A sharp orthomosaic, a clean point cloud, a model that looks exactly the same as the site. None of that addresses the question that matters before any design work begins: is this dataset fit for purpose?
A Beautiful Orthomosaic Is Not an Engineering Survey
Visual fidelity and engineering readiness are two entirely separate criteria, evaluated by two fundamentally different facets of the dataset. An orthomosaic can be sharp, well-blended, and constructed correctly on a geometrical basis, and still not contain the information that an engineer is going to use when designing a site. That's where the DTM comes in; treating the quality of a visual layer as an indicator of engineering readiness is the crux of this article's issue.
What Information Does an Engineer Actually Need?
Depending on the project, the list can resemble "a survey" more closely, but it still breaks down into more specific disciplines: horizontal position, elevation, breaklines (the lines that define changes in grade, such as the edges of roads and the tops of ditches - a smooth surface model will erase them), contours, existing ground surface, structures, utilities and other ground features, and an established coordinate and reference system. A drone survey that fails to deliver any one of these components is not an incomplete product; it's a product that didn't meet specifications from the beginning.
Orthomosaic vs DSM vs DTM
This is the single greatest source of confusion for non-GIS clients, and it needs to be addressed specifically: the orthomosaic is a corrected image, a continuous surface of photographic data that represents the appearance of the site. The DSM is the elevation surface, calculated from the sensor data, which includes everything that the sensor picked up. The DTM is the same surface with everything that isn't the bare earth removed, for engineering purposes. Engineers design using the DTM, and clients unfamiliar with the distinction sometimes expect the orthomosaic to be an engineering-grade surface model because it looks neater.
Why a DSM Is Not Automatically a DTM
Generating a DTM from a DSM involves filtering out objects that aren't the bare earth, and that process has specific limitations. In thick canopy or dense vegetation, photogrammetry-based ground filtering cannot penetrate the cover to see what's on the ground, because the camera cannot see through the leaves. Published research on the topic shows that in dense vegetation, a photogrammetry-based DTM has to be either calibrated or augmented with ground survey data, because the filtering process cannot create data that wasn't captured in the first place. A DSM that looks fine can generate a DTM with unnoticeable flaws when there's thick cover on the ground.
Where Drone Surveys Work Extremely Well
None of the preceding points are an argument against drone surveying as a practice - they're only arguments in favor of understanding the strengths and limitations of the method. Drone surveying is ideal for construction progress, landform mapping, stockpile volume calculations, earthwork takeoffs, corridor and road surveys, general site documentation, and large-area mapping where terrestrial methods would be significantly slower or more labor-intensive. These are the applications where it's not only acceptable, but often superior, to use a drone instead of traditional methods.
Where You Need to Be Careful
The same technology has specific limitations: dense vegetation blocks out the field of view, water surfaces do not reconstruct the same way that solid ground does, vertical formations are more challenging to capture with standard flight patterns and require additional photography, underground utilities are not visible to any optical or photogrammetric method at all, and projects that require extremely precise setting out or extremely tight elevation tolerances often demand greater accuracy than a drone alone can reliably deliver without additional controls.
Drone Survey vs Conventional Total Station/GNSS Survey
The question isn't which method is better, but which data source is appropriate for a given project. A drone survey covers a large area quickly and captures a continuous surface, whereas a conventional total station or GNSS survey covers several points at a time but measures each one with extreme precision, including those that an optical camera would have trouble seeing, such as cracks in pavement or the undersides of structures. Published accuracy ratings reflect that division: a standard RTK/PPK drone survey is accurate to approximately ±10-20mm, while a drone survey combined with a ground control network is accurate to approximately ±5-10mm, which is the threshold for most detailed design and stakeout work. Either figure is well within the capabilities of a conventional total station or GNSS receiver, but the drone and ground control method is appropriate for projects that demand absolute accuracy across a large area, rather than specific points of interest.
The Hybrid Approach
There is no final or complete answer, but the strongest performance is generally achieved by applying multiple methods, each to their greatest advantage. A drone survey covers the large area, RTK/GNSS and ground control points verify that coverage, and conventional total station work fills in the details where a drone camera has structural limitations. It's not that drone surveying is less valuable, but that each method has appropriate applications, and the most competent results are obtained by using the most appropriate method in each circumstance.
What Should Be Included in an Engineering-Ready Drone Deliverable?
A genuinely complete set of engineering-ready data extends well beyond the orthomosaic. As a bare minimum, that should include the orthomosaic, the DSM, the DTM, and contours, the point cloud, a 3D mesh if applicable, CAD/GIS-ready layers instead of a photogrammetry-specific project file, an actual, documented accuracy rating (something that can be proven, rather than a stated claim), a coordinate system, metadata, and sufficient documentation of the survey control and checkpoints to allow the client to understand the data they're receiving. All of these elements serve a specific purpose in the entire processing pipeline that produces the final output.
The Final Question: "Is This Dataset Fit for My Purpose?"
And that is the final takeaway: the question that the client needs to be asking, and which the author hopes the reader will ask before a design team begins working, is not "does this look good", and not even "does this meet the GSD specification" (a topic that is also covered in its entirety in why GSD doesn't mean accuracy). A drone survey can be technically perfect and visually stunning, and fail to meet the requirements of an engineering specification because, ultimately, it doesn't contain the information that that specification demands. The distinction between what a drone survey can and cannot do is more valuable than any detail on the orthomosaic, and that is the topic that this article has been about. If the distinction is unclear, or if drone surveying, conventional surveying, or a combination of the two is appropriate for a given project, then that is the planning conversation that topographic survey and drone imaging and processing services exist to have before any fieldwork begins.
