Spatial pattern analysis of flavescence dorée repartition in vineyards from the Bordeaux region

Abstract

The devastating incidence of flavesence dorée on vineyards throughout the world motivates for a better comprehension of this epidemic disease. In this study, we highlight the characteristic spatial non-random distribution of flavescence dorée diseased plants on a set of 7 vineyards from the Bordeaux region plot. In a first time, we propose a simple statistical framework using Monte-Carlo simulations in order to assess the randomness of the disease repartition. Several statistics are considered such as the mean distance to the nearest diseased neighbor or divergence (using Kullback-Leibler dissimilarity symmetric variant) of the distance histogram to the average distance histogram of random simulations. The performance of these statistics is first evaluated on a set of generated repartitions at different randomness levels using ROC curves as a visual representation of the risks associated with the test. Histogram comparison was found to be more effective and robust for the detection of non-random configurations. The proposed algorithm is then used on real data, showing significant aggregations patterns and edge effect on some of the plots for flavesence dorée diseased plants but also uprooted plants.