The study took place at two sites: i) “Ojos de Agua” forest, of 2,357.62 ha (6°50’50.99”S, 76°27’52.24”W; 382 m a.s.l.; mean annual temperature: 25.0°C; annual precipitation: 1167 mm; mean annual relative humidity: 73%; soil type: Eutric Cambisol), and ii) “El Quinillal” forest, of 10,557.07 ha (7°2’0.00”S, 76°19’52.42”W; 309 m a.s.l.; mean annual temperature: 25.5°C; annual precipitation: 1278 mm; mean annual relative humidity: 74%; soil type: Eutric Leptosol), which is located on the right bank of the Huallaga River (Fig. 1). Both forests are considered tropical dry forests and the tree species ‘Manchinga’ (Brosimum alicastrum (Swartz)) and Quinilla (Manilkara bidentata (A. DC.) Chev.) dominate. In recent years, they are affected by climate change and extensive corn crops (Vallejos-Torres et al., 2021). Both forests had the three plant covers subject of this study: primary forest (trees of approximately 200 years), intervened forest (intervention done approximately 50 years ago), and deforested forest (trees were cut down 10 years ago) (Fig. 1) Both sites were approximately 50 km from each other, and within each site, each plant cover type was 500-1000 m apart from each other.

Soil sampling took place in April 2023. It was carried out with a shovel and a metal bar. Pits were made that were 1 m deep and 1 m wide. Before making the pits, the site was cleaned of weeds and leaf litter. Samples were taken from 0 to 100 cm because several previous studies (Xie et al., 2023; Ryzhova et al., 2023; Zhao et al., 2022), found significant C stock variation up to that depth. To study the vertical variation of SOC at the two sites, we selected a total of 24 plots (10 m × 10 m each), following the methodology of Yu et al. (2019) and distributed in three vegetation covers, these being primary forest, intervened forest, and deforested forest (Table 1). The distance between coverages was 221 m, 1298 m and 2551 m minimum, average and maximum, respectively. In each cover, 4 randomly distributed plots were installed, the distance between them was minimum, average and maximum of 76 m, 350 m and 882 m (Fig. 1). Pits were made for study and sample collection at five soil depths for each plot along 0–20, 20–40, 40–60, 60–80, and 80–100 cm, with a total of 120 soil samples analyzed.

The cylinder (5.5 cm diameter, 5 cm height) method (Blake & Hartge, 1986) was used to estimate soil bulk density (BD) (in g cm^-3), as it follows:

BD: Wd/V…………….................................…..……(1)

where: Wd: weight of the (oven-dried) soil sample (g), and V: sampled soil volume (cm³). SOC concentration was estimated by wet oxidation following the Walkley & Black (1934) method; SOC stocks were calculated as it follows:

SOC (t ha^-1) = OC × De × BD….….….….….….… (2)

where: OC: organic carbon content in the soil (%), De: sampling depth (cm), and BD: bulk density (g cm^-3).

Soil erodibility (K) was estimated following Williams et al. (1984) model, as it follows:

where: SAN: sand content (%); SIL: silt content (%); CLA: clay content (%); SOC: soil organic carbon (%); and =1-SAN/100.

Soil organic carbon density (SOCD) at each soil depth (0–20, 20–40, 40–60, 60–80, and 80–100 cm) was calculated using the following formula:

SOCD = H_ha x BD_ha x SOC_ha x (1-C_ha)/100

where: H_ha: soil thickness (cm), BD_ha: soil bulk density (g cm^-3), SOC: soil organic carbon (t ha^-1), and C_ha: percentage of the soil volume with a fraction >2mm. All values in this formula were transformed to hectares.

All statistical tests were run in R Studio (R Core Team, 2024). The effects of vegetation cover in SOC at different depths were evaluated by analysis of variance (ANOVA) at a significance level of 5%, while means comparison was performed with Tukey's test using the agricolae package in R Studio (Mendiburu, 2010). The categorical variable (cover type) was coded numerically following the methodology established by Yu et al. (2019): 0= Primary forest, 1= Intervened forest, and 2= Deforested forest. The correlation between SOC, vegetation cover type, and soil characteristics was evaluated by the Pearson’s correlation test (α = 0.05), using the base R cor function. To evaluate the response variables that helped explain SOC content at different depths, a principal component analysis (PCA) together with a multiple linear regression (MLR) analysis (Kaiser, 1960) were run.

