Modified Hot-Spot analysis for spatio-temporal analysis: a case study of the leaf-roll virus expansion in vineyards

Given a set of geo-referenced data points, the Getis–Ord Gi* statistic identifies hot-spots of points with values higher in magnitude than one might expect by a random chance. This tool works by looking at each data feature and its neighboring features in comparison to the overall spatial distribution of the phenomenon explored. If the difference between the local sum for a feature and its neighbors is highly larger than expected (the overall sum) a hot-spot is accepted. Leaf-roll virus (LRV) in vineyards appears in clusters which expand from year to year when no pest control is carried out. Exploring the spatio-temporal expansion of hot-spots of the LRV is limited with the Gi* statistics since relative hot-spots are accepted according to the infestation level of a specific year. A modified Gi* was developed which identifies year-to-year hot-spots which are relative to a year of reference. LRV symptoms were mapped yearly in a vineyard from 2005 to 2010. The Gi* indicated for a northern hot-spot only in 2007 and a southern one in 2009. Using the modified Gi* with 2005 as a year of reference, the northern and the southern clusters were identified in 2006 and 2007, respectively.