Aerial images were taken in June 2014 with a multispectral VIS/NIR camera of the canopy from 14 kale (Brassica oleracea var. acephala L.) fields in Canterbury, New Zealand before this forage was grazed by cows. Images were taken at 716m and at 1,372m flying altitude. Calculating the Green Normalised Difference Vegetation Index (GNDVI) from green and NIR channels proved to be the best representation of yield (dry matter) variation from manual biomass cuts in these fields. Several hundreds of individual images covering parts of the fields were semi-automatically stitched to composite images covering full blocks of fields.
Highest coefficients of variation (CV) of GNDVI values in a field are linked with low yield averages, often found at vary patchy fields (CVs of 20%). High yielding fields were less patchy and had CVs of less than 8%. A non-linear calibration curve was derived from the presented data. This functional relationship can explain 70% of the variance of the measured biomass yield data with reflection data of the canopy from these fields. This explaining power does not change when data from aerial images from higher altitudes were analysed. This independency of the preliminary model from height will allow using such an approach with standard high resolution cameras from various platforms (e.g. conventional aircraft, UAV/RPAS).
Grouping the GNDVI data also allows delineating zones of similar yield levels within the forage fields. Such zoning enables farmers to adapt fertilizer application to the yield expectation of such zones or to manage the feed provision for their grazing cows in a spatial variable way across and between fields. The zones can be used for directing the manual sampling of biomass in cases when farmers deem estimations of biomass yield from aerial imagery to be inaccurate. For all these steps higher resolutions - associated with lower flying altitudes - are necessary.[Show full abstract]
![[PDF]](/themes/LURA/images/mimes/b_pdf.png "PDF file"){kind=small}
