Using Virtual Drones to Mitigate the Bias Introduced by Sensor Wavelength Approximations in Crop Monitoring with Drones

Remote sensing based on the reflectance of light at certain wavelengths enables the calculation of various vegetation indices (VIs) as proxies for agronomic variables. However, drone-mounted sensors have a limited number of bands, so the wavelengths defining VIs often have to be modified in line with sensor characteristics. This article addresses the problem of such wavelength shift based on experimental agronomic measurements and on reflectances acquired by both multispectral spectrophotometers and drone-mounted sensors. We demonstrate that wavelength shift can significantly affect VIs, particularly those using the red-edge band, compared to a multispectral reference. In the worst cases, the drone’s VI was not even correlated with its multispectral target. We therefore propose a calibration method using a “virtual drone” simulated from a complete dataset obtained by multispectral measurements in order to use sensors with a limited number of bands. Virtual drones can guide the choice of drone sensors, depending on the features to estimate, or facilitate the intercalibration of sensors for comparisons of the results of the literature studies. This study aims at providing the agronomist community with a method for intercomparing VIs acquired by drones.