Comparison of Soil Organic Carbon Measurement Methods

Abstract

To enhance agricultural soil health and soil organic carbon (SOC) sequestration, it is important to accurately measure SOC. The aim of this study was to compare common methods for measuring SOC in soils in order to determine the most effective approach among different agricultural land types. The measurement methods of loss-on-ignition (LOI), automated dry combustion (Dumas), and real-time near-infrared spectroscopy (NIRS) were compared. A total of 95 soil core samples, ranging in clay and calcareous content, were collected across a range of agricultural land types from forty-eight fields across five farms in the Southwest of England. There were similar and positive correlations between all three methods for measuring SOC (ranging from r = 0.549 to 0.579; all p < 0.001). On average, permanent grass fields had higher SOC content (6.6%) than arable and temporary ley fields (4.6% and 4.5%, respectively), with the difference of 2% indicating a higher carbon storage potential in permanent grassland fields. Newly predicted conversion equations of linear regression were developed among the three measurement methods according to all the fields and land types. The correlation of the conversation equations among the three methods in permanent grass fields was strong and significant compared to those in both arable and temporary ley fields. The analysed results could help understand soil carbon management and maximise sequestration. Moreover, the approach of using real-time NIRS analysis with a rechargeable portable NIRS soil device can offer a convenient and cost-saving alternative for monitoring preliminary SOC changes timely on or offsite without personnel risks from the high-temperature furnace and chemical reagent adopted in the LOI and Dumas processes, respectively, at the laboratory. Therefore, the study suggests that faster, lower-cost, and safer methods like NIRS for analysing initial SOC measurements are now available to provide similar SOC results as traditional soil analysis methods of the LOI and Dumas. Further studies on assessing SOC levels in different farm locations, land, and soil types across seasons using NIRS will improve benchmarked SOC data for farm stakeholders in making evidence-informed agricultural practices.

1. Introduction

In recent decades, agricultural soils have become degraded due to land practices such as heavy cultivation [1], inefficient soil and forage management [2,3], continuous anthropogenic inputs of chemical fertilisers and pesticides [4], soil erosion, and nutrient loss [5]. These practices lead to further loss of soil organic carbon (SOC) that is essential to sustain both land biodiversity and habitats in recent years [6]. The organic carbon component in soil belongs to carbon derivation from soil organic matter (SOM) of terrestrial soils and plays a pivotal part in the carbon cycle, ecosystem functions, and agriculture worldwide [7,8,9]. To monitor and quantify SOC storage and sequestration for soil health at a landscape scale, a robust, inexpensive, and widely applicable method that can be used by landowners to make timely land management decisions is required.

Previous studies promoted different methods for measuring SOC to monitor soil health and potential carbon sinks [10], such as Walkley-Black (WB) [11], loss-on-ignition (LOI) [12], automated high-temperature dry combustion (Dumas) [13], and near-infrared spectroscopy (NIRS) [2]. These techniques require different amounts of soil sample and associated preparation time and cost. Both WB and LOI methods allow SOC estimation from organic matter content [14]. The WB method involves an acid dichromate oxidation procedure of the organic matter to record the SOC recovery rate, which further produces hazardous chemical waste like chromium (Cr) for detrimental environmental threats [5,14]. The LOI method involves decomposition of organic matter under high thermal combustion at a temperature between 350 and 650 °C [15]. Similar to the LOI method, the Dumas method first quantifies carbon dioxide (CO₂) produced by ignition between 950 °C and 1200 °C, followed by gas separation through chromatography and thermo-conductivity to measure the total carbon content from a soil sample using an electronic elemental analyser [9]. This method has a high precision measurement level but is costly [1]. The use of NIRS has become increasingly more mobile, using a compact spectrometer to measure real-time SOC from fresh soil cores in a field. The NIRS method consists of a spectrum of infrared energy reflected from the sample illuminated during scanning, with SOC estimates being the average of four or five scans on the same soil sample. The method considerably reduces analysis cost [3] and processing and measurement time of samples [16]. Among the mentioned key methods, WB, Dumas, and LOI are laboratory methods involving differing amounts of processing time, cost, and expertise for SOC measurements [14], whereas WB is a relatively hazardous method with waste generation compared to LOI or Dumas. Further, although LOI and Dumas involve soil sample preparation such as oven-drying, grinding, sieving, and weighing [5], they can both run initial SOC analysis in a shorter timeframe of a few hours’ time after the sample preparation with low risk to personnel health safety [3]. NIRS analysis is considered to be a simple, quick, and affordable method to determine field SOC contents without involving laboratory steps for sample preparation before the measurements. To provide unbiased benchmarks of SOC measurements in Gloucestershire for this pilot study, the current study compared the three most common methods for measuring SOC using LOI, NIRS, and Dumas.

