Effects of sample pre-treatments on the analysis of liquid organic manures by visible and near-infrared spectrometry

Abstract

Proper application of a fertilizer requires precise knowledge of its nutrient composition. In the case of liquid organic manures (LOM), this information is often lacking due to heterogeneous nature of these fertilizers. Published “book values” of nutrient contents present the average from a wide range of possible nutrient characteristics, but usually differ considerably from the concentration in a particular manure. Thus, chemical analyses are recommended before applying the specific LOM. Unfortunately, this is usually too costly and time-intensive in practical farming. On-farm analysis by optical spectrometry in the visible and near-infrared (Vis-NIR) range is considered as an efficient alternative. However, calibration of Vis-NIR spectrometry for LOM is challenging as shown in many studies. One reason is LOMs’ tendency to rapidly segregate into a fuzzy continuum with liquid and solid properties. By separating LOM into well-defined liquid and solid phases and measuring them separately, calibration of Vis-NIR spectrometry might be improved. In this study, the effects of four sample pre-treatment techniques on the prediction accuracy of macronutrients (N, P, K, Mg, Ca, S), micronutrients (B, Mn, Fe, Cu, Zn), dry matter and pH of LOM using visible and near infrared spectrometry were comprehensively investigated. The concentrations were referred either to wet basis or to dry matter basis. For the study, a total of 163 samples, separated in two similar LOM sets (pig, cattle, digestates), were either dried, filtered, or centrifuged and always compared to non-treated samples. The experiments demonstrate that in comparison to raw samples (Ø r2 = 0.85) neither filtering (Ø r2 = 0.76 for filtrates and Ø r2 = 0.71 for filter residues), centrifugation (Ø r2 = 0.59 for supernatants and Ø r2 = 0.79 for pellets), nor drying (Ø r2 = 0.74) revealed to be a helpful preparation step significantly improving prediction results, independent from referring to wet or dry basis concentrations.