Soil organic carbon content decreased with increasing soil depth (Fig. 2). The mean SOC content in the primary forest was 36.7±9.5 t ha^-1, for the intervened forest was 29.5±7.6 t ha^-1, and for the deforested forest, it was 22.8±7.6 t ha^-1, decreasing with increasing soil depth. In the superficial soil layer (0-20 cm), SOC content in the primary forest had a value of 79.5±21.3 t ha^-1, while in the deep layer (80-100 cm), it was 11.1±3.5 t ha^-1. The SOC content of the superficial soil layer (0-20 cm) in the intervened forest was 58.5±11.8 t ha^-1 and in the deep layer (80-100 cm) was 12.8±2.4 t ha^-1. Meanwhile, the SOC of the superficial soil layer (0-20 cm) in the deforested forest had a value of 41.8±10 t ha^-1, while in the deep layer (80-100 cm), it was 7.4±5.3t ha^-1 (Fig. 3). The mean SOC value at the superficial (0-20 cm) layer was 59.9±21.3 ha-1 and decreased at greater depth with values of 10.4±4.3 ha^-1 (80-100 cm). With increasing soil depth, it was observed that the coefficient of variation (CV) decreased from 35.6% in the 0-20 cm soil layer to 19.8% in the 80-100 cm soil layer (Fig. 2). The SOC content in the three forest canopies showed significant differences for the first and second soil depths, with the primary forest having the highest SOC. In general, in the first 0-40 cm, the SOC content decreased rapidly with soil depth, showing significant differences between plant covers, but below 40 cm, it did not show significant differences (Fig. 3).

The cover type and SOC had a significant, negative correlation at soil depths from 0 to 40 cm and 60 to 80 cm soil depth (Table 2). Overall, most tested variables were not correlated with SOC for the first 40 cm soil depth, but after this depth, variables like clay, silt, and soil erodilibity presented some significant correlations (Table 2).

The main variables that contributed to explain the SOC in the superficial horizon (0-20 cm) were cover type, soil density, and soil erodibility with 13.8%, 13.4%, and 12.6%, respectively (Fig. 4). In the 20-40 cm horizon, the main variables were sand content, soil erodibility, and cover type with 14.9%, 13.5%, and 11.3%, respectively. In the 40 to 60 cm horizon, the main variables were soil erodibility, silt content, and clay content with 14.6%, 13.4%, and 13.1%, respectively. For the 60 to 80 cm horizon, the main variables were cover type and sand content with 13.2% and 12.9%, respectively. And for the 80 to 100 cm horizon, sand and silt content with 13.7% and 12.2%, respectively. It is generally evident that SOC was mostly explained by soil erodibility, followed by cover type and texture with an average of 12.8%, 11.53%, and 11.08%, respectively. Soil water content presented the lowest contribution, with an average of 7.3 % (Fig. 4).

The SOC obtained in this study was 7.6±5.1 t ha^-1 in the primary forest, 6.2±3.6 t ha^-1 in the intervened forest, and 4.7±2.7 t ha^-1 in the deforested forest when evaluated between 0 to 100 cm depth. SOC density presented significant differences in the layers from 0 to 40 cm, but not from 40 to 100 cm (Fig. 5 A). The highest SOCD content was recorded in the surface layer from 0 to 20 cm with an average of 11.69 t ha^-1, representing 40.4% of the total SOC stock in the entire profile (0–100 cm). In the 20 - 40 cm soil layers, the average SOCD was 7.2 t ha^-1, representing 23.5% of the total SOC stock. In the 40 to 60-cm soil layers, the average SOCD was 5.7 t ha^-1, representing 17.7% of the total SOC stock. In the 60 to 80-cm soil layers, the average SOCD was 3.8 t ha^-1, representing 11.3% of the total SOC stock. In the 80–100 cm soil layers, the mean SOCD was 2.4 t ha^-1, representing 7% of the total SOC stock in the entire 0–100 cm soil profile (Fig. 5 B).

The results found are close to those reported by Solis et al. (2020), who showed 87 t C ha^-1 in a coffee system with Inga trees. Plant cover profoundly affects soil carbon stocks (Arasa-Gisbert et al., 2018) and soil carbon sequestration, more precisely at the 0-15 cm layer (between 33.2 t ha^-1 and 52.7 t ha^-1) (Boulmane et al., 2010). A possible reason for this is the soil’s microbial communities – which differ among plant covers, and their influence in the cycling and accumulation of SOC (Weverka et al., 2023).

Species-rich plant communities are more productive and exhibit more significant long-term SOC content. Soil microorganisms are essential for converting plant organic matter into SOC; consequently, the greater the canopy cover and height, the greater the SOC content (Siswo et al., 2023). Overall, our findings were consistent with many previous studies showing that vegetation increases SOC (Saíz et al., 2012; Gruba et al., 2015). In our study, primary forests present Manchinga (B. alicastrum) and Quinilla (M. bidentata) trees as dominants with a larger diameter at breast height and greater height, while in intervened and deforested forests there are no trees of these species due to the massive extraction of wood, making M. bidentata a threatened species (Vallejos-Torres et al., 2021). The distribution of C components in the soil surface is largely influenced by the chemical nature of the forest litter from which SOC originates. The species-specific litter quality determines the compositional characteristics of organic matter input to the soil and influences the magnitude of decomposition processes by microorganisms. Leaf litter introduces organic materials into the soil in different quantities and qualities, influencing the formation and stability of the soil C reserve. Dissolved organic matter in forest ecosystems significantly affects soil carbon cycling due to litter decomposition (Morffi-Mestre et al., 2023); therefore, litter-derived dissolved organic carbon is considered an important source of stabilized C in soil (Preusser et al., 2021). Vegetation type and soil depth affect soil carbon distribution by changing soil physical and chemical properties and microbial activity, as shown in previous research (Song et al., 2016). Soil organic carbon content was generally low in deforested forests with weeds due to sparse vegetation, shallower root systems, and lower root biomass, especially evidenced by fewer roots in deeper soils (Jia et al., 2017).

