Research article
Rapid prediction of particulate, humus and resistant fractions of soil organic carbon in reforested lands using infrared spectroscopy

https://doi.org/10.1016/j.jenvman.2017.02.013Get rights and content

Highlights

  • We developed rapid methods to predict TOC, POC, HOC and ROC in reforested soils.

  • TOC fractions were measured using automated wet-sieving and 13C NMR in soil subsets.

  • Robust MNIRS-PLSR models were built for TOC and its fractions and cross-validated.

  • TOC and its fractions were accurately predicted in reforested and agriculture soils.

  • MNIRS suits rapid and accurate assessment of C sequestration in different land uses.

Abstract

Reforestation of agricultural lands with mixed-species environmental plantings can effectively sequester C. While accurate and efficient methods for predicting soil organic C content and composition have recently been developed for soils under agricultural land uses, such methods under forested land uses are currently lacking. This study aimed to develop a method using infrared spectroscopy for accurately predicting total organic C (TOC) and its fractions (particulate, POC; humus, HOC; and resistant, ROC organic C) in soils under environmental plantings. Soils were collected from 117 paired agricultural-reforestation sites across Australia. TOC fractions were determined in a subset of 38 reforested soils using physical fractionation by automated wet-sieving and 13C nuclear magnetic resonance (NMR) spectroscopy. Mid- and near-infrared spectra (MNIRS, 6000–450 cm−1) were acquired from finely-ground soils from environmental plantings and agricultural land. Satisfactory prediction models based on MNIRS and partial least squares regression (PLSR) were developed for TOC and its fractions. Leave-one-out cross-validations of MNIRS-PLSR models indicated accurate predictions (R2 > 0.90, negligible bias, ratio of performance to deviation > 3) and fraction-specific functional group contributions to beta coefficients in the models. TOC and its fractions were predicted using the cross-validated models and soil spectra for 3109 reforested and agricultural soils. The reliability of predictions determined using k-nearest neighbour score distance indicated that >80% of predictions were within the satisfactory inlier limit. The study demonstrated the utility of infrared spectroscopy (MNIRS-PLSR) to rapidly and economically determine TOC and its fractions and thereby accurately describe the effects of land use change such as reforestation on agricultural soils.

Introduction

The soil organic carbon (SOC) pool is one of the largest terrestrial stores of carbon (C), containing about 1500 Gt C (Schlesinger, 1986). Consequently, even small changes in the amounts of soil C in response to land use, management or climate may have large effects on global C cycling and climate change. Reforestation is implemented around the world to sequester C and improve environmental conditions (e.g., water quality and habitat availability, Cunningham et al., 2015b). The total area of world's planted forest in 2010 was estimated to be 264 million ha, making up 6.6% of the world's forest area (FAO, 2010). In Australia, mixed-species environmental plantings are established on previous agricultural land for C sequestration and other environmental outcomes accounting up to 20% of the 1.14 million ha of reforestation between 1990 and 2012 (Paul et al., 2013). Studies have focused on improving measurement and modelling of biomass C following reforestation with environmental plantings (Paul et al., 2015, Perring et al., 2015), and measured associated changes in SOC (Cunningham et al., 2015a).

Conventional measurement of SOC as total organic C (TOC) using chemical oxidation (Walkley and Black, 1934) or dry combustion (Merry and Spouncer, 1988) is inadequate to explain changes in SOC in terms of soil physical, chemical and biological activity. Partitioning TOC into fractions that are related to active, intermediate or slow and passive or inert conceptual pools used in soil C turnover models (e.g., RothC, Jenkinson et al., 1992; CENTURY, Parton et al., 1987), can help elucidate their role in soil processes (Skjemstad et al., 2004). Such fractions may represent labile, humified and inert C with turnover times respectively of the order of annual (<3 year), decadal (20–50 year) and millennial (>1000 year, Jenkinson and Coleman, 1994). Several functional pools are known which are accessible by different fractionation methods. Physical fractions are such as aggregates, particle sizes and density fractions, chemical fractions are usually extracts (DOM, soil microbial biomass, organic matter soluble in alkali and acid, etc.) and also combinations of fractionation methods are used (von Lützow et al., 2007).

Skjemstad et al. (2004) demonstrated that the RothC model could be initialised and changes in soil C stock simulated by replacing the conceptual stocks of resistant plant material (RPM), humus (HUM) and inert organic matter (IOM), respectively, with operationally measured particulate (POC), humus (HOC) and resistant (ROC) organic C. Subsequently, Janik et al. (2007) showed that the concentrations of POC, HOC and ROC in <2 mm sieved soil could be predicted from mid-infrared (MIR) spectra acquired from finely ground samples.