The objective of this study was to compare methods for measuring SOC in soils from different agricultural land types and to examine the statistical relationships between these methods by developing newly predicted linear regression conversion equations according to each land type among the chosen methods.

2. Materials and Methods

2.1. Field Data

The study was carried out across five farms, where four farms were located at the Cotswolds National Landscape (51°48′18″ N, 1°55′11″ W), and one was Hartpury University Farm in the Severn Vale (51°54′43″ N, 2°18′39″ W), with a total soil sampling area of 262 ha across all the farms. The farm soils were predominantly clay sandy loams (Luvisol) with an average pH ranging from 6.7 to 6.8 and clay-like shallow calcareous soils (Leptosol) with an average pH ranging from 7.5 to 7.7, respectively, among three key agricultural land use classes for the University and Cotswolds farms, respectively, based on the NIRS technique with description (

Table 1. Summary of field data among the farms studied in the Hartpury and Cotswold national landscape, UK.

Location	Land Type	n *	pH **	Soil Type [17]	Soil Texture [17]	Description [17]
Hartpury	Arable	7	6.7	Luvisol	Clayey sandy loams	Slightly acid loams ≥ 15% clay with impeded drainage
	Temporary ley	24	6.8
	Permanent grass	16	6.7
Cotswolds	Arable	16	7.6	Leptosol	Calcareous loams	Shallow and limestone or chalk-like soils ≥ 25% clay with freely draining
	Temporary ley	16	7.7
	Permanent grass	16	7.5

* n = number of collected and analysed soil samples. ** pH represents the potential of hydrogen, which was measured by real-time near-infrared spectroscopy (NIRS) analysis.

) [17], where the technique will be presented in Section 2.3. Fields studied were categorised using agricultural land use classes in England and Wales [18], which were ‘arable’ with forage or autumn-sown cereal crops, ‘temporary leys’ with a grass sward between two and four years for forage production, and permanent grassland with pasture of five years or more. Temporary ley fields were mainly used for forage production and permanent grass fields for flexible rotational sheep, cattle, and horse grazing. There were in total 228 soil core samples collected from forty-eight fields across the five farms studied with different agricultural land types in the Southwest of England between May and October 2024, where 95 of the samples (48 from all the Cotswold farms and 47 from the Hartpury farm) were randomly selected for the current study comparing SOC analysis methods as a pilot study in Gloucestershire before further on-farm regional studies on a wider scale. Of the total soil samples collected and analysed, there were 23 from arable fields, 40 from temporary ley fields, and 32 from permanent grass fields (Table 1).

During May 2024, when soil samples were collected at farms in the Cotswolds, the amount of rainfall ranged from 0 to 2.6 mm per day and the average daily temperature from 4 to 25 °C (Figure 1). The average daily temperature for May was noticeably higher than the average for the same month during the previous 10 years, from 2014 to 2023. Total rainfall was also noticeably higher in the same month compared to the previous 10 years. During October 2024, when soil samples were collected at Hartpury University Farm in the Severn Vale, the amount of rainfall ranged from 0 to 2.7 mm per day and the average daily temperature from 8.3 to 16.1 °C. The average daily temperature for October was noticeably higher than the average for the same month during the previous 10 years, from 2014 to 2023. Total rainfall was also noticeably lower in the same month compared to the previous 10 years.

