beams and had no air barrier across the I-beam just below the sloped roof to attic ceiling transition. In some cases this adjoined a dropped ceiling area at the hallway, and in all cases opened into the dropped bathroom and closet ceilings. The lack of a continuous air barrier kept these interior spaces warm (open to the interior partitions below) and "connected" to the roof slope vent and attic spaces.

  3. The flat cap (attic) areas showed conductive and warm air leakage losses at the butt joints in the batts and around the framing members at the flat to slope transitions. Light fixtures, bathroom fans and duct runs also created gaps in the insulation. This was true in both single layer and crossed double layer configurations. The insulation did not extend into the party walls (approx. 10" wide) or out to meet the top of the roof slope insulation. At the intersection of adjoining middle and end units where the roofs overlapped (termed the "cricket" area), the vents of the end unit slopes originated at the unsealed tops of these party walls.

  4. The tops of the boiler chimney chases were open to the attic.

Phase I Repairs

Based on a report of the results of the non-destructive and destructive test procedures, the team reviewed the problems and agreed to make the following repairs:

Cathedral Roof Slope Warming – Phase I Repair Methods

  1. The eave soffits on all units were opened and blocking was installed between the rafters above the walls to eliminate wind washing. Unsheathed walls were covered with rigid foam and air sealed. Narrow empty roof bays at the gable ends, skylight double rafters, and party walls were insulated as far up as possible from the bottom access and sealed to reduce infiltration.

  2. The large zero clearance chimney boxes were opened and the inside of the exterior walls were sheathed and sealed with rigid foam, the roof slopes were insulated and sheathed along the plane of the slope to continue the insulation over the box to the outside wall. These procedures prevented the infiltration of cold air into the box and the flow of warm air up into the roof slope vent chutes.

  3. The interior roof slope perimeter was sealed in the wood paneled units. The skylight casings were removed and the jambs sealed to the rough openings to prevent cold air infiltration at the bottom, and warm air leakage into the roof above them. The beams at the top of the cathedral slopes were air sealed from below.

  4. Insulation defects in the knee wall and boxed chases were repaired. Sheathing was installed and sealed on the second side of each of these one-sided conditions and unblocked floor to floor rim joist areas were sheathed to prevent heat loss in these lower attic spaces from flowing up into the roof vents above. It was recommended that the recessed light in the floor of this attic be changed to a fluorescent surface-mounted fixture.

  5. Additional glazing was added to the outside of one of the test unit skylights. Ice dam formation below the modified and unmodified skylights was monitored to determine if additional R-value would reduce conductive losses enough to eliminate melting on the glass in cold temperatures.

  6. The fan / lights in the Guest Bathroom slopes were reinstalled as surface mount units in the same location. The original "holes" behind these new installations were insulated and the air barrier completed. Rigid foam was used to insulate the roof over the duct in the reduced rafter space above the fan. In the attic the duct was relocated under the insulation until it exited. A wall cap with a damper was installed at the end of the duct run and the soffit vents for four feet on either side of the vent location were sealed to stop warm moist air from re-entering the attic.

Attic Roof Warming – Phase I Repair Methods

Insulated attic hatches with gaskets and draw-down latches were installed.
Foil-faced rigid foam sheathing was installed and sealed to adjoining air barrier materials in the short slopes and flat ceilings above all of the drop ceilings. This insulation and air barrier system was boxed around the I-beam. The top plates of interior partitions inside the dropped ceiling spaces were air sealed. Insulated exterior walls above the ceiling plane were treated similarly.

In the flat ceilings of the attic, air sealing was done at all accessible penetrations under the batt insulation. The air barrier and insulation were extended into the party walls. Gaps in the batt insulation were to be filled with loose-fill insulation. An additional layer of loose fill insulation or a continuous vapor-permeable air barrier film was to be installed on top of the batts to reduce convection in the batts.

The top of the insulated metal furnace flue was sealed to prevent the heat from flowing into the roof slope vent spaces.

Part 1 ] [ Part 2 ] Part 3 ] Conclusion ]

Back to Case Studies