*3.1. Relationship between Vegetation Indices and Wheat Yield Using Sentinel-2 Imagery*

Figure 3 shows correlation matrices for the three different dates of VIs derived from S2 and yield. A high degree of collinearity among the different VIs was observed on the three dates, with Day 2 (GS39-49) showing the strongest correlation between indices with r values above 0.9. In addition to the negative correlation, when compared to the other VI results, the CI index obtained lower r values, ranging from −0.59 to −0.76 (Figure 3, Day 2). Correlations between the VIs of Day 1 (GS30) were slightly lower but remained above 0.8. Furthermore, the correlation between different VIs for Day 3 is not homogeneous (as indicated by the broader color palette of the matrix), and r values varied from 0.97 to 0.58. Given the high degree of collinearity observed among the VIs, some measures were taken to address this issue during the implementation of the different algorithms. One such measure employed was the Lasso correction of the MLR method.

**Figure 3.** *Cont*.

**Figure 3.** Correlation matrix between the different VIs of the three days (Days 1–3). The last column shows the correlation with the wheat grain yield.

In terms of the relationship between the different VIs and yield, the highest values were measured for Day 2 (GS39-49). Except for the negative correlation of CI (−0.65), the values ranged from 0.81 for GRVI to 0.73 for GDVI. In comparison, for Day 3 (GS69-75), the correlations ranged from 0.72 for GNDVI to 0.54 for GDVI. Although the increase in correlation is not significant (−0.66 compared to −0.65), CI was the only VI that increased the correlation. The lowest correlations were found with the VIs of Day 1 (Figure 3), with values ranging from 0.51 (RVI and GRVI) to 0.26 for GDVI. The correlation of CI was inverse (−0.45). Overall, the highest value was obtained with GRVI for all three dates, whereas GDVI exhibited the lowest values.