2.2. Soil Sampling and Preparation

To ensure representative sampling, 5 samples were taken using the systematic randomised sampling method along a W-pattern across each field [19]. Soil core samples were collected to a maximum depth of 30 cm where possible using a 2 cm diameter soil corer. On the same day as sample collection, fresh soil samples were weighed, scanned, and stored for further analysis in a fridge at 4 °C before the oven-drying process for the LOI and Dumas analysis within 7 days as described below. The scanned samples were first oven-dried at 55 °C temperature for 48 h, followed by a thorough homogenisation for each sample using a soil grinder to pass through a sieve with a 2-mm screen before subsamples were taken for LOI and Dumas analysis [12]. Unlike the NIRS analysis, both LOI and Dumas are regarded as ex-situ methods, which involve measuring collected representative soil samples through dry combustion technique under high temperature [10].

2.3. Near Infrared Spectroscopy (NIRS) Method

A mobile NIRS device (spectral range of 1300 to 2550 nanometres (nm)) was used for SOC estimation (Agrocares, series-E; Wageningen, The Netherlands). The Agrocares NIRS scanner is calibrated on more than 18,000 laboratory-based soil measurements across at least 36 countries [20]. The NIRS scanner consists of a spectrum of infrared energy reflected from the sample illuminated during scanning. As heterogeneity is being considered within soil samples [21], the scanner requires multiple spectrum scans of a sample before moving to the next sample scan [2]. It conducts a pre-scan calibration with a standard white plate before scanning the samples, with four replicate scans performed per sample. Obvious substrates, such as rocks, insects, and plant debris, were removed before NIRS analysis.

2.4. Dumas Method

An automated electronic elemental analyser (Enviro TOC; Elementar, Langenselbold, Germany) was used to measure SOC (as a percentage between 0 and 100%) in grounded and sieved samples. The analyser used solid soil samples at a detection limit of 10 µg/g or 10 ppm in high precision (±<0.1%) under sample digestion above 900 °C and up to a combustion point of 1200 °C with one furnace [9,22]. In-house calibration of the analyser tube fillings was performed in trial solid sample mode, including one blank tin foil and five benchmarked elemental soil samples (5 g, 10 g, 15 g, 20 g, and 30 g) wrapped in tin foil [9]. From the top to bottom layers, the tube fillings adjusted from the calibration consist of a protection tube part, ash finger part, aluminium oxide (AI₂O₃) layer (with a bottom of 10 mm), top quartz wool layer (5 mm height), copper oxide (Cu₂O) layer (85 mm height), middle quartz wool layer (5 mm height), quartz chips (40 mm height), and last quartz wool layer (5 mm height). Prior to SOC measurements, a 10 mg solid soil subsample was taken from the dried sample for soil inorganic carbon or carbonates (SIC) removal using hydrochloric acid (HCI), which is a relatively mild acid for removing SIC effectively before the measurements [1]. Each sample was placed in a silver bowl and underwent SIC acidification to remove SIC using one drop of HCI (0.05 mL) with a laboratory heating panel at mild heat to allow safe SIC evaporation inside a fumehood overnight [1]. The acidified samples were then wrapped in tin foil for the measurements. The analyser then operated for around two hours to obtain a measurement of organic carbon based on carbon dioxide emitted at 950 °C [13].

2.5. Loss-on-Ignition (LOI) Method

A 5 g soil subsample was taken from the original soil sample, dried at 55 °C in a drying oven, and then proceeded to LOI analysis using two carbonite furnaces concurrently at 550 °C for three hours [12]. The percentage of organic matter lost to ignition was calculated as the weight loss between the dried soil temperature (55 °C) and the furnace temperature (550 °C) according to the following Equation (1):

$$\mathrm{LOI} \mathrm{organic} \mathrm{matter} (\%) = {\displaystyle \frac{(Weight 55 °C - Weight 550 °C) }{Weight 55 °C}} \times 100,$$

(1)

The percentage of SOC was calculated by assuming 58% to be organic carbon [23].

