Investigating the potential of visible and near-Infrared spectroscopy (VNIR) for detecting phosphorus status of winter wheat leaves grown in long-term trial

The determination of plant nutrient content is crucial for evaluating crop nutrient removal, enhancing nutrient use efficiency, and optimizing yields. Nutrient conventional monitoring involves colorimetric analyses in the laboratory; however, this approach is labor-intensive, costly, and time-consuming. The visible and near-infrared spectroscopy (VNIR) or hyperspectral non-imaging sensors have been an emerging technology that has proved its potential for rapid detection of plant nutrient deficiency and nutrient status monitoring. However, most studies in this respect have focused primarily on nitrogen and few research were conducted to understand the specificity of measuring phosphorus using this technique. In this study, we investigated the potential of leaf spectral reflectance in the visible and near infrared spectral region to predict phosphorus (P) status in winter wheat leaves. The research was conducted in a long-term experiment, which was installed in 1896 at the Gembloux Agro-Bio Tech faculty. The trial includes various fertilization modalities ensuring phosphorus contrast and variability in data acquired. The spectra acquisition and leaves biomass sampling were done synchronously at different stages of the wheat growth cycle. The reflectance measurements were done on the two youngest fully expanded leaves using the ASD FieldSpec4 spectroradiometer. The recorded spectra, between 350 nm and 2 500 nm at a 1 nm interval, were corrected for light scattering using multiple scatter correction (MSC). Results from partial
Proceedings of the 15th International Conference on Precision Agriculture
June 28 – July 1, 2020, Minneapolis, Minnesota, United States page 2
least squares regression (PLSR) with leave-one-out cross-validation (LOOCV) and 4 latent variables provided a root mean square error (RMSEcv) and a determination coefficient (R2cv) at respectively 0.94 mg/g and 0.71. The obtained model predicted leaf phosphorus status with a ratio of standard deviation to RMSEcv (RPDcv) of 1.9. The cross-validation results showed that the developed PLS predictive model has some potential to detect P status in wheat fresh leaves by identifying 2 classes of P and that using Vis-NIR spectroscopy is a practical option to measure leaf phosphorus concentrations.