Assessing silage maize yield and quality variability using NIRS, Sentinel-2, and PlanetScope multispectral imagery: a precision agriculture approach

The current advances in remote sensing and the availability of high-resolution and high-frequency multispectral images from satellite platforms can enhance the implementation of Precision Agriculture (PA). This study used vegetation indices (VIs) derived from Sentinel-2 and PlanetScope images and linear regression models to assess silage maize yield (<i>Zea Mays</i>), moisture, and qualitative spatial variability regarding Neutral Detergent Fiber (NDF), Acid Detergent Fiber (ADF), and starch content at the field scale. Maize was harvested by a combined harvester equipped with a yield monitor and a NIRS sensor on a 13-ha study field in North-East Italy, which collected more than 10,000 georeferenced yield, moisture, and quality observation points. 11 Sentinel-2 and 28 PlanetScope images were collected, covering the whole cultural cycle, from April to August 2020). Green Normalized Difference Vegetation Index (GNDVI) and Normalized Difference Vegetation Index (NDVI) were calculated for each image. Multiple regression models were tested for yield, moisture, NDF, ADF, and starch content. The capabilities of models to assess crop variability in terms of R² were evaluated throughout the growing season. PlanetScope GNDVI was the best predictor of yield variability very early in the season, 63 days after sowing (R²=0.33). PlanetScope GNDVI provided the highest R² value of 0.48 for crop moisture content at harvest day while ADF, NDF and starch content variability was best predicted by PlanetScope NDVI with R2 between 0.13 and 0.23. Sentinel-2 and PlanetScope VI’s showed similar predictive capabilities, with an early season peak for yield and quality parameters between 63 and 79 days after sowing. Based on the results, there is a potential for assessing crop moisture variability in the proximity of harvest, which is crucial for the ensilage process, while the early season yield and quality assessment can provide major information for precise fertilization management.