The Noun plain, in the western region of Cameroon, lies between latitudes 5˚10' to 5˚40' N and longitudes 10˚20' to 10˚50' E (Figure 1). It covers an area of approximately 2554 km² with a strip of 10 - 20 km on either side of the River Noun, with an eastward extension towards the Bamouns [37] .

The plain belongs to the tropical Sudano-Guinean climatic region of the “mountain monsoon” type, but is slightly modified by the double orographic protection of its plateau [38] ; it receives an average annual rainfall of 1983 mm characterised by a unimodal pattern, with a peak in September. The topography of the region varies considerably, ranging from depressions less than 900 m above sea level to high mountains at 2250 m above sea level. Two geological formations make up this area: the Precambrian basement “granite-gneissic”, which is the bedrock, made up of metamorphic rocks and plutonic magmatic rocks, and the recent volcanic formations, made up of basaltic plateau and pyroclastic deposits (ash, lapillis, volcanic bombs) [38] .

Multiple field trips carried out between 2019 and 2022 revealed that the Noun plain is characterised by three main categories of land uses namely; natural formations, crop fields and agroforests. Aerial photographs and satellite images of the study area were used to identify and classify the most common land uses currently found in the plain. The different land use types are briefly described below.

The natural formations are made up of remnants of deciduous seed forests that remain on flooded plains, savannahs that are mainly found at the top of

mountains, at high altitudes and in very hilly areas, and mixed woody plant formations (fallow land that is more than twenty years old). These natural formations are often reserved for cattle breeders in the Noun plain. The dominant species in this group include Ceiba pentandra, Milicia excelsa, Khaya anthotheca, Lovoa trichilioides, Mitragyna ciliate and others. This type of land use covers an average of 172˚177.57 ha (71.01%) of the total area of the landscape.

Agroforests are made up of perennial crops such as coffee (Coffea arabica), cocoa (Theobroma cacao), orchards and palm trees (Raphia farinifera), although these are increasingly in decline. In the 1930s and 1940s, coffee was the main cash crop [39] , grown on two thirds of the plain. This type of land use covers an average of 24˚113.88 ha (9.95%) of the total area of the landscape. The change in the land use system is clearly visible, with a reduction in coffee-growing areas in favour of annual crops, resulting in a management style characterised by diversification [40] .

Crop fields in the Noun plain are characterised by a crop rotation of two annual crop cycles practised in almost the entire study area throughout the year. A maize cycle from mid-March to mid-July and a bean cycle from mid-August to mid-December. This land use type covers around 43˚391.80 ha (17.90%) of the total area of the Noun plain, and includes areas used for rain-fed and irrigated farming, including rural farms. The main crops are cereals—mainly maize (Zea mays), pulses—mainly beans (Phaseolus vulgaris) and groundnuts (Arachis hypogaea) and oilseeds—mainly soybean (Glycine max). The use of chemical fertilizers is practically indispensable on the plains. However, some farmers combine it with animal droppings and manure. The land is usually prepared manually through conventional tillage and mechanically using various machines.

Three study sites were selected using a stratified random sampling technique. Each stratum represents a reference soil group. The representative study sites selected include the Mangoum series (Andosols), the Makeka 1 series (Ferralsols) and the Fossang series (Acrisols) (Table 1).

In the three sites, nine (09) soil profiles were dug and described following standard procedures [41] . A representative soil profile of dimensions 1.5 m × 1 m × 2 m was dug on each type of land use selected [42] . The Dutch knife and auger were used to collect soil samples at three depth intervals (0 - 20 cm; 20 - 40 cm; and 40 - 60 cm). Additionally, undisturbed soil samples were collected using stainless steel Kopecky rings for bulk density determination. A total of 27 soil samples were collected.

The soil samples were air-dried and sieved through a 2 mm sieve for routine laboratory analysis. Particle size distribution was determined by the hydrometer method; bulk density (BD) was measured by the core method; total porosity was

calculated using a mathematical relationship between bulk density and particle density (PD) (Porosity = (1 – BD/PD) × 100, where PD = 2.65). Organic carbon was determined using the Walkley and Black method. All parameters were determined according to [43] . The method used to calculate soil carbon stocks involves measuring the total organic carbon content at different soil depths and transforming these data, taking into account the soil bulk density and coarse element load. As a first step, we determined the organic carbon stock at each depth using the following formula:

SOCS ( t / ha ) = soil   depth × BD × Cconc . × ( 1 − gravel ) (1)

These stock values were used to assess the variation in SOCS with depth.

Where,

SOCS = soil organic carbon stock for each profile representative of the type of land use and soil studied at a given point (expressed in tons C/ha);

Cconc. = soil organic carbon content;

Gravel = volume percentage of coarse fragments (>2 mm).

