The monitoring of the C content of agricultural lands requires long-term experiments, the analysis of a large number of samples, various sampling depths and adequate analytical tools to perform the experiments. Near infrared (NIR) reflectance spectroscopy is a fast, reliable and low-cost analytical tool for the determination of soil C content. The objective of this study was to present NIR spectroscopy technology as a reliable method to monitor C concentration (total C, inorganic C and organic C) of soil samples from a long-term experiment established in 1986 in Cordoba (Spain). The soil is a rain-fed vertisol, characteristic of Mediterranean agricultural systems. The soil samples studied from four different years (1997, 2000, 2003 and 2006) included all of the variabilities existing in the long-term experiment: tillage system (conventional tillage vs no tillage), crop rotations [wheat ( Triticum aestivum L)-chickpea ( Cicer arietinum L), wheat-faba bean ( Vicia faba L), wheat-sunflower ( Helianthus annus L), wheat-fallow and wheat-wheat], N fertiliser rate (0 kg N ha−1, 50 kg N ha−1, 100 kg N ha−1 and 150 kg N ha−1), as well as different soil depths (0–15 cm, 15–30 cm, 30–60 cm and 60–90 cm). Calibration models were obtained with 492 samples from the aforementioned vertisol using visible and NIR spectroscopy (vis-NIR) and, when tested using a validation sample set, returned the values: total C ( r2 = 0.92, SEP = 1 g kg−1, RER = 26, RPD = 4.8); inorganic C ( r2 = 0.95, SEP = 1.5 g kg−1, RER = 26.6, RPD = 4.6) and organic C ( r2 = 0.76, SEP = 0.8 g kg−1 RER = 12.3, RPD = 2). The vis-NIR models developed reliably predicted the total, inorganic and organic C content at plot scale (homogeneous sample set). The advantage deriving from very specific vis-NIR calibrations for prediction of soil C is that one can rapidly and economically predict C in the study area, enabling monitoring to compare soil C evolution depending on time and variables such as tillage, crop rotation and fertilisation. To develop vis-NIR models at local and regional scales would require a more heterogeneous sample set and adequate calibration/validation strategies.
Read full abstract