<p>Geomorphic change detection (GCD), geotechnical engineering, and hazard mapping are analyses that require the lowest possible absolute total error in digital elevation models (DEMs). One of the most common GCD analyses is the quantification of soil erosion. NRTK (Network Real-Time Kinematic) is one of the three primary modes within the broader method of direct georeferencing (DG). NRTK uses a network of multiple GNSS reference stations to provide real-time correction data to a UAV, enabling centimeter-level positioning accuracy. This approach eliminates the need for ground control points (GCPs), reducing both costs and survey time. However, its application in multi-temporal soil erosion analysis remains insufficiently researched. In this paper the NRTK approach in GCD analysis is evaluated using Matrice 210 RTK V2 at two case studies. In addition, the absolute accuracy of the NRTK was tested on three other sites. Although achieved results can be regarded as promising, especially at lower altitudes, this research highlights drawbacks when employing the NRTK in analysing soil erosion measurement. Namely, the use of the DG-based models in GCD analysis generated unreliable results when compared with the reference model derived using the SfM photogrammetry with GCP. In both study sites, the NRTK approach significantly overestimated the amount of accumulated sediment, affected the total net sediment difference, and eliminated a substantial amount of change. Although the NRTK approach shows limitations in reliably quantifying volumetric changes in soil erosion measurements, results indicate that NRTK can be applied for analyzing linear gully headcut retreat rates. For applications where achieving the lowest absolute total error is not a priority, NRTK can be a relatively reliable solution. However, researchers should exercise caution when using it to analyze soil erosion over different time scales, particularly if the rate of morphological change is low.</p>