**5. Conclusions**

Various vegetation indices (SAVI, MSAVI, NDVI, MSAVI2, IPVI, RVI and GEMI), their calculation based on the red and near-infrared radiation derived from the Sentinel−2 imagery, showed similar relationships with grain yield and number of spikes per square meter of winter wheat and triticale as the commonly used NDVI. While comparing the grain yield and the number of spikes from NDVI for all three locations during two seasons, it can be seen that the values of R<sup>2</sup> for the number of spikes per square meter are lower than the values of R<sup>2</sup> for the grain yield. Consequently, the grain yield is a better yield prediction parameter than number of spikes per square meter. In general, the relationship between NDVI values and grain yield was stronger at more advanced growth stages. Depending on the region of Poland, the strongest correlations between NDVI and yield and its main component were obtained for NDVI derived from images obtained at the end of April (beginning of shooting) in southeastern Poland, at the end of May (beginning of heading) in the northeastern part of the country and only at the end of June (at milk maturity) in the central region of Poland. The divergence in these periods may result from the sowing date, but also from di fferent climatic conditions, especially GDD values, in the three micro-climate regions of Poland. The di fferent strengths of the relationships may be caused by the soil water deficiency due to variable weather and soil conditions, depending on the research location, at di fferent growth stages. Within-field water variability was higher, especially on sandy soils (which were common in some parts of the field in location B) at later growth stages when air temperatures were higher and negative balance between rainfall and evapotranspiration was more common than in locations A and C [39]. Moreover, the presence of clouds, which made the registration of useful satellite images impossible on some dates, also contributed to the results obtained in this study. For example, in location C in southeastern Poland, there were no cloudless images available in May in both seasons and, because of this, it was not possible to evaluate the relationships between VIs and yield for some important growth stages of the crop. The proposed estimated time for accurate yield prediction is about 4–6 weeks before harvest (from the beginning of shooting to milk maturity).

**Author Contributions:** Conceptualization, E.P., D.G.; methodology, E.P., D.G., M.S., formal analysis, E.P., D.G., M.S.; investigation, E.P., D.G., M.S., S.S.; data curation, E.P., D.G.; writing—original draft preparation, E.P., D.G., M.S., S.S.; writing—review and editing, E.P., D.G., M.S., D.R., B.B., S.S.; visualization, E.P., D.G.; supervision, D.G.; project administration, D.R., B.B.; funding acquisition, D.R., B.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the European Space Agency, FERTISAT—Satellite-based Service for Variable Rate Nitrogen Application in Cereal Production, gran<sup>t</sup> number 4000118613/16/NL/EM.

**Conflicts of Interest:** The authors declare no conflict of interest.