Statistical analyses were performed using R software version 3.5.1 and Microsoft Excel software 2013 edition for Windows. Data obtained in the laboratory were subjected to an analysis of variance (ANOVA) using the General Linear Model procedure. Means were separated using the Student-Newman Keuls test, based on the calculation of the least significant difference (LSD) at the 5% probability threshold. To ensure that the results obtained follow a normal distribution law, the Shapiro-Wilks normality test was performed, drawing on the work of [44] . To appreciate the extent of variability of the soil organic carbon stocks, a coefficient of variation (CV) was calculated using the formula:

CV = σ / X ¯ × 100 (2)

where, σ: Standard deviation; X ¯ = arithmetic mean. The CV (%) was assessed according to [45] . Based on this, CV values < 15% were considered as least variable; those with CV between 15 and 35% were grouped as moderately variable and those with CV values > 35% indicated high variability.

Soil type and land use are controlling factors of soil organic carbon stocks. Table 4 presents the results of the analysis of variance of the interactive effect of soil type and land use on SOCS. These results show that the combined effect of land

DF: Degree of freedom; CV: coefficient of variation; ST: soil type; LU: land uses; D: depth.

use and soil type did not reveal any significant difference (p < 0.177). Comparison of soil type mean values shows that the highest organic carbon stock was obtained in Andosols (52.0 tC/ha) and the lowest in Ferralsols (46.9 tC/ha). However, this difference was not significant. For the different land use types, the separation of the means indicates that the SOCS is significantly higher in natural formations (58.9 tC/ha) compared to agroforests (49.6 tC/ha) and crop fields (42.0 tC/ha). Figure 2 and Figure 3 show the variations of SOCS with soil type and land use, respectively.

SOCS decreased with depth in all land use types (Table 4). The results show that the organic carbon stock decreased with depth to 33.3 tC/ha in the 40 - 60 cm interval. Analysis of variance revealed a significant difference in SOCS between depths on the one hand (p < 0.00018) and between land-use types on the other hand (p < 0.0045). However, the interactive effect of land use and soil depth on SOC stock was not significant (p < 0.4072). At all three depths studied, SOCS of natural formations was higher than that of agroforests and cultivated fields.

Variations in soil parameters, such as the OC stock in an environment, are influenced by the distribution and interaction of soil-forming factors such as climate, organisms, topography, parent materials and time. The degree of stabilization of SOC also depends on uncontrollable factors, including coarse fragments. The three groups of soils studied are characterized by varying proportions of coarse fragments. The results show that the content of mineral rejections > 2 mm, which is used to calculate the SOCS of fine soil, is higher in Andosols (51.73%). On the other hand, in Acrisols and Ferralsols, it is relatively low (<4%) throughout the Noun plain. On all the data, we obtained a negative and significant correlation at the 0.01% threshold between coarse fragments and the other parameters of the study, namely: fine elements (r = −0.620), apparent density (r = −0.554) and organic carbon stock (r = −0.470). Figure 4 shows the relationship between soil organic carbon stock and coarse elements for the three soil types studied. We note from the different coefficients of determination that the correlation is strong and significant in Andosols (R² > 0.5; r = −0.79; p = 0.0022), but weak and non-significant in Ferralsols and Acrisols (R² < 0.5).

There was no significant difference between SOCS for the three soil types. This result is similar to the findings of Canedoli et al. [25] in western Italy, where no significant differences were observed between the five soil types studied (Cambisols, Leptosols, Podzols, Regosols and Umbrisols) in the Gran Paradiso National Park. However, a study conducted on three soil types (limestone soil, rice paddy soil and yellow soil) in karst regions illustrates the influence of soil type in the variation of SOCS [46] . It was reported that, this difference stems from the fact that the soil formation conditions and processes of the three soil types in the study area were clearly different. A study carried out in southern Benin in a subequatorial climate by Kooke et al. [24] corroborates the above, but of the three soil types studied, hydromorphic soils store significantly more carbon than ferralitic soils and ferruginous soils, i.e. 14.14 tC/ha, 7.45 tC/ha and 5.28 tC/ha, respectively.

The results illustrate a high variability (CV > 30%) of SOCS according to each soil type in the Noun plain. This result suggests that there is considerable heterogeneity within each soil type, probably due to their inherent potential. For Kögel-Knabner and Amelung [3] , these multiple differences are underpinned by factors such as the formation of reactive mineral surfaces and soil aggregation. These results are in agreement with those of Munoz-Rojas et al. [47] and Chevallier et al. [48] carried out in southern Spain and Costa Rica, respectively, where the coefficients of variation are well above 30% within the different soil types studied.

Acrisols and Ferralsols are both formed on granito-gneiss with higher levels of fine elements than Andosols formed on very recent pyroclastic deposits and rich in coarse fragments. In fact, young soils formed on recent volcanic deposits on the plain have low SOCS due to the low quantities of secondary minerals they contain, which gives them a low capacity for adsorption and stabilisation of organic matter. This finding is essentially similar to that made by Munoz-Rojas et al. [47] with coarse-textured arenosols, which showed very low carbon stocks compared to what should be. The same observation was made on three types of soil in the Amazon rainforest, where soils with a high sand content had significantly low SOCS [22] . It should be noted that Andosols developed on tuffs and volcanic ash store more soil organic carbon because they are rich in allophanes, which contribute to the stabilisation of organic matter, than Andosols developed on recent pyroclastic deposits [3] [48] . According to Abera and Wolde-Meskel [49] , the presence of factors such as high fractions of micro-aggregates can provide physical protection for soil organic matter and thus contribute to the slow rate of organic carbon loss.

