DSM vs DTM for Orthorectification

The choice between DSM and DTM depends on the application.

A DSM represents the surface (i.e., the top of the canopy). Since the hyperspectral scanner primarily captures information from the surface, a DSM is, in principle, the most accurate option. However, in agricultural fields, DSM quality is often limited by spatial resolution and typically cannot model individual plants precisely. In addition, artifacts caused by environmental and sensor-related factors, such as wind, LiDAR ranging accuracy, and georeferencing errors, are difficult to avoid during DSM generation. These artifacts can propagate into the orthomosaic, leading to issues such as double mapping (repetitive patterns), pixelation, and discontinuities.

A DTM represents the bare earth. In most agricultural fields, the terrain is relatively flat or gently sloped, so a DTM derived from CSF is usually smooth and stable. Using a DTM often results in orthomosaics with better visual quality (fewer pixelation artifacts, discontinuities, and repetitive patterns), making the image content easier to interpret. The geometric accuracy is expected to be slightly lower, especially near swath edges. However, the impact of this reduced geolocation accuracy on the final product is typically small, since the hyperspectral sensor is a pushbroom scanner that primarily looks nadir and is usually flown with large side overlap.

When using a DTM for orthomosaic, it is generally best to generate it from an early-season flight, when canopy cover is minimal and CSF is more reliable. Assuming the terrain is stable, the same early-season DTM can then be reused for all datasets across the growing season. This also has the added benefit of providing a consistent terrain reference for the entire season, which is sometimes preferred for downstream analysis.

In summary, both DSM and DTM can provide reasonable geometric accuracy for orthomosaics, especially for plot-level statistics. While DSM is theoretically the most accurate, DTM is often preferred for its consistency and better visual quality throughout the season.