Are vegetation indices derived from consumer-grade cameras mounted on UAVs sufficiently reliable for assessing experimental plots?

• Vegetation indices based on consumer-grade cameras mounted on UAVs are as good as ground-based recordings.
• Angular variation in reflectance has to be considered to avoid misleading conclusions.
• Image overlap and stitching software are influential factors affecting vegetation indices.

Recent technological advances in UAV (unmanned aerial vehicle) technology offer new opportunities for assessing agricultural plot experiments using UAV imagery. Vegetation indices (VIs) based on aerial images derived from consumer-grade cameras are a simple and cheap alternative compared to VIs derived from proximal (on-ground) sensors. The objective of this study was to assess whether VIs derived from consumer-grade cameras mounted on UAVs are reliable and whether there are any shortcomings in image acquisition and analysis that need to be addressed before their general application. This objective was investigated using a rotary-wing and a fixed-wing UAV, true colour (RGB) and colour-infrared (CIR) cameras, four different VIs (ExG, NGRDI, NDVI and ENDVI), altitudes in the range of 30–100 m, different ambient lighting conditions and two different software packages for stitching images together. Results were compared with ground-based recordings by consumer-grade cameras and multispectral sensors. Field experiments in cereals were used to evaluate the assessments. The study showed that VIs based on UAV imagery have the same ability to quantify crop responses to experimental treatments as ground-based recordings with cameras and advanced sensors. However, there are shortcomings that need to be taken into consideration: (1) angular variation in reflectance (bidirectional reflectance), (2) stitching and (3) ambient light fluctuations. Bidirectional reflectance was so extensive that it could lead to misleading conclusions in sunny conditions and this effect could be amplified further by stitching. A procedure for avoiding impacts from bidirectional reflectance is demonstrated when plots were cropped from individual images and a procedure is suggested for stitching images. Camera, VIs and image acquisition altitude were of minor importance, but fluctuating ambient lighting conditions is an issue that should be addressed in future studies.