Advanced field methods of carbon (C) analysis should now be capable of providing repetitive, sequential measurements for the evaluation of spatial and temporal variation at a scale that was previously unfeasible. Some spectroscopy techniques, such as laser-induced breakdown spectroscopy (LIBS), have portable features that may potentially lead to clean and rapid alternative approaches for this purpose. The goal of this study was to quantify the C content of soils with different textures and with high iron and aluminum concentrations using LIBS. LIBS emission spectra from soil pellets were captured, and the C content was estimated (emission line of C (I) at 193.03 nm) after spectral offset and aluminum spectral interference correction. This technique is highly portable and could be ideal for providing the soil C content in a heterogeneous experiment. Dry combustion was used as a reference method, and for calibration a conventional linear model was evaluated based on soil textural classes. The correlation between reference and LIBS values showed r = 0.86 for medium-textured soils and r = 0.93 for fine-textured soils. The data showed that better correlation and lower error (14%) values were found for the fine-textured LIBS model. The limit of detection (LOD) was found to be 0.32% for medium-textured soils and 0.13% for fine-textured soils. The results indicated that LIBS quantification can be affected by the texture and chemical composition of soil. Signal treatment was shown to be very important for mitigation of these interferences and to improve quantification.
Increased greenhouse gas (GHG) concentrations in the atmosphere and consequent global warming are of great concern to a large part of the world’s population. Actions to reduce GHG emissions and increase mitigation are expected from governments and those involved in agriculture, which is recognized as one of the major sources of these gases. Soil carbon (C) content is one of the essential parameters used to evaluate the impact of land use change on the basis of soil C stocks and soil C sequestration [
A variety of techniques is used to estimate soil organic matter, and more specifically soil C content [
Laser-induced breakdown spectroscopy (LIBS) has been assessed as an alternative method to quantify total soil C content [
Advanced field methods for C analysis should now be capable of providing repetitive, sequential measurements for the evaluation of spatial and temporal variation at a scale that was previously unfeasible [
Soils were collected from the Southeast Livestock Research Center of Embrapa, located in São Carlos, state of São Paulo, Brazil. These soils were from a livestock area, in a transitional texture area with two major soil types: Red Latosol and Red Alfisol. Soil samples were collected from different areas of treatment management, at depths of 0 to 100 cm; six replicates were collected for each soil management. Considering all of the treatments, the total number of soil samples analyzed was 240.
Soil texture classification of these samples was performed using texture triangle classification, as determined by the pipette method, and based on the United States Department of Agriculture (USDA) system of particle size.
Soil samples were air-dried, crushed and ground in a mortar, then passed through a 0.150 mm sieve. The total C concentration reference was analyzed using a 2400 CHNS/O analyzer series II from Perkin-Elmer. These same homogenized soil samples were also used to produce pellets for LIBS analyses. Pellets were prepared according to Nicolodelli et al. [
The LIBS spectra were obtained using a LIBS2500 (Ocean Optics, USA) system. This system includes seven spectrometers that are capable of producing a resolution of ∼0.1 nm (FWHM) for spectral analysis ranging from 188 to 980 nm, a Q-switched Nd:YAG laser at 1064 nm (Quantel, Big Sky Laser Ultra50), an ablation chamber, a lens for laser focalization, and an optical system to collect the plasma emission and send it to the spectrometers. A laser pulse of 50 mJ energy and duration of 8 ns were used for all measurements. The laser had a fluence of 1.2 × 103 J∙cm−2 and the spot sampling diameter of the soil pellets was 73 mm. The delay time (relative to a Q- switch delay) and integration time used were 10 μs and 2 ms respectively, which are instrumental fixed conditions. The fixed distance between the radiation collecting lens and plasma sampling was approximately 1 cm.
For each sample, 60 measurements were made, and each measurement corresponded to two laser shots, which generated an average spectrum. Before analysis, a laser shot was also used to clean the sample surface.
