Interannual Variation in UAV Vegetation Index Relationships with Grapevine Yield and Quality: A Two-Year Case Study in Madeira

Unmanned aerial vehicle (UAV)-derived vegetation indices are increasingly used to characterize vineyard canopy variability, yet their relationships with grapevine yield and grape quality may vary between growing seasons. This two-year case study evaluated associations between multispectral vegetation indices and grapevine productivity and quality parameters in a commercial Verdelho vineyard on Madeira Island during the 2024 and 2025 growing seasons, which differed markedly in accumulated precipitation (77.3 and 249.1 mm, respectively), while thermal accumulation increased from 1411 GDD to 1525 GDD in 2025. UAV multispectral imagery was acquired at key phenological stages using a MicaSense Altum sensor. NDVI, NDRE, GNDVI, GLI, and NGRDI were evaluated in relation to yield and grape-quality parameters using Spearman rank correlations based on nine independent experimental replicates per season. To support interannual comparison despite unequal UAV survey availability, seasonal analyses were additionally based on four phenologically matched stages represented in both years. Statistical uncertainty and robustness were assessed using bootstrap confidence intervals, leave-one-out sensitivity analysis, Benjamini–Hochberg false discovery rate correction, and treatment-adjusted partial rank correlations. Relatively strong nominal associations were observed in 2024, particularly involving NDRE and GNDVI, whereas fewer nominal date-specific associations occurred in 2025. No nominal associations were detected for the phenologically matched seasonal means in 2025. However, none of the date-specific or seasonal-mean associations remained statistically significant after FDR correction. Treatment adjustment preserved the direction of all nominal associations but reduced their magnitude in most cases. The results indicate contrasting exploratory interannual patterns rather than statistically confirmed temporal instability. UAV-derived vegetation indices provided useful information on vineyard spectral variability, but relationships with yield and grape quality identified within a single season require validation across larger multi-year and multi-site datasets before broader temporal consistency or transferability can be established.