2.6. Statistical Analysis

A linear mixed model (Equation (2)) in IBM SPSS Statistics (version 29.0 for Windows, Armonk, NY, USA) was used to assess differences in SOC among measurement methods and land types:

Y_ijk = µ + M_i × LT_j + F_k + e_ijk,

(2)

Y_ijk is the dependent variable of percentage SOC; µ = overall mean; M_i = fixed effect of method (i = LOI, NIRS, and Dumas); LT_j = fixed effect of land type (j = arable, temporary ley, or permanent grass); F_k = random effect of farm (k = farms A to E); e_ijk = random error term. Pearson correlation coefficient (r) was then used to assess the relationships between the measurement methods in SPSS. Outlier values were considered to be 1.5 times greater or less than the upper and lower quartiles, respectively. Significance was attributed at p < 0.05.

3. Results

The SOC measurements for the NIRS, Dumas, and LOI methods ranged from 0.9 to 5.2%, from 1.9 to 14.2%, and from 4.0 to 20.5%, respectively (Figure 2). The range of SOC measurements by NIRS and Dumas methods were generally similar once outliers for the Dumas method were removed, where the outliers were defined as 1.5 times greater than the upper quartile as shown in the figure. Notably, LOI had more outlier measurements (10% of values) compared to Dumas (5%) and NIRS (0%). The mean and median SOC measurements increased from NIRS, Dumas, and LOI. The mean and median values for NIRS and Dumas were similar (2.7 and 2.5% for NIRS and 4.1 and 3.5% for Dumas, respectively), but much lower than LOI (7.7 and 6.6%, respectively). Overall, the LOI method gave higher SOC measurements compared to both the NIRS and Dumas methods. The predicted mean for the NIRS method was lower than the Dumas and LOI methods (p < 0.001;

Table 2. Predicted means for different soil organic carbon (SOC) measurement methods and arable, temporary ley, and permanent grass fields.

Effect		Mean (%)	S.E. **
Land type *	Arable	4.6 a	±0.390
	Temporary ley	4.5 a	±0.379
	Permanent grass	6.6 b	±0.375
	p value	<0.001
Method	LOI	8.1 a	±0.380
	NIRS	3.1 b	±0.380
	Dumas	4.6 c	±0.380
	p value	<0.001
Land type × method
Arable	LOI	6.8 a	±0.483
	NIRS	3.1 b	±0.483
	Dumas	4.0 bc	±0.483
Temporary ley	LOI	6.9 a	±0.433
	NIRS	2.8 b	±0.433
	Dumas	3.8 c	±0.433
Permanent grass	LOI	10.6 a	±0.450
	NIRS	3.4 b	±0.450
	Dumas	5.9 c	±0.450
	p value	<0.001

* Means for agricultural land type and measure within a column and with different superscript letters (i.e., a, b, c) indicating the SOC predicted means differ significantly and are attributed at p < 0.05. ** S.E. = standard error for each mean.

). Among land types, permanent grass fields had the highest predicted mean compared to arable and temporary ley fields. There was an interaction between land type × measurement method (p < 0.001), where the means for the NIRS and Dumas methods were of closer magnitude among arable, temporary ley, and permanent grass fields compared to LOI.

All correlations between the three measurement methods for SOC measurements among all fields were significant and similar (Figure 3, Figure 4 and Figure 5). The correlations between the three methods for the measurements among permanent grass fields were positive and significant compared to arable and temporary ley fields (Figure 6, Figure 7, Figure 8, Figure 9, Figure 10, Figure 11, Figure 12, Figure 13 and Figure 14), where the correlations of arable and temporary ley fields were of similar magnitude. For arable fields, the correlations of the SOC measurements among the Dumas and LOI methods and the LOI and NIRS methods were positive but not strong (Figure 6 and Figure 7), and the Dumas and NIRS methods were negatively and weakly correlated (Figure 8). For temporary fields, the correlation among the Dumas and LOI methods was positive but not strong (Figure 9), followed by the correlation among the LOI and NIRS methods (Figure 10) and among the Dumas and NIRS methods (both positive; Figure 11). For permanent grass fields, the correlation between the Dumas and NIRS methods was positive and the strongest among all three land types (Figure 14), where the correlation between the LOI and NIRS was strong (Figure 13), and the correlation between the Dumas and LOI was positive but the lowest among the land types (Figure 12).

Predicted conversion equations of linear regression were also generated among the three methods according to all the fields and each land type (

Table 3. Linear regression equations for SOC measurements among the Dumas, LOI, and NIRS methods conversion for all fields and land types across the farms studied.

