S18 - Session O3 - Precision fruticulture in Flanders - A four year study on sensing techniques, data analysis and variable rate applications

Authors: Joke Vandermaesen *, Stephanie Delalieux, Bjorn Rombouts, Yannick Smedts, John Bal, Dany Bylemans, Serge Remy

In recent years, precision agriculture is rapidly advancing. Sensing techniques are becoming more and more available, yielding an immense amount of data. However, how to process, visualize and interpret those data and translate them into variable rate applications, remains a challenge. In order to develop applications of precision fruticulture, two pear orchards were monitored intensively for four consecutive years using soil scans, soil moisture sensors and RGB and multispectral drone imagery. Simultaneously, ground truth data were gathered, including flower intensity, vegetative growth and fruit yield. To evaluate which sensing techniques yield useful information, all data were compared with each other and in time using Principle Component Analysis and linear regression. The most promising correlations were observed for soil EC scans and certain indices derived from drone imagery, i.e. (i) the number of white pixels vs. flower intensity, (ii) seasonal evolution in NDVI vs. fruit yield and (iii) NDRE vs. drought stress. Based on this knowledge, various variable rate applications were tested in the orchard, i.e. the use of (i) soil EC data for variable fertilisation and pruning, (ii) flower intensity mapping for variable chemical thinning and (iii) soil moisture sensing for variable irrigation. Furthermore, a dashboard was developed for visualisation, comparison and interpretation of relevant GIS data, including an automated processing chain for orchard drone imagery.