Thermal imagery: a visual picture and a measurement product are different offers

A thermal image can be useful as a picture and still be unsuitable as a measurement record. The distinction matters when a company buys imagery for industrial assessment, equipment reporting or environmental analysis. A screenshot may show a clear visual pattern without preserving the underlying values, the assumptions used to interpret them or the information needed to repeat the analysis later.

The purchase specification should therefore describe what the customer will receive beyond the image on screen. It should identify the measured or estimated quantity, the accompanying metadata, the supported export format and the limits of the interpretation. These choices determine whether a product is an illustration, an analytical input or a service that can support a documented business conclusion.

A file extension does not define the evidence

FLIR's Exx series image documentation describes a useful example: its saved JPEG can contain radiometric information alongside a conventional image component for ordinary viewing software. The manual also identifies stored information such as date, emissivity and atmospheric temperature. This is a property of the documented product, not a promise that any JPEG with a thermal colour palette contains measurement data.

A buyer evaluating a service should request a representative delivered file and its description. Opening the picture successfully is only the first check. The receiving analyst needs to establish whether the values and relevant metadata remain available in the software the business actually uses.

This issue often appears at a handoff. A specialist may export a report illustration for a commercial colleague, who later forwards it to another analyst expecting further quantitative work. If only the rendered picture travelled, the recipient may need to return to the original provider. Defining the analytical handoff before purchase avoids treating a presentation file as the complete data product.

Name the quantity being sold

Terms such as thermal data, temperature map and radiometric imagery can conceal different processing stages. USGS's Landsat Collection 2 surface-temperature documentation describes an estimated surface-temperature product assembled using thermal and optical observations, emissivity information and atmospheric inputs. It also explains that stored digital values require a scale and offset to express temperature in Kelvin.

That scientific example demonstrates why the contract should identify units and processing level. A stored number is meaningful only with its definition. A displayed number may already incorporate a model or correction. The customer should know whether it is purchasing sensor output, a processed geophysical quantity or the supplier's interpretation of a particular industrial condition.

The distinction changes the acceptance question. A data provider can be responsible for supplying the agreed product correctly while a separate analyst remains responsible for a maintenance recommendation. If the provider sells both, the scope should make that additional analytical responsibility visible and explain the evidence supporting it.

Compare like quantities across time

An illustrative facilities team wants to compare thermal assessments of a civil warehouse over successive reporting periods. Before using the results to assess a refurbishment project, it needs to know whether both deliveries represent the same quantity, comparable areas and a documented processing approach. A difference in the report should be traceable to its inputs and assumptions.

The supplier's role is to explain which comparisons the service supports. If the two datasets require additional interpretation before they can be compared, that work should be priced and assigned. A simple before-and-after layout should not silently imply that all relevant differences have been accounted for.

Record the product version and retain the original deliveries. If processing is revised, the analyst should be able to distinguish a change in the underlying observation from a change in the way the result was produced. The broader requirement to preserve this chain is discussed in derived intelligence data lineage.

Missing values deserve their own explanation

A temperature map with gaps is not necessarily a failed delivery. USGS explains that missing ASTER emissivity information can produce gaps in its Landsat surface-temperature product. This is a documented dependency of that particular product. It illustrates why a service needs an explicit account of missing information rather than a promise that every image will yield a complete analytical layer.

The commercial report should distinguish areas that were not observed, observations rejected on quality grounds and areas for which the chosen estimate could not be produced. Those conditions affect whether another acquisition would help, whether additional processing is appropriate or whether the proposed use is outside the product's scope.

Coverage percentages should use a stated denominator. If an evaluation excludes all difficult areas before calculating its completion rate, the buyer needs to see those exclusions. A provider can legitimately define a supported area or condition; the issue is whether the sales claim, trial report and recurring invoice describe the same scope.

Decide who owns the processing burden

USGS's atmospheric auxiliary data guidance notes that users of completed Collection 2 Level-2 products do not need to download the auxiliary inputs separately. That is a useful example of a provider carrying processing work on the user's behalf while documenting the dependencies.

A commercial customer should make a similar distinction when comparing offers. One quotation may deliver files ready for the agreed analysis. Another may assume that the customer has specialist software, suitable reference information and staff who can prepare the data. Comparing only the access fee hides a potentially large difference in internal effort.

Ask which software functions and support services are required to reproduce the supplied example. If a subscription includes visual inspection but quantitative export requires another licence, include that cost in the evaluation. If the supplier offers an analyst-reviewed report, identify the review scope and turnaround separately from file delivery.

Make the acceptance package useful after the trial

A strong acceptance package contains the original sample data, an example of the normal export, the metadata description and a short explanation of what the customer can infer from each output. It should include at least one case where the result is incomplete or uncertain, because that is where receiving teams often need the most help.

The agreement should also describe corrections. If a delivered dataset is later found to contain an error, the customer needs to know which files are affected, whether replacements are included and how the supplier will identify reports that may need revision. That obligation is more useful than a general assurance about image quality because it addresses the cost of maintaining a recurring analytical service.

The customer should retain the interpretation supplied during the evaluation. A verbal explanation by a specialist can make a demonstration understandable, but future staff will need a durable record. Where the interpretation depends on assumptions supplied by the customer, preserve those assumptions with the result.

Rights to retain and reuse the material also affect its value. A business preparing reports for clients may need to export a figure, preserve numerical data and share the analytical basis with an authorised reviewer. Those are separate uses that should be visible in the agreement; the GAO account of commercial space data adoption barriers shows why data access and practical reuse deserve separate attention.

A credible thermal product makes these boundaries easy to understand. It gives the customer a clear picture when a picture is sufficient, and a documented measurement or estimate when the decision requires more. For the supplier, that clarity creates a defensible distinction between basic imagery access, analytical data delivery and specialist interpretation.