Method Comparison	n *	Land Type	Regression Equations	p Value
SOCDumas vs. SOCLOI	95	All fields	Y = 1.4742 + 0.3423x	<0.001
SOCLOI vs. SOCNIRS	95	All fields	Y = 2.3065 + 2.0322x	<0.001
SOCDumas vs. SOCNIRS	95	All fields	Y = 1.0103 + 1.1665x	<0.001
SOCDumas vs. SOCLOI	23	Arable	Y = 1.544 + 0.3411x	<0.001
SOCLOI vs. SOCNIRS	23	Arable	Y = 4.7323 + 0.6586x	<0.001
SOCDumas vs. SOCNIRS	23	Arable	Y = 4.8309 − 0.3387x	<0.001
SOCDumas vs. SOCLOI	40	Temporary ley	Y = 2.249 + 0.1528x	<0.001
SOCLOI vs. SOCNIRS	40	Temporary ley	Y = 4.8231 + 0.6684x	<0.001
SOCDumas vs. SOCNIRS	40	Temporary ley	Y = 2.9999 + 0.096x	<0.001
SOCDumas vs. SOCLOI	32	Permanent grass	Y = 2.8683 + 0.2531x	<0.001
SOCLOI vs. SOCNIRS	32	Permanent grass	Y = 2.0658 + 2.7403x	<0.001
SOCDumas vs. SOCNIRS	32	Permanent grass	Y = (−0.2478) + 1.9195x	<0.001

* n = number of collected and analysed soil samples.

). There were relatively larger intercept values for the conversion equations between the LOI and NIRS (4.7323 and 4.8231 in arable and temporary ley fields, respectively), and the Dumas and NIRS (4.8309 in temporary ley fields). Such findings suggested that there could be some weight losses on non-SOM content, which will be further presented in the discussion in Section 4.2. The intercept value for the equation between the Dumas and NIRS in permanent grass fields was the lowest (−0.2478) among all the fields and each land type.

4. Discussion

4.1. Estimates of SOC

This current study found similar correlations between the three SOC measurement methods compared for soils from different agricultural land types. The strong correlation between the measurements taken by the Dumas and LOI (r values > 0.5) was similar to those observed in past studies using the same two methods [9,13,24], but differed in higher means (

Table 4. Determination of soil organic carbon (SOC) content according to literature in different land types, soil types, and measurement methods.

Area	Land Type	Soil Type	Method	Mean SOC (%)	Source
Canada	Intercropping land	Sandy loams	LOI 575 °C HCI ** fumigation H2SO3 *** digestion	2.0 NA ## NA ##	[1]
USA	Lawn, turf and parklands	Hapludalfs and Udorthents	LOI 400 °C LOI 550 °C Dumas	3.3 4.0 3.2	[9]
USA	Agronomic trials	Tarboro loamy sand	WB HM * LOI 360 °C Dumas	NA ## NA ## 1.4 0.8	[12]
Xinjiang	Desert land, shrub land, cropland, grassland	Brown soils	WB LOI 375 °C Dumas	NA ## 2.7 <1.5	[13]
Poland	Arable and a wide range of grasslands	Stagnic luvisols, Dystric Skeletic Cambisols, Folic Albic Podzols, and Histic Albic Podzols	LOI 550 °C Dumas	13.4 6.4	[15]
Ecuador	Forest	Volcanic origin soils	LOI 430 °C Dumas	6.3 6.2	[24]
Iran	Cultivated fields of crops, fallow fields, rangelands, and forests	Inceptisols, Entisols, and Aridisols (mostly calcareous soils)	LOI 550 °C LOI 375 °C LOI 360 °C WB Visible-NIRS	2.5 # 1.8 # 1.2 # NA ## NA ##	[25]

* HM is humic matter colorimetry. ** HCI is hydrochloric acid. *** H₂SO₃ is sulfurous acid (sulfuric (IV) acid). ^# Mean SOC (%) derived from 58% of soil organic matter (SOM) content. ^## NA represents not applicable.