Diffuse reflectance mid-infrared spectroscopy (MIRS) is a rapid, non-destructive and low-cost technique, demonstrated to be suitable for routine analysis of a variety of soil properties (Soriano-Disla et al., 2014). The MIRS technique requires minimal sample preparation (i.e. air drying and fine grinding) and no use of hazardous chemicals. MIR spectra and corresponding analytical data in multivariate analyses such as partial least squares regression (PLSR) can be combined to develop prediction models for soil attributes. The predictive ability of MIRS-PLSR techniques for total, organic and inorganic C of soils has been well investigated and reported (Grinand et al., 2012, Madari et al., 2006, Madhavan et al., 2016). However, the use of MIRS-PLSR to predict the concentrations of fractions of TOC is limited (Bornemann et al., 2010, Janik et al., 2007, Zimmermann et al., 2007). MIRS-PLSR prediction models have been developed for total, organic and inorganic C, and TOC fractions (Baldock et al., 2013a), and these models were applied to predict the content of TOC fractions in subsequent agricultural soil C studies (Karunaratne et al., 2014, Rabbi et al., 2014). However, the applicability of such predictive models to soils under woody vegetation has not been investigated. Further, there is increasing interest in reforestation of agricultural lands to mitigate greenhouse gas emissions through sequestering C in woody biomass (Canadell and Raupach, 2008) and in soil (Lal, 2005). Thus, there is a need to develop accurate and efficient measurement techniques and prediction models applicable to reforested land to understand and predict their potential to sequester C and mitigate greenhouse gas emissions.

An extensive study was conducted to investigate the changes in TOC and its fractions following reforestation with mixed-species environmental plantings at 117 sites from temperate, Mediterranean-type and tropical climatic regions of Australia, for the purpose of calibrating a soil C accounting model (FullCAM, Brack and Richards, 2002) developed from the RothC model (Paul et al., 2015), and to provide measurements and predictions of TOC fractions. Prediction models for TOC fractions have been reported in agricultural soils by Baldock et al., 2013a, but these models are unlikely to be accurate for reforested soils because of the difference in plant inputs, chemistry and decomposition rates between land uses, and thereby changes in TOC fractions (Del Galdo et al., 2003, Berthrong et al., 2012, Cunningham et al., 2015a). This warrants a need to develop infrared spectroscopic prediction models for TOC fractions that are suitable for soils under reforestation. This work presents the first attempt to develop prediction models for TOC fractions in a treed ecosystem whereas previous work has focused on agricultural soils (Baldock et al., 2013a). Our objectives were to measure soil TOC fractions (POC, HOC and ROC) in a representative set of reforested soils; develop robust infrared and PLSR prediction models for TOC and its fractions for reforested soils; and predict TOC and its fractions for both environmental planting and reference agricultural soils (i.e. pastures and cropping).

Section snippets

Soils

Soils were collected from 117 sites, each comprising a mixed-species environmental planting paired with an adjacent agricultural land use (Fig. 1). Details of site characteristics and sampling methods are detailed in England et al., 2016 and summarised here. The sites were across southern and eastern Australia (latitude 30.9 to 38.7 °S, longitude 117.4 to 150.3 °E) and covered the range of rainfall zones where environmental plantings occur (380–1150 mm y−1). Generally the sites represented a

Soil TOC fractions from environmental plantings

The concentration ranges of TOC and its fractions in the environmental plantings soils used for fractionation are provided in Table 1. The recovery of TOC after fractionation was similar to that obtained by Baldock et al. (2013b) and ranged between 87 and 112%, with a mean recovery of 94%, and a standard deviation of 5.6%. The absolute deviation of recovery of TOC fractions indicated that for >90% of the samples, TOC recovery was within ±2 g kg−1 °C with the variation in absolute recovery

Conclusions

We developed accurate MNIRS-PLSR predictive models for TOC and TOC fractions under reforested environments for application on a national/continental scale. The study demonstrated the potential of MNIRS-PLSR technique to rapidly and cost-effectively predict soil TOC and its fractions and thereby assess the magnitude of changes following reforestation. The technique has the potential for use in regional to global carbon accounting in reforested ecosystems. There is a need to build a broad

Acknowledgements

The study was funded by the Australian Government's Filling the Research Gap program. TRC is supported by an Australian Research Council (ARC) Future Fellowship (FT120100463). MPP was partially supported through an ARC Laureate Fellowship to R. Hobbs. We acknowledge M. Bullock, T. Fairman, R. Greene, P. Holliday, R. Law, J. McGowan, T. Morald, T. Murshed, A. Schapel, G. Szegedy and A. Wherrett for provision of soil samples and field and laboratory support.

References (50)

  • B.E. Madari et al.

    Mid- and near-infrared spectroscopic assessment of soil compositional parameters and structural indices in two Ferralsols

    Geoderma

    (2006)
  • D.B. Madhavan et al.

    Mid-infrared spectroscopy for rapid assessment of soil properties after land use change from pastures to Eucalyptus globulus plantations

    J. Environ. Manage

    (2016)
  • K.I. Paul et al.

