[In-site total N content prediction of soil with Vis/NIR spectroscopy]

The Vis/NIR spectroscopy as an efficient tool to predict within-filed soil properties is significantly valuable when establishing agricultural field trials and in precision farming. The object of the study was to investigate the feasibility and possibility of using transformed in-site spectra by relative transformation method (RTM) to prediction soil properties. One hundred and three samples of paddy and fluvo-aquic soil in central china were collected. The in-site moisture (R(w)) and dried (R(d)) Vis/NIR spectra were measured by ASD field handHeld analyzer. The spectral characteristics of two kind soils were analyzed comparatively. The Rw spectra were transformed by RTM into R(n), which were of similar information content and charatistics with R(d). The first derivatives of three spectra revealed that the method could reduce the water disturb on and noise in R(w) Vis/NIR spectrum. The PLS regession model was applied to predict total nitrogen (TN) respectively using R(w), R(d) and R(n) as predictor. The models with Rw predicted TN respectively of paddy, fluvo-aquic and all samples with poor adjusted r2 (< 0.5), while R(d) with good adjusted r2 0.70, 0.88 and 0.71 and R(n) 0.53, 0.62 and 0.64. The result showed that the RTM was efficient to enhance analysis and prediction of soil properties using Vis/NIR spectrum measured on the spot. The combination of PLS and RTM could help implemention of real-time analyzing soil properties using Vis/NIR spectrum.