Infrared (IR) spectroscopy has been widely used as one of the most important analytical techniques in many research fields for decades. Environmental samples, such as waste materials and soils are very complex, consisting of a wide range of different organic compounds and making them hard to analyse. A relatively new spectroscopic technique called Fourier transform infrared photoacoustic spectroscopy (FTIR-PAS) has gained increasing interest for characterising the chemical composition of environmental samples and predicting the usefulness of samples for bioenergy production, fertilization, or other uses. FTIR-PAS is a rapid and non-destructive method, and it is especially suitable for opaque samples (such as manure, biochar, and soil, etc.). FTIR-PAS has been investigated as an alternative method for estimating the usefulness or quality of the samples as conventional laboratory assays are usually time-consuming, tedious, and relatively expensive. Through multivariate data analysis, various important properties of the samples can be extracted and predicted by FTIR-PAS.
The overall objectives of this thesis are to use FTIR-PAS for characterisation of organic wastes and soil, and for assessment of their quality with different purposes. Three manuscripts are presented in this thesis.
Paper 1 identified the unique part of FTIR-PAS spectrum characterising permanganate oxidizable carbon (POXC) which is independent from soil organic carbon (SOC). POXC is a fraction of SOC, which is considered labile and sensitive to changes in land use. The objective of Paper 1 was to extract the information in the FTIR-PAS spectra that is unique to POXC (independently from SOC), and to characterise the chemical components related to the POXC. A soil sample set (575 samples) was collected from four different countries (Laos, Malaysia, Peru, and Thailand), and a high correlation between SOC and POXC was observed (R2=0.84). The predictions of both POXC and SOC in soil samples using FTIR-PAS combined with partial least squares regression (PLSR) were successful, regarding the coefficient of variation (R2) and the root mean square error (RMSE) of cross validation. However, due to the high correlation between POXC and SOC, it is difficult to know if the POXC prediction is directly associated with POXC itself or based on its strong correlation with SOC. Therefore, a method extracting the unique FTIR-PAS spectral features of POXC (independently of SOC) was applied. The results showed the POXC prediction was mainly based on its high correlation with SOC. Around 11% of the spectral information used to predict POXC was identified as unique and four dominant peaks were found to represent the POXC fingerprint.
In Paper 2, the feasibility of using FTIR-PAS to predict the biochemical methane potential (BMP) of various urban organic waste (UOW) was investigated. The spectra of UOW samples were recorded and PLSR models predicting BMP were calibrated. Different transformations of the spectra were applied to obtain the best performance of the BMP models. The overall best BMP prediction achieved an R2 of 0.86 and RMSE of 59.3 ml CH4/g VS in a test set validation with a standard normal variate (SNV), detrending (DT), and first derivative transformation. The results also showed that easily degradable compounds such as aliphatics most likely in lipids and amides in proteins were found to be positively associated with BMP, while hemicellulose, lignin, and other aromatics correlated negatively with BMP.
Paper 3 consists of two parts. In the first part, FTIR-PAS was applied to predict water extractable phosphorus (WEP) of various digestate samples collected from both laboratory-scale digesters and biogas plants. WEP is useful for estimating the plant available phosphorus (P) in waste and therefore can characterise the usefulness of digestate as a P fertilizer. The results showed that FTIR-PAS was capable of predicting WEP of digestates with an R2 of 0.80 and RMSE of 0.78g/kg in cross validation, which can be characterised as a moderately successful model with a ratio of performance to deviation (RPD) of 2.26. In the second part, the digestates from the biogas plants were used to amend three different soils and then the P availability in digestate-amended soils was investigated. The P availability of the amended soils was assessed using the diffusion gradients in thin films (DGT) technique. The technique determines the concentration of phosphate at the interface of the DGT devices (CDGT) which acts as an infinite sink. The results show that the CDGT correlates well with plant availability. In addition, the FTIR-PAS spectra of the amended soils were recorded. Afterwards, the recorded FTIR-PAS spectra were used to calibrate a prediction model of CDGT. An R2 of 0.70 and a RMSE of 134.09 μg P L-1 were obtained as the best performance in cross validation. There is a clearly potential of using FTIR-PAS to predict P availability in digestate-amended soils with a moderately precise model (RPD of 1.86).