). The current study also had higher SOC means, particularly in the LOI (7.7%) under similar sample ignition heat (550 °C in this study; 575 °C in [1]) compared to the means and the drying temperature in Wotherspoon et al. [1], who studied similar soil and land types (2.0% derived from 58% of SOM content by LOI). This suggests that the LOI can combust more than just organic matter contents from soil samples, leading to overestimation of SOC in clayey sandy soils containing particularly low SOC content mainly due to carbon bounding [14], such as Luvisol in the University farm (Table 1), where more moisture can be retained, causing higher water loss during the LOI process [5,15]. This also further implies that the adapted 58% general assumption of SOC from OM content may require adjustment [13]. As the study analysed soil samples across two locations but with different types of soils (clayey sandy loams for the university farm and calcareous soils for the Cotswolds farms) and sampling months in Gloucestershire, UK, it is also important to consider other likely determining factors for the overestimation by the LOI, especially for land type (arable, temporary ley, and permanent grass fields in this study) [3], seasons [19], soil type, and depth [17] for longitudinal SOC studies. Bojko and Kabała (2014) [15] indicated similar SOC results in mixed land types (arable lands and grasslands), where LOI provided a much higher SOC content of 13.36% (in this study it was 9.1%) compared to the Dumas of 6.43% (in this study it was 5.5%), but differed in soil types, which did not contain any carbonates during SOC analysis. Mozaffari et al. (2022) [25] showed a much lower mean SOC content under similar land types by the LOI at 550 °C (2.5%) compared to this present study (8.1%). Given the soil sampling depth is a significant difference between the past study (0–20 cm) and this current study (0–30 cm), it is suggested that soil depth can be one of the influential factors for soil heterogeneity during the LOI process [26], for example, contents of soil nutrients, moisture, pH, and SOM [5,13], particularly in relatively calcareous soil sites.

4.2. Measurement Methods of SOC

On the other hand, most recent studies have only focused on comparing SOC measurements among different ex-situ laboratory methods (WB, LOI, and Dumas) [1,8,12,14,27] but did not use in-situ methods like the NIRS in the study. van der Voort et al. [2] and Kok et al. [20] adopted the NIRS method for SOC estimates with similar farm size, soil types, and land uses as this study but did not consider using any ex-situ methods for cross-validation on the measurements. The authors found that the research on assessing SOC measurements was limited to either in situ or ex situ methods only, without providing a more comprehensive evaluation of the research. As shown in the current study, the LOI method indicated a greater variation in SOC measurements than the NIRS method, suggesting that the SOC values tend to be overestimated by LOI. This overestimation may be due to thermal conductivity variation among different soil types and mass in the LOI analysis during the high-temperature ignition process [10,28]. Furthermore, as the SOC content for LOI was calculated based on the conventional assumption of 58% organic material being organic carbon, this assumption can be influential when estimating SOC contents by LOI among soils with different clay and SOM contents and land uses [23,29].

Several studies comparing SOC measurement methods have tended to investigate soils with higher organic matter content and low clay content [9,11,12,13,24] than in the current study. This can contribute to smaller weight loss with LOI when studying the soils with more clay content [15]. However, the soil samples in the current study contained a high clay content (≥15% overall). This shows that the weight and structural losses from clay fraction, mineral-organic complexes, and other soil components (calcium carbonates) are more likely during the dehydroxylation process of LOI. Such weight losses can cause overestimation on SOMs for higher SOC measures by LOI than by NIRS in general [1,5]. Further, although the NIRS technique allows us to scan and analyse fresh samples safely and instantly with minimal soil structural impact (i.e., gently spread the soil sample in the scan plate) [20], the lower SOC measurements of this technique compared to both the LOI and Dumas techniques are very likely to be the result of diluting the scans through the spectrum reflectance of particles or residues from sandy or even calcareous soils, for instance, small rocks, grass roots, and insects for the lowest SOC means, and the minimum and maximum contents measured among all the measurement methods in terms of land types x methods [30,31].

