engleski

Hyperspectral sensing of Stagnosols properties under variable mineral nitrogen fertilization

Spectral data contain information on soil organic and mineral composition which can be useful for monitoring of soil quality, especially in domain of precision farming. The objective of the research was to evaluate the possibility of using hyperspectral VNIR spectroscopy for farm-scale prediction of soil properties. Two hundred air-dried and grinded soil samples taken from the field experiment of 4 ha area after winter wheat harvest in 2010 at depth of 30 cm were scanned with FieldSpec®3 spectroradiometer (wavelength range: 350-2500 nm ; ASD Inc., USA) to identify differences in the spectral response of the soil treated with ten different rates of mineral nitrogen fertilizer (0-300 kg N ha-1). Principal component analysis (PCA) revealed certain delineation between higher- and lower-N level treatments conditioned by differences in soil pH, texture and soil organic matter composition after a long-term mineral fertilization under variable climate regime. Partial least square regression (PLSR) was used to build prediction models of total carbon content (TC %), total nitrogen content (TN %) and soil pH based on the visible and near infrared spectra. Very strong correlation and low root mean square error were obtained between predicted and measured values for the calibration and validation dataset, respectively (TC %: R2=0.91 and R2=0.89, RMSEC=0.119 and RMSEV=0.127 ; TN %: R2=0.91 and R2=0.90, RMSEC=0.011 and RMSEV=0.012 ; pH: R2=0.82 and R2=0.65, RMSEC=0.405 and RMSEV=0.572). Results indicated that hyperspectral VNIR spectroscopy is applicable method for more efficient analysis of soil quality properties under intensive crop production.

soil spectroscopy; visible/near infrared reflectance; principal component analysis; partial least square regression; fertilization

nije evidentirano