Refining the methods for determining soil organic carbon stocks in South Africa

Abstract

The aim of this research was to refine the methods for determining soil organic carbon stocks (SOCS) in South Africa. There were a few gaps that this study tried to address that the Verified Carbon Standard methodology, and the Gold Standard methodology are not being clear on or specific about. These gaps have a direct impact on the South African industry and includes: (i) the uncertainty regarding which analytical method should be used for determining soil organic carbon (SOC) content in South Africa, (ii) the unavailability of a near-infrared (NIR) calibration algorithm for SOC content in South Africa, (iii) the uncertainty on how to map SOCS at field scale in South Africa, and (iv) the uncertainty of using either a modelling or re-measuring approach be for quantifying SOCS at field scale in South Africa. Key objectives of this study included to: (i) evaluate the preferred method to be used for determining SOC content and the application for South Africa conditions, (ii) develop a NIR calibration algorithm for SOC content in South Africa on three different scales, (iii) determine how to map SOCS specifically at field scale in South Africa, and finally (iv) compare SOCS quantification approaches specifically at field scale in South Africa. The following paragraphs indicate the methodology for achieving the objectives, as well as the main findings and suggestions for the way forward. Accurate quantification of SOC content is essential for the assessment of carbon credits. In South Africa, the standard methodologies for carbon credit assessment does not specify which analytical method should be used for determining SOC content. The study aimed to determine which analytical method should be used for determining SOC content for the assessment of carbon credits. Secondly, it determined whether pedotransfer functions could be used for transferring SOC content values between methods, especially for transferring historic determined SOC content to more recently determined values using more modern methods. Two-hundred-and-twenty topsoil (0-30 cm) samples were collected from five catchments in South Africa and analysed for SOC content with the three analytical methods: Walkley-Black wet-oxidation (WB), total dry combustion (TDC) and loss-on-ignition (LOI). The study found that the TDC method provided the most reliable SOC content values and should still be considered the preferred method for determining SOC content for the assessment of carbon credits in South Africa. The WB method should be avoided if a soil is expected to have a high SOC content, while the LOI method could still be used for determining SOM, however, this method should be avoided when determining SOC content. The study also reached the second aim by successfully creating pedotransfer functions between all three methods. However, only the WB and TDC methods had a very strong relationship (R² = 0.91) and showed that accuracy start to decrease significantly above 2.5% SOC content. Therefore, the pedotransfer function (SOCWB = -0.157 + 0.895 x SOCTDC - 0.0149 x SOCTDC² - 0.000606 x SOCTDC³) could be used for transferring SOC content values with SOC content up to 2.5%. Near-infrared (NIR) spectroscopy has emerged as an easy, rapid and cost-effective alternative for SOC analysis and the accounting of carbon credits. South Africa currently lacks a calibration algorithm for predicting SOC content from NIR spectroscopy. This study aimed to develop a NIR spectroscopy calibration algorithm for SOC content, specific to South Africa. Soil samples were collected from 2 fields and 3 catchments across South Africa. These samples were analysed using the total dry combustion (TDC) method and scanned with a NIR spectrometer. Sixty NIR calibration algorithms were developed on a regional scale. The impact of methodological parameters, such as sample state, sampling design, processing and machine learning models, on the root mean square error (RMSE) of the validation statistics was also assessed. Although 60 regional-scale calibration algorithms were developed, none were suitable (RMSE = 0.39 % and RPIQ > 2) for SOC content prediction, which was attributed to the small sample size (n = 238). However, local calibration models for the Tsitsa catchment and Ottosdal fields presented great accuracy (RMSE < 0.1 and RPIQ > 1.5) that can be used for future SOC content prediction. The study found that the open spectral library global prediction model poorly predicted SOC content using local data (RMSE = 1.23 % and R² = - 0.83). This was attributed to South African samples being underrepresented in the global dataset. Sample state and sampling design were the most influential parameters affecting RMSE. To develop a national calibration algorithm, effort should be placed on developing accurate calibration algorithms for smaller areas that could be added to the national spectral library. Reliable SOCS maps are important for accurately assessing carbon credits. To address this gap, the study aimed to determine whether conventional mapping using ordinary kriging (OK) or digital soil mapping (DSM) with machine learning (ML) should be used for mapping SOCS at field scale in South Africa. Fifty samples from three depths, 0-5, 5-15 and 15-30 cm were collected from two farms in South Africa and analysed for SOC content using the TDC method, while for dry bulk density (ρb) a pedo-transfer was created for predicting the ρb at unsampled locations. Maps for SOC content, ρb and SOCS were created using OK and DSM with ML and assessed with evaluation statistics including root mean square error (RMSE) and Lin's concordance correlation coefficient (ρc). A calculate-first approach and map-first approach were used for calculating the total SOCS for each study site. Findings of this study demonstrated that DSM with ML should be used rather than OK for mapping SOCS at field scale in South Africa. Utilising DSM with ML will lead to more accurate and reliable SOCS maps. However, the results also indicated that the SOC content, ρb and SOCS maps for DSM with ML need to be improved. To increase the accuracy of maps generated using DSM with ML, additional soil samples could be collected, and field scale covariates should be incorporated. The results were inconclusive regarding whether the calculate-first approach or the map-first approach should be used when utilising OK or DSM with ML. Although both approaches have several advantages and disadvantages, more research is needed to determine which approach should be used when utilising OK or DSM with ML. However, the study concluded that the calculate-first approach should be endorsed, due to this approach being simpler and might lead to more reliable and accurate SOCS estimations. There is an ongoing debate about the most appropriate approach for quantifying changes in soil organic carbon stocks SOCS. This study aimed to compare two SOCS quantification approaches, modelling and re-measure, on field scale in South Africa. The baseline assessment (2022) for two fields in South Africa was established beforehand and afterwards, fifty samples from three depth increments, 0-5, 5-15 and 15-30 cm were collected, and the SOCS calculated for two study sites (2024) using the re-measure approach and the modelling approach using the Rothamsted carbon (RothC) model. The sampling cost was also estimated against the value of sequestered SOCS (2022-2024) and compared between the two approaches. The main finding from the study indicated that it was inconclusive if a modelling or re-measure approach should be used for quantifying SOCS on field scale in South Africa. The spatial variation of SOC was the main factor influencing the total SOCS, sequestered SOCS and carbon credit values. The results indicated that the modelling approach using the RothC model slightly underestimated the total SOCS and sequestered SOCS when utilising all 50 observations for both study sites. The industry-represented, "One-observation", did not have a consistent trend and showed overprediction and underprediction for both study sites using the different approaches. This had significant financial implications using both approaches. The study also found that the spatial variation of the SOC must be taken into account at the onset of measuring the SOC content in year one, independent of whether the SOCS would be modelled or re-measured. The carbon sequestration industry must take the model limitations (such as field scale dynamics and sensitivity to short-term changes) into account, and find ways, such as an improved sampling design to address these accuracies of RothC at field scale.