4.3. Sample Size and Land Types of Soil Samples

In addition, required soil sample masses among all three methods differed for SOC measurements, which may contribute to the result variations among the methods. The NIRS method requires whole fresh soil samples for scanning, while the LOI requires a smaller sample mass of 5 g each, but the Dumas requires a much smaller subsample mass of 10 mg each. As no laboratory intervention is needed during NIRS analysis, there is no significant weight and structural loss from the collected samples. However, whether the small subsample size of the LOI and Dumas being analysed could truly be sufficient as representative of the samples still remains a concern and requires further exploration [8,26]. Further, SOC means from all the discussed studies (ranging from 1.3 to 6.3%) with the different land uses were all below 7% and were similar to those in the current study among arable, temporary ley, and permanent grass fields (ranging from 4.5 to 6.6% by land type). Based on the similar results across land types, the current study reflects that agricultural land use type can have a considerably significant and correlative effect on binding and sequestering more atmospheric carbon for terrestrial carbon sink exploration [7]. As permanent grass fields had much higher SOC contents (6.6%) compared to the arable (4.6%) and temporary ley fields (4.5%; all p < 0.001 by land types), this reflects that the soils in permanent grass fields are less disturbed compared to cultivated arable and temporary ley fields [4]. This also suggests that both arable and temporary ley fields can have a 2% difference, which equates to a 50% increase in organic carbon stock, resulting in soils that more closely mimic soils managed under permanent grassland management practices or adjustments [28], such as mild tillage practice, use of organic fertiliser, and growth of native plant species [6,7].

4.4. Recommendations on Alternative Method for SOC Measurement

Based on the similar correlation results of SOC estimates measured across all the measurement methods, real-time NIRS measurements tend to be more cost-effective, time- and labour-saving, and favourable for the preliminary stage of large-scale land monitoring in collecting benchmarked topsoil SOC estimates compared to both the Dumas and LOI methods in the current study. As NIRS does not necessarily require large sample sizes, sample preparation, or treatment under high temperature in a laboratory as done with LOI and Dumas methods [15], the overall physical soil structure of fresh soil samples can be preserved as much as possible with less variation when rapidly, frequently, and safely assessing preliminary SOC measures in bulk sample quantities at different on-farm locations [32]. Hence, the direct, instant, and multiple scanning nature of using a rechargeable and portable NIRS soil scanner for whole fresh soil samples on or off-site can save more time and labour costs from sample preparation, including drying and sieving [20], and also reduce both physical and chemical hazards from high-temperature furnaces and SIC acidification reagents like HCI involved in LOI and Dumas, respectively [2,8]. In this connection, this study recommends that NIRS can be a safe and practical alternative for large-scale SOC analysis among land types [33], with LOI and Dumas as more expensive methods that can help validate SOC measures using NIRS, especially for unique and even extreme land and soil types, for example, peatland soils, humus-grassland soils, or heavy clay lowland soils [34,35].

Given the current study only analysed the five farms across Gloucestershire, UK, further regional and even national studies using NIRS among different locations, land, and soil types are needed.

5. Conclusions

The current study showed that LOI, Dumas, and NIRS methods for measuring SOC can be used across different agricultural land types. Similar and positive correlations between SOC methods were observed. The LOI method is very likely to overestimate the SOC estimates when compared to the other two methods studied. On the other hand, the NIRS method may not pick up all the organics or inorganics from the soil samples due to obstructions in the samples, such as small rocks, plant debris, and insects, leading to underestimation on the SOC estimates. Hence, NIRS has to be mindfully performed with removal of visual obstructions before sample scanning as much as possible for better SOC measurements. In addition to the methods among land types, the study has also evaluated that the permanent grass fields store higher carbon stock compared to the arable and temporary ley fields, with arable and temporary fields having the potential to increase SOC storage by about 2% on average. New predictive linear regression equations for SOC content (%) conversion among all three methods for all the fields and each land type are generated for on-farm benchmarking use. The correlations of the conversion equations among the methods in permanent grass fields are significant compared to those in arable and temporary ley fields. This study further demonstrates that the real-time NIRS measurement method can provide safe, frequent, and even non-invasive field SOC estimates for large-scale fresh soil sample analysis across farms and land types. The method can serve as an effective and preliminary strategy for collecting benchmarked SOC data that can aid farmers and landowners in making timely agricultural and land practice management decisions towards soil health and SOC enhancement.