A method that produces a plausible result has not automatically demonstrated that it works for its intended users. Validation requires a defined question, suitable reference evidence, and an evaluation design that reflects how the method will be used. This perspective examines which observations should be compared, how they should be sampled, and which errors matter in practice. It also considers the difference between evidence used to develop a method and evidence reserved for evaluating it. The objective is a test that can reveal weaknesses as well as confirm useful performance.
This working guide considers near-infrared observations of vegetation and land in turbid coastal waters. The immediate decision is to identify coastal sampling priorities. Useful reference evidence includes water samples, surface conditions and acquisition-time context. The case is an illustrative planning context, with assumptions to check against the actual sensor, landscape and project requirements.
The setting: Turbid coastal waters
Turbid coastal waters contain changing mixtures of suspended material, dissolved substances, biological components and surface reflections. Their image appearance may vary over short distances and between acquisitions. A monitoring design should identify which variable or pattern it aims to investigate rather than treating every colour difference as a direct water-quality measurement. Coastal geometry and the proximity of land also influence which parts of an image provide a defensible observation of water.
Review atmospheric conditions, cloud, shadow, sun glint and land adjacency before comparing spectral measurements. Shallow-bottom contributions can matter in some places, while deeper areas may present a different interpretation problem. A relationship developed for one water type should not be transferred automatically to another. Retain the valid observation mask and the dates supporting each map so that users can distinguish measured patterns from periods without suitable optical evidence.
Field validation should sample the range of conditions in which a model or interpretation rule will operate. Record collection time, location and the measurement method for the target variable, together with relevant surface conditions. Samples and imagery separated by a changing coastal event may not be comparable. A useful product can guide sampling or describe a consistent spectral pattern, with stronger numerical water-quality claims introduced only when their calibration and independent evaluation support the intended application.
Understand what contributes to the NIR response
Near-infrared wavelengths lie outside the range of human vision, but they are important in optical observations of vegetation and land. A near-infrared image should not be confused with a thermal image. Reflected sunlight in this region and emitted thermal radiation describe different physical processes. This distinction matters when communicating crop information: a bright or dark NIR response is not a direct temperature reading. Its interpretation depends on the surface, the observation geometry and the particular wavelength interval measured by the instrument.
Vegetation can show a strong near-infrared response because of interactions within leaves and canopies. At landscape scale, the observed signal also changes with canopy cover, leaf arrangement, shadows and the proportion of visible soil or water. A field with a different crop age may therefore differ substantially from another field without either being abnormal. NIR is useful partly because it adds contrast beyond visible appearance, but that contrast still needs a meaningful comparison. The analytical task determines which associated measurements and field records are required.
Use NIR in a traceable analytical chain
When combining NIR with another band in an index or model, check that the channels refer to comparable ground locations and acquisition conditions. Differences in band resolution, alignment or spectral response can influence the result. Preserve reflectance scale factors, quality flags and any processing applied before the combination. Where resampling is necessary, choose a documented common grid and recognise that the operation changes representation rather than increasing the original observation detail. These controls are particularly important near field edges and in small management units.
An application should demonstrate the benefit of the NIR measurement using evidence appropriate to its stated purpose. For vegetation monitoring, that might mean a relationship with independently recorded canopy conditions or a reliable separation of relevant land-cover classes. For water or soil interpretation, a different protocol may be needed. Review errors as well as successful examples and retain cases that challenge the initial explanation. A clear observation chain makes NIR a useful part of a broader decision process without assigning the band capabilities it does not independently provide.
Design an evaluation that reflects deployment
A model evaluation should resemble the conditions in which the model will be used. Start by identifying which relationships the development data share: nearby locations, repeated observations of the same unit, common collection teams, or measurements from a single period. These relationships can make records less independent than they appear. An evaluation that ignores them may test the ability to recognise familiar conditions rather than the ability to operate in a new setting.
Choose a split strategy that matches the intended transfer. Holding out locations, seasons, operational units, or acquisition conditions can answer different questions. Document which question each split is designed to address, and keep the held-out evidence separate while selecting features and tuning the method. Repeatedly inspecting the same evaluation set to improve a model gradually makes that set part of development, even if it retains its original label.
Report results at a scale that is meaningful to the user. A strong average can hide poor performance for important subgroups or unusual conditions. Examine error patterns and retain examples of both successful and unsuccessful predictions. Where the evidence covers only a limited range of conditions, describe that range explicitly. The evaluation then becomes a practical guide to suitable use, further testing, and the collection of additional reference observations.
Budget the complete processing workflow
A processing budget should cover the complete path from source observations to a maintained product. Data access and computing are only part of that path. Include preparation, geographic updates, reference collection, expert review, delivery, and support where those activities belong to the service. A calculation that runs cheaply on a small demonstration can still require substantial effort when the surrounding work is repeated across many locations.
Measure representative tasks rather than assuming that every observation costs the same amount to process. Record data volume, processing time, retry frequency, manual interventions, and the conditions that create unusually demanding cases. Separate fixed setup work from work that grows with the number of observations or users. These records support decisions about batching, caching, storage policy, and the point at which a different processing design becomes worthwhile.
Connect the budget to a defined level of service. Faster delivery, more frequent observation, broader coverage, or more detailed review can each change the resources required. Describe these trade-offs before promising a delivery pattern that the workflow has not demonstrated. Review the budget using operational evidence and preserve enough detail to understand changes between processing versions. The aim is a sustainable service whose resource choices remain aligned with the information that users value.
Bring the evidence back to the decision
The validation record should explain the population represented by the test, the reference observations used, and the limitations of the comparison. Report meaningful errors alongside useful successes, and retain examples that challenge the method. Treat wider deployment as a new opportunity to evaluate transfer across relevant conditions. A well-designed validation programme gives users a defensible basis for deciding where the method is useful, where additional evidence is needed, and where another approach may be appropriate.
Sources & further reading
Explore the bhumisat spacecraft concept and our vision towards 2030, or return to the journal.

