Hyperspectral Remote Detection as an Alternative to Correlate Data of Soil Constituents

Imaging spectroscopy is a proven technology for identifying and mapping the constituents of soils based on their spectral signatures, the variation of their reflectance or emissivity across different wavelengths, in the visible and infrared bands. Despite the great potential for use of hyperspectral images, there is a lack of previous work that examines the hyperspectral imaging sensors and spectroradiometers, especially in controlled laboratory configurations. Besides that, AisaFENIX is the only sensor of this model available for research in Brazil. In our country there is no experience related to this sensor, being a previous and necessary step of the use of the sensor in aero-surveys. We need to understand the operation and behavior of this sensor. In addition, Brazil is a hot and humid country, with no robust weather models that are in keeping with the local need. This work aims to correlate soil data obtained in the laboratory by two different hyperspectral sensors, AisaFENIX and spectroradiometer, in different bands of the electromagnetic spectrum. Soils were sampled in Mandaguacu, Parana, an area of approximately 200 ha, totaling 207 georeferenced samples. The soil reflectance laboratory data were obtained using the Fieldspec 3 Jr spectroradiometer sensor, which collects a spectral range of 350–2.500 nm (vis-SWIR), and the reflectance values of the hyperspectral images were acquired with the AisaFENIX sensor, which collects a spectral range of 350–2.500 nm. The correlation between the sensors was satisfactory (R2 0.97) in the corresponding iron range (500–640 nm), and for the corresponding bands of kaolinite (2.200 nm), gibbsite (2.265 nm), water (1.400 and 1.900 nm), the R2 values found were 0.61, 0.65, 0.63 and 0.64 respectively. The lowest correlation coefficients found were for the bands of organic carbon (600–1.100 nm), goethite (480 nm), hematite (530 nm) and specific goethite (450–480 nm), with R2 values less than 0.41. The t-test showed that the correlations were significant (p