Publications des agents du Cirad


NIRS for predicting quality indexes in the organic fertiliser industry

Thuriès L., Davrieux F., Bastianelli D., Bonnal L., Oliver R.. 2005. In : 12th International Conference on Near Infrared Spectroscopy, Auckland, New Zealand, 9-15th April 2005. s.l. : IM Publications, 1 p.. International Conference on Near Infrared Spectroscopy. 12, 2005-04-09/2005-04-15, Auckland (Nouvelle-Zélande).

Combined with the ability of improving the major soil properties, the use of composts and organic fertilisers can reduce the cost of fossil-fuel consumption for chemical-fertiliser synthesis and concomitant emission of green house gases. Nevertheless, composts and organic fertilisers may vary according to the sources of raw materials, their elaboration process and producers. In such conditions, it can be very useful to have global estimators of organic fertilisers quality. The C/N ratio can provide information on the capacity of an organic input to be transformed in humus, but it is not sufficient in some cases. The Lignin/Nitrogen ratio is another parameter used in modelling the transformations of organic materials. Tr is an index of the content in carbon 'resistant to degradation on the long term'. This parameter in currently being evaluated in a normalisation procedure. Quality indicators based on chemical (carbon and nitrogen) and biochemical (e.g. lignin) analyses are expensive and time-consuming. Thus we attempted to predict some quality estimators as the C/N and Lignin/N ratios, and the Tr index directly by NIRS. The raw materials studied originated from industrially pre-processed plant residues and other tropical plant residues, potentially utilisable in composting. The measured parameters were Tr (n=122) and Lignin/N (n=76) and Cestim/N (n=270) which is a C/N ratio estimated from the Organic Matter and Total Kjeldahl Nitrogen. Due to the heterogeneity of fresh materials, samples were dried (40°C) and ground (<1 mm sieve) before being scanned on a NIRS 6500 (Foss NIRSystems) in ring cups. Spectra were corrected with SNVD 2,5,5 (WIN-ISI) mathematical pre-treatment and calibrations were performed using a modified partial least square regression (mPLS, WIN-ISI). Despite a wide diversity of raw materials, the equations presented interesting Standard Errors of Calibration (SEC= 5.59; 1.83; 1.71), and determination coefficients (R2= 0.96; 0.94; 0.88) for Tr, Lignin/N, and Cestim/N, respectively. Standard Errors of Cross Validation (SECV= 5.96; 2.36; 1.94 for Tr, Lignin/N, and Cestim/N, respectively) were acceptable with reference to the nature of the variables. The corresponding SD/SECV ratios (RPDcv= 4.7; 3.1; 2.5) were satisfactory for Tr and Lignin/N, while Cestim/N estimation was less accurate. It can be concluded that the Tr and Lignin/N indexes, used as quality estimators of compost and organic fertilisers seem to be valuably estimated by NIRS. The Cestim/N index can be used as a rough classification tool in absence of true C reference values. (Texte intégral)
Communication de congrès

Agents Cirad, auteurs de cette publication :