Emissivity and Evidence: One Winter of Thermal Imaging a 1960s Semi Through Its Retrofit

Thermal imaging is the most persuasive instrument in domestic retrofit and the easiest one to misread. A camera returns an apparent surface temperature calculated from received radiance and a set of assumptions that the operator supplies; it does not measure heat loss, U-values or air movement. Between October 2025 and April 2026, we followed a 1960s semi-detached house in the Midlands through three thermal surveys and a retrofit programme of cavity wall insulation, replacement glazing and targeted draught-proofing. A blower door, fourteen surface thermocouples and four humidity loggers were used throughout to arbitrate between what the images suggested and what the fabric was actually doing.

Method, and why the camera needed calibrating first

The camera used was a 320 x 240 microbolometer with a stated accuracy of ±2 °C or ±2 per cent, whichever is greater. Every survey was conducted with emissivity set to 0.95, reflected apparent temperature measured by the crumpled-foil method, and a spot check against a type K thermocouple taped to the same surface. The reporting convention followed BS EN 13187, which is explicit that a thermogram is a qualitative record of surface temperature distribution rather than an energy measurement.

Two conditions governed the survey design. First, a temperature difference of at least 10 °C across the envelope, and ideally 15 °C, was required for meaningful contrast; an October afternoon with the heating off and the sun on the south elevation produced nothing usable. Second, surveys were taken at least four hours after sunset to allow the external leaf to shed solar gain, because a wall that has absorbed heat all day will read warm long after the heating has gone off and can be mistaken for an insulated one.

Survey one: fields, junctions and the difference between them

The pre-retrofit survey took place on a still February night with an internal air temperature of 19.8 °C and an external air temperature of 2.1 °C. Wall field temperatures on the north elevation averaged 14.2 °C, giving an internal surface resistance consistent with the 1.2 to 1.6 W/m²K expected of an unfilled 1960s cavity — that is, a surface some 5.6 °C below room air, which is exactly the deficit a householder feels as radiant draught.

The more useful information was at the junctions. At the wall-to-floor junction along the front elevation, the surface fell to 10.8 °C, a 3.4 °C depression relative to the field. Around the first-floor lintels, two of which had been formed in situ without a thermal break, the depression was 2.6 °C. The party wall junction, continuous from footings to roof, showed the familiar vertical signature that imaging surveyors learn to expect in terraced and semi-detached stock. None of these were artefacts; each was confirmed by contact sensor and each was present in the blower-door survey, where the pressure differential concentrated cold air at the same coordinates.

Air: the number the pictures cannot give you

A single blower-door test at 50 Pa before the works returned 9.8 air changes per hour, equivalent to a leakage area of roughly 640 cm² at 4 Pa. After the windows, the trickle vents and the loft hatch were addressed, the same test returned 5.1 ACH at 50 Pa. That halving of infiltration mattered more to comfort than any single thermogram in the set, and it is invisible in every image we took.

The lesson is worth stating plainly: a thermal camera shows you where surface temperature departs from expectation. It does not show you where air is moving, only where moving air has cooled a surface. The two are related but not identical, and a cold streak at a skirting is as likely to be a joist end as a draught.

Survey two: the low-e glazing trap

The replacement units were argon-filled, soft-coat low-emissivity double glazing with a stated centre-pane U-value of 1.2 W/m²K. Imaging them with the camera set to 0.95 produced a centre-pane reading of 12.1 °C against a room temperature of 20.4 °C, implying a U-value of roughly 6 W/m²K — an absurd result. The thermocouple bonded to the same pane read 17.6 °C.

The discrepancy is not a fault; it is emissivity. Soft-coat low-e glass has an emissivity closer to 0.1 in the mid-infrared, so it reflects the room rather than emitting its own temperature, and a camera assuming 0.95 will systematically under-read. The correct approach is to set the value for the material, accept a noisier image, or tape a patch of high-emissivity material to the glass — the standard workaround, and one that ruins the appearance of the glazing for anything else on the wall.

The same trap appears on foil-faced insulation boards, stainless-steel flues and anything with a factory-applied reflective finish. In each case the camera reports the reflected environment, not the object.

Condensation risk: what moved, and what did not

Using the measured surface and air temperatures, we calculated the temperature factor f_Rsi at each location and compared it with the 0.75 threshold used in ISO 13788 for dwellings with normal occupancy. Before the works, the wall-to-floor junction sat at 0.62 and the lintel soffits at 0.68. After cavity fill and draught-proofing, the junction rose to 0.71 and the lintels to 0.78 — better, but one of them still below the threshold on the coldest night recorded, when the external temperature fell to −3.4 °C.

Meanwhile, relative humidity indoors rose. The kitchen hygrometer moved from a weekly mean of 62 per cent before airtightness work to 68 per cent afterwards, with peaks above 75 per cent during cooking. The retrofit did not create condensation; it relocated it, from draughty window reveals that dried quickly to a colder corner where air now lingers. The extract fan, rated at 15 l/s, was doing about 9 l/s in situ. That is not a thermal imaging finding, but it is the finding that matters.

What the images got right, and what they invented

Reflections from a radiator appeared as a warm patch on the opposite wall and could easily have been logged as insulation failure. Wet brickwork after rain read cold for an hour because of evaporative cooling. A curtain drawn across a reveal created a convective plume that looked exactly like a structural cold bridge. Each of these has a signature, but the signatures are learned rather than obvious, and none of them are visible to a householder looking at a phone screen.

Closing the loop

The camera earns its place. It found the lintels, the party wall junction and a section of cavity that had never been filled because of a blocked injection hole — all confirmed later by drilling and borescope. It is a hypothesis generator of exceptional speed. But the referee remains the pressure test, the contact sensor and the humidity log, because those measure the quantities the retrofit is actually trying to change. A thermogram is an argument, not a verdict.