Technological advances continue to improve our ability to detect landforms and landscape changes. Remote sensing can provide geomorphological information at larger scales than previously possible but interpreting this information can be more challenging than for field data. Measuring and mapping roughness elements, such as boulders and large wood, is essential for understanding geomorphic processes and restoration in many landscapes where these are abundant. Mapping roughness elements from aerial orthoimagery is common but could produce different results than field measurements due to the 2D nature of the imagery and the variable ability to detect boulders through water and vegetation. We compared measurements of river boulders from aerial imagery collected by uncrewed aerial vehicles (UAVs) to direct measurements in the field. We surveyed boulder size, density and spatial distributions using both approaches at eight river reaches in northern Sweden. We found that the density, coverage and size of boulders mapped from UAVs were strongly correlated with those from field measurements, giving confidence in UAV methods. However, the UAV approach consistently resulted in fewer boulders (30% lower density), lower boulder coverage and smaller boulders (7% smaller mean b-axis) compared to field measurements. The difference between field and UAV measurements was strongly associated with river bankfull depth. Therefore, we conclude that UAV measurements should be restricted to sites with low depth or high visibility through the water column and where bias in boulder detection with depth is not likely to influence study conclusions. In reaches with many boulders, overlap of boulders also likely reduces the suitability of aerial imagery. We conclude that aerial imagery has high potential for mapping landforms in rivers but is not directly equivalent to field studies and the implications of hiding by water, sediment and vegetation need to be considered.