Epsilon Archive for Student Projects

Abstract

Ley fields’ dynamics can be spatially heterogeneous, and an understanding of within-field yield variability is important for improving fertiliser efficiency leading to sustainable forage production. This thesis assessed spatial variability in Swedish ley fields by comparing field-based dry matter yield estimates with Sentinel-2-derived yield-zone classifications, spectral indices, soil properties, meteorological data and topographic features. High- and low-yield zones had been pre-selected using multi-year Sentinel-2 Normalized Difference Vegetation Index (NDVI) medians. The results showed that the NDVI-based classification was mostly supported by estimated yield in ley fields. Most fields yielded higher than average yields in expected high-yield zones. However, the potential of this yield-classification varied between fields and was less effective in some cases. Other spectral indices showed plausible relationships with estimated yield depending on the farms’ location and forage growth stages. Soil properties, topography and meteorological conditions provided useful field-specific context, but no single factor consistently explained the observed yield patterns across all sites. Overall, the study suggests that NDVI-based yield-zone classification can be a useful starting point for expected spatial productivity patterns in ley fields but should be combined with repeated field measurements and environmental data for more reliable interpretation. In addition, more advanced multivariate or machine-learning regression methods could be a better option for measuring within-field biomass variability.