The Acquisition of Spectral Reflectance Measurements under Field and Laboratory Conditions as Support for Hyperspectral Applications in Precision Farming

Spectral reference measurements in different resolutions are required for many applications in remote sensing. Although, theoretically, in-situ measurements represent the conditions at the time of data acquisition by airborne or satellite sensors in an optimum way, they are strongly influenced by illumination, moisture, geometry, and surface roughness. To minimize these effects laboratory measurements of soil samples are used since they allow to better control the measurement conditions. Therefore, many spectral libraries are based on laboratory measurements and many empirical relations for determining spectrally derived parameters depend on data acquired under laboratory conditions. In this study the reflectance of soil samples under field and laboratory conditions were compared under controlled illumination and geometry conditions. Differences were mainly caused by soil aggregates, which reduce the spectral albedo in the wavelength domain between 600 and 2400 nm. This affects the estimation of the inorganic and organic carbon content from the reflectance data using a partial least square (PLS) regression. Several spectra pre-treatment techniques were tested to reduce the soil aggregate influence and to optimize the modeling results.