Monitoring soil substrate influence in vineyards using Sentinel-2 time series and land surface phenology

This work explores the potential of Sentinel-2 time-series imagery to investigate the differences in vineyards associated with varying soil substrates. The investigation was carried out by selecting vineyards in the Piana Rotaliana wine growing area of the autonomous province of Trento (Italy). Consistent time-series datasets were defined by selecting vineyards trained under the same trellis system and avoiding plantation renewal between 2017 and 2023. To extract Land Surface Phenology (LSP) the best configuration between vegetation indices and fitting methods for deriving LSP features was investigated. The results indicate that Enhanced Vegetation Index 2 (EVI2) provides better stability with respect to Normalized Difference Vegetation Index (NDVI). For each vineyard statistical metrics were derived like LSP, Growing Season (GS; good observation within the LSP time period) and off Season (good observation outside the LSP time period) metrics. Eight datasets were defined and a Random Forest model was employed to assess classification accuracy and evaluate its stability across the years. Findings suggest that the substrate effect can be distinguished both from off Season and Growing Season metrics. Vegetation water content indices, such as the Global Vegetation Moisture Index (GVMI), emerged as the most effective and temporally stable predictors. The off Season datasets provided generally better results with respect to GS datasets when the model of one year was tested over the remaining years. The results provide valuable insights into the potential influence of soil characteristics on grapevine response over time, by highlighting the water retention capacity of the substrate and the vineyard response