A comparison of canopy and soil proximal sensing to implement selective harvesting in viticulture

Viticulture is often characterized by high variability within vineyards for pedological features and vine vigor. Moreover, climate change, which is altering the seasonal weather cyclicity, is an increasingly marked variability factor that affects vine responses. The knowledge of variability factors and its effects are an indispensable prerequisite for the proper implementation of selective harvesting. This objective is achievable by proximal sensing systems that allow monitoring variability and identifying homogeneous zones classified on the basis of their potential functionality (e.g., soil water retention, grapevine vigor, etc.). To implement selective harvesting, two different methods of proximal sensing for crop (NDRE, normalized difference red edge) and soil (ECa, apparent electrical conductivity) were used. Collected data were analyzed and interpolated by k-means clustering to make thematic maps. A subdivision of two zones per plots was identified. Grape sampling was performed, and the main qualitative and quantitative parameters were analyzed. The overall results showed that NDRE grapevine vigor maps followed fairly well the soil spatial variability trend. However, the zoning process into two main zones at low and high vigor – comparing the maps obtained at different depth levels of ECa with NDRE – generated spatial surface variations between 12 and 33%. Therefore, the combination of several variables in the geospatial model is preferable to generate prescription maps for selective harvesting in viticulture.