Evaluating Vegetation Index Mapping over a Vineyard from an Alternative Low-Cost Platform

Résumé

Low-cost manned aerial platforms serve as a new remote sensing tool capable to assist in decision making for the sustainable management of extensive vineyard areas. In this case a PPG (powered paraglider) trike was chosen, providing a great flexibility to quickly perform viticultural mapping in high spatial, temporal and spectral resolution using an onboard Condor MS4 (Visible – Near Infra-Red) multispectral camera with four bands (Red-Green-Blue-NIR). At the same time to the aerial data acquisition, a radiometric campaign on field was carried out to radiometrically calibrate the sensor through the vicarious method based on reflectances. Thus, low-cost calibration targets were positioned in the study area and characterized by the spectroradiometer, taking the Ground Sample Distance into account. In order to characterize the bidirectional reflectance effects within the field of view of the camera, a high number of samples were acquired over the targets. The aerial data were registered simultaneously to a Landsat 8 image acquisition.A photogrammetric workflow was developed to process the multispectral data in order to generate a four-band mosaic image required for GIS integration. This orthoimage was spectrally converted into georeferenced vegetation indices (VIs) maps related to vineyard vigour and health. The Normalized Difference Vegetation Index (NDVI) was selected as the most used index, and other considered for their normalization to brightness and soil color variation like the Soil Adjusted Vegetation Index (SAVI), the Modified Soil Adjusted Vegetation Index (MSAVI) and the Generalized Soil Adjusted Vegetation Index (GESAVI). The article also analyzed the propagation of uncertainties upscaling the physical variables from the very high resolution of aerial data to medium resolution of Operational Land Imager on board Landsat 8.This paper proves the great potential of low-cost remote sensing platforms to collect multispectral images for precision viticulture applications. Moreover, the proposed framework through close-range photogrammetric techniques demonstrates that it is highly feasible to generate quantitative mapping products such as vineyard VI maps to crop assessment. Lastly, the VI map analysis coming from the calibration coefficient variations in relation with the different spatial resolution of the imagery shows the quantity of the dependence of these VIs to the radiometric factors