A convincing image can still contain measurements that are unsuitable for a particular task. Quality control asks whether the inputs, transformations, and outputs meet the requirements of the intended application. This perspective follows evidence through the workflow, from acquisition metadata and reference observations to the final interpretation. It encourages teams to define acceptance conditions before seeing the result, rather than adjusting them to favour a preferred conclusion. The purpose is a repeatable way to distinguish usable observations, observations requiring qualification, and observations that should be excluded from a decision.
This working guide considers measuring forest cover without confusing canopy and land use in seasonal dry woodlands. The immediate decision is to distinguish seasonal change from disturbance. Useful reference evidence includes leafing stage, disturbance evidence and land-management records. The case is an illustrative planning context, with assumptions to check against the actual sensor, landscape and project requirements.
Define what the map measures
Forest cover is not a single self-evident quantity. One project may need the fraction of a pixel occupied by tree crowns; another may need a categorical forest mask based on canopy cover, height, and minimum area. A land use inventory may treat temporarily cleared forest differently from a map of visible tree cover. Write the definition before collecting labels, and identify how plantations, orchards, mangroves, sparse woodland, and urban trees will be handled. A consistent definition matters as much as a sophisticated classifier when comparing places or years.
Optical observations provide information about foliage, moisture, and seasonal behavior, but similar vegetation types can have overlapping spectra. A green pixel does not establish tree height or distinguish a natural forest from a plantation. Time series can capture leaf emergence, dry season persistence, and disturbances that a single image misses. Radar adds sensitivity to vegetation structure and moisture, although relationships vary with wavelength, geometry, and forest conditions. Combining observations can improve discrimination when each input has a clear role and compatible spatial support.
Make estimates comparable and testable
Prepare a stable analysis grid, inspect geolocation, and apply consistent atmospheric and quality screening to optical inputs. Choose seasonal windows that suit the local ecology rather than assuming one calendar works everywhere. When clear observations are sparse, report the observation count and the age of the composite. Train with examples spanning forest types, canopy densities, topography, and confusing nonforest classes. Keep independent assessment locations separate from training areas, especially where neighboring pixels share the same stand.
Evaluate omission and commission errors for forest and relevant subclasses. A high overall accuracy can conceal poor performance in a rare forest type or along fragmented edges. For area reporting, use a probability sample and an estimator appropriate to its design rather than treating classified pixel counts as exact. Compare annual products with the same class definitions and processing assumptions. Abrupt canopy loss can indicate harvesting, fire, storms, or conversion, and the image alone may not identify the cause. Preserve that distinction in summaries, and accompany trend estimates with uncertainty and a description of what canopy recovery does and does not establish.
The setting: Seasonal dry woodlands
Seasonal dry woodlands can change markedly in appearance as leaf cover, understory vegetation and soil exposure vary through the year. A sparse-looking canopy during one period may be part of an ordinary seasonal cycle rather than evidence of degradation. Interpretation should begin with the local seasonal context and the intended observation variable. Mapping woody cover, reviewing a disturbance alert and studying vegetation recovery require related but distinct forms of evidence.
Build comparisons using appropriate dates and quality-screened measurements, and examine whether an apparent change persists beyond a short seasonal fluctuation. Ground vegetation can contribute strongly where tree cover is open. Rainfall timing, grazing or other land management may alter that contribution without changing the woody component. A report should retain the observed area and acquisition dates so that a user can distinguish a current measurement from an inference made across a long image gap.
Field validation should record tree and ground-layer conditions separately where that distinction matters to the application. Include the leafing stage, signs of recent disturbance and the management context of each sampled area. A decline in a vegetation measure cannot identify its cause by itself. Evaluate mapped changes against representative observations across different canopy densities and seasonal states, and communicate where the available imagery supports a broad landscape interpretation rather than a precise inventory of individual trees.
Maintain the geographic units of analysis
Many observation services summarise measurements within a field, management block, catchment, forest unit, or other mapped boundary. The meaning of that summary depends on whether the boundary accurately represents the unit being discussed. Start with a clear source for the geometry and record its date, coordinate reference, and intended use. A boundary drawn for visual orientation may require additional checking before it becomes a measurement unit.
Inspect overlaps, gaps, narrow fragments, and features that may have changed since mapping. A geometrically valid polygon can still describe an outdated operational unit. Ask the responsible team how divisions, mergers, new infrastructure, and management changes are recorded. Where pixels or reference measurements cross a boundary, define how they are assigned or excluded so that results remain comparable between analysts and processing runs.
Version the boundaries alongside the observations. If a unit changes, preserve a relationship between the previous and revised geometry rather than replacing the history without explanation. Comparisons through time should identify whether they use a constant spatial unit or the operational units valid at each date. This makes it easier to distinguish a change in the measured surface from a change introduced by the geography used to summarise it.
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.
Bring the evidence back to the decision
Quality control is most useful when its results travel with the information being delivered. Preserve the reasons for accepting or rejecting an observation, and expose limitations that could change the user's interpretation. Review recurring failures as clues about the design of the workflow. A clear quality record supports consistent comparisons between operators and processing versions, while helping the team decide whether a problem originates in acquisition, preparation, interpretation, or the underlying observation requirements.
Sources & further reading
Explore the bhumisat spacecraft concept and our vision towards 2030, or return to the journal.



