Robust site-specific radiometric calibration of PlanetScope NDVI using UAS-based reference data for precision agriculture

Index images provide opportunities for monitoring phenotypic characteristics such as plant height and disease symptoms, as well as the detection of pest presence. This allows for targeted application of agrochemicals and the implementation of preventive measures, leading to increased yields and reduced environmental impact. The study focuses on the radiometric inconsistencies between satellite-based imagery and high-resolution unmanned aerial system (UAS) imagery. While PlanetScope provides high temporal resolution, its spectral response often deviates from ground-truth UAS flight missions due to varying atmospheric conditions and sensor specifications. A robust site-specific calibration workflow using Quantile-Based Piecewise Linear Normalization and residual-based outlier mitigation was used. The method was tested across three different crop types: barley, sunflower and maize. Initial results showed significant systematic offset. Following the application of the proposed robust normalization model, the root mean square error (RMSE) was reduced by up to 39%, while maintaining high coefficients of determination. The study demonstrates that site-specific non-linear calibration is essential for reliable crop health monitoring, particularly in heterogeneous fields where soil background and canopy shadows significantly affect the spectral signature satellite data, enabling accurate multi-temporal analysis and better-informed decisions for variable-rate fertilization and yield prediction.