Development of a Rapid Analysis Method for Soil Carbon, Nitrogen and Organic Matter for South African Forestry Soils

“With the advent of carbon accounting, there is a higher demand for soil organic matter analysis, however, the number of samples required for monitoring and the cost of current analysis makes this activity prohibitively expensive.”

The state of soil organic matter (SOM) in plantations is of growing concern as we understand its role in soil and plantation health, and carbon sequestration. With the advent of carbon accounting, there is a higher demand for SOM analysis, however, the number of samples required for monitoring and the cost of current analysis makes this activity prohibitively expensive. Hence, new methods for rapid, cost-effective and high-throughput analysis need to be developed.

Currently, total soil organic matter is measured using loss-on-ignition (LOI) while total carbon and nitrogen is measured using combustion-gas-analysis methods by CNS analysers. These methods are robust, reliable and reproducible, however they are expensive and have limited throughput. For reference, the cost of analysis for total carbon and nitrogen range from R250 – R450 per sample which becomes expensive at scale and impedes comprehensive assessment of the site studied.

Near infrared reflectance spectroscopy (NIRS) measures the interaction of light with the sample in the NIR range. This interaction creates a NIR signature or spectrum which is then modelled and trained against reference data of the sample to develop a calibration model which then is used to predict specific properties of a sample. The use of NIR predictive models offer non-destructive, rapid, cost-effective and high-throughput analysis of samples, which addresses the shortcomings of current SOM analysis methods. NIR can be used as an alternative tool, allowing for analysis of a greater number of samples and therefore improving site coverage.

At the ICFR Forestry Research and Analytical Laboratory (FRAL) a project was initiated to develop and test NIRS as an alternative method for SOM analysis. This was made possible by our access to samples from various projects covering several forestry regions, stored at the ICFR soil archive. Soil samples were selected to represent the lithologies most commonly encountered in the KZN and Mpumalanga region, with SOM content ranging from low to high.

Methods used to gather reference data are currently used at FRAL for routine SOM analysis.

“The use of NIR predictive models offer non-destructive, rapid, cost-effective and high-throughput analysis of samples, which addresses the shortcomings of current SOM analysis methods”

Total SOM was measured using LOI while total carbon and nitrogen was measured using a Leco Trumac CNS analyser. The NIR spectra of all soil samples were captured using the Bruker MPA NIRS. Three models were created using partial least squares regression to predict total SOM, carbon and nitrogen content respectively. The performance of the model was assessed by its root mean squared error of prediction (RMSEP), whereby the lower the error the more accurate the model was. The R2-value was measured to determine the linearity of the predictions.

Training models for total SOM, carbon and nitrogen had good performance metrics, viz. RMSEP < 1.5% and R2 > 89% showing good correlation with the reference data. The models were assessed against a NIR validation dataset and showed similar levels of performance to that of the training data. This indicated that the models reliably interpreted the NIR signature of the sample and predicted the SOM properties adequately, motivating that NIR SOM analysis can be developed into a tool for routine soil
analysis.

The promising performance metrics of the models warrant discussion about the best practices to deploy them. Considering that these models are predictive, their performance is associated with the sample types used to train them. Additionally, factors such as site lithology, climate and silvicultural practices would have a confounding effect on the performance of the models and need to be considered. Therefore, we propose that the models are calibrated for the sites they are intended to be used on to account for these confounding factors.

TAKE HOME MESSAGE
The study showed that NIR can successfully measure SOM properties from plantation soil samples and can be applied as a site-specific analysis tool. By complementing existing methods with NIR analysis, the cost of analysis could decrease, allowing for a greater number of samples to be submitted and facilitating a more comprehensive site assessment. These models are under active development for improved model performance and flexibility using samples from the Long-Term Site Monitoring project to become representative of the entire Forestry Industry.

“The study showed that NIR can successfully measure SOM properties from plantation soil samples and can be applied as a site-specific analysis tool”

Source: FSA

Back to NEWS:
BUSINESS  l  CLIMATE CHANGE  l  CONSTRUCTION  l  EDUCATION  l  ENVIRONMENT  l  FORESTRY ENGINEERING  l  FORESTRY  l  GOVERNMENT  l  INTERNATIONAL  l  LAND  l  RECYCLING l  RESEARCH  l  ROOFING  l  SHORT HAUL  l  SILVICULTURE  l  SOCIAL RESPONSIBILITY  l  TRANSPORT  l  TREATMENT  l  TRANSPORT  l  VALUE ADDING