Our land-use results showed a significant correlation at soil depths from 0-40 cm and from 60-80 cm depth and of soil erodibility with SOC between 0-80 cm depth, similar to the results found by Yu et al. (2019). The lower soil erodibility in the primary forest can be attributed to the greater amount of soil organic matter (SOM) compared to intervened and deforested forests. Land use type can affect soil properties and plant community characteristics, likely affecting soil erodibility (Chen et al., 2023). In turn, soil erodibility is related to the granulometry, structure, and stability of soil aggregates, which indicates that soils with higher silt content are more erodible than clay soils. Meanwhile, SOM also plays a vital role in soil erodibility by maintaining the stability of soil aggregates (Deng et al., 2018). In our results, the clay fraction did not show a significant correlation with SOC. At the same time, there was a negative correlation of SOC with the sand fraction, results consistent with Zhong et al. (2018). The physicochemical properties of soil are interrelated and affected by land use and management activities (Thabit et al., 2023). Primary forests influence carbon contents due to the large amounts of organic matter associated with the thickest soil fraction.

Soil organic carbon is affected by many related factors and is regulated in complex ways, with spatial differences both higher and lower in the soil column. Land use and soil factors significantly influenced SOC content. In this study, the effects of soil water content on SOC content also decreased with soil depth, influenced by the vegetation cover (Wang et al., 2022). The physicochemical properties of top soils and deep soils were significantly different. Therefore, the regulatory factors and mechanism of SOC varied between soil layers. Generally, the accumulation of SOC on the surface results from interactions between abiotic processes regulated by environmental factors and biotic processes regulated by microbes. Because the surface soil contains a large amount of plant litter, sufficient water and air on the surface are also conducive to increase soil microbial activity (Zhang et al., 2021). The SOC changed significantly in the different forest types. This can be attributed to the increase in organic materials (litter and roots), as Xing et al. (2023) indicated that the highest SOC content found in the first soil horizons in primary forests is due to the presence of roots.

In this study we found a significant negative correlation between soil bulk density and SOC, indicating that bulk density does not influence the leaching of surface SOC to deep soil layers since low SOC in deep soils is related to the high density in tropical forests (Yang et al., 2016). In our study we found some contrasting results with Jia et al. (2017), who investigated carbon stocks in different vegetation covers in deep soils and indicated that land use significantly affects deep SOC. In addition to land use type, soil factors also significantly impact the vertical distribution of SOC (Jiang et al., 2017). Soil moisture content positively correlated with SOC; therefore, SOC decomposition and soil C content were associated with changes in soil environment and soil microbial biomass due to soil moisture variation.

In our study, 37.80% of SOCD and 40.37% of SOC were concentrated between 0-20 cm soil depth, and this decreased with soil depth in such a way that 7.88% of SOCD and 7.05% of SOC occurred in the 80-100 cm depth profile in the forests studied. This fraction distribution is controlled by factors such as humidity and bulk density along soil depths (Zhuo et al., 2022). It is known that soils with high SOCD contain a high accumulation of organic matter (Arunrat et al., 2020) and, therefore, high carbon content, as found in our study. The vegetation cover type also influenced SOCD; primary forests with dense vegetation cover have the highest SOCD, followed by intervened forests, and forests deforested with cultivars (Zhu et al., 2021). Studies have shown that an increase in vegetation cover, such as in a primary forest, could facilitate carbon accumulation in the soil and, therefore, increase SOCD (Gong et al., 2017). This can be attributed to (i) an increase in plant root productivity, (ii) a reduction in SOC loss by effectively blocking wind erosion, and (iii) the accumulation of litter on the soil surface. Generally, the increase in vegetation is followed by an increase in litter (Tian et al., 2022). Yu et al. (2019) found that the SOC content of each vegetation type ranged in the following order: forest land, cropland, and grassland. This is corroborated by Zhao et al. (2019), who found that humidity is one of the most critical factors that control the variations of SOCD studied between 0 cm to 100 cm soil depth.

In summary, this study of vertical variation of carbon at depths of 0-100 cm showed that vegetation cover significantly affects soil carbon stocks, more significantly at the 0-20 cm layer. Primary forests present Manchinga (B. alicastrum) and Quinilla (M. bidentata) trees as dominants with a larger diameter at chest height and greater height, while in the intervened and deforested forests, there are no trees of these species. The lower soil erodibility found in the primary forest can be attributed to a greater soil organic matter (SOM) content compared to intervened and deforested forests. Land use affects soil properties and plant community characteristics, which are likely to affect soil erodibility. In this study, the effects of soil water content on SOC content also decreased with soil depth, influenced by the vegetation cover index and soil bulk density. The types of vegetation cover also influenced SOC density. Above all, primary forests with dense vegetation cover have the highest SOC density, followed by intervened forests and forests deforested with cultivars.