Simulating multispectral MSI bandsets (Sentinel-2) from hyperspectral observations via spectroradiometer for identifying soybean cultivars

Abstract Monitoring soybean areas by remote sensing is extremely useful, especially in Brazil, which has a large territorial extension and where soybean cultivation has spread to all regions of the country. In this sense, the development of remote sensing techniques that enable the quantification and discrimination of soybean areas and now in cultivated cultivar level is of crucial importance for the soybean production chain in Brazil. This study aimed to discriminate soybean cultivars as a function of different hyperspectral bands using the sensor-system MSI-Sentinel-2 (Vis-NIR-SWIR) as a simulation and sample sizes using multivariate statistics to determine if the specific bands of this sensor are capable of performing such discrimination. Four soybean cultivation areas in the Midwest region cultivated with four cultivars (BMX Potencia, NA5909, Don Mario, and FT Campo Mourao) were assessed. Spectral readings from each sample soybean leaf were performed, and a total of 2400 vegetation spectral readings were obtained. Data were composed of 28 bands and 22 reflectance factor height (RID) values for each soybean cultivar. Multivariate statistical analysis was performed to verify the association between soybean cultivars and their relationship with hyperspectral bands, as well as to verify the possibility of cultivar differentiation based on hyperspectral bands. The results obtained demonstrated to be possible discriminate soybean cultivars by using multivariate techniques applied to multi and hyperspectral data. The bands that contributed significantly (>5%) to cultivar differentiation in order of importance were: B26, B27, A17, A21, A20 and A14. Discriminant analysis was efficient in the cultivar classification, and canonical variable analysis revealed bands associated with specific discrimination of each cultivar. Bands that most contributed to cultivar discrimination were also identified for MSI orbital sensor.