LIBS spectra have a small number of representative C signals that are intense and have high transition probabilities; in the case of the soil matrix, the most intense spectral lines suffer from interference caused by the concomitant emission of elements commonly present in Brazilian soils, mainly Fe, interfering C emission line at 247.86 nm, and Al, interfering C emission line at 193.03 nm. Carbon quantification using the C line at 247.86 nm without proper correction to eliminate or reduce Fe interference is difficult [
Absolute relative errors were individually calculated from each prediction model and the LOD was estimated. The direct analysis of solid to LOD calculation is not trivial and not standardized because the blank (sample without the analyte) is difficult obtain [
Considering the spectral resolution of 0.1 nm, it is expected that the C emission line at 193.03 nm is affected spectral interference from Al lines: Al (II) (ionic) at 193.04 nm, and Al (I) (atomic) at 193.16 and 193.58 nm [
Emission spectrum of a Brazilian tropical soil sample, from 190 to 980 nm (a); The C and Al lines at approximately 193 nm used in this work are also shown (b)
Lower correlation obtained between C emission intensity at 193.03 nm measured by LIBS and C (%) amount determined by CHNS from soil samples
normalization of the intensity of the emission line of C (I) at 193.03 nm. For calibration, the ratio of the intensity values of lines 193.03 nm (a.u.)/193.58 nm (a.u.) was correlated with the C content values that were previously determined by the CHNS analyzer.
Previous papers by our group [
The spectral correction in LIBS measurements is important in order to minimize matrix effects and to mitigate the Al interference [
Comparison of the predicted results of the validation samples with LIBS and the reference values obtained by CHNS is shown in
. Soil C content determined by CHNS analyzer (average of two measurements) at a soil sampling depth of 0 - 5 cm, and C content difference (in percentage) calculated from different field replicates
C content value ± standard deviation | Maximum difference in percentage | ||||||
---|---|---|---|---|---|---|---|
Pasture system | Replicate1 | Replicate 2 | Replicate 3 | Replicate 4 | Replicate 5 | Replicate 6 | |
Management 1 | 5.0 ± 0.1 | 2.19 ± 0.01 | 1.2 ± 0.1 | 2.07 ± 0.02 | 1.6 ± 0.1 | 1.37 ± 0.04 | 76% |
Management 2 | 1.83 ± 0.01 | 2.43 ± 0.02 | 2.68 ± 0.02 | 2.02 ± 0.01 | 1.84 ± 0.02 | 1.38 ± 0.03 | 48% |
Management 3 | 2.41 ± 0.03 | 2.81 ± 0.01 | 2.71 ± 0.04 | 2.5 ± 0.03 | 2.13 ± 0.03 | 3.01 ± 0.00 | 30% |
Management 4 | 1.06 ± 0.01 | 1.51 ± 0.00 | 1.6 ± 0.0 | 1.87 ± 0.03 | 1.62 ± 0.02 | 1.2 ± 0.0 | 42% |
Management 5 | 3.34 ± 0.03 | 2.17 ± 0.01 | 2.34 ± 0.04 | 1.74 ± 0.02 | 2.36 ± 0.01 | 2.24 ± 0.07 | 49% |
. Calibration equations estimated for soil data as a function of soil texture and for all soils combined. Linear determination coefficient (r2) was obtained for each calibration equation
Soil texture | Clay amount g∙kg−1 | Calibration equation | r2* |
---|---|---|---|
No division | 160 - 530 | 0.67 | |
Fine-textured soil | 430 - 530 | 0.84 | |
Medium-textured soil | 160 - 350 | 0.70 |
*p < 0.05.
Predicted C concentration for soil samples versus C concentration measured by CHNS technique. (a) All samples; (b) Medium-textured soils; (c) Clay soils
of mean absolute relative error related to the soil texture models were 21% for medium-textured soils and 14% for fine-textured soils. Better correlation and smaller errors were found in the fine-textured soil LIBS model, with higher concentrations of C. In addition, the LOD was 0.13% for fine-textured soils and 0.32% for medium- textured soils.
The results indicated that LIBS quantification can be affected by the texture and chemical composition of soil. Soils rich in Fe prevent the use of the spectral line at 247.86 nm in portable systems with a spectral resolution of 0.1 nm. The C emission line at 193.03 nm is interfered with by the Al emission line, so in soils with a high Al concentration it is necessary to perform spectral correction to obtain calibration models. It was shown that portable systems with low spectral resolution associated to calibrated models (taking into account soil texture and Al line interference as proposed by this study) could be used to predict soil C content for medium-textured soils (with mean error of 21% and LOD of 0.32%) and for fine-textured soils (with mean error of 14% and LOD of 0.13%).
In conclusion, this study showed the feasibility of using LIBS, even with low resolution, for soil C quantification. This method could be useful for carbon counting, for producing survey maps and for providing precise data collection from different experimental areas or management systems. Considering the heterogeneous nature of the soil, the availability of a technique that allows the study of spatial variability of soil C stocks may greatly improve modeling studies for management systems and forecast scenarios. Thus, these results can be used as a basis for the future development of portable systems.
The authors express their gratitude for the grant from FAPESP (2010/09211-6), CNPq (560292/2010-4) CAPES, Embrapa and Pecus Research Network.
*Corresponding author.