The Conditions That Make or Break a Drone Thermal Inspection

A drone thermal inspection fails quietly. The flight goes fine, the imagery looks plausible, the report gets written — and none of it means anything, because the survey was flown in conditions where the faults could not show themselves. Nobody notices until someone acts on a finding that was never there, or misses one that was.

This is the part of thermal work that gets least attention in quoting conversations. The camera is the easy part. The conditions are what decide whether the survey is interpretable.

A thermal camera measures difference, not condition

A radiometric thermal sensor records apparent surface temperature. It does not detect a cracked cell, a loose lug or a wet roof. It detects that one patch of surface is at a different temperature than the patch beside it, and a human then decides what that difference means.

That has a hard consequence: if the asset is not under load, there is no difference to see. A solar string that is not producing does not heat up at a fault. A breaker carrying no current does not run hot at a bad connection. A roof with no sun on it and no temperature gradient across it has nothing to reveal about trapped moisture. The conditions create the signal. The camera only records it.

So the first question on any thermal scope is not which sensor. It is: when will this asset actually be under enough thermal load to show us what we are looking for, and will the weather let us fly then?

Solar: irradiance is the entry condition

For a solar farm thermal survey, irradiance — the solar energy landing on the plane of the modules — is what puts the array under load. Below a certain level of sunlight, the modules simply are not producing hard enough for a failing cell, a bypassed substring or a failing junction box to separate itself thermally from its neighbours.

The industry convention for utility-scale aerial thermography is a minimum irradiance threshold measured in the plane of the array, and it is a floor, not a target. More is better, and the number should be measured on site rather than inferred from a weather app, because a hazy Alberta afternoon can look bright to a human and read well below threshold on an irradiance meter.

Practical consequence for scheduling: Alberta's usable solar thermal window is narrower than people assume. It is a midday band, not a whole day, and it shrinks considerably outside the summer months. A survey booked for a specific date has to carry the possibility of being moved.

Wind is the biggest destroyer of a solar thermal survey

If there is one condition that quietly ruins more thermal surveys than anything else, it is wind. A hot spot on a module is a small area of elevated surface temperature. Moving air convectively cools that surface — and it cools the hot spot faster than it cools the panel around it, because the temperature gradient there is steeper.

The result is not that the imagery gets noisy. It is that the anomaly shrinks toward the background and, past a certain wind speed, disappears into it. The survey looks clean. The array is not.

This is why an operator saying "we can fly in that wind" is answering the wrong question. Most survey aircraft can be flown in wind well beyond the point where the thermal data stops being interpretable. Flyability and data validity are two separate limits, and the data limit is the lower one.

On Alberta sites this is the condition most likely to move a booking. A steady chinook afternoon can be a beautiful day to fly and a useless day to inspect.

Cloud: the failure mode that looks like a finding

Broken cloud is more dangerous than solid overcast. Overcast is obvious — irradiance is under threshold, nobody flies, nobody is fooled.

Intermittent cloud is the trap. A cloud shadow crossing an array during a flight means different rows were imaged at different points in their heating and cooling curves. Panels that were shaded two minutes ago are still recovering; panels in continuous sun are at steady state. The imagery then contains real, measurable temperature differences that have nothing whatsoever to do with module condition.

Those differences are exactly what an anomaly-detection workflow is built to flag. A survey flown through passing cloud will generate findings. They will just be findings about the weather.

The same logic applies after any interruption: an array needs time under stable conditions to return to thermal steady state before the imagery means anything again.

Time of day, and what it does to reflection

Time of day sets three things at once: how much irradiance is available, how much of the sky is reflecting off the glass, and how the operator has to fly to stay out of the reflection.

Module glass is reflective in the thermal band. At low sun angles the camera sees the sky, the surroundings and sometimes the aircraft itself reflected in the panel surface, superimposed on the temperature signal it is trying to record. Flight geometry manages this — the camera is kept off the specular angle — but geometry cannot rescue a survey flown at the wrong hour.

Building-envelope and roof thermography sit at the opposite end of the day. There the useful load is the temperature difference between inside and outside, and the useful window is usually after sunset, once absorbed solar energy has bled off the surface and the materials are releasing stored heat at different rates. Flying a roof at noon mostly measures which parts of it the sun has been on.

Electrical and mechanical: load is the condition

For substations, switchgear, motors and other inspection targets, weather matters less and operating state matters more. A resistive connection only runs hot when it is carrying current, and the temperature rise scales with how hard it is working.

Inspecting a lightly loaded asset produces small, ambiguous differences. Inspecting the same asset near normal operating load produces the separation that makes a finding defensible. That means the flight has to be scheduled around the plant, not around the drone crew — and it means the report should record what the load actually was, because a temperature rise with no stated load is not a comparable number.

Wind still applies here. Outdoor equipment cools convectively the same way modules do.

Record the conditions in the report

Whatever the asset, the conditions the survey was flown in belong in the deliverable, not in someone's memory. At minimum:

• Date, start and end time of the thermal flight
• Irradiance measured on site for solar work, and how it was measured
• Wind speed and direction during the flight
• Cloud state, and whether any interruption occurred
• Ambient temperature
• Operating state or load for electrical and mechanical assets
• Camera and its calibration status, and the emissivity assumption used

This is what makes a thermal report auditable and comparable year over year. Two surveys of the same array flown in different conditions are not directly comparable, and without the conditions recorded there is no way for anyone downstream to know that.

The honest answer is sometimes "we reflew it"

A thermal survey flown outside its window should be reflown, not interpreted. That is an uncomfortable conversation when a crew is already mobilised and the client wants a result, but the alternative is worse: a report that assigns findings to conditions, or misses real faults because the wind flattened them out.

When we quote thermal work we say up front that the date is weather-dependent and what specifically would move it. It makes scheduling less tidy. It is the only way the number at the end of the report means anything.

If you are scoping a thermal inspection — solar, electrical or envelope — and want a straight answer on what window your site actually needs, get in touch.