A number of studies have already been carried out on the relationship between fine element content and soil carbon content. For example, the work of Pallo et al. [27] in southern Burkina Faso showed a significant relationship between carbon stock and particles (clay, clay/silt and sand/clay). To date, very little research has focused on the interdependence between SOCS and the coarse fragments used to calculate these stocks. The work carried out in the Noun plain indicates that, overall, there is a negative and highly significant relationship (r = −0.470**) between coarse fragments and SOCS. Results corroborates with those obtained by Zhang et al. [46] in the mountainous karst basin in China, where the correlation between gravel content and SOCS revealed a significant difference at the 0.01 threshold with a correlation coefficient greater than 50% (r = −0.61**). This is also the case for the results of the work carried out by Pallo et al. [27] between coarse fragments and SOC content, where the coefficient of determination admittedly showed a weak interdependence (R² = 0.26**), but it was still very significant.

Another highly revealing fact from the results of this study carried out in the Noun plain is that the coarse fragment content of Andosols is greater than 50% and the interdependence with carbon stock is highly significant (R²= −0.67**). On the other hand, for Ferralsols and Acrisols, the proportion of coarse fragments is less than 5% and the interdependence is low and not significant. This result is in line with that of Tsozué et al. [31] where the soils studied are characterised by gravel contents of less than 3% and which did not reveal any significant difference with the other parameters of the study. However, other works reveal a significant correlation at the 0.01% threshold between SOCS and coarse fragments with a content greater than 10% [50] . Coarse fragments in a soil influence the storage of organic carbon in that soil by reducing the useful storage volume of the soil [25] .

SOCS in the Noun plain showed a reduction with depth in all nine profiles characterised for this study. The highest average carbon stock was obtained at a depth of 0 - 20 cm (66.3 tC/ha) and the lowest at a depth of 40 - 60 cm (33.3 tC/ha). The result corroborates those reported by Ali et al. [51] at the same depth with the same intervals (20 cm) and that of [50] in north-eastern China. This decrease in OC stock from surface to depth is also in line with the work of Saha et al. [52] , who showed that the variation in total SOCS at different depths is significantly greater at the surface (0 - 15 cm). Similar results have been reported in Egypt [47] , in a subhumid climate in Algeria [53] , in the upper Magou basin in Benin [15] , and on Mount Bambouto in West Cameroon [31] , where SOCS decreases with soil depth. Organic carbon in the first metre of soil plays a predominant and crucial role in soil functioning, protection, conservation and sustainable management [54] [55] , especially when it is bound to fine soil particles. According to the research results of Atchada et al. [15] on the influence of land use patterns on soil organic carbon stocks, there is a strong relationship between the content of fine elements and the soil organic carbon stock according to the four types of land use in their study. They conclude that an increase in the proportion of fine elements favours an increase in organic matter content.

Land use is an important factor controlling the storage of OC in soils, as it affects the quantity and quality of litter input, the rate of litter decomposition and the stabilisation processes of organic matter in soils. The results show that the distribution of SOCS depends on the type of land use (Table 4 and Figure 2). In fact, the highest SOCS was obtained in natural formations (58.9 tC/ha) and the lowest stock was obtained in crop fields (42.0 tC/ha). These results are in agreement with those reported in North-West India [52] , in North-East Pakistan [51] , in Yunnan Province in China [56] , in the Cerrado of Brazil [57] , and in the mountainous area of North-West Cameroon [6] . This significant variation in SOCS with respect to land use type observed in the Noun agricultural plain is similar to the findings reported by De Blécourt et al. [16] in a study on the impact of land use change in Namibia and Zambia.

Figure 2 shows that, regardless of soil type, SOCS in natural formations and agroforest are higher than those in crop fields. This is due to the presence of litter produced by the trees, which is permanently recycled and contributes to the storage of SOC in the first few centimetres of soil. In addition, anthropogenic activities of an agricultural nature in the Noun plain contribute to the continuous removal of crop residues from surface soils, either by export or by burning, which could have led to a lower accumulation of OC in the surface soils of cultivated land. This is in line with the work of Gmach et al. [57] , whose study focused on assessing different land uses in the Cerrado that could restore soil organic matter levels. This observation was also made by Tegha et al. [58] , who found in their results that SOC content and stocks were highest in forest land and lowest in arable land, and asserted that converting forest land to small-scale farming increases soil mineralisation, which leads to a reduction in SOC and, consequently, soil degradation. This degradation is the result of tillage, which weakens soil aggregates and thus causes mineralization of soil organic carbon [33] [59] .