In recent years, unmanned airborne systems (UAS) have gained a lot of interest for their potential use in precision agriculture. While the imagery from near-infrared (NIR) enabled off-the-shelf cameras included in UAS can be directly used to facilitate crop scouting, the application in quantitative analyses remains cumbersome. The ultimate goal is to calculate (nitrogen) prescription maps from vegetation indices obtained from UAS imagery, but two main issues hamper this workflow: (1) the derivation of surface reflectance values from camera digital numbers, and (2) the design of vegetation indices that deal with the specific characteristics of UAS imagery. The former issue needs to be resolved to enable multi-temporal analyses (and even to account for within-flight image to image variations due to scattered clouds), but traditional approaches cannot be used given the lack of spectral sensor calibration data to convert digital numbers (DN) to at sensor radiance and the lack of ancillary data to convert radiance to surface reflectance. The second issue is important to consider since UAS images are characterized by a very high ground sample distance (GSD) and are acquired under a wide range of solar zenith angles. The resulting irregular shadows from buildings, trees and even small crops over and between individual plants that are prominently visible within individual images are not eliminated by atmospheric correction. Moreover, these shadows cause artifacts in conventional vegetation indices (such as the normalized difference vegetation index, NDVI) that have been designed to work at the canopy level at GSDs and scales where individual plants and shadows are hardly visible or significant. Here, we visualize the effects described above and describe an empirical approach to determine if, and how these issues can be solved. For the derivation of surface reflectances, we present a correlation study of vegetation indices derived from UAS imagery (Trimble UX5) with NDVI values of natural and manmade features from conventional remote sensing platforms and RTK-georeferenced measurements with a terrestrial close range active crop NDVI sensor (Trimble GreenSeeker), in both homogenous overcast and blue sky conditions. We use the empirical line method on targets with known reflectance properties to test the influence of different normalization approaches and evaluate different vegetation indices that are statistically and analytically related to the NDVI. Based on the results, we discuss the overall feasibility of deriving prescription maps from UAS imagery, and the importance of filter choice and design of the vegetation index to obtain results with an acceptable correlation to conventionally obtained NDVI maps in multi-temporal datasets covering realistic ambient conditions over a crop test field.