Quantifying the content and composition of soil carbon in the laboratory is time-consuming, requires specialised equipment and is therefore expensive. Rapid, simple and low-cost accurate methods of analysis are required to support current interests in carbon accounting. This study was completed to develop national and state-based models capable of predicting soil carbon content and composition by coupling diffuse reflectance mid-infrared (MIR) spectra with partial least-squares regression (PLSR) analyses. Total, organic and inorganic carbon contents were determined and MIR spectra acquired for 20 495 soil samples collected from 4526 locations from soil depths to 1 m within Australia’s agricultural regions. However, all subsequent MIR/PLSR models were developed using soils only collected from the 0–10, 10–20 and 20–30 cm depth layers. The extent of grinding applied to air-dried soil samples was found to be an important determinant of the variability in acquired MIR spectra. After standardisation of the grinding time, national MIR/PLSR models were developed using an independent test-set validation approach to predict the square-root transformed contents of total, organic and inorganic carbon and total nitrogen. Laboratory fractionation of soil organic carbon into particulate, humus and resistant forms was completed on 312 soil samples. Reliable national MIR/PLSR models were developed using cross-validation to predict the contents of these soil organic carbon fractions; however, further work is required to enhance the representation of soils with significant contents of inorganic carbon. Regional MIR/PLSR models developed for total, organic and inorganic carbon and total nitrogen contents were found to produce more reliable and accurate predictions than the national models. The MIR/PLSR approach offers a more rapid and more cost effective method, relative to traditional laboratory methods, to derive estimates of the content and composition of soil carbon and total nitrogen content provided that the soils are well represented by the calibration samples used to build the predictive models.

No abstract

Soil organic carbon (OC) exists as a diverse mixture of organic materials with different susceptibilities to biological decomposition. Computer simulation models constructed to predict the dynamics of soil OC have dealt with this diversity using a series of conceptual pools differentiated from one another by the magnitude of their respective decomposition rate constants. Research has now shown that the conceptual pools can be replaced by measureable fractions of soil OC separated on the basis of physical and chemical properties. In this study, an automated protocol for allocating soil OC to coarse (>50 µm) and fine (≤50 µm) fractions was assessed. Automating the size fractionation process was shown to reduce operator dependence and variability between replicate analyses. Solid-state 13C nuclear magnetic resonance spectroscopy was used to quantify the content of biologically resistant poly-aryl carbon in the coarse and fine size fractions. Cross-polarisation analyses were completed for coarse and fine fractions of 312 soils, and direct polarisation analyses were completed for 38 representative fractions. Direct polarisation analyses indicated that the resistant poly-aryl carbon was under-represented in the cross-polarisation analyses, on average, by a factor of ~2. Combining this under-representation with a spectral analysis process allowed the proportion of coarse- and fine-fraction OC existing as resistant poly-aryl C to be defined. The content of resistant OC was calculated as the sum of that found in the coarse and fine fractions. Contents of particulate and humus OC were calculated after subtracting the resistant OC from the coarse and fine fractions, respectively. Across the 312 soils analysed, substantial variations in the contents of humus, particulate, and resistant carbon were noted, with respective average values of 9.4, 4.0, and 4.5 g fraction C/kg soil obtained. When expressed as a proportion of the OC present in each soil, the humus, particulate, and resistant OC accounted for 56, 19, and 26%, respectively. The nuclear magnetic resonance analyses also indicated that the use of a 50-µm sieve to differentiate particulate (>50 µm) from humus (≤50 µm) forms of OC provided an effective separation based on extents of decomposition. The procedures developed in this study provided a means to differentiate three biologically significant forms of soil OC based on size, extent of decomposition, and chemical composition (poly-aryl content).

Two long-term field trials in South Australia were used to detect and characterise changes in soil biological properties that were a consequence of different agricultural management. The properties examined were total bacteria, fungi, and actinomycetes; total pseudomonads; cellulolytic bacteria and fungi; mycorrhizal fungi; plant root pathogens (Gaeumannomyces graminis var. tritici, Rhizoctonia solani, Pythium irregulare); bacterial-feeding protozoa; soil mesofauna (collembola and acari); earthworms; microbial biomass; C and N mineralisation; in situ CO2 respiration; cellulose decomposition; and soil enzyme activity (peptidase, phosphatase, sulfatase). The sensitivity of these biological properties was assessed to tillage (no-tillage v. conventional cultivation), stubble management (stubble retained v. stubble harvested), crop rotation (continuous wheat v. wheat-sown pasture), and N fertilisation (nil v. 80 kg N/ha applied during the crop phase). Tillage, stubble management, crop rotation, and N fertilisation significantly (P<0.01) affected C mineralisation and microbial biomass. Tillage with stubble management significantly affected root pathogenic fungi, protozoa, collembola, earthworms, and cellulose decomposition. Crop rotation affected mycorrhizal fungi, protozoa, and soil peptidase activity, and N fertiliser had a significant effect on mycorrhizal fungi, protozoa, and cellulose decomposition. As these biological properties are responsive to agricultural management, they may have potential as bioindicators. Total bacteria, fungi, and actinomycetes, cellulosedecomposing bacteria and fungi, soil phosphatase and sulfatase activity, and N mineralisation were less affected by these treatments and may therefore have limited potential as bioindicators.

No abstract