    Economic and employment implications of a carbon market for integrated farm forestry and biodiverse environmental plantings

    Land Use Policy

    (2013)
  • K.I. Paul et al.

    Improved models for estimating temporal changes in carbon sequestration in above-ground biomass of mixed-species environmental plantings

    For. Ecol. Manage

    (2015)
  • M.P. Perring et al.

    Soil-vegetation type, stem density and species richness influence biomass of restored woodland in south-western Australia

    For. Ecol. Manage

    (2015)
  • S.M.F. Rabbi et al.

    The relationships between land uses, soil management practices, and soil carbon fractions in South Eastern Australia

    Agric. Ecosyst. Environ.

    (2014)
  • M. von Lützow et al.

    SOM fractionation methods: relevance to functional pools and to stabilization mechanisms

    Soil Biol. Biochem.

    (2007)
  • M. Zimmermann et al.

    Quantifying soil organic carbon fractions by infrared-spectroscopy

    Soil Biol. Biochem.

    (2007)
  • ASTM International

    ASTM D6122-06, Standard Practice for Validation of the Performance of Multivariate Process Infrared Spectrophotometers

    (2006)
  • J.A. Baldock et al.

    Predicting contents of carbon and its component fractions in Australian soils from diffuse reflectance mid-infrared spectra

    Soil Res.

    (2013)
  • J.A. Baldock et al.

    Quantifying the allocation of soil organic carbon to biologically significant fractions

    Soil Res.

    (2013)
  • S.T. Berthrong et al.

    Soil C and N changes with afforestation of grasslands across gradients of precipitation and plantation age

    Ecol. Appl.

    (2012)
  • L. Bornemann et al.

    Particulate organic matter at the field scale: rapid acquisition using mid-infrared spectroscopy

    Soil Sci. Soc. Am. J.

    (2010)
  • J.G. Canadell et al.

    Managing forests for climate change mitigation

    Science

    (2008)
  • S.C. Cunningham et al.

    Reforestation with native mixed-species plantings in a temperate continental climate effectively sequesters and stabilizes carbon within decades

    Glob. Change Biol.

    (2015)
  • Cited by (23)

    • Carbon dynamics in tree plantings: How changes in woody biomass impact litter and soil carbon

      2022, Forest Ecology and Management
      Citation Excerpt :

      Soil carbon (0–30 cm) studies providing the best constraint to the calibration were those that included measurement of fractions of SOC, including particulate (POC or RPM), humus (HOC or HUM), and resistant (ROC or IOM) organic carbon. These fractions were measured using calibrated mid-infrared spectroscopy (MIRS) and Partial Least Squares Regression models (Baldock et al. 2013a; Baldock et al. 2013b; Madhavan et al., 2017). This was done for 46% of the dataset collated, with the remainder having only total organic carbon (TOC) measured, generally using a CNS analyser (LECO Corporation, St. Joseph, MI, USA) (Table S6).

    • Highly decomposed organic carbon mediates the assembly of soil communities with traits for the biodegradation of chlorinated pollutants

      2021, Journal of Hazardous Materials
      Citation Excerpt :

      Diffuse reflectance spectra in the mid and near-infrared (MNIRS) spectra (7800 −450 cm–1 at 8 cm–1 resolution) were acquired for all samples using a PerkinElmer Frontier FT-NIR-MIR Spectrometer (PerkinElmer Inc., Waltham, MA, USA) equipped with a KBr beam-splitter, a DTGS detector and AutoDiff automated diffuse reflectance accessory (Pike Technologies, Madison, WI, USA). Spectra were pre-processed as described in Madhavan et al. (2017) and concentrations of total organic C (total-C), resistant organic C (resistant-C), humic organic C (humic-C) and particulate organic C (particulate-C) were predicted from MNIRS spectra of the Australian Soil Carbon Research Project (ScaRP) (Baldock et al., 2013) using partial least squares (PLS) regression. From these spectra and methodology, resistant-C was defined as the proportion of char and lignin C (enriched in poly-aryl C) in both fine and coarse fractions, and considered to be biologically resistant.

    • Soil organic carbon and aggregation in response to thirty-nine years of tillage management in the southeastern US

      2020, Soil and Tillage Research
      Citation Excerpt :

      Management changes are reflected faster in the active than passive SOC fractions (Kaiser et al., 2014; Pandey et al., 2014; Six et al., 1999; Wander and Bidart, 2000). Thus, partitioning total SOC into different fractions can illustrate management-induced changes in C accrual in a more realistic manner (Gabarrón-Galeote et al., 2015; Madhavan et al., 2017; Skjemstad et al., 2004). In general, active SOC fraction is closely associated with soil health and agronomic benefits (Lal, 2006; Quiroga et al., 2006) while passive fraction contributes to long-term C sequestration (Kumar et al., 2014).

    View all citing articles on Scopus
    View full text